The research of the bayes decision-making model in fire alarm system based on normal distribution
Wang Li-ming, Shao Ying, Xiaoling Yan, Shan Yong · 2008
Currently in fire alarm system, detector bring on high failing alarm rate and high mistaking alarm rate within nearby territory value because of some kinds of reason. In this article, a solution is proposed that the detector characteristic values taken a kind of pattern to carry on the decision-making classification. The Bayes decision-making method is used in the classified process. Pattern classification of alarm is simplified to two values classification in continual condition. Single vector decision-making function based on multidimensional normal distribution is proposed according to the feature of the fire detector data. At the same time the classification differentiation function is produced based on the principle that the classification average risk is the smallest. Finally the classification distinguish function is simplified in different condition. The experiment indicates that new solution is found in this article that it can effectually reduce failing alarm rate and mistaking alarm rate. The conclusion indicated that using the pattern classification method is possible to achieve the quite good effect in the detector critical partial value.