Дисертації з теми "Predicting Innovation"

Disruptive Innovation theory explains how industry entrants can defeat established firms and quickly gain a significant share of their key markets, in spite of the fact that incumbents tend to be significantly more experienced and better resourced. The theory has been criticized for being underspecified: whilst the general mechanics of the phenomenon of disruptive innovation are clear, it has not been established which individual variables are essential to the process and which ones are merely ancillary. As a consequence, to date it has not been possible to build a predictive model on the basis of the theory managers can use to assess the disruptive potential of their own and their competitors' innovation projects. In this research project the predictive power of each of the main variables that are mentioned in the literature has been assessed on the basis of a qualitative analysis of five real world case studies. Only variables for which information can be collected using publicly available data before disruption happens have been retained. By clarifying the detail of disruptive innovation theory, this study has been able to address a key issue in the debate, namely, whether products that are more expensive and more complex than the market standard can ever be classified as 'disruptive innovations' or whether they should always be regarded as 'high-end anomalies'. In this study two distinct disruptive innovation strategies have been identified based on the current phase of the product life cycle, the current focus of mainstream demand and the market segments first targeted when coming to market. The first strategy entails growing an existing market by moving the focus of demand on to a secondary market driver as soon as customers begin to lose their willingness to pay a premium for upgrades in the performance areas they historically used to value. Early on in the product life cycle, disruptors can conquer the mainstream market 'from above' with products that are different and more reliable or more convenient but not simpler or cheaper. The second strategy creates a new separate market by offering a radically new type of additional functionality. Over time the new market replaces the old market. These products are likely to be expensive because of their small production run and difficult to use because they are the first models of their kind. High-end customers constitute a natural foothold market for these products as they are wealthy and highly skilled.

Instructional technologies can be effective tools to foster student engagement, but university faculty may be reluctant to integrate innovative and evidence-based modern learning technologies into instruction. It is important to identify the factors that influence faculty adoption of instructional technologies in the teaching and learning process. Based on Rogers' diffusion of innovation theory, this quantitative, nonexperimental, one-shot cross-sectional survey determined what attributes of innovation (relative advantage, compatibility, complexity, trialability, and observability) predict the probability of faculty adopting the audience response system (ARS) into instruction. The sample for the study consisted of 201 faculty who have current teaching appointments at a university in the southeastern United States. Binary logistic regression analysis was used to determine the attributes of innovation that predict the probability of faculty adopting the ARS into instruction. The data indicated that the attributes of compatibility and trialability significantly predicted faculty adoption of ARS into instruction. Based on the results of the study, a professional development project that includes 3 full days of training and experiential learning was designed to assist faculty in adopting ARS into instruction. Because the current study only included the faculty at a single local university, future studies are recommended to explore a more holistic view of the problem from different institutions and from other stakeholders who may contribute to the process of instructional technology adoption. The project not only contributes to solving the local problem in ARS adoption, but it is also instrumental in promoting positive social change by fostering evidence-based teaching strategies and innovations that maximize student learning.

Knowing what to sell, when to sell, and to whom to sell is essential buyer behavior insight to allocate scarce marketing resources efficiently and effectively. Applying the theory of relationship marketing (Morgan and Hunt 1994), this study seeks to investigate the link between commitment and trust and firm adoption of radically innovative information technology (IT). The construct of radical innovation is operationalized through the use of cloud computing. A review of the vast scholarly literature on radical innovation diffusion and adoption, and modeling techniques used to analyze buyer behavior is followed by empirical estimation of each of the radical innovation adoption questions of purchase timing, brand choice, and purchase amount. Then, the inefficiencies in the independent model process are highlighted, suggesting the need for an integrated model. Next, an integrated model is developed to link the purchase timing, brand choice, and purchase amount decisions. The essay concludes with insight for marketing practitioners on the strength of the factors of commitment and trust on adoption of radical innovation, an improved methodology for the business-to-business marketing literature, and potential further research paths.

The purpose of the current study was to examine the level of support for innovation and individual creativity as potential mediators of relationship between managerial task performance, managerial contextual performance, organizational financial performance and employee performance in non-profit sport organizations in Turkey. For the purpose of the study, 721 volunteer managers and employees from 21 Department/School of Physical Education and Sport (D-SPES) and 23 Province Directorates of Youth and Sport (PDYS) were participated in the study. Individual Creativity Scale, Support for Innovation Scale, Managerial Task Performance Scale, Managerial Contextual Performance Scale, Organizational Financial Performance Scale, and Employee Performance Scale were used for data collection. Results of the Structural Equation Modeling (SEM) analysis revealed that the model adequately describes the data for the sample and the fit indices were all within the acceptable thresholds. The model accounted for 68% variance in support for innovation, 0.7% variance in individual creativity and 44% variance in employee performance. These results suggested that support for innovation and individual creativity significantly mediate the effects of managerial task performance, managerial contextual performance and organizational financial performance on employee performance.

This study examines the deployment of Wireless E9-1-1 Phase One and Wireless E9-1-1 Phase Two as a diffusion of innovation. The research method used in this study is a cross-sectional study employing secondary data in a discriminant function analysis. The study population is Virginia units of local governments (95 counties and 39 cities) that had not deployed Wireless E9-1-1 Phase One or Wireless E9-1-1 Phase Two as of January 1, 2001. The period of time included in this study is from 2001 to 2006. The purpose of the study is to assess the overall accuracy of the three principle theories of policy innovation adoption: diffusion, internal determinants, and unified theory, which are variations of the fundamental diffusion theory, in predicting the deployment of wireless E9-1-1 by Virginia units of local government. This assessment was conducted by identifying Virginia specific variables from models associated with these policy innovation theories to determine the best performing model for the deployment of Wireless E9-1-1 throughout the Commonwealth of Virginia. The Virginia specific variables utilized in this study are: Wealth, Population, Fiscal Health, Dedicated Funding, Financial Dependency, Urbanization, Regionalism, and Proximity to Interstate. Dedicated Funding and Regionalism had the largest absolute size of correlation among the predictor variables for the deployment of Wireless E9-1-1 Phase One and Wireless E9-1-1 Phase Two, thus generating the best performing model. This information will provide the basis from which to develop a statewide comprehensive policy and plan for Next Generation 9-1-1 and will help provide an answer to the question of when and how governments get involved in designing and implementing a 9-1-1 emergency service network.

In recent decades, the use of power converters has become very popular in the field of electric drives. Several control techniques have been proposed for power converters and every year, the ongoing research and the always more powerful microprocessors, lead to new high performance solutions. Despite this, since the output of the worldwide research often results in complex and hardly applicable solutions, other well-established techniques, such as linear and hysteresis control with pulsewidth modulation, are still the main choice in a great number of industrial applications. The reasons of their leadership can be found considering the characteristics of these methods: on one side simplicity of comprehension and implementation and, on the other, sufficiently good performance and robustness. Due to these relevant features, despite there is still extensive room for improvements, it is not painless to propose solutions that can be attractive for people working in industry to compete with, and possibly to replace, traditional methods. Desirably, a control algorithm for electric drives has to be simple and easily understandable. Besides, it has to be suitable for real-time applications. Robustness and reliability, beyond that performance, have to be guaranteed since the nature of the different applications, e.g. home appliances and automotive. In this perspective, Predictive Control could represent a candidate to introduce improvements and gains in the aforementioned industrial applications. Predictive control is a wide class of controllers that uses the model of the system for the prediction of the future behavior of the controlled variables. This information is used by the regulator in order to obtain the optimal actuation, according to a predefined optimization criterion represented by a cost function. This control techique is based on concepts that are extremely simple and intuitive and besides, depending on the type of predictive control, the implementation can also be simple. In parcticular, Finite Control Set allows considering the discrete nature of the power converter and results in an extremely simple implementation. Beyond simplicity, other advantages can be recognized. First, with predictive control it is possible to avoid the cascaded structure obtaining a very fast transient response. Besides, nonlinearities can be included in the model avoiding the need of linearizing the model for a given operating point and improving the operation of the system for all conditions. Finally, it is possible to include limitations of the variables when designing the regulator. The aim of this thesis is to study Predictive Control applied to the current control of synchronous reluctance machines, analysing and addressing some open research topics regarding this kind of control. In particular, two main aspects are studied, namely the need of a precise knowledge of the machine model and the possibility to drive a synchronous reluctance machine along the Maximum Torque per Ampere, the Flux Weakening and the Maximum Torque per Voltage operations. The performance are strictly related to the accuracy of the model used for the prediction. In case of parameters mismatch or variation, rather than other model inadequacies, the prediction could be affected causing a worsening of the overall behaviour of the drive. The first part of this work is commited to study this aspect, analysing the effects of mismatches and variations focusing in particular on the detrimental effects of iron saturation. A novel model-free solution is presented to overcome the limitations given by an inadequate model. This method allows achieving good reference tracking and limited current ripple in every working condition. Besides, it presents great advantages in terms of simplicity: no additional hardware and no complicated calculations are required. The design is effortless since there are no gains, thresholds and so on, that have to be tuned. This technique could be used to develop an universal drive, meaning that completely different machines could be controlled with exactely the same algorithm, without self commissioning or identification procedures. Thanks to the aforementioned features, this techique could allow the spread of predictive control in industrial applications. In order to fully exploit the characteristics of the drive while assuring the lowest power losses in every working condition, a proper control algorithm has to be used. In the second part of this work, a predictive regulator able to track the most suitable trajectory depending on the machine operation is presented. In particular, the Maximum Torque per Ampere, the Flux Weakening and the Maximum Torque per Voltage trajectories are considered. The proposal is a combination of predictive control and hysteresis control, since its aim is to keep the current error within a certain hysteresis band, and it allows combining the benefits of the two control techniques. This study is carried out considering Predictive Current Control for Synchronous Reluctance machines. This kind of machine has been considered since it is of great interest due to the fact that it features high power density, superior reliability, high efficiency and it is cost effective due to the absence of permanent magnets and circuits in the rotor. Besides, since its significant iron saturation, its control represents a challenge (in particular) for predictive control schemes and for this reason it is a perfect case study.
Negli ultimi anni, l'utilizzo di convertitori di potenza in applicazioni di azionamenti elettrici è diventato molto diffuso. Diverse tecniche di controllo per convertitori di potenza sono state proposte e ogni anno, i risultati della ricerca e gli sviluppi di microprocessori consentono di raggiungere performance sempre maggiori. Nonostante ciò, poichè gli output della ricerca sono spesso soluzioni complesse e di difficile implementazione, le soluzioni più usate in ambito industriale rimangono quelle ormai consolidate, come il controllo lineare ed il controllo ad isteresi. Un algoritmo di controllo per un azionamento elettrico dovrebbe essere semplice e di facile compresione. Inoltre dev'essere adatto ad applicazioni real-time. Robustezza ed affidabilità, oltre che alle performance, devono essere garantite, in particolare in applicazioni come gli elettrodomestici e l'automotive. Alla luce di ciò, il Controllo Predittivo rappresenta un valido candidato per introdurre vantaggi e miglioramenti in ambito industriale. Questa tecnica di controllo sfrutta un modello del sistema per predire il comportamento futuro delle variabili controllate. Questa informazione è utilizzata per scegliere l'azione di controllo migliore in base ad un criterio di ottimalità predefinito. Questo tipo di controllo è basato su idee che sono concettualmente semplici e intuitivi. Inoltre, l'implementazione della versione Finite Set risulta particolarmente facile. Oltre alla semplicitò gli altri vantaggi sono la possibilità di evitare la struttura in cascata (tipica del controllo lineare), le nonlinearità e le limitazioni possono essere direttamente incluse nel modello. Lo scopo di questa tesi è di studiare il controllo predittivo applicato al controllo di corrente di una macchina Sincrona a Riluttanza, analizzando ed affrontando alcune tematiche ancora aperte. In particolare, due aspetti sono considerati: la necessità di conoscere in modo preciso il modello della macchina e la possibilità di controllare la macchina lungo le traiettorie di MTPA, Flux-Weakening e MTPV.

