Applied Research On Life Prediction Of Food Processing Control System

The food processing system is a distributed network structure which composed of multiple coupled sub-systems.This network production structure has become the mainstream production mode in the Industry 4.0 era.The food processing systems have characteristics that are not available in traditional production systems,such as real-time production,structural distribution,and multi-scale nodes.Therefore,the operation and maintenance of system characteristics has become an emerging research topic.With the rapid development of industrial technology,the coupled production system has banned the traditional production unit,and the resulting data has increased sharply.In the field of fault diagnosis,deep learning highlights its computational advantages.Although deep learning algorithms have achieved great accuracy results,the research on the characteristics of the system is relatively insufficient.In response to these problems,this article mainly studies fault prediction based on deep learning,which mainly includes the following aspects:(1)This paper proposes a food processing system fault prediction method based on regularized ISU-LSTM(Insert Sparse Unit-Long and Short Term Memory,ISU-LSTM).In response to the real-time requirements in the current fault prediction field,the insert sparse unit is designed to replace the forget gate in the traditional network,and the insert sparse unit long and short-term memory network(ISU-LSTM)is established.The network effectively sparse the network structure and meets the fault real-time prediction requirement.In order to deal with the constantly updated sensor data,the algorithm in this paper continuously optimizes the mean square error,realizes the continuous update of network parameters and dimensions and effectively avoids the neural network model training from falling into local optimization.(2)This paper proposes a food processing system fault prediction method based on Residual-Simple Recurrent Unit(Res-SRU).In view of the problem that multi-node data is difficult to generalize and calculate in food processing control system,the equivalent mapping function in the residual network is pioneered into the multi-layer Simple Recurrent Unit.In order to deal with the problem that adjacent nodes have the characteristics of coupling work,the multi-node collaborative algorithm is used to diagnose corresponding faults of adjacent nodes.The information of all fault nodes in the area is obtained according to the Euclidean distance range and probability criterion.The hierarchical residual SRU network is used as the main structure of the prediction model,so that the multi-node fault data can also be effectively trained.(3)This paper designs a health management system called PHM(Prognostic and Health Management,PHM).Based on the main research algorithm of this paper,PHM takes food processing system equipment as the application research object and proposes Parameter Update Optimization(PUO)algorithm,including data processing,health management center and interactive applications.Three modules realize the function of monitoring the status of the actual food processing system and effectively predict the accurate RUL(Remaining Useful Life,RUL)value.The fault alarm is issued before the fault occurred and the fault decision evaluation was given.Finally,the core function module was verified and tested.Aiming at the real-time,multi-node,and distributed production characteristics of the cyber-physical fusion system and combined with the advantages of deep learning to process large data sets,this paper proposes a LSTM fault prediction method based on the sparse ISU module.This fault prediction method is improved by adding a regularized sparse module.In this paper,the calculation speed of neural network is improved by adding a regularized sparse module.In addition,an SRU network with residual connections is proposed to predict the remaining useful life of the food processing control system.Finally,the two networks are integrated into the PHM system of the health management center.The result shows that the PHM system can obtain accurate prediction results.