Research On Grain Loss Monitoring Method Of Rice Combined Harvester Based On WOA-BP

With the acceleration of agricultural modernization in China,combine harvester has become an important agricultural equipment in modern agriculture.The performance indexes of combine harvester include grain loss rate,impurity rate,crushing rate and operation efficiency.The monitoring of these parameters is very important for adjusting the operation parameters of combine harvester and improving the quality of grain harvest.Therefore,real-time monitoring of grain loss rate in the harvest process of combine harvester is one of the important topics.At present,some foreign combine harvesters have installed grain loss rate monitoring devices,but the domestic grain loss rate monitoring device is still in the research stage,and the monitoring device has the problem of insufficient real-time performance and accuracy to be improved.For the monitoring method of grain loss rate,domestic scholars have proposed algorithm models such as decision tree,support vector machine and BP neural network.These models are susceptible to over-fitting and local extremum,which reduces the accuracy and response speed of monitoring results.In view of the above problems,this paper proposes a monitoring method of rice grain loss rate based on WOA-BP,aiming to improve the accuracy and response speed of real-time monitoring of grain loss rate during the operation of combine harvester,and provide a basis for drivers to adjust the operation status of combine harvester and intelligent control of combine harvester in time.The main research work of this paper is as follows :(1)Research on the monitoring principle,overall scheme and hardware construction of rice grain loss rate.Firstly,the factors of grain cleaning loss and the monitoring principle of grain loss rate of combine harvester were analyzed.Secondly,a real-time monitoring device for grain loss rate is built.The device is mainly composed of piezoelectric ceramic sensor module,charge amplification circuit,bandpass filter circuit,AD conversion module,screen display module,embedded processor module and CAN bus communication module.Among them,the piezoelectric ceramic sensor module is located at the tail of the combine harvester cleaning screen to receive the material signal,which is input into the embedded processor after charge amplification,band-pass filtering and AD conversion.The processor runs the WOABP algorithm model and other processing programs to calculate the grain loss rate and displays it through the UI interface.At the same time,the monitoring data is transmitted to the host computer NVIDIA Jetson TX2 through CAN bus communication and stored in the SD card for offline data analysis.(2)The establishment of rice grain recognition algorithm model based on WOABP.Firstly,the collected material impact signal data is preprocessed,including time domain and frequency domain feature extraction and normalization,and trained by WOA-BP algorithm model.The results show that the accuracy of grain signal recognition in the test data set is 95.93%.Secondly,compared with the three kinds of grain loss rate monitoring algorithm models proposed by scholars,including decision tree,support vector machine and BP neural network model,the performance evaluation was carried out on the same data set by using five performance indexes of accuracy,precision,recall,F1-score and training time.The results showed that the five indexes of WOA-BP were 96.33%,96.10%,96.73%,96.42% and 3.89 s,respectively,indicating that the comprehensive classification effect of WOA-BP model was better.Finally,by drawing the ROC curve of WOA-BP and calculating the AUC value,the results show that it is 96.11%,indicating that the WOA-BP algorithm model has good classification performance.(3)Software design of rice grain loss rate monitoring device.The program of the grain loss rate monitoring device is designed by modularization.The program software mainly includes the main program design,interface display program,AD conversion program,WOA-BP model program,calculation of grain loss rate program,data storage program and CAN bus communication program.Through the research and design of each related program,the function of the grain loss rate monitoring device is realized.(4)Experimental verification of grain loss rate monitoring device.Firstly,the process of grain cleaning loss was simulated on the indoor conveyor belt test bench,and the height and angle of the sensor module were analyzed and tested.The results showed that the test effect was better under the condition of 56 cm from the desktop of the conveyor and 45°.Under this condition,the indoor test was carried out with different conveying speeds and different mass ratios of grains and impurities.The results showed that the relative error of the monitoring results was less than 8.5%when the conveying speed was 1.3-2.1m/s,and the proportion of impurities increased at different conveying speeds.That is,when the mass ratio of grains to impurities is1/2,1/2.5 and 1/3,the relative error shows an increasing trend.Compared with the indoor test results of the grain loss rate monitoring device based on decision tree and SVM,the results show that the relative errors of each device are less than 8.5%,12%and 15% respectively,and the average monitoring periods are 3s,2s and 4s respectively.The device studied in this paper has lower relative error and shorter monitoring period,which is in line with the expected results.Finally,the field experiment was carried out.The results showed that when the average operating speed of the combine harvester was 0.8 ～ 1.2m/s,the relative error of the monitored loss rate was less than 10.5%.Compared with the field test results of the SVM-based grain loss rate monitoring device studied by predecessors,the results show that the relative errors of each device are less than 10.5% and 16.5% respectively,and the average monitoring periods are 3s and 4s respectively.The device studied in this paper improves the monitoring accuracy and reduces the monitoring period,which has high practicability.The results of indoor and field experiments show that the designed grain loss rate monitoring device has high precision and good real-time performance,which provides technical support for online monitoring and intelligent control of grain loss rate of combine harvester.