Dissertations / Theses on the topic 'Neural fuzzy network'
To see the other types of publications on this topic, follow the link: Neural fuzzy network.
Create a spot-on reference in APA, MLA, Chicago, Harvard, and other styles
Consult the top 50 dissertations / theses for your research on the topic 'Neural fuzzy network.'
Next to every source in the list of references, there is an 'Add to bibliography' button. Press on it, and we will generate automatically the bibliographic reference to the chosen work in the citation style you need: APA, MLA, Harvard, Chicago, Vancouver, etc.
You can also download the full text of the academic publication as pdf and read online its abstract whenever available in the metadata.
Browse dissertations / theses on a wide variety of disciplines and organise your bibliography correctly.
1
Brande, Julia K. Jr. "Computer Network Routing with a Fuzzy Neural Network." Diss., Virginia Tech, 1997. http://hdl.handle.net/10919/29685.
Full textAbstract:
The growing usage of computer networks is requiring improvements in network technologies and management techniques so users will receive high quality service. As more individuals transmit data through a computer network, the quality of service received by the users begins to degrade. A major aspect of computer networks that is vital to quality of service is data routing. A more effective method for routing data through a computer network can assist with the new problems being encountered with today's growing networks.
Effective routing algorithms use various techniques to determine the most appropriate route for transmitting data. Determining the best route through a wide area network (WAN), requires the routing algorithm to obtain information concerning all of the nodes, links, and devices present on the network. The most relevant routing information involves various measures that are often obtained in an imprecise or inaccurate manner, thus suggesting that fuzzy reasoning is a natural method to employ in an improved routing scheme. The neural network is deemed as a suitable accompaniment because it maintains the ability to learn in dynamic situations. Once the neural network is initially designed, any alterations in the computer routing environment can easily be learned by this adaptive artificial intelligence method. The capability to learn and adapt is essential in today's rapidly growing and changing computer networks. These techniques, fuzzy reasoning and neural networks, when combined together provide a very effective routing algorithm for computer networks.
Computer simulation is employed to prove the new fuzzy routing algorithm outperforms the Shortest Path First (SPF) algorithm in most computer network situations. The benefits increase as the computer network migrates from a stable network to a more variable one. The advantages of applying this fuzzy routing algorithm are apparent when considering the dynamic nature of modern computer networks.Ph. D.
2
James, Keith. "Online adaptive fuzzy neural network automotive engine control." Thesis, Loughborough University, 2011. https://dspace.lboro.ac.uk/2134/9089.
Full textAbstract:
Automotive manufacturers are investing in research and development for hybridization and more modern advanced combustion strategies. These new powertrain systems can offer the higher efficiency required to meet future emission legislation, but come at the cost of significantly increased complexity. The addition of new systems to modernise an engine increases the degrees of freedom of the control problem and the number of control variables. Advanced combustion strategies also display interlinked behaviour between control variables. This type of behaviour requires a more orchestrated multi-input multi-output control approach. Model based control is a common solution, but accurate control models can be difficult to achieve and calibrate due to the nonlinear dynamics of the engines. The modelling problem becomes worse when some advanced combustion systems display nonlinear dynamics that can change with time. Any fixed model control system would suffer from increasing model/system mismatch. Direct feedback would help reduce a degree or error from model/system mismatch, but feedback methods are often limited by cost and are generally indirect and slow response. This research addresses these problems with the development of a mobile ionisation sensor and an online adaptive control architecture for multi-input multi-output engine control. The mobile ionisation system offers a cheap, fast response, direct in-cylinder feedback for combustion control. Feedback from 30 averaged cycles can be related to combustion timing with variance as small as 0.275 crank angle degrees. The control architecture combines neural networks and fuzzy logic for the control and reduced modelling effort for complex nonlinear systems. The combined control architecture allows continuous online control adaption for calibration against model/plant mismatch and time varying dynamics. In simulation, set point tracking could be maintained for combustion timing to 4 CAD and AFR to 4, for significant dynamics shifts in plant dynamics during a transient drive cycle.
3
Gabrys, Bogdan. "Neural network based decision support : modelling and simulation of water distribution networks." Thesis, Nottingham Trent University, 1997. http://ethos.bl.uk/OrderDetails.do?uin=uk.bl.ethos.387534.
Full text
4
Wang, Ziqing. "Fuzzy neural network for edge detection and Hopfield network for edge enhancement." Thesis, National Library of Canada = Bibliothèque nationale du Canada, 1999. http://www.collectionscanada.ca/obj/s4/f2/dsk1/tape7/PQDD_0005/MQ42458.pdf.
Full text
5
Hudgins, Billy E. "Implementation of fuzzy inference systems using neural network techniques." Thesis, Monterey, California. Naval Postgraduate School, 1992. http://hdl.handle.net/10945/23919.
Full text
6
Campbell, Jonathan G. "Fuzzy logic and neural network techniques in data analysis." Thesis, University of Ulster, 1999. http://ethos.bl.uk/OrderDetails.do?uin=uk.bl.ethos.342530.
Full text
7
Ara?jo, J?nior Jos? Medeiros de. "Identifica??o n?o linear usando uma rede fuzzy wavelet neural network modificada." Universidade Federal do Rio Grande do Norte, 2014. http://repositorio.ufrn.br:8080/jspui/handle/123456789/15249.
