Dissertationen zum Thema „Radiation early warning network“

Excessive droughts on the African continent have caused the Swedish Meteorological and Hydrological Institute to launch a program of gathering data in hopes of producing models for rainfalls and droughts. A sensor capable of gathering such data has already been chosen, however there remains the problem of conveniently retrieving data from each of the sensors spread over a large area of land. To accomplish this goal, a small, cheap and efficient wireless capable module would need to be used. A possible candidate is the new WiFi-module from Espress if designated ESP8266. It is an extremely cheap and versatile wireless SoC that is able to perform the task of a wireless communications adapter for the sensor unit. The point of this thesis is to investigate the suitability of IEEE 802.11 for the task, and produce a piece of firmware for the ESP8266. The firmware shall enable it to be attached to a sensor and operate as a wireless mesh node in a self-organizing WLAN sensor network, enabling data retrieval via WiFi multi-hop deliveries.

Bien que le niveau de rayonnement soit une préoccupation sérieuse qui nécessite une surveillance continue, de nombreux systèmes existants sont conçus pour effectuer cette tâche. Radiation Early Warning System (REWS) est l'un de ces systèmes qui surveille le niveau de rayonnement gamma dans l'air. Un tel système nécessite une intervention manuelle élevée, dépend totalement de l'analyse d'experts et présente des lacunes qui peuvent parfois être risquées. Dans cette thèse, l'approche RIMI (Refining Incoming Monitored Incidents) sera introduite, qui vise à améliorer ce système pour gagner en autonome tout en laissant la décision finale aux experts. Une nouvelle méthode est présentée qui aidera à changer ce système pour devenir plus intelligent tout en apprenant des incidents passés de chaque système spécifique
Although radiation level is a serious concern that requires continuous monitoring, many existing systems are designed to perform this task. Radiation Early Warning System (REWS) is one of these systems which monitors the gamma radiation level in the air. Such a system requires high manual intervention, depends totally on experts' analysis, and has some shortcomings that can be risky sometimes. In this thesis, the RIMI (Refining Incoming Monitored Incidents) approach will be introduced, which aims to improve this system while becoming more autonomous while keeping the final decision to the experts. A new method is presented which will help in changing this system to become more intelligent while learning from past incidents of each specific system

Published Article
Early indication of possible drought can help in developing suitable drought mitigation strategies and measures in advance. Therefore, drought forecasting plays an important role in the planning and management of water resource in such circumstances. In this study, a non-linear streamflow forecasting model was developed using Artificial Neural Network (ANN) modeling technique at the Melka Sedi stream gauging station, Ethiopia, with adequate lead times. The available data was divided into two independent sets using a split sampling tool of the neural network software. The first data set was used for training and the second data set, which is normally about one fourth of the total available data, was used for testing the model. A one year data was set aside for validating the ANN model. The streamflow predicted using the model on weekly time step compared favorably with the measured streamflow data (R2 = 75%) during the validation period. Application of the model in assessing appropriate agricultural water management strategies for a large-scale irrigation scheme in the Awash River Basin, Ethiopia, has already been considered for publication in a referred journal.

This work deals with two main fields: a) The design, built, installation, test, evaluation, deployment and maintenance of Seismological Network of Crete (SNC) of the Laboratory of Geophysics and Seismology (LGS) at Technological Educational Institute (TEI) at Chania. b) The use of Wavelet Transform (WT) in several applications during the operation of the aforementioned network. SNC began its operation in 2003. It is designed and built in order to provide denser network coverage, real time data transmission to CRC, real time telemetry, use of wired ADSL lines and dedicated private satellite links, real time data processing and estimation of source parameters as well as rapid dissemination of results. All the above are implemented using commercial hardware and software which is modified and where is necessary, author designs and deploy additional software modules. Up to now (July 2008) SNC has recorded 5500 identified events (around 970 more than those reported by national bulletin the same period) and its seismic catalogue is complete for magnitudes over 3.2, instead national catalogue which was complete for magnitudes over 3.7 before the operation of SNC. During its operation, several applications at SNC used WT as a signal processing tool. These applications benefited from the adaptation of WT to non-stationary signals such as the seismic signals. These applications are: HVSR method. WT used to reveal undetectable non-stationarities in order to eliminate errors in site’s fundamental frequency estimation. Denoising. Several wavelet denoising schemes compared with the widely used in seismology band-pass filtering in order to prove the superiority of wavelet denoising and to choose the most appropriate scheme for different signal to noise ratios of seismograms. EEWS. WT used for producing magnitude prediction equations and epicentral estimations from the first 5 secs of P wave arrival. As an alternative analysis tool for detection of significant indicators in temporal patterns of seismicity. Multiresolution wavelet analysis of seismicity used to estimate (in a several years time period) the time where the maximum emitted earthquake energy was observed.

Artificial Neural Networks (ANNs) have been comprehensively researched, both from a computer scientific perspective and with regard to their use for predictive modelling in a wide variety of applications including hydrology and the environment. Yet their adoption for live, real-time systems remains on the whole sporadic and experimental. A plausible hypothesis is that this may be at least in part due to their treatment heretofore as “black boxes” that implicitly contain something that is unknown, or even unknowable. It is understandable that many of those responsible for delivering Early Warning Systems (EWS) might not wish to take the risk of implementing solutions perceived as containing unknown elements, despite the computational advantages that ANNs offer. This thesis therefore builds on existing efforts to open the box and develop tools and techniques that visualise, analyse and use ANN weights and biases especially from the viewpoint of neural pathways from inputs to outputs of feedforward networks. In so doing, it aims to demonstrate novel approaches to self-improving predictive model construction for both regression and classification problems. This includes Neural Pathway Strength Feature Selection (NPSFS), which uses ensembles of ANNs trained on differing subsets of data and analysis of the learnt weights to infer degrees of relevance of the input features and so build simplified models with reduced input feature sets. Case studies are carried out for prediction of flooding at multiple nodes in urban drainage networks located in three urban catchments in the UK, which demonstrate rapid, accurate prediction of flooding both for regression and classification. Predictive skill is shown to reduce beyond the time of concentration of each sewer node, when actual rainfall is used as input to the models. Further case studies model and predict statutory bacteria count exceedances for bathing water quality compliance at 5 beaches in Southwest England. An illustrative case study using a forest fires dataset from the UCI machine learning repository is also included. Results from these model ensembles generally exhibit improved performance, when compared with single ANN models. Also ensembles with reduced input feature sets, using NPSFS, demonstrate as good or improved performance when compared with the full feature set models. Conclusions are drawn about a new set of tools and techniques, including NPSFS and visualisation techniques for inspection of ANN weights, the adoption of which it is hoped may lead to improved confidence in the use of ANN for live real-time EWS applications.

