Fault Diagnosis Of Scraper Conveyor Start-Stop Condition Based On Distributed Deep Neural Network

As a bridge between shearer and hydraulic support in fully mechanized face,scraper conveyor has complex and bad working conditions.Once scraper conveyor is shut down due to malfunction,it is difficult and costly to equipment maintenance,and it also poses a serious threat to coal mine output and miners’ life safety.At present,scraper conveyor fault diagnosis often adopts dynamic modeling,signal analysis,SVM and artificial neural network methods,which need a lot of deep professional knowledge and artificial experience,and there are certain limitations.In addition,with the development of "intelligent mine" construction,the increase of sensor measurement points,sampling frequency and data management model upgrade of scraper conveyor lead to massive accumulation of monitoring data,which leads to the traditional deep learning model based on centralized cloud computing not only takes up a lot of communication resources,but also consumes too much time in the data transmission link of scraper conveyor fault diagnosis,and it is difficult to meet the real-time requirements of diagnosis.To solve the above problems,this paper proposes a fault diagnosis method of scraper conveyor under start-stop condition based on distributed deep neural network(DDNN).The main research contents are as follows:(1)Failure mechanism analysis of scraper conveyor under start-stop condition.Based on parameters and load characteristics in scraper conveyor start-stop condition,the mapping relationship between five kinds of start and stop faults and working parameters of scraper conveyor,such as heavy load start,start power imbalance,overload stop,chain stop and repeated start,is analyzed,which provides a more reliable theoretical basis for the selection of scraper conveyor start-stop working condition fault diagnosis indicators.(2)Data collection,analysis and processing.Based on the failure mechanism of the scraper conveyor start-stop condition,the nose-tail current and speed values reflected by current sensor,voltage sensor and frequency sensor in frequency converter are selected as fault diagnosis data,which are preprocessed by data filling,data fusion,image data transformation and other technologies.(3)Design of fault diagnosis method for scraper conveyor under Start-stop Condition.Firstly,deep neural network was used to automatically extract characteristics from monitoring data of scraper conveyor by data fusion and data-to-image conversion.Secondly,a branch introducing convolutional bag-of-features(CBo F)was added to deep neural network model,which was divided into the shallow part of edge and the deep part of cloud by branch point.Finally,the data transmission path is shortened and the scraper conveyor’s precise and agile fault diagnosis is realized,by using the cloud-edge collaborative reasoning method.(4)Development of fault diagnosis system for scraper conveyor start-stop Condition.Develop a scraper conveyor start-stop working condition fault diagnosis system with good man-machine interface,which can provide monitoring of scraper conveyor operation data,real-time fault analysis,diagnosis result storage and visualization,data query,technical parameter display and other functions.