Full textAbstract:
Made available in DSpace on 2014-12-17T14:55:19Z (GMT). No. of bitstreams: 1
JoseMAJ_TESE.pdf: 3560157 bytes, checksum: 2f20316c7b980a74bdb7b82e97e3bb43 (MD5)
Previous issue date: 2014-03-24Conselho Nacional de Desenvolvimento Cient?fico e Tecnol?gico
In last decades, neural networks have been established as a major tool for the identification of nonlinear systems. Among the various types of networks used in identification, one that can be highlighted is the wavelet neural network (WNN). This network combines the characteristics of wavelet multiresolution theory with learning ability and generalization of neural networks usually, providing more accurate models than those ones obtained by traditional networks. An extension of WNN networks is to combine the neuro-fuzzy ANFIS (Adaptive Network Based Fuzzy Inference System) structure with wavelets, leading to generate the Fuzzy Wavelet Neural Network - FWNN structure. This network is very similar to ANFIS networks, with the difference that traditional polynomials present in consequent of this network are replaced by WNN networks. This paper proposes the identification of nonlinear dynamical systems from a network FWNN modified. In the proposed structure, functions only wavelets are used in the consequent. Thus, it is possible to obtain a simplification of the structure, reducing the number of adjustable parameters of the network. To evaluate the performance of network FWNN with this modification, an analysis of network performance is made, verifying advantages, disadvantages and cost effectiveness when compared to other existing FWNN structures in literature. The evaluations are carried out via the identification of two simulated systems traditionally found in the literature and a real nonlinear system, consisting of a nonlinear multi section tank. Finally, the network is used to infer values of temperature and humidity inside of a neonatal incubator. The execution of such analyzes is based on various criteria, like: mean squared error, number of training epochs, number of adjustable parameters, the variation of the mean square error, among others. The results found show the generalization ability of the modified structure, despite the simplification performed
Nas ?ltimas d?cadas, as redes neurais t?m se estabelecido como uma das principais ferramentas para a identifica??o de sistemas n?o lineares. Entre os diversos tipos de redes utilizadas em identifica??o, uma que se pode destacar ? a rede neural wavelet (ou Wavelet Neural Network - WNN). Esta rede combina as caracter?sticas de multirresolu??o da teoria wavelet com a capacidade de aprendizado e generaliza??o das redes neurais, podendo fornecer modelos mais exatos do que os obtidos pelas redes tradicionais. Uma evolu??o das redes WNN consiste em combinar a estrutura neuro-fuzzyANFIS (Adaptive Network Based Fuzzy Inference System) com estas redes, gerando-se a estrutura Fuzzy Wavelet Neural Network - FWNN. Essa rede ? muito similar ?s redes ANFIS, com a diferen?a de que os tradicionais polin?mios presentes nos consequentes desta rede s?o substitu?dos por redes WNN. O presente trabalho prop?e uma rede FWNN modificada para a identifica??o de sistemas din?micos n?o lineares. Nessa estrutura, somente fun??es waveletss?o utilizadas nos consequentes. Desta forma, ? poss?vel obter uma simplifica??o da estrutura com rela??o a outras estruturas descritas na literatura, diminuindo o n?mero de par?metros ajust?veis da rede. Para avaliar o desempenho da rede FWNN com essa modifica??o, ? realizada uma an?lise das caracter?sticas da rede, verificando-se as vantagens, desvantagens e o custo-benef?cio quando comparada com outras estruturas FWNNs. As avalia??es s?o realizadas a partir da identifica??o de dois sistemas simulados tradicionalmente encontrados na literatura e um sistema real n?o linear, consistindo de um tanque de multisse??es e n?o linear. Por fim, a rede foi utilizada para inferir valores de temperatura e umidade no interior de uma incubadora neonatal. A execu??o dessa an?lise baseia-se em v?rios crit?rios, tais como: erro m?dio quadr?tico, n?mero de ?pocas de treinamento, n?mero de par?metros ajust?veis, vari?ncia do erro m?dio quadr?tico, entre outros. Os resultados encontrados evidenciam a capacidade de generaliza??o da estrutura modificada, apesar da simplifica??o realizada
8
Huang, Ju-Yi, and 黃朱瑜. "RBF Based Neural Fuzzy Network." Thesis, 1999. http://ndltd.ncl.edu.tw/handle/14967502279792003745.
Full textAbstract:
碩士國立中興大學
機械工程學系
87
Abstract In this thesis, a new neural fuzzy configuration that combines the RBF neural network structure and fuzzy logic theory is proposed. In this new neural fuzzy structure, the conventional six layers neural fuzzy network is simplified to a four layers neural fuzzy network. For single input problem, this new network structure is a kind of RBF neural network. When a multi-inputs problem is applied, it functions similar a conventional neural fuzzy network. Computer simulation results show that the proposed new neural fuzzy scheme can be successfully applied to the nonlinear function approximation and classification problems. To fulfill the on-line training requirement, an efficient heuristic learning rule is included. Experimental results show that the proposed approach can be successfully applied to the precise regulating and tracking problems of an AC servo motor system. For real industrial application, a systematic approach to achieve global optimal CMP process is carried out. In this new approach, orthogonal array technique in the Taguchi method is adopted for efficient experiment design. The RBFNF neural-fuzzy is then used to model the complex CMP process. Signal to Noise Ratio (S/N) Analysis technique used in the conventional Taguchi method is also implemented to find the local optimal process parameters. Successively, the global optimal parameters are acquired in terms of the trained RBFNF network. In order to increase the CMP throughput, a two-stage optimal strategy is also proposed. Experimental results show that the two-stage strategy can perform better then the original approach even though the process time is reduced by 1/6.
9
Lo, Fang-Chun, and 羅方鈞. "Fuzzy Neural Network Based Face Recognition." Thesis, 2011. http://ndltd.ncl.edu.tw/handle/31012850840479560921.
Full textAbstract:
碩士國立暨南國際大學
電機工程學系
99
In recent years, image processing and robot vision technology continues to development, and the human face recognition has become an important technology; therefore, this thesis presents a human face recognition system based on fuzzy neural network. The local feature is used as the facial feature parameters.The face recognition method which is presented consists of two steps. First, using the Scale Invariant Feature Transform (SIFT) method to find the features from the pictures. The Scale Invariant Feature Transform (SIFT) method has four parts, Scale-space extrema detection, Keypoint localization, Orientation assignment, Keypoint descriptor. In the second step, we take all features as FNN inputs to make the final decision. In order to training fast that we use the method of binary tree, and build a binary tree. The experiment results display that the method can identification human face efficiently.
10
Hsu, Chung-Pin, and 徐崇濱. "Fuzzy Neural Network Based Face Locating." Thesis, 2010. http://ndltd.ncl.edu.tw/handle/14074241759215438268.
Full textAbstract:
碩士國立暨南國際大學
電機工程學系
98
This thesis proposes a robust human face locating based on color with fuzzy neural network. First, the color of pictures all transform from RGB color space to HSV color space and only hue and saturation are preferred to the needed color information to eliminate the effect of illumination. Then, get the color information of holes, which are included by segmented skin color, as features for fuzzy neural network. Use those features to classify eyes, mouths and other parts. This kind of way can reduce the region of color that needed to classify for improving the accuracy of recognition. At the same time, rotate the angle of hue belonging to eyes and mouths by 180 degree for getting the average values more accurate. After the eyes and mouths are recognized, pick the most possible relative positions for getting the most accurate human face location. Because this method simply uses the color for judgment, it needs not rely on the integrity of facial shape. Therefore, even if there are other skins connecting with face or something in front of face cause the face shape or facial feature shape to be changed, this method can still be used. As a result, shooting angle need not be completely right in the face and this method does not need continuous images to be discriminated.
11
Li, Zhao Ji, and 李昭冀. "Neural fuzzy controller design with two-level neural network." Thesis, 1995. http://ndltd.ncl.edu.tw/handle/58558698163485044944.
Full text
12
Xue, Kuo Qiang, and 薛國強. "An intelligent sales forecasting system through artificial neural networks and fuzzy neural network." Thesis, 1996. http://ndltd.ncl.edu.tw/handle/07455980576654976365.
Full text
13
Tsai, Chiachih, and 蔡嘉志. "Applications of Wireless Sensor Networks Based on Fuzzy Neural Network." Thesis, 2012. http://ndltd.ncl.edu.tw/handle/18594029970285141084.