Chi ha la responsabilità di gestire un’area protetta non solo deve essere consapevole dei problemi ambientali dell’area ma dovrebbe anche avere a disposizione dati aggiornati e appropriati strumenti metodologici per esaminare accuratamente ogni singolo problema. In effetti, il decisore ambientale deve organizzare in anticipo le fasi necessarie a fronteggiare le prevedibili variazioni che subirà la pressione antropica sulle aree protette. L’obiettivo principale della Tesi è di natura metodologica e riguarda il confronto tra differenti metodi statistici multivariati utili per l’individuazione di punti critici nello spazio e per l’ordinamento degli “oggetti ambientali” di studio e quindi per l’individuazione delle priorità di intervento ambientale. L’obiettivo ambientale generale è la conservazione del patrimonio di biodiversità. L’individuazione, tramite strumenti statistici multivariati, degli habitat aventi priorità ecologica è solamente il primo fondamentale passo per raggiungere tale obiettivo. L’informazione ecologica, integrata nel contesto antropico, è un successivo essenziale passo per effettuare valutazioni ambientali e per pianificare correttamente le azioni volte alla conservazione. Un’ampia serie di dati ed informazioni è stata necessaria per raggiungere questi obiettivi di gestione ambientale. I dati ecologici sono forniti dal Ministero dell’Ambiente Italiano e provengono al Progetto “Carta della Natura” del Paese. I dati demografici sono invece forniti dall’Istituto Italiano di Statistica (ISTAT). I dati si riferiscono a due aree geografiche italiane: la Val Baganza (Parma) e l’Oltrepò Pavese e Appennino Ligure-Emiliano. L’analisi è stata condotta a due differenti livelli spaziali: ecologico-naturalistico (l’habitat) e amministrativo (il Comune). Corrispondentemente, i risultati più significativi ottenuti sono: 1. Livello habitat: il confronto tra due metodi di ordinamento e determinazione delle priorità, il metodo del Vettore Ideale e quello della Preminenza, tramite l’utilizzo di importanti metriche ecologiche come il Valore Ecologico (E.V.) e la Sensibilità Ecologica (E.S.), fornisce dei risultati non direttamente comparabili. Il Vettore Ideale, non essendo un procedimento basato sulla ranghizzazione dei valori originali, sembra essere preferibile nel caso di paesaggi molto eterogenei in senso spaziale. Invece, il metodo della Preminenza probabilmente è da preferire in paesaggi ecologici aventi un basso grado di eterogeneità intesa nel senso di differenze non troppo grandi nel E.V. ed E.S. degli habitat. 2. Livello comunale: Al fine di prendere delle decisioni gestionali ed essendo gli habitat solo delle suddivisioni naturalistiche di un dato territorio, è necessario spostare l’attenzione sulle corrispondenti unità amministrative territoriali (i Comuni). Da questo punto di vista, l’introduzione della demografia risulta essere un elemento centrale oltre che di novità nelle analisi ecologico-ambientali. In effetti, l’analisi demografica rende il risultato di cui al punto 1 molto più realistico introducendo altre dimensioni (la pressione antropica attuale e le sue tendenze) che permettono l’individuazione di aree ecologicamente fragili. Inoltre, tale approccio individua chiaramente le responsabilità ambientali di ogni singolo ente territoriale nei riguardi della difesa della biodiversità. In effetti un ordinamento dei Comuni sulla base delle caratteristiche ambientali e demografiche, chiarisce le responsabilità gestionali di ognuno di essi. Un’applicazione concreta di questa necessaria quanto utile integrazione di dati ecologici e demografici viene discussa progettando una Rete Ecologica (E.N.). La Rete cosi ottenuta infatti presenta come elemento di novità il fatto di non essere “statica” bensì “dinamica” nel senso che la sua pianificazione tiene in considerazione il trend di pressione antropica al fine di individuare i probabili punti di futura fragilità e quindi di più critica gestione.
Who has the responsibility to manage a conservation zone, not only must be aware of environmental problems but should have at his disposal updated databases and appropriate methodological instruments to examine carefully each individual case. In effect he has to arrange, in advance, the necessary steps to withstand the foreseeable variations in the trends of human pressure on conservation zones. The essential objective of this Thesis is methodological that is to compare different multivariate statistical methods useful for environmental hotspot detection and for environmental prioritization and ranking. The general environmental goal is the conservation of the biodiversity patrimony. The individuation, through multidimensional statistical tools, of habitats having top ecological priority, is only the first basic step to accomplish this aim. Ecological information integrated in the human context is an essential further step to make environmental evaluations and to plan correct conservation actions. A wide series of data and information has been necessary to accomplish environmental management tasks. Ecological data are provided by the Italian Ministry of the Environment and they refer to the Map of Italian Nature Project database. The demographic data derives from the Italian Institute of Statistics (ISTAT). The data utilized regards two Italian areas: Baganza Valley and Oltrepò Pavese and Ligurian-Emilian Apennine. The analysis has been carried out at two different spatial/scale levels: ecological-naturalistic (habitat level) and administrative (Commune level). Correspondingly, the main obtained results are: 1. Habitat level: comparing two ranking and prioritization methods, Ideal Vector and Salience, through important ecological metrics like Ecological Value (E.V.) and Ecological Sensitivity (E.S.), gives results not directly comparable. Being not based on a ranking process, Ideal Vector method seems to be used preferentially in landscapes characterized by high spatial heterogeneity. On the contrary, Salience method is probably to be preferred in ecological landscapes characterized by a low degree of heterogeneity in terms of not large differences concerning habitat E.V. and E.S.. 2. Commune level: Being habitat only a naturalistic partition of a given territory, it is necessary, for management decisions, to move towards the corresponding administrative units (Communes). From this point of view, the introduction of demography is an essential element of novelty in environmental analysis. In effect, demographic analysis makes the goal at point 1 more realistic introducing other dimensions (actual human pressure and its trend) which allows the individuation of environmentally fragile areas. Furthermore this approach individuates clearly the environmental responsibility of each administrative body for what concerns the biodiversity conservation. In effect communes’ ranking, according to environmental/demographic features, clarify the responsibilities of each administrative body. A concrete application of this necessary and useful integration of ecological and demographic data has been developed in designing an Ecological Network (E.N.).The obtained E.N. has the novelty to be not “static” but “dynamic” that is the network planning take into account the demographic pressure trends in the individuation of the probable future fragile points.