Biomechanical compression studies at different temperatures were conducted to correlate the biomechanical response of raw bovine Longissimus dorsi muscles varying in USDA Quality Grade with overall sensory tenderness scores. Phase 1 assessed the biomechanical properties of raw 2.54 cm3 samples obtained with a Texture Analyzer fitted with a 10 cm diameter platen which applied a constant strain of 3% for four minutes. Muscle specimens were arranged with fibers in parallel and perpendicular orientations to the applied force and tested at 0, 2, 4, 6, 8 and 10°C. Initial stiffness, final stiffness and energy dissipated of raw steak cubes with fiber orientation in parallel and perpendicular fiber orientations were calculated using the models and technique of Spadaro (1996) and correlated to overall sensory tenderness scores for each compression temperature. All compression values had higher correlation coefficients with overall sensory tenderness than did Warner-Bratzler Shear Force (WBSF). Of the prediction equations developed, it was concluded that samples compressed perpendicularly at 2ºC were better predictors of overall sensory tenderness (R2 = 0.77) than WBSF (R2 = 0.11). Phase 2 assessed the biomechanical properties of raw steaks (2.54 cm thick) using a 2 mm diameter stainless steel probe in lieu of the platen and compressing samples 0.635 cm for 0.25 sec at -6.6, 4.4 or 10°C. Initial stiffness (ISTFPR), final stiffness (FSTFPR) and energy dissipated (EDPR) of raw intact steaks were calculated using a modification of the models and technique of Spadaro (1996) and correlated to overall sensory tenderness scores for each compression temperature. ISTFPR, FSTPF and EDPR values regressed against overall sensory tenderness produced higher R-square values (R2 = 0.71 at 4.4ºC and R2 = 0.70 at 10ºC) than prediction equations using WBSF (R2 = 0.65). The significance of this study was that sensory tenderness could be predicted rapidly and more accurately on intact raw loin samples using a nondestructive probe measurement than could be predicted with WBSF. This innovative technique could potentially be used as a selection tool to ensure beef tenderness, be integrated into an on-line USDA Quality Grading system and be utilized as a powerful non-destructive research technique.

Organic ?-conjugated solids are promising candidates for new optoelectronic materials. The large body of evidence points at their advantageous properties such as high charge-carrier mobility, large nonlinear polarizability, mechanical flexibility, simple and low cost fabrication and superior luminescence. They can be used as nonlinear optical (NLO) materials with large two-photon absorption (2PA) and as electronic components capable of generating nonlinear neutral (excitonic) and charged (polaronic) excitations. In this work, we investigate the appropriate theoretical methods used for the (a) prediction of 2PA properties for rational design of organic materials with improved NLO properties, and (b) understanding of the essential electronic excitations controlling the energy-transfer and charge-transport properties in organic optoelectronics. Accurate prediction of these electro-optical properties is helpful for structure-activity relationships useful for technological innovations. In Chapter 1 we emphasize on the potential use of the organic materials for these two applications. The 2PA process is advantageous over one-photon absorption for deep-tissue fluorescence microscopy, photodynamic therapy, microfabrication and optical data storage owing to the three-dimensional spatial selectivity and improved penetration depth in the absorbing or scattering media. The design of the NLO materials with large 2PA cross-sections may reduce the optical damage due to the use of the high intensity laser beams for excitation. The organic molecules also possess self-localized excited states which can decay radiatively or nonradiatively to form excitonic states. This suggests the use of these materials in the electroluminescent devices such as light-emitting diodes and photovoltaic cells through the processes of exciton formation or dissociation, respectively. It is therefore necessary to understand ultrafast relaxation processes required in understanding the interplay between the efficient radiative transfer between the excited states and exciton dissociation into polarons for improving the efficiency of these devices. In Chapter 2, we provide the detailed description of the various theoretical methods applied for the prediction as well as the interpretation of the optical properties of a special class of substituted PPV [poly (p-phenylene vinylene)] oligomers. In Chapter 3, we report the accuracy of different second and third order time dependent density functional theory (TD-DFT) formalisms in prediction of the 2PA spectra compared to the experimental measurements for donor-acceptor PPV derivatives. We recommend a posteriori Tamm-Dancoff approximation method for both qualitative and quantitative analysis of 2PA properties. Whereas, Agren's quadratic response methods lack the double excitations and are not suitable for the qualitative analysis of the state-specific contributions distorting the overall quality of the 2PA predictions. We trace the reasons to the artifactual excited states above the ionization threshold. We also study the effect of the basis set, geometrical constraints and the orbital exchange fraction on the 2PA excitation energies and cross-sections. Higher exchange (BMK and M05-2X) and range-separated (CAM-B3LYP) hybrid functionals are found to yield inaccurate predictions both quantitatively and qualitatively. The failure of the exchange-correlation (XC) functionals with correct asymptotic is traced to the inaccurate transition dipoles between the valence states, where functionals with low HF exchange succeed. In Chapter 4, we test the performance of different semiempirical wavefunction theory methods for the prediction of 2PA properties compared to the DFT results for the same set of molecules. The spectroscopic parameterized (ZINDO/S) method is relatively better than the general purpose parameterized (PM6) method but the accuracy is trailing behind the DFT methods. The poor performances of PM6 and ZINDO/S methods are attributed to the incorrect description of excited-to-excited state transition and 2PA energies, respectively. The different semiempirical parameterizations can at best be used for quantitative analysis of the 2PA properties. The ZINDO/S method combined with different orders of multi-reference configuration interactions provide an improved description of 2PA properties. However, the results are observed to be highly dependent on the specific choice for the active space, order of excitation and reference configurations. In Chapter 5, we present a linear response TD-DFT study to benchmark the ability of existing functional models to describe the extent of self-trapped neutral and charged excitations in PPV and its derivative MEH-PPV considered in their trans-isomeric forms. The electronic excitations in question include the lowest singlet (S1) and triplet (T1†) excitons, positive (P+) and negative (P-) polarons and the lowest triplet (T1) states. Use of the long-range-corrected DFT functional, such as LC-wPBE, is found to be crucial in order to predict the physically correct spatial localization of all the electronic excitations in agreement with experiment. The inclusion of polarizable dielectric environment play an important role for the charged states. The particle-hole symmetry is preserved for both the polymers in trans geometries. These studies indicate two distinct origins leading to self-localization of electronic excitations. Firstly, distortion of molecular geometry may create a spatially localized potential energy well where the state wavefunction self-traps. Secondly, even in the absence of geometric and vibrational dynamics, the excitation may become spatially confined due to energy stabilization caused by polarization effects from surrounding dielectric medium. In Chapter 6, we aim to separate these two fundamental sources of spatial localization. We observe the electronic localization of P+ and P- is determined by the polarization effects of the surrounding media and the character of the DFT functional. In contrast, the self-trapping of the electronic wavefunctions of S1 and T1(T1†) mostly follows their lattice distortions. Geometry relaxation plays an important role in the localization of the S1 and T1† excitons owing to the non-variational construction of the excited state wavefunction. While, mean-field calculated P+, P- and T1 states are always spatially localized even in ground state S0 geometry. Polaron P+ and P- formation is signified by the presence of the localized states for the hole or the electron deep inside the HOMO-LUMO gap of the oligomer as a result of the orbital stabilization at the LC-wPBE level. The broadening of the HOMO-LUMO band gap for the T1 exciton compared to the charged states is associated with the inverted bond length alternation observed at this level. The molecular orbital energetics are investigated to identify the relationships between state localization and the corresponding orbital structure. In Chapter 7, we investigate the effect of various conformational defects of trans and cis nature on the energetics and localization of the charged P+ and P- excitations in PPV and MEH-PPV. We observe that the extent of self-trapping for P+ and P- polarons is highly sensitive on molecular and structural conformations, and distribution of atomic charges within the polymers. The particle-hole symmetry is broken with the introduction of trans defects and inclusion of the polarizable environment in consistent with experiment. The differences in the behavior of PPV and MEH-PPV is rationalized based on their orbital energetics and atomic charge distributions. We show these isomeric defects influence the behavior and drift mobilities of the charge carriers in substituted PPVs.
Ph.D.
Doctorate
Physics
Sciences
Physics

Ma thèse de doctorat traite de la sélection de projets d'innovation en entreprises, en utilisant les marchés de prédiction comme mécanisme de sélection alternatif. Le processus d'innovation et son évaluation sont des activités ayant des répercussions sur la croissance et le développement. L’évidence montre que les méthodes habituelles d'évaluation et de sélection de projets d’innovation, comme le processus en entonnoir, ne sont pas rentables. Proposer une méthode plus efficace contribuera de manière significative à une meilleure allocation des ressources. Dans la première partie de ma thèse, je teste les prévisions du marché de prédiction contre celles des experts. Dans la deuxième, j'examine les aspects comportementaux de la prise de décision sur le marché de prédiction entrepreneurial, notamment comment le biais d’optimisme influence les décisions des traders. J’ai mené pour ces parties des expériences avec des sujets humains. Dans la troisième partie, j'examine les propriétés et éléments clés des marchés de prédiction et fourni une chronique et une classification d’articles sur les contributions les plus importantes de la littérature dans ce sujet
My PhD thesis deals with selection of corporate and entrepreneurial innovation projects, using prediction markets as an alternative selection mechanism. Innovation process and its evaluation are two very important economic activities with repercussions for growth and development. Available evidence strongly suggests that conventional evaluation and selection methods, such as development funnel in corporate setting or decisions of Venture Capital firms in entrepreneurial one, do not yield cost-effective results. Coming up with an efficient and cost-effective method would contribute significantly to better resource allocation and social welfare. In the first part of the thesis, I test the prediction market predictions against experts’. In the second part, I examine behavioral aspects of decision-making in entrepreneurial prediction market setting, particularly how optimism bias influences traders’ decisions in prediction market. I conducted experiments with human subjects for the first two parts. In the third part of the thesis, I examine the most important elements and properties of prediction markets and provide a survey of most important contributions to prediction market literature, together with the classification and list of articles in major categories

Digital Clay represents a new type of 3-D human-computer interface device that enables tactile and haptic interactions. The Digital Clay kinematics structure is computer controlled and can be commanded to acquire a wide variety of desired shapes (shape display), or be deformed by the user in a manner similar to that of real clay (shape editing). The design of the structure went through various modifications where we finally settled on a crust matrix of spherical joint unit cells. After designing the kinematics structure, the next step is predicting the deformation of the crust matrix based upon a handful of inputs. One possible solution for predicting the shape outcome is considering minimizing the potential energy of the system. In this thesis two methods will be introduced. The first method will be an abstract model of the crust where the energy is calculated from a simplified model with one type of angular springs. The second method is the actual manufacturable crust model with two types of angular springs. From the implementation of these two methods, the output will be center-points of the unit cells. From the center-points, one can also calculate the joint angles within each unit cell.

In 2006, the University of British Columbia (UBC) School of Nursing (SON) received UBC Senate approval to limit admission to their baccalaureate nursing program to third-year, with advanced standing entry. The judicious selection of nursing students is increasingly salient in light of the looming retirement of 'baby-boomer' nurses, the goal of minimizing attrition from educational programs, and the resource-intensive admission procedure currently utilized at UBC, in which the applicants' academic and non-academic backgrounds are evaluated. A retrospective, correlational study of all 1,343 applicants to the program, for entry between January 2003 and January 2006, was undertaken. A series of multiple, logistic, and ordinal regression analyses were conducted to determine whether the current selective admission tools used by the SON added any valuable information over that obtained through the consideration of admission grade point average (GPA), in the selection of students who would be both academically and clinically successful in the program. The admission tools included a supplemental application, comprised of a personal statement and structured resume, which outlined the applicant's leadership, volunteer and employment experiences, motivation for choosing a career in nursing, and ability to work with others. Interviews were conducted of selected applicants to further evaluate applicants' suitability for nursing, particularly with respect to their clinical aptitude. The study variables included the applicants' demographics, their supplemental application and interview scores, and their admission GPA, as well as the outcome variables of interest: offers of program admission and in-program course grades. Although admission GPA was consistently predictive of the students' success, neither the supplemental application nor the interview scores were found to have predictive utility. The supplemental application had particularly poor inter-rater reliability. The variables found to besignificantly predictive of the students' cumulative average grade for selected nursing courses were gender, visible minority status, and admission GPA, accounting for approximately 27% of the variance in grades (pseudo-R²= 0.27). The results provide little evidence to justify continuation of the admission process currently in place.
Applied Science, Faculty of
Nursing, School of
Graduate