Full textAbstract:
博士國防大學理工學院
國防科學研究所
100
Due to immense potential applications, wireless sensor networks (WSNs) have attracted research interests in recent years, including remote environmental monitoring, data fusion, sensing (temperature, pressure, speed) and military applications. This dissertation applies the fuzzy logic and neural network technologies to a monitored area which deployed miniature wireless sensor nodes. With the advantages of inherent accuracy and simplicity, the fuzzy logic and neural network technologies manifests the effectiveness on the environmental monitoring and control applications of wireless sensor networks. First, we apply the fuzzy technology to control the air-conditioning strength and blade angle of a car conditioner to equalize the comfortable temperature in the front- and rear-seat areas. The wireless nodes equipped with temperature sensor are installed to gather temperature information and then transmit this information to the central control terminal which executes the fuzzy inference control logic. The experiments show that the fuzzy technology would greatly improve the response for the automotive control and smart computation in the wireless sensor network systems. And then we develop a novel fuzzy logic algorithm to the remote environmental monitoring applications. Through a simple and effective fuzzy logic algorithm, every interesting node in the monitored area can be effectively calculated. This novel algorithm manifests their simplicity and accuracy and its performance characterized by root mean square error is better than the one with the standard Mamadni fuzzy logic method. Our study focuses on two particular neural network models, back-propagation network (BPN) and general regression neural network (GRNN) for the temperature prediction in a monitored factory. The prediction accuracy of these two models is evaluated by practical monitored data. We found that the model based on GRNN can accelerate the learning speed and rapidly converge to the optimal regression surface with large number of data sets. With the simulation results, we can show that the model based on GRNN effectively improve the predictability of the one based on BPN. Finally, we combine the genetic algorithm (GA) and the radial basis function (RBF) neural network in study of event detection for factory monitoring. As we know, the center of RBF, the width of RBF and output weight of RBF have a great influence on the performance of RBF neural network. In this study, we apply genetic algorithm to determine these parameters to improve the performance of the event detection. The experiments indicate that the GA-RBF algorithm is better than the traditional BPN and RBF neural network algorithms in both speed and precise of convergence. In this work, we find a responsive and effective algorithm in the WSN applications by integrating fuzzy theory and neural network technology. The combination of fuzzy theory and neural network technology should be a powerful strategy for the various WSN applications.
14
Hui, Leu Bing, and 呂炳輝. "A Fuzzy Neural Network Model for Revising Imperfect Fuzzy Rules." Thesis, 1994. http://ndltd.ncl.edu.tw/handle/54434906960677904661.
Full textAbstract:
碩士國立臺灣科技大學
工程技術研究所
82
In this thesis, a Knowledge-Based Neural Network model, named KBFNN, is proposed. The initial structure of KBFNN is constructed by existing partial fuzzy rules. The domain theories are represented by fuzzy rules and revised by neural network training. To construct KBFNN by fuzzy rules, two kinds of fuzzy neurons are proposed. They are S-neurons and G-neurons. The S- neurons perform similarity measure to compute the firing degrees of fuzzy rules. The G-neurons carry out the defuzzification of inference results. For processing fuzzy number efficiently, the LR-type fuzzy numbers are used. An Inverted Pendulum System and a Knowledge-Based Evaluator are used to illustrate the workings of the proposed model. However, in many problems, the initial fuzzy rules might not exist or might be hard to acquire. To deal with these problems, another fuzzy neural network model, called Fuzzy BP, with fuzzy inference is proposed. It performs nonlinear mapping between fuzzy input vectors and crisp outputs. A fuzzy neuron which performs fuzzy weighted summation, defuzzification, and nonlinear mapping is proposed. One sample problem, the Knowledge-Based Evaluator is considered to illustrate the working of Fuzzy BP, and the experimental results are very encouraging.
15
Chen, Kuan-Ting, and 陳冠廷. "Training algorithm for fuzzy embedded neural network." Thesis, 1997. http://ndltd.ncl.edu.tw/handle/73122431557855669591.
Full text
16
Wang, Ja-Ping, and 王家屏. "Adaptive Control Based On Fuzzy Neural Network." Thesis, 1996. http://ndltd.ncl.edu.tw/handle/21465185633766857736.
Full textAbstract:
碩士國立交通大學
控制工程系
84
In this thesis, we investigate the possible application of fuzzy neural network to adaptive regulation control. We propose a control scheme including FNNC (fuzzy neural network controller) and FNNI (fuzzy neural network identifier) to adaptive regulation control. FNNC is discussed using input- ouput linearization on technique and is analyzed using Lyapunov theorem. FNNI is used for the backpropagation of errors through the plant to the controller.In particular, dynamic plant sensitivity is provided by the FNNI to adjust theparameters of FNNC. Several simulation results shows the merit of applying FNNto the adaptive control.
17
Chiang, Ten-Wen, and 蔣天文. "Fuzzy Neural Network Application in Automobile Light." Thesis, 1994. http://ndltd.ncl.edu.tw/handle/01900854715845753051.
Full textAbstract:
碩士大葉大學
電機工程研究所
82
As a car driver, you might have been troubled by the following problems: You forget to turn on the head light because you start your car in a light place at night. You forget to switch the head light from high to low beam when meeting another car coming in the counter direction. You do not know the right time to turn on the head light or fog lamp while driving in the rain, fog or twilight. When you drive up or down a slop or make turn, the head light can not focus on tthe way efficiently. Since the situations mentioned above might be dangerous to a driver, it is necessary to improve the automobile light system. This paper will try to present a Fuzzy Neural Network control system as a better solution. The Fuzzy Neural Network control system combines Fuzzy Theory with Neural Network. It has the function of human thinking and human nervous system without human control. The system can turn on and adjust the light at the right time by means of sensoring the surroundings. Therefore, it can avoid the faults caused by human control and ensure safer driving.
18
Wu, Chin-Hao, and 吳晉豪. "Fuzzy Neural Network Based Human Motion Recognition." Thesis, 2011. http://ndltd.ncl.edu.tw/handle/17399538588396196085.
Full textAbstract:
碩士國立暨南國際大學
電機工程學系
99
The human motion recognition and robot vision have more and more applications in recent years. This thesis presents that a human motion recognition method based on fuzzy neural network. Here, the local feature is used as the descriptor. The human motion recognition method is consisted of three steps. In the first step, the features are found by the space-time interest point detector in the image sequences. The idea of space-time interest point detector is based on the extension of Harris interest point operator. The image values are found in the local spatiotemporal volumes that have large variations along both the spatial and the temporal directions. The coordinates of image values are the space-time interest points (STIP). They will correspond to the local spatiotemporal neighborhoods with non-constant motion in the image sequences. In the second step, the corresponding motion events are matched in the image sequences. The motion information is captured by the histogram of optical flow in a STIP local neighborhood. In the third step, the values of histogram of optical flow are used as the fuzzy neural network inputs to make the detection. However, a fuzzy neural network classifier is used to deal with the multiclass problem. The detection rate is not good. For this reason, thesis uses the architecture of fuzzy neural network of centered binary tree to achieve the high classification efficiency.
19
Liao, Wei-Sheng, and 廖偉盛. "Fuzzy Neural Network Control for Underwater Vehicles." Thesis, 2009. http://ndltd.ncl.edu.tw/handle/26431665147822802884.
Full textAbstract:
碩士元智大學
電機工程學系
97
This thesis proposes a fuzzy neural network (FNN) controller for autonomous underwater vehicles (AUVs). In order to track signals or specific desired yaw angle, backpropagation algorithm is presented to tune the weights, means, and variances in the fuzzy neural network. Backpropagation algorithm can reduce the program operation time and increases the reaction speed. The simulation results illustrate that the proposed controller has good tracking performance without large control efforts. Finally, this method is applied to control a small autonomous underwater vehicle which travels in a narrow and harsh environment. The proposed fuzzy neural network controller lets the autonomous underwater vehicle have the functions of tracking and positioning. The simulation results verify that the proposed fuzzy neural network controller is effective.
