The Research And Practice Of Deep Learning On Fire Alarm Intelligent Management

There are many brands and models of the fire automatic alarm master(hereinafter called the alarm master),and the alarm master of different brands or of same brands with different models has different data protocols,which increases the difficulty of the information integration of smart fire.And there will still be new protocols which are externally shown as fire,fault,shielding or interconnection and other fire incident information.For detection or environmental factors,which is not conducive to the optimization of fire resource allocation.The intelligent analysis of the protocol of the alarm master and the intelligent judgment of the false alarm information can well promote the healthy development of smart fire.This thesis first analyzes the external factors affecting the detection devices and puts forward the theoretical system of fire intelligent management according to which a complete system of fire intelligence analysis is established to achieve the data aggregation of the alarm protocol information and the external influencing factors and realize standardized processing to generate standard integrated fire incident information.On this basis,a SSANNSMFA model based on deep learning mechanism is proposed as the core module of fire intelligent analysis system.According to the characteristics of integrated fire incident information,this model achieves the deep learning of the fire protocol data of known detection devices and the quick and efficient identification of the fire protocol data of unknown detection devices by combining the Stacked Sparse Autoencoder Neural Network(SSANN)and the regularized softmax regression framework into a new deep neural network model.The calculation and derivation of the regression residual of SSANN and softmax and the experiment show that the SSANNSMFA model has higher classification accuracy and false alarm resolution for the feature extraction of the protocol data of the detection devices.Finally,the model is applied to fire information integrated network management platform.Through theoretical analysis and practice,it is shown that the fire analysis intelligent model proposed in this thesis can achieve the intelligent analysis of alarm protocol and solve the problem of some false alarms.At the same time,the model can also be applied to other areas of smart fire,such as the drawing recognition and inspection robots.In addition,it can be applied to other industrial control protocol data identification such as the PLC.However,the model still needs to be further verified and improved due to the small number of real fire alarms in practice.