Practitioners of Forecast-based Action (FbA) argue that a humanitarian response able to utilize forecasts to accurately predict disaster, secure funding, and take action before the onset of a crisis will benefit donors and beneficiaries alike. In search of effective and efficient early-action regimes, a number of major humanitarian actors are developing FbA projects of various designs, predominantly in response to natural disaster and famine. While numerous organizations and institutions have expressed interest in developing FbA mechanisms, the tool has only been applied in a limited capacity to the humanitarian needs generated by armed conflict. This research seeks to understand whether a scalable FbA approach can be developed to stage principled, anticipatory humanitarian action in response to situations in which rigorous evaluations predict the likelihood of imminent armed conflict. The hypothesis is that the application of FbA to armed conflict is possible, but due to the complex political nature of conflict, implementing organizations should try to focus on creating mechanisms managed by humanitarian actors and, in so far as possible, be insulated from outside influence. This research is the first academic work to specifically investigate the application of FbA to armed conflict. Following an extensive review of current FbA mechanisms and conflict early warning practices, this research concludes that a conflict-centered FbA system akin to the automated FbA systems in use today to respond to natural disaster and famine is possible, but that the endeavor presents many practical and conceptual barriers to implementation. In particular, diffuse models such as the Start Fund offer a hopeful glimpse at a type of horizontal, member-driven FbA mechanism that is both highly context-sensitive and relatively insulated from outside influence. Such a design, however, features notable and inherent limitations in its ability to reliably and accurately predict the outbreak of conflict and respond in a manner that minimizes regretful actions.