20
HUANG, CHIH-CHIANG, and 黃志強. "Apply Fuzzy Neural Network to Combined Forecasts." Thesis, 2008. http://ndltd.ncl.edu.tw/handle/05824304000518221804.
Full textAbstract:
碩士國立臺灣科技大學
工業管理系
96
Demand planning in many industries exists uncertainty. To reduce the costs and increase benefits, the accuracy of demand forecasting becomes an important task. This research investigates the policy of demand planning through many kinds of forecasting methods, it will improve the performance in production schedule and productivity supply and reduce bullwhip effect. Combined forecasts method is to combine different forecasting methods. Many experts point out that combined forecast is more useful than any individual forecasting methods in prediction performance. In addition, nonlinear combined forecast is better than linear combined forecast. We use 11 groups of ATM cash demand at random within the territory of England as the target of prediction, by combining two individual forecasting methods’ predicted value to reach stability and accuracy of carrying on the demand while planning for this industry and prove that the nonlinear combined method is more apparent on the result that is predicted. To estimate the parameters of linear combined forecasts we use adaptive set of weights, k method and linear composite for nonlinear combined forecasts, we use the adaptive fuzzy neural networks to train and study and the results show that this method provides the most suitable weights for combine forecasts.
21
Chang, Chun-Lung, and 張俊隆. "A Fuzzy Cellular Neural Network Integrated System." Thesis, 2005. http://ndltd.ncl.edu.tw/handle/68668749123456924858.
Full textAbstract:
博士國立交通大學
電機與控制工程系所
94
It is widely accepted that using a set of cellular neural networks (CNNs) in parallel can achieve higher-level information processing and reasoning functions either from application or biologics points of views. Such an integrated CNN system can solve more complex intelligent problems. In this thesis we propose two novel frameworks for automatically constructing a multiple-CNN integrated neural system in the form of a recurrent fuzzy neural network. The systems, called recurrent fuzzy CNN (RFCNN) and recurrent fuzzy coupled CNN (RFCCNN), can automatically learn its proper network structure and parameters simultaneously. The structure learning includes the fuzzy division of the problem domain and the creation of fuzzy rules and CNNs. The parameter learning includes the tuning of fuzzy membership functions and CNN templates. In the RFCNN/RFCCNN, each learned fuzzy rule corresponds to a CNN. Hence, each CNN takes care of a fuzzily separated problem region, and the functions of all CNNs are integrated through the fuzzy inference mechanism. A new on-line adaptive ICA (independent component analysis) mixture-model technique is proposed for the structure learning of RFCNN/RFCCNN, and the ordered-derivative calculus is applied to derive the recurrent learning rules of CNN templates in the parameter-learning phase. The proposed RFCNN/RFCCNN provides a solution to the current dilemma on the decision of templates and/or fuzzy rules in the existing integrated (fuzzy) CNN systems. The capability of the proposed RFCNN and RFCCNN are demonstrated and compared on the real-world defect inspection problems. Experimental results show that the proposed scheme is effective and promising.
22
Chuorng, Feng-Shou, and 莊豐收. "Object extraction using a fuzzy neural network." Thesis, 1994. http://ndltd.ncl.edu.tw/handle/09120154157225247821.
Full text
23
XU, SHUN-TANG, and 許順鏜. "Fuzzy logic controller as a neural network." Thesis, 1992. http://ndltd.ncl.edu.tw/handle/60162101651486012092.
Full text
24
Xu, Shun-Tang, and 許順鏜. "Fuzzy logic controller as a neural network." Thesis, 1992. http://ndltd.ncl.edu.tw/handle/76502445639528045442.
Full text
25
鄭宇修. "Fuzzy Neural Network Software Effort Estimation Model." Thesis, 2002. http://ndltd.ncl.edu.tw/handle/72947934981716460651.
Full textAbstract:
碩士國立臺灣科技大學
資訊管理系
90
Under-estimating the effort needed for software development may cause the sacrifice of the software quality or seriously lead to the failure of the software development project because of the insufficient distribution of the allocated resources. However, over-estimating the software development effort may also cause the problem of the inefficient usage of allocated resources or lose the chance of gaining the software project in the price bidding because of allocating too much resource. Therefore, it is an important research topic to precisely estimate the software development effort. Recently some scholars adopt the techniques of Artificial Intelligence to build up the software development effort estimation model. Neural Network and Fuzzy Logic are the most widely used techniques. A model established by Neural Network has the advantages of making the model output gradually closer to the real value by the processes of training data and learning. However, the precision of the model established by Neural Network is greatly affected by the quality of training data. Meanwhile, the content of the model cannot be interpreted by model’s users. A model established by Fuzzy Logic has the advantages of easily optimization and verification via adjusting the fuzzy rules of the models. However, the fuzzy rules are required to define by the domain experts who have different degrees of subjectivity. We found that these two techniques have the complementary prosperity and thus proposed an Fuzzy Neural Network Software Effort Estimation Model which has not only the advantages of both two techniques but also of releasing the limitations of each of the two techniques. Based on an empirical study, our proposed model performed better than existing software effort estimation models in literature on both the estimative precision and explanation ability of the model.
26
Li-Hsin, Lai, and 賴立新. "SYSTEMATIC MODELING APPROACH OF FUZZY NEURAL NETWORK." Thesis, 1998. http://ndltd.ncl.edu.tw/handle/90717783755662283727.
Full textAbstract:
碩士大同工學院
電機工程研究所
86
In this thesis, we propose a systematic method to construct a fuzzy neural network and then apply the proposed network to track a signal or control an unknown plant. The presented method integrates the learning ability of neural network and the advantage of fuzzy logic controller to handle the nonlinear system modeling problems. To model a system, we first apply the grey relational mountain method to find the clustering centers of training data. The grey relational mountain method, based on the grey relational analysis and the mountain method, provides a simple scheme to find the proper clustering centers that have higher relation to the training data. Then we adopt these clustering centers as the initial centers of membership functions in fuzzy neural network. After constructing the architecture of fuzzy neural network, the proposed method can adjust the scaling factors of input/output variables and tune the membership functions by the back propagation algorithm. Finally, the proposed fuzzy neural network is applied to track a signal and control an unknown or complex ill-defined system. Of the learning ability and self-tuning ability, the performance of tracking and on-line control will be confirmed by simulationresults.
27
Yang, Neng-Jie, and 楊能傑. "An Optimal Recurrent Fuzzy Neural Network Controller." Thesis, 2002. http://ndltd.ncl.edu.tw/handle/22893053061456487124.
