论文部分内容阅读
报警系统失效主要包括漏报、误报,对系统进行失效概率预测,可以帮助判断设备质量优劣,评估系统效能。利用Matlab软件编程,通过神经网络预测失效概率。根据不同场所正在使用的火灾报警器的失效数据作为原始数据,归纳总结失效原因,建立事故树,结合专家打分法与模糊理论得到网络的输入值与输出值。通过网络训练,得到可以对系统失效概率进行预测的RBF神经网络,测算效率大幅提高。以70组不同品牌、用途的火灾报警系统作为算例,通过训练数据,最终达到输入底事件发生概率可直接输出顶事件发生概率的目的。结果表明,RBF神经网络相较于BP网络与事故树算得的失效概率具有更高的拟合度,RBF神经网络模型在进行系统失效概率预测时具有可靠性。
The failure of the alarm system mainly includes omission, false alarm and prediction of the failure probability of the system, which can help determine the quality of the equipment and evaluate the system performance. Using Matlab software programming to predict failure probability by neural network. Based on the failure data of fire alarm which is being used in different places as the original data, the cause of failure is summarized and the accident tree is established. The input and output values of the network are obtained by combining the scoring method of experts and fuzzy theory. Through the network training, the RBF neural network, which can predict the probability of system failure, is obtained, and the calculation efficiency is greatly improved. Taking 70 groups of fire alarm systems with different brands and purposes as an example, the probability of the occurrence of the top event can be output directly by training the data to finally reach the probability of occurrence of the bottom event. The results show that the RBF neural network has a higher fitting degree than the failure probability calculated by the BP network and the fault tree, and the RBF neural network model is reliable in predicting the system failure probability.