In the last 50 years, flooding has figured as the most frequent and widespread natural disaster globally. Extreme precipitation events stemming from climate change could alter the hydro-geological regime resulting in increased flood risk. Near real-time precipitation monitoring at local scale is essential for flood risk mitigation in urban and suburban areas, due to their high vulnerability. Presently, most of the rainfall data is obtained from ground‐based measurements or remote sensing that provide limited information in terms of temporal or spatial resolution. Other problems may be due to the high costs. Furthermore, rain gauges are unevenly spread and usually placed away from urban centers. In this context, a big potential is represented by the use of innovative techniques to develop low-cost monitoring systems. Despite the diversity of purposes, methods and epistemological fields, the literature on the visual effects of the rain supports the idea of camera-based rain sensors but tends to be device-specific. The present thesis aims to investigate the use of easily available photographing devices as rain detectors-gauges to develop a dense network of low-cost rainfall sensors to support the traditional methods with an expeditious solution embeddable into smart devices. As opposed to existing works, the study focuses on maximizing the number of image sources (like smartphones, general-purpose surveillance cameras, dashboard cameras, webcams, digital cameras, etc.). This encompasses cases where it is not possible to adjust the camera parameters or obtain shots in timelines or videos. Using a Deep Learning approach, the rainfall characterization can be achieved through the analysis of the perceptual aspects that determine whether and how a photograph represents a rainy condition. The first scenario of interest for the supervised learning was a binary classification; the binary output (presence or absence of rain) allows the detection of the presence of precipitation: the cameras act as rain detectors. Similarly, the second scenario of interest was a multi-class classification; the multi-class output described a range of quasi-instantaneous rainfall intensity: the cameras act as rain estimators. Using Transfer Learning with Convolutional Neural Networks, the developed models were compiled, trained, validated, and tested. The preparation of the classifiers included the preparation of a suitable dataset encompassing unconstrained verisimilar settings: open data, several data owned by National Research Institute for Earth Science and Disaster Prevention - NIED (dashboard cameras in Japan coupled with high precision multi-parameter radar data), and experimental activities conducted in the NIED Large Scale Rainfall Simulator. The outcomes were applied to a real-world scenario, with the experimentation through a pre-existent surveillance camera using 5G connectivity provided by Telecom Italia S.p.A. in the city of Matera (Italy). Analysis unfolded on several levels providing an overview of generic issues relating to the urban flood risk paradigm and specific territorial questions inherent with the case study. These include the context aspects, the important role of rainfall from driving the millennial urban evolution to determining present criticality, and components of a Web prototype for flood risk communication at local scale. The results and the model deployment raise the possibility that low‐cost technologies and local capacities can help to retrieve rainfall information for flood early warning systems based on the identification of a significant meteorological state. The binary model reached accuracy and F1 score values of 85.28% and 0.86 for the test, and 83.35% and 0.82 for the deployment. The multi-class model reached test average accuracy and macro-averaged F1 score values of 77.71% and 0.73 for the 6-way classifier, and 78.05% and 0.81 for the 5-class. The best performances were obtained in heavy rainfall and no-rain conditions, whereas the mispredictions are related to less severe precipitation. The proposed method has limited operational requirements, can be easily and quickly implemented in real use cases, exploiting pre-existent devices with a parsimonious use of economic and computational resources. The classification can be performed on single photographs taken in disparate conditions by commonly used acquisition devices, i.e. by static or moving cameras without adjusted parameters. This approach is especially useful in urban areas where measurement methods such as rain gauges encounter installation difficulties or operational limitations or in contexts where there is no availability of remote sensing data. The system does not suit scenes that are also misleading for human visual perception. The approximations inherent in the output are acknowledged. Additional data may be gathered to address gaps that are apparent and improve the accuracy of the precipitation intensity prediction. Future research might explore the integration with further experiments and crowdsourced data, to promote communication, participation, and dialogue among stakeholders and to increase public awareness, emergency response, and civic engagement through the smart community idea.
Negli ultimi 50 anni, le alluvioni si sono confermate come il disastro naturale più frequente e diffuso a livello globale. Tra gli impatti degli eventi meteorologici estremi, conseguenti ai cambiamenti climatici, rientrano le alterazioni del regime idrogeologico con conseguente incremento del rischio alluvionale. Il monitoraggio delle precipitazioni in tempo quasi reale su scala locale è essenziale per la mitigazione del rischio di alluvione in ambito urbano e periurbano, aree connotate da un'elevata vulnerabilità. Attualmente, la maggior parte dei dati sulle precipitazioni è ottenuta da misurazioni a terra o telerilevamento che forniscono informazioni limitate in termini di risoluzione temporale o spaziale. Ulteriori problemi possono derivare dagli elevati costi. Inoltre i pluviometri sono distribuiti in modo non uniforme e spesso posizionati piuttosto lontano dai centri urbani, comportando criticità e discontinuità nel monitoraggio. In questo contesto, un grande potenziale è rappresentato dall'utilizzo di tecniche innovative per sviluppare sistemi inediti di monitoraggio a basso costo. Nonostante la diversità di scopi, metodi e campi epistemologici, la letteratura sugli effetti visivi della pioggia supporta l'idea di sensori di pioggia basati su telecamera, ma tende ad essere specifica per dispositivo scelto. La presente tesi punta a indagare l'uso di dispositivi fotografici facilmente reperibili come rilevatori-misuratori di pioggia, per sviluppare una fitta rete di sensori a basso costo a supporto dei metodi tradizionali con una soluzione rapida incorporabile in dispositivi intelligenti. A differenza dei lavori esistenti, lo studio si concentra sulla massimizzazione del numero di fonti di immagini (smartphone, telecamere di sorveglianza generiche, telecamere da cruscotto, webcam, telecamere digitali, ecc.). Ciò comprende casi in cui non sia possibile regolare i parametri fotografici o ottenere scatti in timeline o video. Utilizzando un approccio di Deep Learning, la caratterizzazione delle precipitazioni può essere ottenuta attraverso l'analisi degli aspetti percettivi che determinano se e come una fotografia rappresenti una condizione di pioggia. Il primo scenario di interesse per l'apprendimento supervisionato è una classificazione binaria; l'output binario (presenza o assenza di pioggia) consente la rilevazione della presenza di precipitazione: gli apparecchi fotografici fungono da rivelatori di pioggia. Analogamente, il secondo scenario di interesse è una classificazione multi-classe; l'output multi-classe descrive un intervallo di intensità delle precipitazioni quasi istantanee: le fotocamere fungono da misuratori di pioggia. Utilizzando tecniche di Transfer Learning con reti neurali convoluzionali, i modelli sviluppati sono stati compilati, addestrati, convalidati e testati. La preparazione dei classificatori ha incluso la preparazione di un set di dati adeguato con impostazioni verosimili e non vincolate: dati aperti, diversi dati di proprietà del National Research Institute for Earth Science and Disaster Prevention - NIED (telecamere dashboard in Giappone accoppiate con dati radar multiparametrici ad alta precisione) e attività sperimentali condotte nel simulatore di pioggia su larga scala del NIED. I risultati sono stati applicati a uno scenario reale, con la sperimentazione attraverso una telecamera di sorveglianza preesistente che utilizza la connettività 5G fornita da Telecom Italia S.p.A. nella città di Matera (Italia). L'analisi si è svolta su più livelli, fornendo una panoramica sulle questioni relative al paradigma del rischio di alluvione in ambito urbano e questioni territoriali specifiche inerenti al caso di studio. Queste ultime includono diversi aspetti del contesto, l'importante ruolo delle piogge dal guidare l'evoluzione millenaria della morfologia urbana alla determinazione delle criticità attuali, oltre ad alcune componenti di un prototipo Web per la comunicazione del rischio alluvionale su scala locale. I risultati ottenuti e l'implementazione del modello corroborano la possibilità che le tecnologie a basso costo e le capacità locali possano aiutare a caratterizzare la forzante pluviometrica a supporto dei sistemi di allerta precoce basati sull'identificazione di uno stato meteorologico significativo. Il modello binario ha raggiunto un'accuratezza e un F1-score di 85,28% e 0,86 per il set di test e di 83,35% e 0,82 per l'implementazione nel caso di studio. Il modello multi-classe ha raggiunto un'accuratezza media e F1-score medio (macro-average) di 77,71% e 0,73 per il classificatore a 6 vie e 78,05% e 0,81 per quello a 5 classi. Le prestazioni migliori sono state ottenute nelle classi relative a forti precipitazioni e assenza di pioggia, mentre le previsioni errate sono legate a precipitazioni meno estreme. Il metodo proposto richiede requisiti operativi limitati, può essere implementato facilmente e rapidamente in casi d'uso reali, sfruttando dispositivi preesistenti con un uso parsimonioso di risorse economiche e computazionali. La classificazione può essere eseguita su singole fotografie scattate in condizioni disparate da dispositivi di acquisizione di uso comune, ovvero da telecamere statiche o in movimento senza regolazione dei parametri. Questo approccio potrebbe essere particolarmente utile nelle aree urbane in cui i metodi di misurazione come i pluviometri incontrano difficoltà di installazione o limitazioni operative o in contesti in cui non sono disponibili dati di telerilevamento o radar. Il sistema non si adatta a scene che sono fuorvianti anche per la percezione visiva umana. I limiti attuali risiedono nelle approssimazioni intrinseche negli output. Per colmare le lacune evidenti e migliorare l'accuratezza della previsione dell'intensità di precipitazione, sarebbe possibile un'ulteriore raccolta di dati. Sviluppi futuri potrebbero riguardare l'integrazione con ulteriori esperimenti in campo e dati da crowdsourcing, per promuovere comunicazione, partecipazione e dialogo aumentando la resilienza attraverso consapevolezza pubblica e impegno civico in una concezione di comunità smart.