The most important income for the pig industry in Italy arises from the processing of meat in high-added-value typical products, mainly dry-cured hams having Protected Designation of Origin (PDO) label. Even through use of purebred pigs is not an exception, raw thighs are largely obtained from crossbred (CB) animals, which are slaughtered at heavy body weight (BW) (nearly 165 kg) and advanced age (270 d or more). Breeding programs of pig lines linked to dry-cured ham production should focus on enhancement of growth performance, carcass traits and qualitative and technological properties of green thighs, in order to comply the rules dictated by the disciplinary of production. Moreover, the breeding goal (aggregate genotype) must be defined at the level of the “CB slaughter animal”, so that the unfavourable effect on selection response due to genetic differences between purebreds (PB) and crossbreds, but also to environmental differences between nuclei and commercial farms, is alleviated. Given that breeding values (EBV) estimated using purebred phenotypes are often poorly related to the breeding goal defined at crossbred levels, the approach called “combined crossbred-purebred selection” is used in traditional genetic evaluation. It consists in estimation the EBV for the relevant traits using both PB and CB information. This approach is troublesome because it requires the availability of crossbred animals (half-sibs of the PB breeding candidates) which must be specifically reared as tested animals, and the phenotypic records of traits difficult and expensive to measured (as ham quality traits and curing losses). Moreover, EBV for members of a PB full-sibs family are identical when no PB individual information is available. The main aim of this thesis was to provide new knowledge about innovative traits important in heavy pig industry and to propose new statistical models and genetic evaluation procedures to enhance the prediction of breeding values for carcass traits, and qualitative and technological properties of raw thighs. The studies were carried out in collaboration with an Italian breeding company, Gorzagri s.s. (Fonzaso, Italy). It aims at producing boars (C21 line) and gilts (Goland line) used in commercial farms as sires and dams of CB pigs which are reared for production of PDO dry-cured hams. The first part of this thesis aimed to investigate the prevalence and the genetic determinism of boar taint (BT), a new trait related to both meat quality and animal welfare which is becoming increasingly important for pig industries in EU countries after the decision to ban surgical castration of piglets from 1 January 2018, and hence, the need to find alternative to reduce BT in entire male pigs. The objectives of the study were: to evaluate the prevalence of BT in intact male pigs at 160 and 220 d of age, and to estimate the genetic parameters for BT compounds at the two different ages and the genetic relationships with growth traits. Contents of androstenone (AND), skatole (SKA) and indole (IND) have been quantified by HPLC with fluorescence detection in biopsy samples of adipose tissues collected in-vivo from the neck area of 500 C21 entire male pigs at 160 and 220 d of age. In addition, for 100 of the investigated animals, BT compounds were quantified also in a sample of subcutaneous fat of raw thighs collected at slaughterhouse. Contents of BT compounds measured in intact male pigs at 220 d were higher than those found at 160 d of age, as well as the percentage of samples that exceeded the sensory thresholds discriminating tainted from untainted carcasses. Thus, the prevalence of BT is expected to be greater in mature and heavy pigs in respect to young and light pigs. The high phenotypic correlations between the contents of BT compounds measured in backfat and ham subcutaneous fat suggest that BT might be relevant also in PDO dry-cured hams. Medium-high heritabilities were found for contents of BT compounds at both ages, indicating that reduction of BT by means of genetic selection seems a valid alternative to surgical castration in pigs slaughtered at heavy BW. Genetic correlations between the contents of BT compounds measured at the two different ages were moderate to high. However, Spearman’s rank correlations revealed that the breeding candidates ranked differently when measures of BT compounds taken at the different ages are used. Finally, weak genetic correlations were found between BT compounds and growth traits, indicating that selective breeding to decrease the contents of AND, SKA and IND is expected to have trivial effects on growth performance. In the second section of this thesis the contribution of social genetic effects on variation of carcass and ham quality traits was investigated. In the current breeding program of C21 line the impact of the genetic effect of an individual on phenotypes of its pen mates (called social genetic effects or heritable social effects) is neglected. If present, social genetic effects are part of the total heritable variance and they affect the response to selection. The aims of this study were to estimate (co)variance components for body weight adjusted at 270 d (BW270), carcass and ham quality traits using direct and competitive models, and to compare the ability of such models to fit the data. The study was carried out on 9,871 CB finishing pigs raised in social groups containing from 4 to 7 individuals (6.1 pigs per group on average). Four sequential univariate animal models were compared through likelihood ratio test. REML estimates of covariance components were obtained for BW270, carcass backfat depth and lean meat (CLM), iodine number (IOD) and linoleic acid content of raw ham subcutaneous fat, subcutaneous fat depth in the proximity of semimembranosus and quadriceps femoris muscles, and linear scores for ham round shape (RS), subcutaneous fat (SF), and marbling. Model comparison based on likelihood ratio test revealed that the model accounting for heritable social effects was significantly better than the best direct model for BW270, CLM, IOD, RS, and SF. The contribution of social genetic effects to the total heritable variance was large for CLM and BW270, whereas the one for ham quality traits was lower. The correlation between direct and social additive genetic effects was positive for BW270, but it was not significantly different from zero, suggesting independence between direct and social genetic effects for this trait. In contrast, a negative genetic covariance between direct and associative components was found for CLM, IOD, RS, and SF, which reduced the total heritable variance exploitable for genetic selection. The results obtained in this study suggest that social genetic effects affect variation in traits relevant for heavy pigs used in dry-cured hams manufacturing and, hence, the procedures for estimation of breeding values should take heritable social effects into account. The third part of this thesis aimed to investigate new methods including genomic information in order to enhance the EBV prediction for traits related to carcass, ham quality and manufacturing of dry-cured hams. Genomic selection (GS) approach might overcome some of the drawbacks of current procedures for genetic evaluation of C21 line based on sib-testing. The objective of the first study regarding GS was to develop and to investigate GS procedures based on single step BLUP (SSBLUP) methodology for genomic evaluation of PB breeding candidates of C21 boar line for traits relevant for dry-cured ham production. Observations on BW270, carcass and ham quality traits were recorded from 11,488 CB finishing pigs. In addition, for 1,878 of the investigated animals, phenotypes for weight losses occurring during salting, resting and curing production stages were available. To constitute the reference population, 1,088 CB pigs, 136 nucleus boars (C21_NB: sires of CB tested pigs) and 500 C21 half-sibs of CB animals (C21_HS: half-sibs of CB tested animals) were genotyped for 8,826 single nucleotide polymorphisms using GGP Porcine LD Chip. Traditional evaluation was performed using BLUP methodology, whereas genomic evaluation was carried out using SSBLUP method which combine pedigree and genomic information in order to estimate genomic breeding values (GEBV) for both genotyped and ungenotyped animals. Pearson’s correlation coefficients between EBV and GEBV (r_(EBV,GEBV)) were calculated in order to compare traditional and genomic evaluation. In addition, the 500 C21_HS with genotypic information were used to validate the prediction equation. High r_(EBV,GEBV) were observed for all the investigated traits for both CB animals, C21_NB, suggesting that genetic merit estimated using GS was consistent with breeding values prediction from traditional evaluation. Positive and moderate to high r_(EBV,GEBV) were observed for C21_HS used in validation population. Differences in ranks of breeding candidates were found when they were based on traditional EBV or GEBV. In contrast to EBV, GEBV differed for members of a PB full-sibs family when they had genomic information. Thus, GS allows to choose the best animals based on “individual” genetic merit rather than “family” merit, increasing the accuracy of selection and reducing the rate of inbreeding. In addition, GS procedures might simplify genetic evaluation of PB breeding candidates of C21 boar line for traits difficult to measure. As found in the second part of this thesis, competitive models provided a better fit than classical direct model for BW270, CLM, IOD, RS and SF. Hence, for these traits, traditional genetic evaluation accounting for heritable social effects was compared to SSBLUP analysis performed using a competitive model. The dataset used in this study was the same described above. For animals in validation population, positive and moderate to high Spearman’s correlation coefficients between direct EBV and direct GEBV were found for all the investigated traits, as well as between total breeding values and total genomic breeding values (TGBV). Hence, differences in ranks of breeding candidates were observed even when social genetic effects were included in the genomic evaluation. In fact, GS procedures accounting for heritable social effects allow to obtain individual TGBV rather than “family” TBV, increasing the benefit of GS for genetic evaluation of PB boar lines linked to dry-cured ham production.
I profitti più rilevanti per l’industria suinicola italiana derivano dalla trasformazione della carne in prodotti tipici di alto valore economico, principalmente prosciutti crudi con Denominazione di Origine protetta (DOP). Le cosce fresche destinate alla produzione dei prodotti DOP derivano principalmente da soggetti ibridi che vengono macellati a pesi elevati (circa 165 kg) e ad età avanzate (almeno 270 giorni). I programmi di selezione delle linee suine destinate alla produzione dei prosciutti crudi dovrebbero focalizzarsi sul miglioramento delle prestazioni di crescita, della qualità della carcassa e delle caratteristiche qualitative e tecnologiche delle cosce fresche, cosicché i requisiti dettati dai disciplinari di produzione siano soddisfatti. In questo scenario gli obiettivi di selezione devono essere definiti a livello di soggetto ibrido, piuttosto che a livello di linea pura. I valori genetici (breeding values, EBV) stimati utilizzando le informazioni fenotipiche dei soggetti di linea pura sono poco correlati con gli obiettivi di selezione definiti al livello dei soggetti ibridi. Per questo la valutazione genetica dei candidati riproduttori di linea pura attualmente avviene attraverso un approccio che combina informazioni derivanti sia da soggetti puri che da soggetti ibridi. Questo approccio presenta alcune criticità dovute alla necessità di allevare appositamente soggetti ibridi (mezzi fratelli dei candidati riproduttori di linea pura) per ottenere i rilievi fenotipici per caratteristiche che alcune volte sono difficili da misurare (come i caratteri di qualità della coscia o i cali di stagionatura). Inoltre, quando nella valutazione genetica non sono incluse informazioni fenotipiche rilevate sugli animali di linea pura, i EBV per i candidati riproduttori che appartengono alla stessa famiglia di fratelli pieni sono identici. L’obiettivo principale della tesi è stato quello di fornire nuove conoscenze su caratteri innovativi per l’industria del suino pesante e di proporre nuovi modelli statistici e procedure di valutazione genetica per migliorare la predizione dei EBV per caratteri legati alla qualità della carcassa e delle cosce fresche destinate alla produzione dei prosciutti crudi DOP. Gli studi qui presentati sono stati condotti in collaborazione con un’azienda suinicola italiana, Gorzagri s.s (Fonzaso, Italia). Essa ha come obiettivo la produzione di verri (linea C21) e scrofe (linea Goland) usati negli allevamenti commerciali come padri e madri di soggetti ibridi destinati alla produzione dei prodotti DOP. La prima parte della tesi si proponeva di studiare l’incidenza e il determinismo genetico dell’odore di verro, un nuovo carattere associato sia alla qualità della carne che al benessere animale, il quale sta diventando sempre più rilevante per l’industria suinicola italiana ed europea dopo la decisione di abbandonare la castrazione chirurgica dei suinetti maschi a partire dal 1 gennaio 2018. Questo impone la necessità di trovare valide alternative a questa pratica per ridurre l’odore di verro nei suini maschi interi. Allo scopo di indagare la possibilità di sfruttare la selezione genetica per limitare l’odore di verro nel suino pesante, gli obiettivi dello studio sono stati: valutare la prevalenza dell’odore di verro in suini maschi interi di 160 e 220 giorni d’età e stimare i parametri genetici per i composti responsabili del problema alle due diverse età nonchè le relazioni genetiche che intercorrono tra questi e i caratteri legati all’accrescimento. I contenuti di androstenone (AND), scatolo (SCA) e indolo (IND) sono stati quantificati tramite cromatografia liquida ad alta prestazione in microcampioni di tessuto adiposo prelevati in-vivo dalla regione del collo di 500 suini maschi interi della linea C21 al raggiungimento dell’età di 160 e 220 giorni. Inoltre, per 100 di essi, i composti sono stati quantificati anche in un campione di grasso di copertura della coscia prelevato in sede di macellazione. Le concentrazioni dei tre composti misurati nel grasso di animali di 220 giorni sono risultati superiori a quelli riscontrati a 160 giorni di età, così come la percentuale di campioni che superano i limiti di accettabilità utilizzati per discriminare le carcasse affette o meno dall’odore di verro. Questo suggerisce che l’incidenza del problema dell’odore di verro è maggiore nei suini più pesanti e maturi da un punto di vista sessuale. Inoltre, un’elevata correlazione è stata riscontrata tra il contenuto di AND, SCA e IND nel grasso dorsale e nel grasso di copertura delle cosce, suggerendo che il problema può essere rilevante anche nei prosciutti crudi DOP. I composti responsabili dell’odore di verro hanno mostrato un’elevata ereditabilità ad entrambe le età, confermando che la selezione genetica potrebbe essere una valida soluzione per ridurre l’odore di verro anche in suini macellati a pesi elevati. Le correlazioni genetiche tra AND, SCA e IND misurati alle due diverse età sono risultate di grado moderato o elevato. Tuttavia, sono state riscontrate delle differenze nelle classifiche dei candidati riproduttori quando i fenotipi misurati alle due diverse età sono stati usati per la stima dei EBV. Infine, le correlazioni genetiche tra AND, SCA e IND e caratteri legati all’accrescimento sono risultate modeste, suggerendo che la riduzione della concentrazione di questi composti nel tessuto adiposo tramite selezione genetica non dovrebbe influenzare le prestazioni di crescita in suini non castrati. Nella seconda parte di questa tesi è stato esaminato in contributo degli effetti sociali sulla variabilità di caratteri legati alla qualità della carcassa e della coscia in suini pesanti. Le attuali procedure di valutazione genetica della linea C21 non tengono in considerazione l’influenza che il genotipo di un animale ha sul fenotipo dei suoi compagni di gruppo (effetto sociale). Se presenti, gli effetti sociali contribuiscono alla varianza ereditabile e quindi possono influenzare la risposta alla selezione. Lo studio si è quindi proposto di stimare le componenti di varianza per il peso a 270 giorni di età (PESO270), caratteri di qualità della carcassa e della coscia utilizzando modelli genetici “classici” e modelli competitive (che includono gli effetti sociali), nonché determinare la bontà dell’adattamento ai dati di questi modelli. Lo studio è stato condotto su 9,871 suini ibridi Goland allevati in gruppi contenenti da 4 a 7 individui, con dimensione media del gruppo pari a 6.1 suini. Quindi, quattro modelli sequenziali sono stati utilizzati per la stima delle componenti di varianza. Il confronto dei modelli ha rivelato che il modello competitive mostra un miglior adattamento ai dati rispetto ad un modello che considera solo gli effetti genetici additivi per il PESO270, la percentuale di carne magra in carcassa (CM), il numero di iodio, la globosità e lo spessore del grasso sottocutaneo misurati sulla coscia. Gli effetti sociali contribuiscono in modo considerevole alla varianza ereditabile per il PESO270 e CM, mentre hanno un effetto meno rilevante per gli altri caratteri esaminati. La correlazione tra la componente genetica additiva e quella sociale è risultata negativa per i caratteri legati alla qualità della carcassa e della coscia, causando una diminuzione della varianza ereditabile sfruttabile per la selezione di questi caratteri. In conclusione, lo studio ha dimostrato che gli effetti sociali contribuiscono alla variabilità di caratteri rilevanti nel suino pesante destinato alla produzione dei prosciutti crudi DOP e, quindi, le attuali procedure per la valutazione genetica della linea C21 dovrebbero includere anche la componente sociale. La terza parti della tesi ha riguardato lo studio di nuove procedure di valutazione genetica basate sulla selezione genomica (SG). L’obiettivo è quello di migliorare la predizione dei EBV per caratteri importanti nel suino pesante superando alcune delle criticità dell’attuale sistema di valutazione genetica della linea C21 basato su un programma di sib-testing. In un primo lavoro, sono state studiate procedure di SG basate sulla metodologia single-step BLUP (SSBLUP). Le informazioni fenotipiche per caratteri riguardanti la qualità della carcassa e della coscia sono state registrate in 11,488 suini ibridi Goland. Per 1,878 animali sono stati inoltre rilevati i cali di peso della coscia durante le diverse fasi di trasformazione in prosciutto crudo. Allo scopo di costituire la popolazione di riferimento per lo sviluppo dell’equazione di predizione dei breeding values genomici (GEBV), 1,088 soggetti ibridi, 136 verri nucleo di linea C21 (C21_VN) e 500 soggetti C21 mezzi fratelli degli animali ibridi (C21_MF) sono stati genotipizzati per 8,826 polimorfismi a singolo nucleotide utilizzando il DNA chip GGP Porcine LD Chip. La valutazione genetica tradizionale è stata condotta sfruttando la metodologia BLUP, mentre la valutazione genomica si è basata su un approccio SSBLUP. L’equazione di predizione dei GEBV è stata validata utilizzando i 500 C21_MF con genotipo come popolazione di validazione. La valutazione genetica tradizionale e quella genomica sono state confrontate tra loro attraverso la correlazione tra EBV e GEBV (r_(EBV,GEBV)). Elevati coefficienti di correlazione tra EBV e GEBV sono stati osservati sia per i soggetti ibridi genotipizzati che per gli individui C21_VN, suggerendo che le stime dei GEBV sono simili a stime accurate dei EBV ottenute con metodi tradizionali. Per quanto concerne i soggetti inclusi nella popolazione di validazione, r_(EBV,GEBV) sono risultate di entità minore rispetto ai due gruppi considerati in precedenza. Differenze nelle classifiche dei candidati riproduttori sono state osservate quando esse erano basate sui EBV piuttosto che sui GEBV. Tali differenze sono attribuibili all’abilità della metodologia SSBLUP di stimare GEBV individuali, anziché EBV identici per soggetti che appartengono alla stessa famiglia di fratelli pieni (quando essi hanno informazione genomica). Quindi l’utilizzo di un approccio di SG permette di scegliere i soggetti migliori in base a meriti genetici individuali piuttosto che “familiari”, limitando il tasso di consanguineità e aumentando l’accuratezza della selezione. Inoltre, i risultati ottenuti rivelano che l’implementazione di procedure di SG nel programma di selezione della linea C21 consentirebbero di introdurre tra gli obiettivi di selezione anche caratteri di difficile misurazione. Come descritto in precedenza, gli effetti sociali contribuiscono alla variabilità di caratteri rilevanti per il suino pesante e che modelli competitive forniscono un miglior adattamento ai dati per i caratteri PESO270, CM, il numero di iodio, la globosità e lo spessore del grasso sottocutaneo misurati sulla coscia. Quindi, per questi caratteri, sono state analizzate procedure di SG basate sulla metodologia SSBLUP che consideravano, oltre agli effetti genetici diretti, anche gli effetti sociali. Le informazioni fenotipiche e genomiche utilizzate sono le medesime descritte per il precedente studio. Differenze nelle classifiche degli animali inclusi nella popolazione di validazione sono state riscontrate sia quando esse erano basate su EBV e GEBV riferiti alla sola componente genetica diretta, che quando sono stati utilizzati i valori genetici e genomici totali. Quindi l’implementazione procedure di SG nel programma di selezione della linea C21 che utilizzano modelli competitive consentirebbero di sfruttare i vantaggi della SG già descritti in precedenza e contemporaneamente stimare più correttamente la varianza ereditabile totale considerando anche la componente associativa.

