Feature Extraction And Abnormal Condition Identification Based On Deep Learning And Multi-Source Information For Fused Magnesia Furnace

The mechanism of the production process of the fused magnesia furnace is complex,involving a variety of physical and chemical reactions,so it is difficult to establish an accurate model,and the identification of the abnormal working conditions of the fused magnesia furnace mainly depends on the artificial method.The industrial environment is adverse and labor intensity is high.Most of the operators rely on experience operation.It’s difficult to identify the change of working conditions in time and accurately,which is very unfavorable to production efficiency and production safety.In order to solve this problem this thesis uses image,sound and current information in the melting process of the fused magnesia furnace,with a method based on the deep learning,to extract and fuse the features of multi-source information in the melting process.Based on fused information,this thesis proposed a method to recognize abnormal conditions in the melting process of the furnace.This method can help to realize the automatic identification of abnormal working conditions of the furnace and reduce the labor intensity of workers.It can also provide the basis for self-healing control,improve production efficiency and ensure production safety.The main contents of this thesis are as follows:(1)A deep variational autoencoder based feature extraction method is proposed for image information in the melting process of the furnace.(2)A Fourier transform and finite impulse response digital filter based feature extraction method is proposed for audio information in the melting process of the furnace.(3)A segmental calculation of root mean square value based feature extraction method is proposed for current information in the melting process of the furnace.(4)Multi-source information is fused to identify abnormal conditions of the furnace using deep neural network.