碩士
國立交通大學
土木工程系所
105
Earthquake Early Warning [EEW] has become one of the forefront researches in seismology and signal processing in recent years. The key reason for this popularity is related to the great deal of application this research has in present day society since earthquakes are still responsible for a large percentage of property damages and lives lost. At present it is still beyond our understanding to predict an earthquake before it occurs, however EEW is the closest option we have to help save lives and reduce losses. Earthquake early warning hence is an important tool for modern day earthquake prone societies like Taiwan, Japan and the US. Lots of resources are being spent on the development of different types of EEW system networks of varying types. Current systems may be operated by both government of private sectors. Most if not all government systems are mainly based on traditional seismic stations positions at different geographical locations, these sensors alert distant sensors and main observation station when an event is triggered, giving warning of the incoming propagation of an earthquake wave. More advanced methods being develop utilized by private companies and researchers use smartphones as sensors with installed applications that feeds back acceleration data of the smartphone to a centralized network that triggers when an earthquake event is detected. The latter of these methods are the focus of this research document. This is the first step in a multi-phase development a smartphone based seismic detection network using MATLAB in Taiwan. This initial step is to show that a smartphone can be used to measure earthquake intensity and issue an alert to a single user using the MATLAB platform. The smart phone used in this research is a Samsung Galaxy S6 device. Other tools used includes installation of the MATLAB Mobile app on the smartphone, a computer with MATLAB software and a router. Tests were performed at the National Chiao Tung University Structural Engineering Building and at the National Center for Research on Earthquake Engineering in Taiwan. One of the main goals in this research was to investigate alternative methods to trigger an earthquake alarm apart from using the more popular STA/LTA algorithm however the main objective was to be able to create a MATLAB script that can alert a single user when an earthquake is occurring. To do this an earthquake event was simulated by placing the smartphone on a shaking table while running the written MATLAB code. The result of which allowed the researcher to remotely detect an earthquake motion and receive and alert that an earthquake event was being propagated. Final results showed PGA (Peak Ground Acceleration) values measured were within eighty-five (85%) of input signal’s peak acceleration amplitudes. Reaction time from the start of the event to reporting of an event is estimated to be around ten (10) seconds on average. On this basis one can develop future mechanisms to create a smartphone seismic network locally which can be regionally applied.

The thesis is about rockfall early warning with a micro-seismic monitoring network, that includes rockfall events detection, classification, localization, and a method for large rock fall early warning.

The main topic of this analysis is the development and testing of possible strategies to seismic risk mitigation through the declaration of alert massage, ongoing of earthquake, before that the strong sacking occur. With this aim, a single and multi-stations approach are investigated performing statistical test to proof the rapidity and reliability of alert. The first step of this study is explained in the chapter 1, where an on-site EEWS approach, called SAVE system, is implemented and tested. The system is a basic on-site single-station approach able to quickly characterize the earthquake in term of local intensity, magnitude and distance classification. The estimates provided by the system are affected by rather large uncertainties but the methodology is rapid, so to allow for an effective activation of automatic procedures for risk mitigation, also at the target sites very close to the source. The low computational cost of this algorithm makes it suitable for embedded solutions in a standalone seismic station, and a prototype was developed and described in the final chapter. In chapter 2, a refined and more accurate on-site system is proposed. Unlike the SAVE methodology, this P-wave based approach is aimed at providing a reliable overcoming prediction of the critical intensity level at recording site. The use of advanced data processing techniques, jointed with a multi-parameter approach, provides a more reliable prediction of maximum local intensity expected at site. This approach is useful for the rapid and strictly constrained local intensity prediction. Chapter 3 finally describes an integrated methodology aimed to interpolate the information obtained from the individual stations and map the potential damage zone (PDZ) in real time. Further details on this methodology, called Quake-up, are discussed in the same Chapter. In chapter 4, technical details about the project development of seismic station, called MOMA, and improvement of the spectral response of on-board geophone were provided.

碩士
逢甲大學
資訊電機工程碩士在職專班
95
Most Enterprises encounter the financial crisis can often be offered from the financial ratio in the financial report. Most investors often select the stock by the Earning per Share (EPS) in the stock market but in fact, the companies that the earning per share capacity is low even if crisis Company, per share earning capacity a well-off company of the financial affairs, need assessment in many aspects and consider in the financial crises of enterprises. This research will combine every financial ratio, through the analytic approach of the neural network technology, find out and solve the route and build and construct a set of financial crisis precaution model beastly. In documents, the general financial crisis precaution model, all regard more apparent financial ratio as parameters, and then make use of these parameters to construct the financial crisis precaution model. Channel the contingent financial ratio into the financial crisis precaution model in this research, as the main research approach. This research will be regarded as the sample of studying with the non- financial type stock of the listed company, Six great 33 kinds of financial ratio of an index in the financial statement are regarded as the parameter, the ones that analyse and find out the financial ratio and financial crisis are related through correlate analysis and regression analysis, extracting out the financial rate regards as the parameter of studying, hive off with kinds of neural method of network the materials finally, by explaining the information of hiving off ,predict that the financial management state of listed company judges whether to worth being invested in or not. Shown by the result of study , considering MSE, R-value, crisis incidence and classification correct rate of training samples and test sample, extract out nine financial rates to construct the early warning system of Financial Distress , its accuracy of predicting the financial crisis is had, train samples has 97.42% of the correct rates of classification (The normal company is 98.85%, the crisis company is 95.98%), testing samples has 74.49% of the correct rates of classification (The normal company is 71.43%, the crisis company is 77.55%).