Full textAbstract:
碩士中原大學
電機工程研究所
90
In this thesis, an optimal recurrent fuzzy neural network controller is by an adaptive genetic algorithm. The recurrent fuzzy neural network has recurrent connections representing memory elements and uses a generalized dynamic backpropagation algoruthm to adjust fuzzy parameters on-line. Usually, the learning rate and the initial parameter values are chosen randomly or by experience, therefore is human resources consuming and inefficient. An adaptive genetic algorithm is used instead to optimize them. The adaptive genetic algorithm adjust the probability of crossover and mutation adaptively according to fitness values, therefore can avoid falling into local optimum and speed up convergence. The optimal recurrent fuzzy neural network controller is applied to the simulation of a second-ordeer linear system, a nonlinear system, a highly nonlinear system with instantaneous loads. The simulation results show that the learning rate as well as other fuzzy parameters are important factor for the optimal design. Certainly, with the optimal design, every simulation achieve the lowest sum of squared error and the design process done automatically by computer programs.
28
Lin, Tay-Shang, and 林泰祥. "Computing the Integrals Using Fuzzy Neural Network." Thesis, 1997. http://ndltd.ncl.edu.tw/handle/19009516173650070839.
Full textAbstract:
碩士國立交通大學
控制工程系
85
In this thesis, we propose a method of usung a fuzzy neural network(FNN)system to map and compute the untegrals of an unknown function. The parameters of the FNN are learned by back-propagation algorithm. We only integrate the membership function to achieve the goal of integrating the unknown function.Finally, we apply this proposed method to solve a convolution problem and constant-coefficient differential equation problems.
29
Chiu, Yi-Feng, and 邱一峰. "STUDY ON SELF-CONSTRUCTING FUZZY NEURAL NETWORK CONTROLLER USING RECURRENT NEURAL NETWORK LEARNING STRATEGY." Thesis, 2013. http://ndltd.ncl.edu.tw/handle/38808034711756082416.
Full textAbstract:
碩士大同大學
電機工程學系(所)
101
In this thesis, the self-constructing fuzzy neural network controller (SCFNN) using recurrent neural network (RNN) learning strategy is proposed. For back-propagation (BP) algorithm of the SCFNN controller, the exact calculation of the Jacobian of the system cannot be determined. In this thesis, the RNN learning strategy is proposed to replace the error term of SCFNN controller. After the training of the RNN learning strategy, that will receive the relation between controlling signal and result of the nonlinear of the plant completely. Moreover, the structure and the parameter-learning phases are preformed concurrently and on-line in the SCRFNN. The SCFNN controller is designed to achieve the tracking control of an electronic throttle. The proposed controller, there are two processes that one is structure learning phase and another is parameter learning phase. The structure learning is based on the partition of input space, and the parameter learning is based on the supervised gradient-decent method using BP algorithm. Mahalanobis distance (M-distance) method in this thesis is employed as the criterion to identify the Gaussian function will be generated / eliminated or not. Finally, the simulation results of the electronic throttle valve are provided to demonstrate the performance and effectiveness of the proposed controller.
30
Tu, Jian-Chung, and 涂畯程. "An Interactively Recurrent Functional Neural Fuzzy Network with Fuzzy Differential Evolution." Thesis, 2013. http://ndltd.ncl.edu.tw/handle/31099884657922841540.
Full textAbstract:
碩士國立勤益科技大學
資訊工程系
101
In this thesis there are three major contributions. First of all, a modified Differential Evolution (DE) approach named “Fuzzy Differential Evolution (FDE)” is proposed, where the fuzzy theory is applied to updated parameters of the Recurrent Neural Fuzzy Network (RNFN). Lastly, a novel Neural Fuzzy Network, i.e., named Interactively Recurrent Functional Neural Fuzzy Network (IRFNFN) is designed. In other words, the main idea of the proposed FDE is to combine traditional DE with fuzzy sets, i.e., considering the diversity of individuals in the DE as input pattern of fuzzy sets. Via the process of fuzzy inference, it will be able to adaptively learn the corresponding parameters. The elites learning strategy is also included in the proposed FDE in order to find out better solutions. After that, a RNFN with the proposed FDE is conducted to verify the efficiency and performances. Lastly, a novel neural fuzzy network called IRFNFN is proposed, in order to achieve the aim of that every rule node has relationships with the others. Simulation results indicated that the proposed IRFNFN can effectively reduce the number of required rule nodes and enhance the learning process.
31
Wu, Wen-Shiang, and 吳文翔. "ADAPTIVE FUZZY CONTROL BASED ON FUZZY NEURAL NETWORK FOR NONLINEAR SYSTEMS." Thesis, 2007. http://ndltd.ncl.edu.tw/handle/62406727886315562549.
Full textAbstract:
碩士元智大學
電機工程學系
95
This paper presents a fuzzy controller to control the plant and design an adaptive fuzzy neural network to identify the system response and provide a group of parameters to adjust the fuzzy controller output, and to guarantee the system robustness. So, the system has robust performance with external disturbance. Finally, a two-link robotic manipulator and a planetary gear type inverted pendulum are investigated. The experiment results show that the proposed method is very effective.
32
Sung, Yu-Min, and 宋裕民. "Credic Identification By Using Fuzzy Neural Network Theory." Thesis, 2002. http://ndltd.ncl.edu.tw/handle/33768491197526394487.
Full textAbstract:
碩士國立交通大學
電資學院學程碩士班
90
This thesis has been emploied （Support Vector Machine，SVM） and （Self-cOnstructing Neural Fuzzy Inference Network，SONFIN）together as an analytic system of credit card approval evaluation. Some of the customer’s basic information are chosen as the input of the network; then the credit summary was got as the output of the network. There are two goals of the simulation of the credit card approval : one is using SVM from the bank’s point of view, to judge the approval of credit card by analyzing applicants’ information ; the other is using SONFIN from the accuracy of the judgment to applicants’ annually breach. The SVM provides an architecture to extract support vectors for generating fuzzy IF-Then rules from the training data set , and a method to describe the fuzzy system in terms of kernel functions. Thus , it has the inherent advantage that the model does not have to determine the number of rules in advance, and the overall fuzzy inference system can be represented as series expansion of fuzzy basis functions. Moreover ,the SONFIN contains five-layers of constructed network. The summary of the five-layer constructed network is as follows：1st layer：input layer；2nd layer：linguistic label layer, such as：large, small, etc；3rd layer：forms the formula of precondition layer of the fuzzy rule ；4th layer： consequent layer ；5th：layer : output layer. Finally, the simulation result can be the reference for financing institutes to evaluate customers’ financial status and taken into account the risk and given credit.
33
Hsu, Jang-Pong, and 許振鵬. "Modeling and Applications of Fuzzy Neural Network Systems." Thesis, 1998. http://ndltd.ncl.edu.tw/handle/29845499773992259444.
