A Study on Intelligent System of Multi-Sensor Fire Predication

指導教授:黃淳德[[abstract]]人類本身就是最好的火警感測器,經由視覺、嗅覺、觸覺配合經驗累積，為保護自身安全能綜合研判局勢，當機立斷作出決策並採取行動，但無法以人類進行專注且長時間監控，情緒反應及持久耐力是一大考驗，硬體系統設備沒有情緒及倦勤缺點但如何仿效人類高智慧綜合資訊研判決策行為，是當今所有科技領域共同探索的課題，本文之火警預測以發展具人類智慧行為提出方案，令火警智慧系統對採樣數據進行分析研判，並作出決策動作，對環境誤報因素作出自我學習及修正功能，以期火警發報的正確性，達減少誤報率，即早預警之目的。 火災信息具非線性特徵以人工智慧數據處理方法，所建構的火災預測系統具自學習和自適應功能，為本文火災探測技術研究方向。以感測器完整的反應環境信息，為火災探測提供確切即時的決策數據。在探討傳統火災探測器動作原理及火災探測演算法中得知，人工智慧火災探測演算法的應用對提高火災預測正確率，降低誤報率，均優於前者，以類神經網路及模糊推理技術來框架火災預測系統，結合多感測器傳達實際情況，融合火災信息特徵，調整權重適應環境變化，經由模擬驗證該方法所建演算法在多感測器火警預測證實預期成果。[[abstract]]Humanity himself is the best fire sensor. By sight, smell or touch, most of all, accompany with the experiences, people can study and come to decision the overall situation of fire to take proper action to protect themselves. Besides the emotion, the limitation of stamina makes people cannot prolong monitoring for signal topic either. The artificial systems complement the shortages mentioned above, but how to emulate the intelligences of human become the important tasks nowadays. In this study, an intelligent prediction of fire detect system was proposed, it was trained by fire messages and can make accurate decision while fire occurred. Also, this intelligent system equips the achievements of less influence of circumstances, high accuracy and low misreport. The main purpose of this study is to construct an intelligent system of fire detection with self-learning and adaptive to deal the fire messages which with the characteristics of nonlinearity. The fire sensors provide correct and real time message to make fire prediction. The constructed intelligent system must offer higher accuracy and lower misreport than traditional separated system. In this study, the artificial neural networks, fuzzy inference, multi-sensors and introduce the concepts of fusion were combined to form this kind of system with the fitting of goals. Through simulation, the correct rate of the fire detection is improved and the false positive rate is reduced also