- Author
-
Hosokawa, M.
- Title
- Study of Fire Judging Rule Using Fuzzy Reasoning: Application of Neural Networks.
- Journal
-
Report of Fire Research Institute of Japan,
No. 78,
33-40,
September 1994
- Keywords
-
fire detection systems
|
signal detection
|
neural networks
- Abstract
- [ABSTRACT IN ENGLISH] An important problem of conventional fire detection systems is the high incidence of false alarms caused by steam in bathrooms, dust, tobacco smoke, gases produced in cooking and spray mist. In these circumstances, we have developed a new intelligent fire detection system including a new fire detector with three sensors of smoke, heat, and CO gas. In addition, a new signal processing technology called Fuzzy Expert System is intorduced into the system. The membership functions in the system are classified to 4 categories for flaming fires, smoldering fires, steam, and tobacco smoke. The system can decrease the false alarms drastically and quickly and display the details of fire rooms. However, in order to increase the identification ability between fires and false alarms, there are some difficulties to adjust Fuzzy Membership Functions. The objective of this study is to develop the fire judging rule using fuzzy reasoning which reduces effort for adjustment. For this purpose, neural networks are applied to the fuzzy reasoning. Neural networks learnvariety of data which are realized in experimental room for flaming fire, smoldering, steam, tobacco and dust. As a result, the networks work as usual fuzzy membership functions. It has been found that the proposed fire judging rule can reduce the difficulties of adjustment and get visual rules which a human being has.