Full textAbstract:
博士國立成功大學
資訊工程學系
86
This dissertation presents two fuzzy neural network models and theirapplications in the fields of speech recognition, image recognition,information retrieval, and machine control. The first model is aConnectionist Fuzzy Classifier (CFC). This model realizes the "weightedEuclidean distance" fuzzy classification procedure in a massivelyparallel manner, and employs a hybrid supervised/unsupervised learningscheme to organize reference samples. By CFC and its extensive versions,three applications including the speech recognition, the color-blindnessplate recognition, and the information retrieval on Internet have beenexperimented and good results were gotten. The second model is a three-layered parallel fuzzy inference mechanism, called RFNN-DPS (ReinforcementFuzzy Neural Network with Distributed Prediction Scheme). This modelrealizes parallel linguistic knowledge reasoning and performs reinforcementlearning with a novel distributed prediction scheme. RFNN-DPS consists ofonly one network without an additional predictor for predicting theexternal reinforcement signal, and the internal reinforcement information isdistributed into rule nodes. According to the experimental results of thetruck backer-upper control, the truck-and-trailer backer-upper control, thecart-pole balance, and the beam-and-ball balance, RFNN-DPS showed theadvantages of simple network structure, fast learning speed, and explicitrepresentation of rule reliability.
34
Lin, Chia-Yang, and 林家陽. "Multi-lead ECG Classification Using Fuzzy Neural Network." Thesis, 2000. http://ndltd.ncl.edu.tw/handle/92890304759874654877.
Full textAbstract:
碩士中原大學
電子工程學系
88
The electrocardiogram (ECG) is one of the major tools for cardiac diagnosis. Some cardiac symptoms can be identified from particular single channel ECG while others may rely on multichannel ECG. Since quite often there is no physician on the scene for long-term monitoring, the automatic ECG diagnosis is essential. We use the multichannel ECG features for diagnosis to increase the accuracy of automatic diagnosis system and reduce the burden on clinicians. As a preprocessing step, we extract ECG features from different channels by using the fact that the slopes of waveform turning points have a zero-crossing characteristic. After feature extraction, we perform ECG cardiopathy classification by using a fuzzified artificial neural network called radial basis function (RBF) network. We compare the recognition accuracy of the networks when single channel features and two-channel features are used. We also modify the network architecture to see how this change can affect the recognition accuracy. In this research, we use MIT/BIH arrhythmia database for training and testing the network. In addition, a set of in-house clinical data obtained from Taipei Veterans General, Chang-Gung Memorial, and Tri-Service General Hospitals is also used. The data include Normal, left bundle branch block (LBBB), right bundle branch block (RBBB) and premature ventricular contraction (PVC) in lead Ⅱ and lead V1 ECG signals. Our experiment results show 17.4%, 18.3%, and 10.5% gain in average recognition accuracy when two-channel features are used simultaneously for the MIT/BIH training set, testing set, and the in-house data, respectively.
35
Hu, Hsiang-Fan, and 胡湘帆. "Counterpropagation Fuzzy-Neural Network for Stream Flow Estimation." Thesis, 1998. http://ndltd.ncl.edu.tw/handle/62069583158537656141.
Full text
36
Li, I.-Hsun, and 李宜勳. "A Merged Fuzzy Neural Network: Analysis and Applications." Thesis, 2007. http://ndltd.ncl.edu.tw/handle/8mga7j.
Full textAbstract:
博士國立臺灣科技大學
電機工程系
95
To solve learning problems with vast numbers of inputs, this dissertation proposes a novel learning structure merging a number of small fuzzy neural networks (FNNs) into a hierarchical learning structure called a merged-FNN. In this dissertation, the merged-FNN is proved to be a universal approximator. Two different applications demonstrate that the merged-FNN has greater potential than the neural network and traditional FNN in real systems. One is the system identification and the other is the controller design for nonaffine nonlinear systems. In the system identification, the computing approach uses the merged-FNN using B-spline membership functions (BMFs) with a reduced-form genetic algorithm (RGA). The reduced-form genetic algorithm is employed to tune all free parameters of the merged-FNN, including both control points of the B-spline membership functions and weights of the small fuzzy neural networks. For a practical application, a battery state-of-charge (BSOC) estimator, which is a twelve input, one output system, in a lithium-ion battery string is proposed to verify the effectiveness of the merged-FNN. From experimental results, the learning ability of the newly proposed merged-FNN with RGA is superior to that of traditional neural networks with back-propagation learning. In the aspect of the controller design, we propose an observer-based adaptive fuzzy-neural controller for nonaffine nonlinear systems, structured by the merged-FNN to substantially reduce the number of adjustable parameters and the computation time of the controller. The traditional direct adaptive fuzzy-neural control scheme for nonaffine nonlinear systems has a vast number of free parameters if many inputs (linguistic terms) and membership functions of the fuzzy-neural network (FNN) are required. This leads to the problem of a huge computation time. Spending so much computation time adjusting these parameters results in a serious controller time-delay problem. To solve this problem, the traditional FNN is replaced by the merged-FNN to form an observer-based adaptive controller. We prove that the merged-FNN can take the place of the traditional fuzzy-neural networks under some assumptions while maintaining the property of stability. Moreover, the adaptive scheme using the proposed merged-FNN guarantees that all signals involved are bounded and the output of the closed-loop system asymptotically tracks the desired output trajectory. From experimental examples, the proposed merged-FNN has far fewer parameters than the traditional FNN, and the computation time is significantly reduced.
37
Chang, Chih-Chie, and 張智傑. "Gear Fault Diagnosis by Using Fuzzy Neural Network." Thesis, 2005. http://ndltd.ncl.edu.tw/handle/z49un2.
Full textAbstract:
碩士中原大學
機械工程研究所
93
This paper applies fuzzy neural network (FNN) in the fault diagnosis of the gear-rotor system. According to the document and experiment data, the relationship between fault and frequency spectrum is built up as the rule of approximate reasoning diagnosis and the training data of NN. In this paper, gears which have the four typical faults will be taking into vibrating examining : (1)Gear skew (2)Shaft not parallel (3) Tooth breakage (4)wear. Picking up the characteristic signals by using the technique of analysing spectrums . After classifing by membership function, and detect this classified data in NN. Comparing with the detect result of FNN, NN, and approximate reasoning. Verifying that FNN could make rational and accurate diagnosis from the complicated spectrum.
38
Yang, Tsung-Han, and 楊宗翰. "Fuzzy Neural Network Synchronous Control of Gantry Stage." Thesis, 2011. http://ndltd.ncl.edu.tw/handle/a885gc.
Full textAbstract:
碩士國立臺北科技大學
自動化科技研究所
99
To improve a single-axial driving force, many high-load stages start adopting gantry stage architecture in different industry application. If the synchronous error among dual-drive servo systems at a high-speed motion is too large, then the mechanical coupling force yielded on both servo systems will result in a mechanical deformation or damage. Therefore, it is an important issue to find a way to drive the stage to achieve a synchronous motion effectively and precisely. For single-axial servo motor, the existing parameter variations and external disturbance will degenerate the control performance. First, the model reference adaptive control (MRAC) is proposed to compensate the parameter uncertainty, yielded by the inaccuracy of parameter identification, and external disturbance. Since the two axes of the gantry stage are asynchronous, it is difficult to model the complicated non-linear behavior of mechanical coupling phenomenon. This thesis proposes a fuzzy neural network (FNN) compensation control and an on-line learning algorithm to overcome the aforementioned problem. The fuzzy neural network performs a reasonable fuzzy partition to both synchronous position and velocity errors between two dual-drive servo systems and transmits the parted signals to generate the compensated force via neural network reasoning, and the compensated force is fed back to the controller of each axis. The on-line learning algorithm adjusts the connected weighting of the neural network by using a supervised gradient descent method, such that the defined error function(E) can be minimized. Finally, two kinds of low-speed and high-speed sinusoidal position commands are designed for the experiments, and the experimental results show that the proposed MRAC and FNN control scheme are feasible to improve the single-axis and synchronous control, respectively.