碩士
國立臺灣科技大學
營建工程系
105
Natural disasters such as typhoons, floods and earthquakes occur frequently in Taiwan. A signal of warning in advance of the disaster is preferred since such message may help us reduce the damage. Big data from network intelligence resources flourishes in recent years. This research attempts to explore the feasibility of the technology that is used in disaster prevention. The collected data are divided into PTT Gossiping and three news websites (China times、Udn news and Apple daily). In order to examine and analyze the collected data, three representative events were selected as case studies, including the disastrous earthquake on February 6th, 2016, flooding in Taiwan Taoyuan International Airport on June 2nd, 2016 and Typhoon Nepartak in July 8th, 2016 Jieba is used to define keywords and ARIMA is sued to build a model for predicting the collected data. Based on results found here, some observation are described as follows: 1. Both PTT Gossiping and NEWS websites reflect the occurrence of disaster correctly, 2. Because earthquake covers greater area compared to that of the flooding event occurred in airport, information from PTT Gossiping can reflect hazard event earlier, 3. On the hand, flooding only influences its nearby area, so the news websites can have an earlier response, 4.ARIMA is able to predict the degree of disaster discussion on the internet community network.

博士
國立清華大學
工業工程與工程管理學系
98
Many studies have showed the human errors were the major reason to cause of the accident. In fact, human errors may result in tangible or intangible cost loss and influence of system safety. Especially in nuclear power plant (NPP), serious human failure would damage to operation safety. Thus, no matter what the roles they are, such as operation or maintenance, each of the people in NPPs is played as a vital role. Previous studies indicate that about 30 to 50 percent accidents were result from human failure. Hence, the aim of this dissertation would design two early warning methods in control room and maintenance department. The proposed early warning methods in this study could decrease human error in both control room and maintenance environment. First of all, this study applied the concepts of the Shewhart control chart to design a pre-alarm system for NPP control room. Two pre-alarm types were designed to compare with the original system, and all participants were requested to monitor each simulated system under both normal and abnormal states. The tasks for the participants included shutting down the reactor, searching for procedures, monitoring system parameters and executing secondary tasks. In each trial, the task performance, mental workload and situation awareness (SA) of the participants were measured. Respecting to maintenance, currently, on-line maintenance for NPPs is performed quite often while the system is in operation. The limited maintenance time very often bring heavy mental workload to their engineers. Therefore, according to the factors affecting the mental workload, this dissertation would construct a predictive mental workload model while maintaining digital systems in NPP. Through predicting mental workload, the manager can organize the human resources for each daily task to sustain the appropriate mental workload as well as improve maintenance performance. Finally, to verify the feasibility of the proposed scheme, this dissertation further applies GMDH into IC packaging plant to develop a suitable pre-alarm system. The results indicated (1) that participants had lower mental workload, but equal SA, when monitoring the system with either type of pre-alarm designs, (2) that lower alarm frequency and higher secondary task performance were obtained with the pre-alarm design, (3) that the proposed model is expected to provide the supervisor a reference value of engineers’ mental workload and the prediction ability of the model was high. (4) GMDH can construct a reliable pre-alarm system to reduce operators’ visual fatigue and raise yield rate in IC packaging plant.

碩士
大葉大學
電機工程學系碩士班
91
The fuzzy controller, the core of the ASDL broadband network failure early warning system, is the subject of this study which develops on the basis of the fuzzy theory. The basic framework of the controller includes two inputs, incoming and outgoing traffic values from the DSLAM(Digital Subscriber Line Access Multiplexer) of ADSL network equipment towards the Remote terminal of ADSL Tranceiver Unit(ATU-R), one output set has two respective modes of different traffic type, network communication status and subscriber circuit board status. According to statistical analysis, the membership function defines the traffic value of the inputs into three fuzzy subsets of HIGH, MED, LOW, and the operation status of the output into three fuzzy subsets of BUSY, ACT, and DOWN. The system first receives DSLAM-to-ATU-R incoming and outgoing traffic values, then feeds the data into the fuzzy controller, and finally uses the relative relationship between the incoming and outgoing traffic values to reason the output result applying fuzzy rules. According to the definition of the membership function, an output result of BUSY indicates system in high traffic, ACT indicates system in operation, and DOWN indicates a failure condition. The failure early warning system of this study is mainly designed to monitor an abnormal status of the upstream equipment of DSLAM—ATM, and ISP, and the IP layer of network communications. It provides speedy failure information and alerts to network administrators through an early warning mechanism in order to elevate the efficiency of repair and ensure a good communication quality. In addtion, the failure early warning system of this study is to make improvement on failure detection function to ensure a reduced time for failure discovery and increased client satisfaction, in unusual situations when DSLAM system’s fail to alarm of subscriber circuit board irregularities.

