Electrical Fire Detection Based On LSTM Network And Fuzzy Reasoning With Multi-source Data Fusion

With the acceleration of urbanization in China,people gradually gather from scattered residential areas to major cities and towns.High-rise buildings are the main way to solve the problem of population explosion,but in recent years,high-rise buildings frequently generate gas fires,resulting in serious casualties and property losses.How to detect the electric fire effectively and accurately and avoid the huge losses caused by it in time has become a hot issue that scholars at home and abroad have jointly studied.However,unlike traditional fires,electrical fires are caused by electrical faults caused by electrical lines,which have the characteristics of high concealment,long development cycle and high harmfulness.It is difficult for traditional fire detection methods to detect or detect them with low accuracy.Therefore,starting from the mechanism of electric fire and the form of fire source,this paper analyses and points out the problems of low detection accuracy and single condition of alarm decision-making in the existing methods of electric fire detection,and then proposes a multi-sensor data fusion electric fire detection model based on LSTM network and fuzzy reasoning technology to solve these problems,so as to realize the accuracy of electric fire detection finally.Hierarchical output of identification and alarm decision-making.To this end,the specific work of this paper is summarized as follows.In order to solve the problem of low detection accuracy caused by the lack of effective use of time series characteristics of detection data in existing electrical fire detection methods,LSTM,a long-term and short-term memory network,which is good at processing time series data,is introduced into the electrical fire detection model.Firstly,the characteristics of electrical fire state changing with time during the development of electrical fire are analyzed,that is,the fire state at the preceding moment will affect the fire state at the following moment.Then,a feature layer recognition model based on LSTM network is established to update the recognition results of the detection model through the comprehensive judgment of the historical state of the detection sequence data and the current input state.Finally,the electrical fire status which can best represent the real situation of the target detection area in this period is obtained,which improves the recognition accuracy of electrical fire.Aiming at the problem that the decision-making conditions are too single and the output decision-making is difficult to characterize the severity of electrical fire in the process of alarm decision-making of existing electric fire detection methods,this paper introduces two decision-making factors,i.e.the abnormal duration of residual current signal in the target detection area and the fire protection grade of buildings,at the decision-making level of the detection model,and combines the existing smoldering fire and open fire probability to model.Fuzzy reasoning fusion,by virtue of the unique advantages of fuzzy reasoning technology in dealing with uncertain problems,obtains the final decision-making results,improves the rationality of electric fire alarm decision-making,and avoids the shortage or waste of fire power input.Finally,on the national standard test data sets of open fire SH4,smoldering fire SH1 and typical interference signals in kitchen,the detection model proposed in this paper is simulated.Experiments show that the model can identify several states of electrical fire accurately and steadily,and output reasonable alarm decisions according to the environmental characteristics of the target detection area.