39
Tao, Yu-lung, and 陶雨龍. "An Improved Fuzzy Neural Network and Its Application." Thesis, 2004. http://ndltd.ncl.edu.tw/handle/57549132352229624577.
Full textAbstract:
碩士中原大學
工業工程研究所
92
Fuzzy Neural Networks have been successfully applied to extract knowledge from data in the form of fuzzy rules. However, the drawback with the fuzzy neural approach is that the fuzzy rules induced by the learning process are not necessarily understandable. The purpose of this thesis is thus to improve and evaluate two kinds of fuzzy neural network based on Takagi-Sugeno fuzzy inference system. Specificially speaking, this study investigates the adaptive network-based fuzzy inference system (ANFIS) and an improved the fuzzy nueral network (FNN). The proposed FNN uses fuzzy c-means for clustering data set, while most of the fuzzy neural networks including the ANFIS, use the sum of all existing rules as output. The first-order rule is utilized in the FNN. The proposed models are applied to a feedback control system and an empirical study with four cases is conducted. The four experiments are (i) a cart and pole system, (ii) a ball and beam system, (iii) a chaos production forecasting system and (iv) an inventory control system. Experiment outcomes revealed that the two models can precisely generate human-understandable fuzzy rules with good interpretability. The advantage and disadvantage of the FNN and ANFIS are discussed as well.
40
Chou, Shu-Yu, and 周舒玉. "IMAGE RESTORATION BY USING GENETIC FUZZY NEURAL NETWORK." Thesis, 1996. http://ndltd.ncl.edu.tw/handle/59505902464022710815.
Full textAbstract:
碩士大同工學院
電機工程學系
84
In this thesis, we propose a method to restore gray level images degraded by a known shift-invariant blur function and noise by using a genetic fuzzy neural network. The genetic algorithms are useful for some ill understood or highly irregular systems since the expert experience and knowledge are difficult to get. We use the optimal searching ability of genetic algorithms to generate the optimal rules and scaling factors of a fuzzy controller without the aid of expert experience or knowledge.The restoration procedure consists of two stages: restore images and enhance the effect of boundary. During the first stage, restore gray level images degraded by a known shift-invariant blur function and additive noise using genetic fuzzy controller, and genetic theory is used to find the fittest membership function of fuzzy model. Finally, enhance the boundary effect of the restoration images using the fuzzy neural network.
41
Chang, Wen-Bin, and 張文賓. "Neural-Network-Based Linear Combination Fuzzy Logic Controller." Thesis, 1993. http://ndltd.ncl.edu.tw/handle/37314354910006924949.
Full textAbstract:
碩士國立臺灣科技大學
工程技術研究所
81
To design a fuzzy logic controller without depending on control engineering and expert's experience, we implemented Sugeno's fuzzy logic control rule on multi- layer feedforward neural network. This neural network can be trained by backpropa- gation learning algorithm, so the parameters of fuzzy logic controller such as parameters of membership function and para- meters of consequence part of rules are designed by learning. The advantages of this research method are (1) design a fuzzy logic controller without depending on expert's experience (2) the trained fuzzy logic controller has strong noise resistant. A fuzzy car running example is presented to illusrate the per- formance and applicability of the proposed neural network.
42
Chang, Ting-Kang, and 張庭綱. "FUZZY NEURAL NETWORK STUDY USING DIFFERENTIAL EVOLUTION ALGORITHM." Thesis, 2010. http://ndltd.ncl.edu.tw/handle/73101702481839796098.
Full textAbstract:
碩士大同大學
電機工程學系(所)
98
A differential evolution (DE) algorithm based fuzzy neural network (FNN) (DEFNN) controller is proposed in this thesis. DEFNN controller is composed of an FNN identifier, a DE estimator, a computation controller, and a hitting controller. There are two main learning phases in DEFNN controller – the training phase and the online phase. The training phase is utilized to find the best preset parameters of DEFNN controller. In this thesis, several parameters such as the learning rates of the back-propagation (BP) algorithm, the parameters of error term which are used in BP algorithm, the initial values of the FNN identifier and some preset parameters of DEFNN controller are provided by DE estimator. After the best preset parameters are obtained, DEFNN controller will be active online. In the online phase, the FNN identifier is used to identify the unknown terms of the nonlinear system dynamic. The BP algorithm is adopted to update the parameters of the FNN identifier to achieve favorable approximation performance. Then the computation controller is designed to calculate the outputs of the FNN identifier. Finally, the hitting control which is utilized to eliminate the uncertainties and external disturbances of the nonlinear system combine with the output of computation controller to form the main control effort. The results of the simulations are implemented to verify the effectiveness of the proposed controller.
43
Lin, Wen-Yang, and 林文揚. "A Self-Organizing Fuzzy Rule-Based Neural Network." Thesis, 2008. http://ndltd.ncl.edu.tw/handle/13240710955983335036.
Full textAbstract:
碩士國立中興大學
電機工程學系所
96
This thesis proposes a new fuzzy neural network, the Self-Organizing Fuzzy Rule -Based Neural Network (SOFRBNN). Based on Takagi-Sugeno-Kang (TSK) type fuzzy rules, the network consists of two parts, the antecedent and the consequent parts. A Kalmam filer trained Neural Network (KFNN) constitutes the consequent part of each rule, where a KFNN comprises three layers. There are no rules in SOFRBNN initially as rules are on-line generated according to training data. Once a new rule is generated, a new KFNN is generated accordingly in the consequent part. For KFNN parameter learning, the parameters between layers two and three are tuned using Kalman filter, while the parameters between layers one and two are tuned using gradient descent algorithms. The antecedent part parameters in SOFRBNN are also tuned using gradient descent algorithms. SOFRBN characterizes concurrent structure and parameter learning, and good learning performance accompanies this characteristic. This thesis performs simulations on chaotic signal prediction and nonlinear system identification to verify SOFRBNN performance.
44
Chuang, Kai-Hsiang, and 莊凱翔. "Identification of fMRI Signal Using Fuzzy Neural Network." Thesis, 1997. http://ndltd.ncl.edu.tw/handle/33318825486856816899.