This study employs duality theory to model the dairy industry. Supply and demands for milk, cull cows, feed, labor and veterinary services were simultaneously estimated using Weighted Least Squares. Elasticities and partial adjustments were obtained for the Nation and the following regions, Appalachia, Cornbelt, Northeast, Pacific, Southern Plains and Upper-Midwest. Predictions for the change in quantity of goods demanded and supplied were made assuming a parallel shift in the supply of milk and demand for feed. In conclusion, predictions on the impact of bovine Somatotropin are made assuming all results are correct.

The main objective of this thesis was to assess the infrared spectroscopy for the prediction at individual level of “new phenotypes” related to the technological properties of the cow milk, testing classic and innovative statistical approaches and evaluating the genetic parameters for a possible inclusion of the predicted traits in the selection indices as indirect selection method. A total of 1,264 individual milk samples were used for an individual model cheese making procedure and 7 new cheese-making related traits were obtained: 3 measures of cheese yield as percentage of processed milk (%CYs; fresh cheese yield, total solids cheese yield, water retained in the curd) and 4 measures of milk nutrients retained in the curd or lost in the whey (%RECs; fat, protein, total solids and energy). The traditional milk coagulation properties (rennet coagulation time, RCT; curd firming time, k20; curd firmness at 30 and 45 min, a30 and a45 respectively ) were also measured using a Formagraph (Foss Electric A/S, Hillerød, Denmark) in a curd firmness (CF) testing time of 90 min. Using all the 360 information of the CF test recorded for each sample over the 90 min, some new modeled parameters were also obtained (modeled rennet coagulation time, RCTeq; asymptotical potential value of CF at an infinite time, CFP; curd-firming rate constant, kCF; curd-syneresis rate constant, kSR; maximum level of CF, CFmax; time at which CF attains the maximum level, tmax;). For each sample two Fourier-transform infrared (FTIR) spectra were collected with a MilkoScan FT6000 (Foss Electric, Hillerød, Denmark) over the spectral range from 5,000 to 900 wavenumber × cm-1, and averaged before data analysis. A first chemometric process was carried out, using the WinISI II software (Infrasoft International LLC, State College, PA) in which the partial least square regression (PLS) models are implemented, for the prediction of %CYs and %RECs. High prediction accuracies were found except for the fat recovery. In order to improve the prediction accuracy, Bayesian models, commonly used for genomic data, were tested and compared with PLS models. The results have shown that for those traits that are difficult to be predicted, the Bayesian models perform better than PLS. Using an external validation procedure, the PLS was used for the prediction of %CYs and %RECs, while the BayesB model was used for the prediction of MCP and CF modeled parameters. In both cases the prediction accuracy found in validation, ranged from low to moderate. The genetic parameters of the predicted were estimated through a bivariate Bayesian analysis and linear models. Despite the low-moderate prediction accuracy in validation, the heritabilities of the predicted values were similar or higher than those of the corresponding measured values. The indirect selection of the studied traits was assessed through the genetic correlations between measured and predicted values, and the results shown that even when the coefficient of determination for the validation was moderate, the genetic correlations between predicted and measured values were always higher than the phenotypic correlations, and in the majority of cases near or higher than 90%. The calibrations developed for the %CYs and %RECs have been used to obtain the predictions on a population data set consisting of about 200,000 spectra of individual milk samples of Holstein, Brown Swiss and Simmental dairy cows. The genetic parameters of the predicted traits were estimated and the heritability values were comparable to those of the measured traits. The genetic correlations of %CYs and %RECs with milk production and composition provide evidence that the current selection paradigm used in dairy cattle may have a limited effects on the technological parameters. Milk protein and fat content do not explain all the genetic variations of %CYs and (in particular) %RECs, thus, these traits could be directly selected to improve the cheese making aptitude of milk and its correlated economic value
L’obiettivo principale di questa tesi è stato quello di valutare l’efficienza della spettroscopia a infrarosso per la predizione, a livello individuale, di “nuovi fenotipi” che descrivono le proprietà tecnologiche del latte bovino. Sono stati testati approcci statistici di calibrazione classici e innovativi, e sono stati inoltre stimati e valutati i parametri genetici delle predizioni ottenute per verificarne la possibile inclusione negli indici di selezione come metodo indiretto. Su un totale di 1,264 campioni di latte individuale, sono state effettuate le analisi che hanno previsto l’impiego di una procedura standard di micro-caseificazione per la misura di 7 caratteri relativi alla trasformazione casearia, in particolare sono state rilevate 3 misure di resa espresse come percentuale del latte lavorato, (%CYs; resa a fresco, resa in solidi totali, acqua ritenuta nella cagliata) e 4 misure di recupero di nutrienti nella cagliata o persi nel siero (%RECs; grasso, proteina, solidi totali ed energia). Le proprietà di coagulazione tradizionali (tempo di coagulazione, RCT; tempo di rassodamento, k20; consistenza del coagulo a 30 e 45 minuti dall’aggiunta del caglio, a30 e a45 rispettivamente) sono state misurate con un Formagraph (Foss Electric A/S, Hillerød, Denmark) in un test della consistenza del coagulo (CF) di 90 min. Utilizzando tutte le 360 informazioni di CF per campione registrate nei 90 min, sono stati inoltre ricavati, attraverso un modello matematico, dei nuovi parametri (tempo di coagulazione modellizzato, RCTeq; valore asintotico potenziale di CF per un tempo infinito, CFP; costante di rassodamento, kCF; costante di sineresi, kSR; valore massimo di CF, CFmax; tempo necessario affinché CF raggiunga il livello massimo, tmax). Per ogni campione sono stati raccolti due spettri a infrarosso in trasformata di Fourier (FTIR), utilizzando un MilkoScan FT6000 (Foss Electric, Hillerød, Denmark) nel range spettrale compreso tra 5,000 e 900 onde × cm-1, i due spettri sono stati mediati prima delle analisi. Un primo processo di calibrazione è stato effettuato per la predizione di %CYs e %RECs, utilizzando il software WinISI II (Infrasoft International LLC, State College, PA) in cui sono implementati dei modelli basati sulla partial least square regression (PLS). I risultati ottenuti hanno mostrato ottime accuratezze di predizione tranne che per il recupero di grasso. Per migliorare le accuratezze di predizione, sono stati testati dei modelli Bayesiani, comunemente usati in genomica, e confrontati con la PLS. Dai risultati ottenuti, per alcuni caratteri difficili da predire, si è visto che i modelli Bayesiani hanno delle prestazioni migliori. Utilizzando una procedura di validazione esterna come metodo di valutazione delle prestazioni di calibrazione, la PLS è stata utilizzata per la predizione di %CYs e %RECs, mentre i modelli Bayesiani sono stati utilizzati per la predizione delle proprietà di coagulazione e per i parametri derivanti dalla modellizzazione della consistenza del coagulo. In entrambi i casi i risultati ottenuti, relativi all’accuratezza di predizione, hanno mostrato un’efficienza medio bassa. Inoltre, sono stati stimati i parametri genetici dei valori predetti nel processo di validazione e nonostante la medio-bassa accuratezza delle predizioni, le ereditabilità dei valori predetti sono state simili o più alte dei corrispondenti valori misurati. L’impiego dei valori predetti come metodo di selezione indiretta è stato valutato attraverso la stima delle correlazioni genetiche tra valori predetti e misurati. I risultati hanno dimostrato, anche in questo caso che le correlazioni genetiche erano sempre superiori a quelle fenotipiche e nella maggior parte dei casi vicine o superiori al 90%. Infine, le equazioni di predizione sviluppate per %CYs e %RECs, sono state impiegate per la predizione di questi fenotipi su un set di dati costituito da circa 200,000 spettri di campioni individuali di latte di vacche di razza Frisona, Bruna e Pezzata Rossa italiane. I parametri genetici delle predizioni ottenute per ogni carattere sono stati stimati, dimostrando di essere ereditabili, con valori di ereditabilità simili a quelli dei valori misurati. Le correlazioni genetiche tra i valori predetti di %CYs e %RECs, e quelli relativi ai dati produttivi e di composizione del latte, hanno dimostrato che i modelli di selezione in uso hanno un effetto limitato sul miglioramento dei parametri tecnologici. Proteina e grasso del latte non spiegano tutta la variabilità genetica di %CYs e, in particolare, di %RECs, quindi per il miglioramento dell’attitudine casearia e conseguente valorizzazione economica del latte, questi caratteri andrebbero selezionati direttamente

