Research On Combine Harvester Fault Diagnosis System Based On Deep Learning

In recent years,with the rapid of the number of combine harvesters in China,the problems of the high failure rate of combine harvesters and low harvesting efficiency gradually appear.Therefore,how to intelligently diagnose and monitor the fault of combine harvester has become a research hotspot in the field of fault monitoring of combine harvester.Based on the current technical requirements of fault diagnosis and monitoring of combined harvester,this paper proposes a new intelligent fault diagnosis method for combine the harvester mechanical and hydraulic system based on the combination of improved stacked autoencoder and multi-sensor information fusion technology.The aim is to realize the effective monitoring,intelligent diagnosis,and accurate evaluation of the running state of the combine harvester.In this paper,a realtime fault diagnosis and monitoring system for combine harvester is designed and implemented based on the monitoring objects of mechanical fault and hydraulic fault.Aiming at the high requirements of the fault diagnosis algorithm of the harvester in real-time,sensitivity,accuracy,and anti-noise.An improved Stack Denoising Auto Encoder(SDAE)was proposed based on the traditional Stack Denoising autoencoder,which introduced the Gaussian noise of multiple distribution centers.The performance of the combined harvester was verified by the field fault data samples.Meanwhile,based on the Refined Composite Multiscale Dispersion Entropy(RCMDE),a Refined Composite Multivariate Multiscale Fluctuation Dispersion Entropy(RCMMFDE)is proposed.By synthesizing the information of multiple coarse grain sequences of each channel of multiple vibration signals and using the fine compound method,the method makes it less dependent on the length of time series,and the feature extraction is more stable and reliable,which lays the algorithm foundation for the real-time fault diagnosis and monitoring system of combine harvester.The real-time fault monitoring system of combine studied in this paper takes Stack Denoising Autoencoder and Refined Composite Multivariate Multiscale Fluctuation Dispersion Entropy as the core algorithm,select the combine harvester operation driving speed,feeding auger speed,threshing cylinder speed,hydraulic oil pressure,hydraulic oil temperature,hydraulic oil flow,liquid level and body vibration as the main monitoring objects,and selects the feed quantity,cutting stubble height,yield per hectare and crop height were used as auxiliary parameters for fault monitoring.The fault monitoring system mainly includes vehicle terminal and remote server terminal.The vehicle terminal mainly completes the collection,processing and sending of monitoring data,and completes the harvester fault early warning and alarm locally.The fault monitoring system mainly includes the on-board end and the remote server end.The on-board end mainly completes the collection,processing,and sending of monitoring data,and also completes the local fault warning and alarm of the harvester.The on-board end is divided into operation information collection module,on-board diagnosis module,data display module,and data transmission module.Windows 10 operating system is used as the software running environment of the car terminal.The remote server-side includes a model update module,a data transmission module,and server-side database.The remote server-side completes the training and transmission of the model required by the on-board side.According to the characteristics of the fault data samples of combine,two parallel SDAE networks and one information entropy algorithm are constructed at the data level to extract the monitoring data of different types of sensors.The decision level uses SDAE neural network to integrate the features and diagnose the equipment fault types.Through the field test of the whole system,the results show that the on-board terminal can accurately diagnose the fault state of the combine harvester,alarm,and give corresponding maintenance suggestions.Through the simulation test of the combined harvester fault monitoring system,the test shows the feasibility of the system.