Full textAbstract:
碩士國立臺灣大學
電機工程學系
85
Functional MRI (fMRI), with its high spatial, temporal resolution, non-radioactive, and non-invasive properties, is potential in the fields ranging from physiology, psychology, pharmacology to clinical applications, etc. However, due to its very low signal-to-noise ratio, post processing is required to identify the active regions.Most contemporary post processing strategies require assumed functional response waveform based on the prior knowledge about experimental paradigm to produce effective results. While most studies are performed in steady state, these methods are not very suitable for complex functional experiments and will produce biased result even when brain*s response is simple. This will limit the applications of functional MRI studies. To fully utilize fMRI to cognition applications, one needs to develop a flexible analysis tool for the complicated signal characteristics of brain responses. Utilizing unsupervised clustering network and fuzzy set theory, we have successfully developed a cascade fuzzy neural network, which combines Kohonen clustering network and Fuzzy C-Means algorithm, to analyze fMRI time signal [30]. By comparing the receiver operating characteristic (ROC) analysis results of our proposed method with other two kinds of conventional post processing methods- correlation coefficient analysis and t- statistical parametric mapping - on a series of testing phantoms, we have proved that our method can identify the actual functional response even when the activation area is very small, under noisy conditions, or even when the actual response is deviated from traditionally assumed box-car type.Human studies involving motor cortex activation also show that our method successfully identifies the functional responses waveform and the active regions as well. Furthermore, it can also discriminate responses in gray matter from those possibly coming from venous vessels. And most of all, even when the experimental paradigm is unknown, such that conventional post processing methods are inapplicable, our method is still effective.With this flexible fMRI tool, psycho-physiologist now can go on and proceed complicated one shot human cognition studies. Linguistic studies including Chinese language and Taiwanese dialect study will be performed in the near future.
45
Lin, Rui-Jie, and 林瑞杰. "Additive Gaussian Membership Functions in Fuzzy Neural Network." Thesis, 1999. http://ndltd.ncl.edu.tw/handle/55615490855259126785.
Full textAbstract:
碩士國立交通大學
電機與控制工程系
88
In this thesis, a new method to tune the membership functions of fuzzy neural network (FNN) is presented. First we study the FNN it inherits the property of both fuzzy inference system and neural network. Then we present that any gaussian function can be represented by the linear combination of gaussian functions with small standard deviation. Therefore, it can be substituted for the second layer of FNN (called FNN5). We use the FNN5 to approximate some functions and prove that it is a universal approximator. Furthermore, apply this proposed method to tune PI controller based on gain phase margin (GPM) specifications. Both FNN and FNN5 have high performance by the simulation verification, however FNN5 is more accurate than FNN on fine-tuning.
46
Chao, An-Ming, and 趙安民. "Motor Fault Diagnosis by Using Fuzzy Neural Network." Thesis, 2004. http://ndltd.ncl.edu.tw/handle/2s7wyy.
Full textAbstract:
碩士中原大學
機械工程研究所
92
Using membership function and neural network (NN) develops a motor diagnosis system based on fuzzy neural network (FNN) in this paper. According to literature and experience principle, the relationship between fault and frequency spectrum is built up as the rule of fuzzy diagnosis and the training data of NN. In experiment, to use spectrum analyzer pick up the vibration signal of motor, then transform the signal into frequency domain by fast Fourier transforms and classify it by membership function, and detect this classified data in NN. In this paper, comparing with the detect result of FNN, fuzzy, and NN in two examples prove that FNN is better than other methods for multiple fault diagnosis.
47
Liu, Shin-Jin, and 劉新金. "Fuzzy Neural Network Control for DC-DC Converter." Thesis, 2004. http://ndltd.ncl.edu.tw/handle/92172566881595830505.
Full textAbstract:
碩士元智大學
電機工程學系
92
Intelligence controller is suitable for uncertain nonlinear systems, which are not easy to be realized by the classical design methods. In this thesis, a fuzzy and a neural network controllers for dc-dc converter are investigated. The presented approaches are general and can be applied to any dc-dc converter. A fuzzy and a neural network controllers are designed and are compared with PID controller. The settling time and overshoot for startup and step response have been compared. Finally, simulation and experimental results of buck and boost converters demonstrate the possibility of applying intelligence control method in practical converter design.
48
Cheng, Chih-Yuan, and 鄭智元. "GA-Based Fuzzy-Neural network and Its Applications." Thesis, 2002. http://ndltd.ncl.edu.tw/handle/24820134784112415470.
Full textAbstract:
碩士輔仁大學
電子工程學系
90
In this paper, we use the learning ability of neural networks to build a fuzzy inference system, in which weighting factor values are automatically adjusted by genetic algorithms. Because of the superiority of genetic algorithms in directed random search for global optimization, they are used to obtain a set of optimal weighting values for the fuzzy neural network to approximate functions to desired accuracy. To address the problem of time-consuming evolutionary process in searching for an optimum value, we propose a novel crossover methodology for the genetic algorithm so that the system performance can be improved. Also, the effect of the proposed crossover methodology on searching results is also investigated in the paper. Theoretical justification on the use of the direct adaptive fuzzy controllers using a state feedback approach is valid when all of the system states are available for measurement. However, system states are not always available. It is therefore the objective of this paper to develop a design algorithm of the direct adaptive fuzzy-neural output-feedback controller (DAFOC) for uncertain nonlinear systems under the constraint that only the system output is available for measurement.
49
Tang, Dong-Cheng, and 湯東澂. "Fuzzy Neural Network Based Adaptive Human Skin Classifier." Thesis, 2010. http://ndltd.ncl.edu.tw/handle/73751965203115857372.
Full textAbstract:
碩士國立暨南國際大學
電機工程學系
98
This paper proposes an adaptive human skin classifier. The design of adaptive human skin classifier can further subdivide the skin color block, and reach the effect of system fuzzy studying by fuzzy theory. Using this technology can apply on the skin color segmentation, such as face detection, gesture recognition, and face location. It's major to use various ratio color values of HSV to be the color features which is used to compare the skin and nonskin distribution relationship between the two as identification rule. In the experiment process, first to manually divide pictures into two parts which are skin and nonskin. Then skin and nonskin ratio values can be gained, and can individually discuss about their ratio value relationships. The study is by fuzzy neural network to learn. First,go training in the specific images, and then use trained fuzzy neural network to proceed adaptive skin segmentation in the new images. The experiment results display that the method which is proposed in this paper can segment the skin blocks in image more definitely.
50
Wang, Chung-Hao, and 王仲豪. "STUDY ON SELF-CONSTRUCTING FUZZY NEURAL NETWORK CONTROLLER USING RECURRENT WAVELET NEURAL NETWORK LEARNING STRATEGY." Thesis, 2014. http://ndltd.ncl.edu.tw/handle/66373384738532600320.
Full textAbstract:
碩士大同大學
電機工程學系(所)
102
In this thesis, the self-constructing fuzzy neural network controller (SCFNN) using recurrent wavelet neural network (RWNN) learning strategy is proposed. SCFNN has been proven over the years to simulate the relationship between input and output of the nonlinear dynamic system. Nevertheless, there are still has the drawback of training retard in this control method. The RWNN approach with a widely similar range of nature since the formation of wavelet transform through the dilation and translation of mother wavelet, it has capability to resolve time domain and scaled and very suitable to describe the function of the nonlinear phenomenon. Importing the adaptable of RWNN learning strategy can improve the learning capability for SCFNN controller. The proposed controller has two learning phase, that is structure learning and parameter learning. In the former, Mahalanobis distance method is used as the basis for identify the function of Gaussian is generated or eliminated. The latter is based on the gradient-decent method to update parameters; the both learning phases are synchronized and real-time executed in parallel. In this study, the electronic throttle system as a control plant of nonlinear dynamic in order to achieve the throttle angle control, the simulation shows that the proposed control method has good capability of identification system and accuracy.