The topic of the research in this dissertation is the application of machinelearning in solving problems characteristic to biosystems, with specialemphasis on agriculture. Firstly, an innovative regression algorithm based onbig data was presented, that was used for yield prediction. The predictionswere then used as an input for the improved portfolio optimisation algorithm,so that appropriate soybean varieties could be selected for fields withdistinctive parameters. Lastly, a multi-objective optimisation problem was setup and solved using a novel method for categorical evolutionary algorithmbased on NSGA-III.
Предмет истраживања докторске дисертације је примена машинског учења у решавању проблема карактеристичних за биосистемe са нагласком на пољопривреду. Најпре је представљен иновативни алгоритам за регресију који је примењен на великој количини података како би се са предиковали приноси. На основу предикција одабране су одговарајуће сорте соје за њиве са одређеним карактеристикама унапређеним алгоритмом оптимизације портфолија. Напослетку је постављен оптимизациони проблем одређивања сетвене структуре са вишеструким функцијама циља који је решен иновативном методом, категоричким еволутивним алгоритмом заснованом на NSGA-III алгоритму.
Predmet istraživanja doktorske disertacije je primena mašinskog učenja u rešavanju problema karakterističnih za biosisteme sa naglaskom na poljoprivredu. Najpre je predstavljen inovativni algoritam za regresiju koji je primenjen na velikoj količini podataka kako bi se sa predikovali prinosi. Na osnovu predikcija odabrane su odgovarajuće sorte soje za njive sa određenim karakteristikama unapređenim algoritmom optimizacije portfolija. Naposletku je postavljen optimizacioni problem određivanja setvene strukture sa višestrukim funkcijama cilja koji je rešen inovativnom metodom, kategoričkim evolutivnim algoritmom zasnovanom na NSGA-III algoritmu.

The instantaneous flow-rate in high-pressure pipelines of fluid power systems can be predicted by means of refined numerical models, the performance of which should be validated extensively against experimental pressure transients. However, such an approach in complex hydraulic systems can require an advanced modelling of the hydraulic, mechanical and electronic components and can therefore be cumbersome and ineffective. Experimental analysis is valuable to support the fast and complete evaluation of integrated electronic-hydraulic power systems, i.e., reciprocating pumps equipped with sophisticated regulation devices, gasoline and diesel fuel injection apparatus, anti-lock braking systems and traction-control systems. However, the state of the art only allows the unsteady high-pressure liquid flows in these systems to be characterized in terms of the pressure time-history measured at fixed locations. The situation is particularly critical for volumetric pumps, since the main mission of these components is the high-pressure instantaneous delivered flow-rate and one main aspect concerning the final judgement of their performance is related to the flow-rate ripple, which can only be evaluated approximatel. The selection of the operating principle for the flow-rate measurement in high-pressure pipelines can be considered challenging because miniaturized devices, capable of operating at high-pressure levels and characterized by a superior dynamic performance, are required. Obstruction meters, which relate the fluid velocity to the measured pressure drop across a calibrated orifice, offer steady measurements and are therefore not suitable for transients. On the other hand, positive displacement, turbine, and Coriolis flowmeters cannot generally be used because of their relatively low dynamic response (the cut-off frequency can reach 1 kHz) and their excessive overall sizes. As far as magnetic induction based flowmeters is concerned, they are characterized by high dynamic performance, but these meters require fluids with electrical conductivity as low as 0.1 S/cm, and diesel oil and gasoline both have much lower conductivity than this threshold. Finally, laser Doppler velocimetry-based techniques, which have been applied to the analysis of fluid dynamic transients in gasoline injection systems ([12]), imply a high initial setup cost, and are only applicable for laminar flows. Furthermore, these measurements require a pipe section made of quartz glass and this technique cannot therefore be applied when the pressure is higher than 100 bar. Flow-rate estimations, based on instantaneous pressure measurements, seem to be the most attractive opportunity, due to the superior dynamic response and the miniaturized sizes of the pressure transducers. Having assessed that pressure wave propagation occurs along a single direction, a mathematical relation exists between the pressure and flow-rate time-histories at any pipe section. On the other hand, the flow rate time-distribution cannot be inferred from the pressure time-history, measured at a single location, when the pressure waves travel back and forth through the pipeline, as usually occurs in engineering systems. An innovative methodology has been developed to evaluate the unsteady flow-rate in high-pressure pipelines, on the basis of the signals from two piezoelectric pressure transducers. A first-order non-linear ordinary differential equation was derived from Euler’s momentum balance and mass conservation equations, and was numerically solved to compute the instantaneous mass flow-rate. The flowmeter working principle was validated successfully through a comparison with numerical flow-rate data, which were predicted by means of a reliable one-dimensional model of a CR fuel injection system, and satisfactory accuracy and repeatability of the measurements were proved. However, the flowmeter algorithm required knowledge of the initial datum on the flow-rate, which can be difficult to infer. Furthermore, a complex procedure was required to remove the disturbances of the pressure transducer zero-offset errors from the measured pressure gradient. In particular, an additional piezoresistive transducer was needed to identify the presence of any possible time drifts in the piezoelectric transducer responses and to reference them, thus increasing the production costs of the device. Finally, the numerical discretization of an ordinary differential equation was required to evaluate the instantaneous flow-rate and this does not represent a very robust and efficient solution for a measuring device. During this Ph.D. course, which has been mainly focused on the numerical-experimental analysis of hydraulic power system, an optimized algorithm has been developed to evaluate the flow-rate on the basis of two pressure measurement performed in a high-pressure pipeline, and a definitive design has been proposed for the flowmeter previously studied. The newly designed and validated measuring device has been installed on the delivery pipe of a reciprocating pump in order to investigate the instantaneous flow-rate ripple, to validate a numerical model of the considered pump and to analyse the effect of the pump-integrated flow regulator on its hydraulic performance. A second application that will be analyses in the following sections concerns the installation of the proposed flowmeter to the inlet port of a hydraulic servo-valve. The activity was aimed at validating a numerical model of the test bench used to analyse the hydraulic behaviour of this kind of valve according to the ISO 10770-1 directive. The model has been subsequently used to perform some considerations about the overall hydraulic system. The flowmeter outputs have been also used for the development of a predictive model of a complete Common Rail (CR) injection system which is able to correctly simulate the rail pressure control system and the injector performance in steady-state and transients conditions with a deep degree of detail.

The present thesis investigates innovative energy technologies and control algorithms for enhancing demand-side management in buildings. The work focuses on an innovative low-temperature solar thermal system for supplying space heating demand of buildings. This technology is used as a case study to explore possible solutions to fulfil the mismatch between energy production and its exploitation in building. This shortcoming represents the primary issue of renewable energy sources. Technologies enhancing the energy storage capacity and active demand-side management or demand-response strategies must be implemented in buildings. For these purposes, it is possible to employ hardware or software solutions. The hardware solutions for thermal demand response of buildings are those technologies that allow the energy loads to be permanently shifted or mitigated. The software solutions for demand response are those that integrate an intelligent supervisory layer in the building automation (or management) systems. The present thesis approaches the problem from both the hardware technologies side and the software solutions side. This approach enables the mutual relationships and interactions between the strategies to be appropriately measured. The thesis can be roughly divided in two parts. The first part of the thesis focuses on an innovative solar thermal system exploiting a novel heat transfer fluid and storage media based on micro-encapsulated Phase Change Material slurry. This material leads the system to enhance latent heat exchange processes and increasing the overall performance. The features of Phase Change Material slurry are investigated experimentally and theoretically. A full-scale prototype of this innovative solar system enhancing latent heat exchange is conceived, designed and realised. An experimental campaign on the prototype is used to calibrate and validate a numerical model of the solar thermal system. This model is developed in this thesis to define the thermo-energetic behaviour of the technology. It consists of two mathematical sub-models able to describe the power/energy balances of the flat-plate solar thermal collector and the thermal energy storage unit respectively. In closed-loop configuration, all the Key Performance Indicators used to assess the reliability of the model indicate an excellent comparison between the system monitored outputs and simulation results. Simulation are performed both varying parametrically the boundary condition and investigating the long-term system performance in different climatic locations. Compared to a traditional water-based system used as a reference baseline, the simulation results show that the innovative system could improve the production of useful heat up to 7 % throughout the year and 19 % during the heating season. Once the hardware technology has been defined, the implementation of an innovative control method is necessary to enhance the operational efficiency of the system. This is the primary focus of the second part of the thesis. A specific solution is considered particularly promising for this purpose: the adoption of Model Predictive Control (MPC) formulations for improving the system thermal and energy management. Firstly, this thesis provides a robust and complete framework of the steps required to define an MPC problem for building processes regulation correctly. This goal is reached employing an extended review of the scientific literature and practical application concerning MPC application for building management. Secondly, an MPC algorithm is formulated to regulate the full-scale solar thermal prototype. A testbed virtual environment is developed to perform closed-loop simulations. The existing rule-based control logic is employed as the reference baseline. Compared to the baseline, the MPC algorithm produces energy savings up to 19.2 % with lower unmet energy demand.