碩士
國立宜蘭大學
生物機電工程學系碩士班
106
The purpose of this study was to use long range (LoRa) communication technologies to develop an orchard environment monitoring and early warning system. Remote environmental data acquisition and analysis could be conducted using the system to establish a database of the orchard environment. When environmental hazards were detected, the early warning system would immediately inform the user of real-time situations by sending them emergency alert messages. Three orchards of Shang Jiang pear in Sanxing Township, Yilan County, Taiwan, were used as the experimental sites, in which environmental monitoring stations were built. Identical sets of environmental sensors, including environmental temperature, relative humidity, soil moisture, wind speed, and illuminance sensors, were installed in the monitoring stations. Arduino Uno microcontroller board was used to extract environmental data. Subsequently, a LoRa communication module was used to transmit the environmental parameters to a LoRa base station and upload the data to the Message Queuing Telemetry Transport server. Distances between the orchards and the LoRa base station were 3.15, 6.53, and 8.05 km. LabVIEW 2015 was employed to establish an environmental database and store relevant data for analysis. Node-RED websites were utilized to display environmental information and sent emergency alarm messages. The orchard environmental monitoring and early warning system established based on the two proposed methods can achieve its designated goals. To facilitate effective data collection, the developed system transmitted sensor readings of the environment were digitalized and integrated, and then transmitted through LoRa wireless communication technology to the server end. According to the data sent from the environmental sensors, users observed environmental changes in each of the orchards and compared the geographical locations and microclimate differences among the orchards. The proposed system can be used for long-term data collection in orchard environments to accumulate a considerable amount of environmental data. In the future, the accumulated data can be used to create big data databases of orchard environments and optimize the orchard environment monitoring and early warning system. Then, the Shang Jiang pear trees can be cultivated in an environment most suitable for their growth and bear fruits with improved quality and quantity.

碩士
朝陽科技大學
財務金融系
107
Recently, the rapid growth of financial frauds, such as fraudulent financial reporting and depletion of the companys assets, have resulted in investors losing confidence in the securities market. Therefore, the Securities and Futures Institute implemented the Information Disclosure and Transparency Ranking System to make the information more transparent, and allowed the companies to disclose more information in financial statements before encountering the financial crisis. This study used the listed electronics company in stock exchange market and over-the-counter market as the research object from 2010 to 2018. Respectively, and applied decision tree and deep neural network approaches to establish a financial crisis prediction Model. By adding the information disclosure as a control variable, the empirical results showed that the overall accuracy of decision tree method reached 88.83%,and adding the information disclosure variable reaches 89.39%.The overall accuracy of the deep neural network reached 78.81%, and adding the information disclosure variable reaches 79.80%. Using the decision tree with the True Positive Rate is 89.93%, and adding the information disclosure is 91.18%, the True Positive Rate of the deep neural network is 92.54%, and adding the information disclosure variable reaches 92.79%. Both of decision tree and deep neural network could effectively predict the financial crisis of companies. Especially, the decision tree achieved the best prediction result. If selecting seven important financial ratios from thirteen variables through correlation coefficient, the accuracy rate will be affected. The accuracy rate of the decision tree for empirical results reduced to 87.15% and the deep neural network reduced to 75.98%, but the rules are reduced by half. Therefore, The results show that investors can give priority to these seven ratios, so that investors can find crisis companies early and adjust their investment strategies.

碩士
國立交通大學
管理學院高階主管管理碩士學程
92
The purpose of this research is to construct an early warning model for financial crisis of the listed companies by using artificial neural network (ANN) with back propagation (BP) algorithm. ANN has error tolerance ability, learning ability, high speed computational ability, and high-volume memorizing ability. It also considers both linear and nonlinear relationship at the same time. To compare with the traditional method, we also adopt logistic regression method to build early warning model. Results show that the accuracy rate in forecasting financial crises is superior for ANN model than that of logistic model.

碩士
中華大學
科技管理研究所
94
The credit departments of farmers’ associations play an important role in the community of finance. In recent years, these departments have faced operational crises because of the competitive industry environment and the uneasiness to accumulate net values. Therefore, it becomes a problem in the Framer’s Associations. Furthermore, it also influences the stability of the overall financial system and thus increases the social cost. Traditionally, Discriminant Analysis and Logistic Regression Analysis are the most popular tools which are used to predict financial crises. However these tools have limited themselves to a stricter environment or background, which is lack of adaptability in reality. In this thesis, BPN is used as a prediction tool because it includes parallel processing, inductive reasoning, and learning abilities. In order to verify the accuracy of the BNP were adopted, utilized 24 CAMELS financial ratios of the credit departments, which were absorbed in the last 1 to 3 years as parameters and the results verified that BNP’s accurate prediction rate of financial crises was almost 100%. Using the proposed model, several advantages could achieved (1) the supervisory organization could deal with a proper mechanism supervision that could prevent the waste of resources, (2) the executive could adjust strategy of management, (3) the depositor could reduce loss, and (4) the operation of the agriculture's financial system would be sound at the same time.

碩士
中原大學
土木工程研究所
101
The geological structure of Taiwan is complicated and fragmented, and a significant part of Taiwan is covered by mountains and hills. Due to numbers of population with no more plain, the highly development of slope has become normality. But it caused great loss of many lives and damages to properties when overexploitation of slope coupled with abundant rainfall . In addition, high mountains is also the best-choice locations of radar station and air defense basement, if the road of Combat readiness is damaged by natural disasters, it will cause seriously affected on tasks of national defense. So, the monitoring of slope stability for protecting, avoiding and decreasing natural disaster becomes a very important subject. As technology advances, real-time monitoring of slope stability by using Inclined Wireless Sensor Network is one of the new ideas on combined engineering technology with science. This study is aim to improve reliability and practicality of this application technology by providing suggestions on equipment specifications, procedures of accuracy correction, field provisioning and the management mechanism of slope.

碩士
義守大學
工業管理學系
103
Due to the rapid changes of the overall economic environment, possible financial distress increases in a corporation every year recently. Therefore, how to establish an effective early warning model of a business crisis is a relatively important issue for a corporation. In this thesis, the grey correlation analysis and neural network forecasting models were established to predict possible financial crises of a corporation for early warning. In this research, companies who listed in the Taiwan Stock Exchange and faced financial crisis during 2009 and 2012 were investigated. Other companies in the same industry with good financial conditions were compared with those who had financial crises at the same periods. Financial indicators and corporate governance variables for the last four seasons were studied. The grey relational analysis was used to filter out the most important factors that will affect the company’s financial conditions. Then, two neural networks were trained to find out the best forecasting model for financial indicators and corporate governance variables. Our results showed that the best predictive model was the model only used financial indicators from last season. However, the model incorporated both financial indicators and corporate governance variables may be considered for a long term forecasting.