Alcune aree definite dall'Unione Europea nel contesto delle smart cities and communities si fondono pienamente con i motori elettrici come, per esempio, l'efficienza energetica, le tecnologie a basse emissioni di carbonio e la mobilità. I motori elettrici sono utilizzati in molteplici applicazioni industriali e non, consumando tra il 43% e il 46% dell'energia elettrica prodotta su scala mondiale.Nonostante alcune applicazioni siano contraddistinte da dinamiche elevate, come manipolatori o macchine utensili, la maggior parte di esse sono caratterizzate da basse dinamiche in quanto facenti parte di processi industriali, per esempio pompe, compressori, ventilatori o nastri trasportatori. Si è stimato che il costo dell'intero ciclo di vita di un motore elettrico è ascrivibile per il 92% - 95% all'energia consumata, il che indurrebbe un tempo di ritorno dall'investimento per installazione di un azionamento elettrico minore di due anni. Nonostante il notevole risparmio economico e ambientale ottenibile, è piuttosto sorprendente apprendere che solo il 10% - 15% di tutti i motori industriali siano controllati da azionamenti elettrici. Per quanto riguarda le diverse tecnologie di motori elettrici, i motori sincroni a riluttanza stanno ricevendo una notevole attenzione sia da ricercatori industriali che accademici. Il crescente interesse è principalmente motivato dalle loro intrinseche caratteristiche quali l'alta efficienza, il basso costo e il basso impatto ambientale dovuto alla mancanza di magneti permanenti. Per di più, le loro caratteristiche soddisfano appieno i requisiti imposti dalle smart cities and communities e sono adatti per tutte le applicazione, caratterizzate da una bassa dinamica, viste sopra. Per questi motivi, questa tecnologia di motori può essere posta al centro dei processi di rinnovamento di quelle applicazioni. Vi è ampio consenso sul potenziale incremento delle vendite sia di azionamenti elettrici che di motori sincroni a riluttanza. I motori sincroni a riluttanza sono soggetti a una marcata saturazione magnetica, rendendo i classici modelli a parametri concentrati poco adatti. La prima parte di questa tesi riguarda lo sviluppo di un innovativo modello magnetico per motori anisotropi. Si basa su una rete neurale non tradizionale, chiamata Radial Basis Function. La sua proprietà locale rende questo tipo di rete neurale particolarmente adatta ad un addestramento durante il normale funzionamento del motore. Si propone una completa procedura di design e addestramento della stessa. In particolare vengono fatte alcune considerazione le quali permettono di definire a priori alcuni parametri della rete neurale rendendo il problema di addestramento lineare. Si descrivono due algoritmi di addestramento, il primo veloce ma computazionalmente dispendioso perciò adatto per un'implementazione offline mentre il secondo idoneo ad un addestramento online. Infine, per concludere l'identificazione parametrica del motore, si propone uno schema basato sull'iniezione di una corrente continua il quale permette di stimare la resistenza di statore indipendentemente da tutti gli altri parametri della macchina. L'indipendenza parametrica permette un notevolmente miglioramento nell'accuratezza di stima del modello magnetico ottenuto con la rete neurale. La seconda parte di questa tesi, invece, tratta il controllo del motore e come sia possibile migliorarne le performance utilizzando il modello identificato. Innanzitutto, per incrementarne l'efficienza si presenta un innovativo metodo per trovare la curva a massima coppia per corrente. La tecnica proposta lavora in stretta simbiosi con l'identificazione del modello magnetico in quanto è in grado di capire dove si trova la curva cercata rispetto all'attuale punto di lavoro sfruttando la stima locale dei flussi magnetici. Identificata la direzione di movimento, l'azionamento continuamente muove il punto di lavoro coerentemente. Infine, si propongono tre diversi controlli di corrente pensati per gestire un motore fortemente non lineare, tutti basati sul modello stimato. Il primo è un controllore proporzionale-integrale nel quale i parametri vengono modificati al variare del punto di lavoro con lo scopo di mantenere la dinamica della corrente di motore costante. Il secondo è anch'esso basato su un controllore proporzionale-integrale ma a guadagni costanti accoppiato ad un'azione di feed--forward la quale compensa tutte le non linearità presenti nella mappa magnetica. Infine, il terzo è un controllo predittivo il quale determina direttamente la posizione degli switch tali per cui la funzione di costo è minimizzata. All'interno del controllo, è inserito un vincolo sulla corrente massima e si utilizza un particolare algoritmo per ottenere un lungo orizzonte di predizione. Tutti i metodi presentati nella tesi sono stata verificati attraverso dettagliate simulazioni e prove sperimentali, eccezione fatta per il controllo predittivo il quale è stato testato attraverso simulazioni.
Smart cities and communities are conjugated by European Union in different areas, including energy efficiency, low carbon technologies and mobility which are deeply merged with electric motors. Electric machines are ubiquitous in industry for a wide range of applications, consuming between 43% and 46% of all electricity that is generated in the world. Although some machines are used for high-performance applications, such as robots and machine tools, the majority are used in industrial processes for pumps, compressors, fans, conveyors, and other slower-dynamic applications. It is estimated that 92% - 95% of the life cycle costs of electric motors are associated with the energy they consume, leading to typical payback periods of < 2 years for the installation of an adjustable-speed drive. It is rather surprising to learn that, despite overwhelming evidence of the attainable savings, only 10% - 15% of all industrial motors presently use electronic adjustable speed drives. On the motor side, Synchronous Reluctance (SynR) motors are gaining lots of attention from industrial researchers and academics, due to their inherent characteristics like the high efficiency, the low cost and the low environmental footprint. Their characteristics fully meet the requirements imposed by smart cities and communities and the aforementioned low-dynamics applications, so they could be the heart of the revamping of those plants. There is wide agreement that the potential for future growth in the sales of industrial drives and SynR motors is still very substantial. SynR motors are prone to magnetic saturation, making the classic model with lumped parameters unsuitable. The main part of this thesis concerns the development of a new magnetic model for anisotropic motors, especially for SynR motors. It is based on a special kind of neural network (NN), called Radial Bases Function (RBF) NN, which is particularly advisable for an online updating due to its local property. A complete training procedure is proposed in which some considerations are done to define several NN parameters and to convert the nonlinear training problem into a linear one. Two different training algorithms are presented, the former one is fast but computationally cumbersome then suitable for an offline training while the latter one is lighter then proper for an online training. In order to complete the online parameters identification, a scheme based on a DC current injection is developed to estimate the stator resistance. An exhaustive analysis is carried out to disclose that the proposed method is independent from other motor parameters which is a strength asset in a saturable motor. An accurate stator resistance value improves in turn of the magnetic model. The second part of this dissertation deals with how to exploit an accurate magnetic model to enhance the motor control. In order to improve the efficiency of the motor, exploiting the RBF NN model and the online training algorithm, the Maximum Torque per Ampere (MTPA) curve is found. Starting from a blank NN, it is continuously online trained and a proper algorithm understands where the MTPA curve is respect to the current working point. Afterwards, the drive moves itself towards the actual MTPA. Finally, three different current control schemes tailored for anisotropic motors are presented, all based on the available NN-based magnetic model. The first one is a gain-scheduling PI control where the control gains are accordingly tuned to the working point to keep constant the control bandwidth. The second one is based on a classical PI regulator with a FF action to compensate for all the nonlinearity of magnetic maps. The third one is a constrained direct Model Predictive Control (MPC) where a long prediction horizon is achieved. In order to accomplish a long prediction horizon, the Sphere Decoding Algorithm is properly modified to make it suitable for a nonlinear system. The whole thesis was fully validated through an intensive simulation and experimental stage, except the long--horizon MPC which was tested only by simulation.

The research presented in this thesis involves different aspects related to advanced control methodologies and self-commissioning identification algorithms in modern electrical drives. The theoretical study and the validation of the results obtained were performed in the three years of Ph.D. at the Electric Drives Laboratory in the Department of Management and Engineering of the University of Padova, (VI) Italy. The research topics were mainly three, all related to the implementation and development of advanced controls for electric drives, aimed at a more efficient use of the electric machines in the modern mechatronic applications. The demand of electric drives capable of guarantee high-performance and flexible enough to update in real time the parameters involved in the control algorithm are the motivation of the present research, as well as the meshing or replacement of standard or obsolete control techniques with modern ones, able to fully exploit the new hardware resources. In order to contextualizes and motivate the choice of the present research in the world scenario, a comprehensive bibliographic framework can be found in the introduction of each chapter of the thesis. The part one of the thesis presents two new control architectures for Permanent Magnet Synchronous Motors, that is a type of electric machine notoriously appreciated by both academia and industry for its flexibility of use and controllability. To this aim, in Chap.2 is proposed a non-linear control algorithm for the automatic search of the Maximum Torque Per Ampere (MTPA) operating condition for Permanent Magnet Synchronous Motors with anisotropic structure, to be integrated in a conventional Field Oriented Control scheme. The exhaustive convergence and stability analysis performed in order to derive a new and original tuning method of the controller (proven by numerous experimental evidences) is definitely one of the distinguishing features in this research topic. In parallel to the first topic, for the same type of motor has been investigated and developed (first analytically and then by simulation) a speed and current Direct Predictive Control with Hierarchical decisional structure. Unlike the traditional control techniques, the proposed Direct Predictive Control with modified hierarchical control structure has a faster dynamic and the capability to impose different operating conditions aimed at the energy efficiency optimisation. The on-line execution of the algorithm required for the experimental validation, has become possible thanks to the adoption of a control platform based on FPGA logic (Chap.3). In fact, the processing speed provided by these devices, released from the execution of sequential instructions (typical of the architecture of the microprocessors), ensures an execution time of the algorithm contained in a few us. The part two of the thesis (i. e. Chap.5) presents an innovative technique of parameter identification for induction motors, capable of estimating the parameters of the equivalent inverse-Gamma electric circuit completely at standstill. As known, the saturations in the parameters of the magnetic circuit of the induction motor and the relative nonlinearities, deteriorate the performance of the standard sensored or sensorless vectorial controls. The studied self-commissioning procedure addresses and solves many problems related to the estimate of the non-linearity of the parameters, and then it can be considered as an evolution of the classical identification techniques in the literature. The practical feasibility, doubly validated by numerous experimental tests and by many finite element simulations on three different induction motors, concludes the chapter and proves definitely the method.
La ricerca presentata in questa tesi coinvolge molteplici aspetti che si legano alle più recenti metodologie di controllo studiate per azionamenti elettrici di ultima generazione. Lo studio teorico e la validazione in ambito sperimentale sono il frutto del lavoro svolto nel triennio di dottorato presso il laboratorio di azionamenti elettrici del Dipartimento di Tecnica e Gestione dei Sistemi Industriali dell’Università degli Studi di Padova. I temi di ricerca trattati sono principalmente tre, tutti legati alla realizzazione e allo sviluppo di algoritmi di controllo innovativi, capaci di incrementare l’efficienza e le prestazioni delle macchine elettriche di ultima generazione per applicazioni meccatroniche. Azionamenti elettrici in grado di garantire elevate prestazioni ma sufficientemente flessibili da aggiornare in tempo reale i diversi parametri coinvolti nel algoritmo di controllo sono il filo conduttore e la motivazione della presente ricerca, così come la sostituzione di logiche di controllo standard o obsolete con nuove architetture di controllo capaci di sfruttare le più recenti innovazioni hardware. Al fine di contestualizzare e motivare la ricerca condotta nel panorama mondiale, nell’introduzione di ciascun capitolo è inserito un esaustivo inquadramento bibliograficoinerente inerente il problema affrontato. La prima parte della tesi presenta due nuove architetture di controllo per motori sincroni a magnete permanente, tipologia di macchina elettrica notoriamente apprezzata dal mondo accademico e industriale sia per la sua flessibilità d’uso che per la sua facile controllabilità. In tal senso, nel Capitolo2 è descritto e formalizzato un controllo non lineare per motori sincroni a magnete permanente anisotropi, inseribile in schemi di controllo convenzionali ad orientamento di campo per ottenere la condizione di funzionamento a massima coppia su corrente (MTPA). L’esaustiva analisi di convergenza e stabilità condotta al fine di ottenere un nuovo ed originale metodo per la sintonizzazione del regolatore (comprovato da numerose evidenze sperimentali) è sicuramente una delle caratteristiche distintive per questo ramo della ricerca. Per la stessa tipologia di motore è stato poi sviluppato un controllo predittivo a stati finiti con struttura decisionale gerarchica. A differenza delle tecniche di controllo tradizionali, la soluzione studiata garantisce una dinamica veloce e la possibilità di imporre condizioni operative diverse, volte all’ottimizzazione e all’incremento dell’efficienza energetica. L’esecuzione on-line di tale algoritmo per le verifiche sperimentali si è resa fattibile grazie all’adozione di una piattaforma di controllo basata su logica FPGA (Capitolo3), in quanto la velocità di calcolo offerta da tali dispositivi, svincolata dall’esecuzione sequenziale delle istruzioni tipica dei microprocessori, garantisce tempi di esecuzione dell’algoritmo contenuti a pochi us. Nella seconda parte della tesi (Capitolo5) è presentata un innovativa tecnica di identificazione parametrica per motori asincroni, capace di stimare i parametri del circuito equivalente a Gamma-inverso del motore asincrono, a rotore fermo. Come noto, le saturazioni del circuito magnetico della macchina e le non linearità ad esso associate deteriorano le performances nei normali controlli vettoriali sensored e soprattutto sensorless. Il metodo di identificazione parametrica studiato affronta e risolve molti problemi connessi alla stima delle non linearità dei parametri, configurandosi a tutti gli effetti come un evoluzione delle classiche tecniche di identificazione presenti in letteratura. La fattibilità pratica del metodo, validata con innumerevoli prove sperimentali e simulazioni agli elementi finiti su tre diversi motori ad induzione, conclude il capitolo e prova in modo definitivo la realizzabilità del metodo.

This text focuses on advanced torque control of permanent magnet synchronous motor drives. A novel modular structure is introduced to simplify the design and implementation of Model Predictive Control (MPC). The layout consists of the control and the control framework. The dynamic control is the novel virtual flux controller, which is used to reach desired reference values, and the state observer, which is used to reduce effects of non-modeled system properties. The control framework consists of static mappings to simplify the control problem. Besides the alpha-beta and d-q transformations, a reference generation procedure is used to generate state references based on optimality criteria. Also, the actuation scheme is part of the control framework and defines the available input set and the resulting control properties. The first method actuates directly switch states, i.e. voltage vectors, which yield an integer set named Finite Control Set (FCS). The other method actuates duty cycles via modulation, which yield the Convex Control Set (CCS). A stability analysis is carried out for both, CCS-MPC and FCS-MPC. MPC is called stable, if it is feasible and convergent, which can be ensured using the main MPC stability theorem. However, stringent computation requirements make it difficult to apply the theorem in practice. Thus, the Lyapunov based MPC approach is applied to the motor drive, which provides stability guarantees independent of the prediction horizon. A stability constraint based on control Lyapunov functions (CLF) ensures convergence to the origin and the resulting optimal control problem is shown to be feasible for all time. In other words, a control input can be found at each sampling instant, which satisfies all constraints and yields a stable closed-loop system. The properties of CCS-MPC are derived using a nonlinear controller and the constrained closed-loop system is shown to be stable in the sense of Lyapunov. The stability properties of FCS-MPC are more complex due to the integer input set. Using set-theoretical methods, it is shown that a sufficiently large control error can be steered towards the origin. In other words, the proposed FCS-MPC is shown to be set stable, i.e. the control error is guaranteed to converge to a well-defined neighborhood of the origin. MPC requires that a Constrained Finite Time Optimal Control (CFTOC) problem is solved at each sampling time. Small sampling periods and limited computation capabilities of embedded hardware require the CFTOC to be sufficiently simple, which is achieved using the virtual flux model in the static reference frame. The problem size is contained using a sufficiently small prediction horizon and efficient algorithms are necessary to provide a result within a sampling period. The CFTOC of the proposed CCS-MPC is a (convex) linear or quadratic programming problem, which can be solved using existing efficient algorithms. To provide a minimal approach, an efficient algorithm is introduced to solve the one-step-ahead prediction CFTOC analytically. FCS-MPC results in a mixed integer programming problem and is therefore more difficult to solve with standard numerical methods. In practice, the CFTOC is solved by enumeration, which is combined with branch-and-bound, i.e. branch-and-cut, techniques to improve the computational efficiency. The control algorithms have been developed on a Software-in-the-Loop (SiL) platform based on Matlab/Simulink and the code is implemented without modification on an experimental test-bench. The evaluation confirms the design and implementation of CCS-MPC and FCS-MPC and shows good results in dynamic and steady-state operation. The two MPC approaches have complimentary properties, which can be used to target different applications. CCS-MPC achieves a constant switching frequency and is a promising alternative to proportional-integral (PI) vector control. The concept can be combined with different modulation schemes, e.g. the Symmetric Space Vector Modulation (SSVM) and the Discontinuous Space Vector Modulation (DSVM) are used in this text. FCS-MPC takes the inverter switching into account and achieves an approximately constant switching ripple but a variable switching frequency. The concept is most profitably applied to systems where a high sampling frequency compared to the switching frequency is desired, e.g. high power or servo drives. Moreover, FCS-MPC lacks Pulse Width Modulation (PWM) harmonics in its current spectrum. Consequently, it is advantageous in terms of acoustic noise since emphasized tones are missing. However, the distinguished PWM harmonics of CCS-MPC are simpler to filter. In summary, it can be said that the work on advanced torque control of permanent magnet synchronous motor drives has produced an innovative strategy. The introduction of a new structure has significantly simplified the model predictive control problem, the concept of stability in particular. Moreover, this structure results in the implementation of simple algorithms, which can be computed efficiently.
Il soggetto affrontato dal presente lavoro sono i controlli avanzati di coppia per azionamenti con un motore sincrono a magneti permanenti. A questo scopo, è stata introdotta una struttura modulare che semplifica la progettazione e l’implementazione del controllo predittivo basato su un modello (model predictive control, MPC): lo schema è costituito dal controllo dinamico e dal quadro di controllo. Il controllo dinamico è un regolatore di flusso virtuale, utilizzato per raggiungere un valore di riferimento voluto e un osservatore di stato che serve a ridurre gli effetti delle proprietà non modellizzate del sistema. Il problema del controllo è stato semplificato tramite l’utilizzo di trasformate statiche chiamate quadro di controllo. Accanto alle trasformate alpha-beta e d-q viene usata una procedura per la generazione di riferimenti di stato, basati su un criterio ottimale. Il quadro di controllo contiene anche lo schema di attuazione, che serve per definire l’insieme di ingressi disponibili. Da un lato, il controllore comanda in modo diretto l’accensione e lo spegnimento dei semiconduttori, ovvero i vettori di tensione, ottenendo un insieme finito d’ingressi (Finite Control Set, FCS). Dall’altro lato vengono attuati cicli di accensione (duty-cycles) attraverso una modulazione (pulse width modulation, PWM): ciò risulta in un insieme convesso d’ingressi (convex control set, CCS). È stata eseguita un’analisi di stabilità sia per CCS-MPC sia per FCS-MPC. MPC è stabile, se il problema di controllo ottimale ad esso associato è risolvibile e l’errore di stato converge all’origine. Tale stabilità può essere garantita attraverso il principale teorema di stabilità di MPC. Tuttavia, i requisiti di calcolo restrittivi rendono il teorema difficilmente applicabile nella pratica. Di conseguenza, viene introdotto l’approccio MPC basato su Lyapunov (Lyapunov-based MPC) per gli azionamenti, il quale fornisce garanzie sulla stabilità indipendentemente dall’orizzonte di predizione. Un vincolo di stabilità basato sulle funzioni di controllo di Lyapunov (control Lyapunov function, CLF) assicura la convergenza all’origine ed è stato provato che il problema ottimale di controllo risultante è sempre risolvibile. In altre parole, ad ogni istante di campionamento si può trovare un ingresso che soddisfi tutti i vincoli del sistema e renda stabile il sistema a circuito chiuso. Le proprietà di CCS-MPC vengono ottenute utilizzando un controllo non lineare ed è dimostrato che il sistema vincolato ad anello chiuso è stabile secondo Lyapunov. Le proprietà di stabilità di FCS-MPC sono più complesse a causa dell’insieme non continuo d’ingressi. Utilizzando metodi della teoria degli insiemi si dimostra che un errore di controllo sufficientemente ampio può essere diretto verso l’origine e tenuto in un dintorno dell’origine ben definito. MPC richiede che in ogni istante di campionamento si risolva un problema di ottimizzazione (constrained finite time optimal control, CFTOC). La limitata potenza di calcolo dei microcontrollori e la brevità dei periodi di campionamento richiedono un CFTOC relativamente semplice, che si può ottenere utilizzando un modello di flusso virtuale nel sistema statico di riferimento. Scegliendo piccoli orizzonti di predizione si limita la dimensione del CFTOC, la cui risoluzione necessità di algoritmi efficienti, che permettano di ottenere un risultato all’interno di un periodo di campionamento. Il CFTOC di CCS-MPC è un programma (convesso) lineare o quadratico (linear program, lp; quadratic program, qp) che può essere risolto tramite algoritmi efficienti e noti. Al fine di elaborare una strategia di tipo minimalista, viene introdotto un algoritmo efficiente che risolve analiticamente il problema con un orizzonte di predizione di un passo. Il CFTOC di FCS-MPC è un problema di programmazione lineare o quadratico a numeri misti interi (mixed-integer) ed è quindi più difficile da risolvere con metodi numerici standard. In pratica si calcolano tutte le soluzioni possibili, tra le quali viene scelta la soluzione ottimale. Per migliorare l’efficienza di calcolo si combina quest’approccio con tecniche branch-and-bound e branch-and-cut. Gli algoritmi di controllo sono stati sviluppati su una piattaforma software-in-the-loop (SiL) basata su Matlab/Simulink e il codice di programmazione è stato implementato su un banco di prova sperimentale, senza modifiche. La valutazione approva la progettazione e la realizzazione di CCS-MPC e FCS-MPC e indica buoni risultati sia nell’operazione dinamica che in quella stazionaria. I due approcci MPC hanno proprietà diverse che risultano vantaggiose per applicazioni differenti. CCS-MPC ha una frequenza di commutazione costante ed è un’alternativa promettente al controllo vettoriale proporzionale-integrale (PI). Il concetto può essere combinato con diversi schemi di modulazione, nella fattispecie si usa la modulazione simmetrica di vettori spaziali (symmetric space vector modulation, SSVM) e la modulazione discontinua di vettori spaziali (discontinuous space vector modulation, DSVM). FCS-MPC tiene conto della commutazione dell’inverter e raggiunge all’incirca un ripple di commutazione costante, ma ottiene una frequenza di commutazione variabile. Il concetto è vantaggioso per sistemi dove è richiesta un frequenza di campionamento alta rispetto alla frequenza di commutazione, per esempio azionamenti ad alta potenza o servoazionamenti. Inoltre, lo spettro della corrente di FCS-MPC non contiene armoniche PWM e di conseguenza è vantaggioso in termini di rumore acustico, data la mancanza di toni distinti. Tuttavia, le armoniche PWM distinte di CCS-MPC sono più semplici da filtrare. Si può concludere affermando che lo studio del problema dei controlli avanzati di coppia per azionamenti con un motore sincrono a magneti permanenti, ha portato all’individuazione di una strategia innovativa. L’introduzione di una nuova struttura di controllo ha semplificato notevolmente il problema di controllo predittivo, con particolare attenzione al concetto di stabilità. Inoltre, le implementazioni di tale struttura si sono rivelate particolarmente efficaci su piano computazionale.

Recently, the interest of the automotive market for hybrid vehicles has increased due to the more restrictive pollutants emissions legislation and to the necessity of decreasing the fossil fuel consumption, since such solution allows a consistent improvement of the vehicle global efficiency. The term hybridization regards the energy flow in the powertrain of a vehicle: a standard vehicle has, usually, only one energy source and one energy tank; instead, a hybrid vehicle has at least two energy sources. In most cases, the prime mover is an internal combustion engine (ICE) while the auxiliary energy source can be mechanical, electrical, pneumatic or hydraulic. It is expected from the control unit of a hybrid vehicle the use of the ICE in high efficiency working zones and to shut it down when it is more convenient, while using the EMG at partial loads and as a fast torque response during transients. However, the battery state of charge may represent a limitation for such a strategy. That’s the reason why, in most cases, energy management strategies are based on the State Of Charge, or SOC, control. Several studies have been conducted on this topic and many different approaches have been illustrated. The purpose of this dissertation is to develop an online (usable on-board) control strategy in which the operating modes are defined using an instantaneous optimization method that minimizes the equivalent fuel consumption of a hybrid electric vehicle. The equivalent fuel consumption is calculated by taking into account the total energy used by the hybrid powertrain during the propulsion phases. The first section presents the hybrid vehicles characteristics. The second chapter describes the global model, with a particular focus on the energy management strategies usable for the supervisory control of such a powertrain. The third chapter shows the performance of the implemented controller on a NEDC cycle compared with the one obtained with the original control strategy.

Dans un contexte de changements globaux, les scientifiques et les législateurs requièrent des outils leur permettant de traiter la question de la perte de biodiversité. L'analyse de viabilité de population (PVA) est l'outil principal pour traiter le problème. Cependant, les outils développés dans les années 90 n'intègrent que très peu les récents progrès réalisés en génétic du paysage et sur la compréhension de la dipsersion. Ici, j'ai développé une plateforme de modélisation flexible et modulaire pour réaliser des PVA qui palie à la plupart des limitations des logiciels existants et répondant de ce fait à l'appel fait par Evans et al. (2013) pour développer des modèles prédictifs des systèmes écologiques. MetaConnect est un modèle individu centré, basé sur le déroulement des processus biologiques et principalement basé sur la réalisation d'analyses de viabilités qui peut être utlisé à la fois comme un outil de recherche ou d'aide à la décision. Dans ma thèse, je présente le module central de MetaConnect et sa validation puis présente différentes application de cette plateforme à des fins théoriques et appliquées
In a context of global change, scientists and policy-makers require tools to address the issue of biodiversity loss. Population viability analysis (PVA) has been the main tool to understand and plan for this problem. However, the tools developed during the 90s poorly integrate recent scientific advances in landscape genetics and dispersal. Here, I developed a flexible and modular modeling platform for PVA that addresses many of the limitations of existing software and in this way answer the call made by Evans et al. (2013) for predictive systems ecology models. MetaConnect is an individual-based, process-based and PVA-based modeling platform which could be used as a research or a decision-making tool. In my thesis, I present the modeling base core of MetaConnect and its validation and then present different uses of this platform in theoretical and applied ecology

En nutrition animale, 56% des développements produits aboutissent à des échecs commerciaux (Source: auteur). Une enquête est menée en Europe auprès de cinquante et une personnes travaillant pour vingt-huit entreprises du secteur de l’alimentation animale. Une base de données est constituée à partir d’informations recueillies sur le processus de développement de cent vingt-cinq nouveaux produits techniques parmi lesquels cinquante gros échecs commerciaux, quarante-sept grands succès et vingt-huit succès commerciaux mitigés. L’Analyse en Composantes Principales permet de dépister les treize principaux facteurs explicatifs du résultat commercial. Ces derniers sont liés aux compétences et aux connaissances de l’entreprise et du personnel, au mix-marketing et à l’attention portée au projet de développement par la Direction de l’entreprise.Un modèle économétrique est mis au point à partir des cinq indicateurs suivants (notés sur une échelle de zéro à dix): degré d’innovation du produit et capacité d’adaptation de celui-ci au marché local, implication de la Direction dans le projet de développement produit, connaissance de la concurrence, attention portée au client quant au bénéfice financier que ce dernier peut retirer de l’achat du nouveau produit. Le modèle permet de prédire la performance commerciale du nouveau produit sous forme d’une probabilité de succès commercial. L’emploi de celui-ci et le contrôle des treize principaux facteurs explicatifs du résultat commercial constituent un outil d’aide à la décision en matière de lancement de nouveaux produits en nutrition animale
In animal nutrition, 56% of product developments fail (source: author). An empiric field study is conducted in Europe in collaboration with fifty-one persons working for twenty-eight companies involved in animal nutrition. Information regarding the development process of one hundred and twenty-five new technical products, among them fifty big commercial failures, forty-seven great successes and twenty-eight medium successes, are collected to create a data-base. Thank’s to a “Principal Component analysis”, thirteen explicative factors are detected as responsible of the commercial result. These ones are related to skills and the knowledge status of the company and the employees, to marketing-mix and to the attention paid by the management team to the new product development.An econometric model is developed from the five following indicators recorded in the form of a score from zero to ten: innovation degree and adaptation capacity to the local market, involvement of the management team in the project of product development, knowledge of the competitors, attention to grant to the financial benefit that the client could get in case this one purchases the new product. The model allows to predict the commercial performance in the form of a probability of commercial success. The use of this econometric model together with the control of the thirteen main explanatory factors of the commercial result constitute a tool for decision-making of new product launch devoted to animal nutrition