碩士
國立臺灣大學
地質科學研究所
102
Since there is no practical method for earthquakes prediction today, the main disaster prevention method is based on the seismic design of buildings. Earthquake early warning (EEW) is another effective way to reduce damage in real-time (Kanamori et al., 1997). Because EEW needs to provide reliable message in a short time, it is important to shorten the reporting time window by the improving of data process. In order to provide location and magnitude after an earthquake just happened, a low cost and high density EEW system has been developed and established by using the Palert seismometers in Taiwan (Wu et al., 2013). Due to the distribution of the stations, which detected the signals at first, is poor. It needs more than eight stations to get reliable information. Thus, it shortens the lead time before strong ground shaking. This study use the arrival-time order location (AOL) method, which introduced by Anderson in 1981, to improve the efficiency of Palert EEW system for earthquake location. At the same time, because of Palert has a relatively low signal-to-noise ratio (S/N) in τc (Kanamori, 2005, Wu and Kanamori, 2005a) determinations. So τc approach does not use in the Palet EEW system (Wu et al., 2013). This study try to use the signals stacking small arrays to enhance S/N ratio and try to use τc for magnitude estimation. Results shows that, AOL method can provide a reliable earthquake location by only using four to five stations. It can improve the EEW efficiency. By stacking the signals from small array can also get more accurate magnitude estimation usingτc. So that more information can be provided in on-site EEW warning purpose.

Traditional communication channels like news channels are not able to provide spontaneous information about disasters unlike social networks namely, Twitter. The present research work proposes a framework by mining real-time disaster data from Twitter to predict the path a disaster like a tornado will take. The users of Twitter act as the sensors which provide useful information about the disaster by posting first-hand experience, warnings or location of a disaster. The steps involved in the framework are – data collection, data preprocessing, geo-locating the tweets, data filtering and extrapolation of the disaster curve for prediction of susceptible locations. The framework is validated by analyzing the past events. This framework has the potential to be developed into a full-fledged system to predict and warn people about disasters. The warnings can be sent to news channels or broadcasted for pro-active action.

碩士
國立臺灣海洋大學
應用經濟研究所
92
A series of crises are often arisen if there is something wrong with the management of the community financial, the damage and impact of which to economy is far more serious the that caused by the bankruptcy of a company. A financial warning system for the governing community financial institutions was more important. In the past researches of financial distress prediction, traditional statistical techniques such as multivariate statistical method, Before using the multivariable statistical method. There have been more artificial neural network applications to this field in domestic since 1994. According to those researches, financial distress prediction models build by artificial neural network was more feasible than traditional statistical methods. In this paper applied back-propagation network the build the financial distress prediction models, and to make the function of crisis management mechanism toward the community financial institution in Taiwan, and based on theoretical and legal construction. The predictability comparison provides the highest accuracy for Primitive BPN(81.10％) in the surveillance system, followed by Factory BPN(77.85％) and Ordered Logit(75.9％).

A Very Early–warning Smoke Detection Apparatus (VESDATM) detects the earliest traces of smoke by continuously sampling the air from a designated area. Air sampling is achieved by use of a system of long pipes containing numerous small inlet orifices termed as sampling holes. The air samples are drawn to the detector by means of an aspirator. In spite of the high sensitivity of the detector, much of this advantage can be lost if the smoke transport time within the pipe network is excessive. Consequently there has been a legislation introduced by Standards such as AS 1670 and BS 5839 stating the maximum transport time to be within 60 seconds of entering that extremity of a pipe system of 200 meters aggregate length, and the suction pressure was to be no less than 25 Pascals. Once the pipe network is installed, it is impractical and often impossible to test the transport time and suction pressure drop of every sampling hole in a complex network of pipes. Therefore, a software modelling tool is required to accurately predict these parameters to 90% of measured value with high accuracy. The flow regimes within the sampling pipes proved complex, involving frequent transitions between laminar and turbulent flows due to disturbances caused to the main flow by jet flows from the sampling holes. Consequently, the published equations to determine friction factors does not predict pressure loss and transport time results to an acceptable accuracy for this thesis. Computational Fluid Dynamics simulations were carried out at various magnitudes of disturbances similar to the effects in VESDA pipe network. The data from the CFD were analysed and the results were used as a guide to develop mathematical models to calculate the friction factor in flow regimes where jet disturbances are present. The local loss coefficients of fittings such as bends and couplings were experimentally determined for all types of fittings used in VESDA pipe networks. The local loss coefficients that were determined made significant improvements in calculating pressure losses compared to the results obtained when commonly used loss coefficient values were used. The characteristics of the VESDA aspirators of all models were determined. The experiments were carefully set up to ensure the apparatus did not have any influence on the aspirator performance. Mathematical models were developed for each VESDA model. A relationship between the magnitude of disturbance and the delay it caused for the smoke to travel from one segment to the next was established. From this relationship, a new transport time mathematical model was developed. Validations of all mathematical models were carried out in different pipe configurations. In all cases the results calculated were within 90% or better compared to the measured results.

碩士
國立臺北大學
企業管理學系
92
In recent years, the banks of Taiwan Small and Medium Enterprises have faced some operational risk due to economic recession and traditional industry to move outside. The non-performing loans of Taiwan Small and Medium Exterprises have raised from 17.62% in 2001 to 18.64% in 2002. In order to avoid financial crisis of the banks of Taiwan Small and Medium Enterprises, a set of system with systematical and scientific methods will make the degree of financial loss lower. By integrating data mining and artificial neural network , this research tries to develop the early warning system to detect the critical factors which affect the operational crisis of the Banks. The result of this study show that the early-warning system with artificial neural network surpasses the models with Logit and Probit. Besides, the early-warning system with Artificial Neural Network has high accurate rate. Basically, this study can provide the useful information for related policy decision-making.