In recent years, social media platforms such as Twitter and Facebook have emerged as effective tools for broadcasting messages worldwide during disaster events. With millions of messages posted through these services during such events, it has become imperative to identify valuable information that can help the emergency responders to develop effective relief efforts and aid victims. Many studies implied that the role of social media during disasters is invaluable and can be incorporated into emergency decision-making process. However, due to the "big data" nature of social media, it is very labor-intensive to employ human resources to sift through social media posts and categorize/classify them as useful information. Hence, there is a growing need for machine intelligence to automate the process of extracting useful information from the social media data during disaster events. This dissertation addresses the following questions: In a social media stream of messages, what is the useful information to be extracted that can help emergency response organizations to become more situationally aware during and following a disaster? What are the features (or patterns) that can contribute to automatically identifying messages that are useful during disasters? We explored a wide variety of features in conjunction with supervised learning algorithms to automatically identify messages that are useful during disaster events. The feature design includes sentiment features to extract the geo-mapped sentiment expressed in tweets, as well as tweet-content and user detail features to predict the likelihood of the information contained in a tweet to be quickly spread in the network. Further experimentation is carried out to see how these features help in identifying the informative tweets and filter out those tweets that are conversational in nature.

Innovation is one of the major concerns for technology-driven organizations. Extensive amount of literature is available on ‘managing innovation’ within organizations at individual level as well as firm level. However, ‘innovation’ today is emerging as a ‘team-level’ phenomenon and organizations are increasingly using ‘teams’ as an important tool to harness employee talents.Unfortunately, the existing literature on ‘managing innovation’ at team-level, seems to suffer from several inconsistencies in definition, conceptualization and modelling of the phenomenon. This current research work is an attempt to explore some of these inconsistenciesand improve upon them in order to build long-term ‘innovation capabilities’ in research teams operating in technology-driven organizations. A multivariate cross-level conceptual model has been proposed to identify those "organizational-level" factors which significantly impact ‘innovation capabilities’ of research teams operating in government-funded research organizations in India. This top-down cross-level model assesses the impact of five organizational-level (higher-level) factors (leadership, culture, structure, knowledge, network) on team-level (lower-level) innovation capability through two team-level mediating factors (‘focus’ and ‘intensity’ of team’s efforts towards innovation). Stratified sampling technique was used to collect responses through an online questionnaire from 136 research teams operating in some of the most elite government-funded research organizations pursuing cutting-edge academic or industrial research. Factor Analysis (FA) and Structural Equation Modelling (SEM) was used to identify those factors which tend to significantly influence the ‘innovation capability’ of research teams. Results were validated byusing the‘triangulation’ approach through descriptive case studies. Triangulation helped us in contrasting and comparing the results acrossacademic and industrial research organizations. Findings revealed the composition of ‘team-level innovation capability’ as a bivariate construct constituted of aspects such as ‘manifestation’ and ‘customer-orientation’. Importance of mediators such as ‘team focus’ and ‘team intensity’ while mediating the impact of organizational-level ‘cardinal’ factors such as ‘leadership’ and ‘expertise’ on ‘team-level innovation capability’ was found to be significant. The top-down influence of organizational-level ‘flexibility’ for academic research organizations and ‘knowledge protection’ for industrial research organizations was also found to be significant. Conceptualization and validation of this multivariate cross-level model to predict ‘team-level innovation capability’ can be considered as an important theoretical contribution of this study.Methodologically, it integrates the findings from theoretical literature, statistical analysis and empirical case studies to come up with a robust framework to predict innovation capability at team levels.Construction ofthis robustand empirically validated modelwhich can be used by theTop Management Teams (TMTs) at technology-driven organizations for diagnostic, interventional and organizational developmental purposesmay be considered as one of the major practical contributions of this study.

碩士
國立臺灣大學
資訊管理學研究所
104
There are about 20% coopetition in all alliances and more than 85% of companies who did coopetition is belong to high-tech industry. It shows coopetition plays an important role for contemporary business, especially in high-tech industries. Many researches show that coopetition brings positive effect on innovativeness (Belderbos, Carree and Lokshin, 2004; Quintana-Garcia and Benavides-Velasco, 2004). The appropriate selection of coopetition targets for a given bidder company constitutes a critical first step for an effective coopetition activity. Yet existing studies employ financial and network indicators when constructing inter-firm relationship prediction models (Tsakanos, Georgopoulos, & Siriopoulos, 2007; Schilling & Phelps, 2007). Even though some considered technological indicators, most of them performed qualitative researches (e.g., questionnaire and interviews) rather than quantitative research. For example, Hall & Ziedonis (2001) estimated the patenting behavior by structure questions and a follow-up survey. Due to the importance of coopetition and many researches show the effect of technological indicators on innovation performance, our study developed an automated prediction model for predicting the innovation performance resulting from a coopetition. Our evaluation results, on the basis of the coopetition cases between January 1990 to December 2015 that involve companies in high-tech industries (i.e., ICT, software and pharmaceutical industries). With defined technological, financial and network indicators, we developed an innovation performance prediction model with 0.863 correlation coefficient. We proved that the technological variables are effective for this prediction task, and we investigated the incorporation of network variables successfully improve the prediction effectiveness.

Poster presentation from the Living the Future 6 Conference, April 5-8, 2006, University of Arizona Libraries, Tucson, AZ.
The Access & Delivery team was charged with providing the UA Libraries with a report summarizing recommended long and short-term strategies in directing the library for the future. This poster illustrates the path that the Access & Delivery team took to arrive at recommendations and strategies for the UA Library’s strategic plan.

碩士
國立中央大學
統計研究所
105
The stock returns predictability is always a popular issue in the financial market, and the Bonferroni Q-test proposed by Campbell and Yogo (2006) has been a common and general method. But in recent years, the researcher in Phillips (2014) proposed that this method is not always valid in some situations. Therefore, we need to use the different way to get the relative confidence intervals which are needed in Bonferroni Q-test procedure. We then define an estimator of the nuisance parameter and set a boundary to distinguish the time when to use the different confidence intervals, and from this, we can complete the whole predictive Bonferroni Q-test. Then we use Monte Carlo to progressively verify our composite testing method. Beyond that, the general predictive tests usually have a normal assumption, this assumption is not satisfied the practical financial data. We all know that economic and financial data have high persistent and heavy tail, so we focus on the case that the data are near unit root. And we relax the normal restriction to a heavy-tailed assumption even infinite variance to do the predictive test so that the analysis is more corresponding to the real data. We also can have the result from Monte Carlo that our composite method is valid and precise under the heavy-tailed assumption. According to the empirical analysis using the U.S. equity data, we find reliable evidence for predictability of the earnings-price ratio, and the other predictor all have high persistence and heavy-tailed property. From the empirical results, we can conclude that unlike the normal assumption in the test before, our heavy-tailed assumption in this predictive test is more corresponding to the data.

碩士
元智大學
工業工程與管理學系
104
The compition among the panel glass manufacturers is drastically increasing. The official reconition of its ability to produce a new type of new product is crutial to the panel glass munfacturer. The most time comsuming process during the reconition certifying process is the glass glazing innovation process. Every panel manufactuer has a set of suppliers that jointly develop the glazing material for a new product. The development process can range from 6 weeks to 6 months. Therefore, the speedy joint innovation for glassing is an important competitive edage for a panel glass manufacturer. This study is conduct to develop a systematic way to elvaluate the efficiency of suppliers, and focus on assisting the best joint innovating supplier through the joint innovating process. We find that using tanditional methods such as suppliers’historial perfomuse or objective factors such as work force of R& D department cannot accurately predict the perfomuse for a new produce development speed. This study utilizes the latent class analysis to build a model that can predict weather a specific supplier is suitable to develop the given new product. In the numerical case study, for a given new product, we collected 24 weeks lab data for each work-in-process samples provided by 9 suppliers. We analysis those data with latent class analysis procedure and successively construct a perdiction model. This model allows the panel glass manufacturer to select a group of best suppliers for developing process based on 4 weeks work-in-process sample data.

碩士
國立成功大學
製造資訊與系統研究所
106
Business management involves many aspects. Innovation is the driving force for enterprise progress. Before product innovation, companies must understand consumer needs to provide products that consumers need. Traditional companies understand consumer needs through time-consuming questionnaires and statistical analysis. With the rise of social media and the popularity of smart phones and internet, more and more users will leave their thought on products in the web social media. The product development based on social analysis can only evaluate the current situation of the product in the market, and cannot provide the exact development object for the enterprise reference. Therefore, how to analyze valuable product development objects from social media materials to provide product development and design to enhance the competitive advantage in the product market is an important issue. Therefore, the main purpose of this research is to develop a product innovation prediction mechanism based on product-related E-WOMs to help enterprises quickly grasp product development trends and objects as an important reference for product development. For the above purposes, the main research projects of this study include: (i) designing a social analytics-based product innovation prediction methodology, (ii) developing the enabling technology of the proposed product innovation prediction method and (iii) implementing and verifying social analytic-based product innovation prediction systems.

Predicting innovation is a peculiar problem in data science. Following its definition, an innovation is always a never-seen-before event, making the usual approach of learning patterns from the past a useless exercise. Here we propose a strategy to address the problem in the context of innovative patents, by defining innovation as never-seen-before associations of technologies. We think of technological codes present in patents as a vocabulary and the whole technological corpus as written in a specific, evolving language. We leverage such structure with techniques borrowed from Natural Language Processing by embedding technologies in a high dimensional euclidean space where relative positions are representative of learned semantics. Dynamics on this space predicts specific innovation events, that are tested against null models. These methods provide a completely new way of understanding and forecasting innovation, by tackling it from a revealing perspective and opening interesting scenarios for a number of applications and further analytical approaches.

碩士
國立清華大學
資訊工程學系所
106
Reliability concerns of a chip are worsening due to the downscaling technology. Among the reliability issues, aging effect is the dominant concern since it degrades the circuit performance over time. Traditionally, run-time monitoring approaches are proposed to estimate aging effects. However, such techniques tend to predict and monitor delay degradation status for circuit mitigation measures rather than the chip health condition. In this paper, we propose an aging-aware chip health prediction methodology which adapts to PVT (Process, supply Voltage, and Temperature) variations and workload conditions. Our prediction methodology adopts an innovative on-chip delay monitoring strategy by tracing representative aging-aware delay behavior. The delay behavior is then fed into a machine learning engine to predict the age of a tested chip. Experimental results show that our strategy can obtain 97.12% accuracy with 4.62% area overhead on average compared with Monte Carlo simulation. To the best of the authors’ knowledge, this is the first work that accurately predicts current chip age and provides future chip health information.

碩士
元智大學
工業工程與管理學系
104
In these few years, because of the high amount of companies and products that keep increasing, the understanding of the market trend has become an important issue for companies to survive in the competitive environment. Fortunately, with the development of internet, people are used to share every single thing on the popular social media platform such as Twitter, and we can know their opinion and deduce the market trend by analyzing these data. In this research, we propose a framework called Product Improvement Framework based on Social Media Analysis (PIF-SMA) to extract the data from the popular social media platform and make it useful by using word2vec which is created by Google. Considering that the data extracted from the social network contains mess information, we make the system smarter by processing the data using ontology. Due to the nature of the data we extracted from social network is huge, the framework is built on the Hadoop and Spark platform which is the most popular cluster system platform in the big data field to make the PIF-SMA framework work in higher efficiency even close to real time processing. By analyzing social media, the PIF-SMA framework proposes to provide an easy way for company not only to understand the future of market trend but also to know the functions which customers will care about in the product.

碩士
世新大學
資訊管理學研究所(含碩專班)
102
The study displays the result that the prediction system created by adaptive fuzzy neural network is better than the prediction system created by back-propagation neural network. Experimental results show that a single indicator SelfPoint1 have the best predictive power, dual technical indicators MACD + SelfPoint1 have the best predictive ability and multiple technical indicators MACD + W% R + SelfPoint1 have the best predictive ability, most indicators profit forecast accuracy and rewards relevant, high degree of accuracy of technical indicators, finally, Use neural network is better than the no neural network.