Research On Classification Method Of Excitation Signal Of Engineering Vehicle EECU Based On Convolutional Neural Network

With the rapid development of the infrastructure construction industry,the demand for construction vehicles has increased year by year.Because the Engine Electronic Control Unit(EECU)has been in harsh working environment such as severe vibration,high temperature and electromagnetic interference for a long time,it causes frequent failures.The excitation signal is the key to ensure the normal startup and operation of the EECU,and it has various and complex characteristics.As a result,its identification is difficult and the efficiency of EECU fault diagnosis is reduced.Because the characteristics of the excitation signal generated by the crankshaft and the camshaft are the most complex and diverse,and play a vital role in the electronic control system.Therefore,in view of the above problems,this paper mainly takes the crankshaft and camshaft sensors as the research objects,and conducts identification and classification research on the excitation signals generated by them:1.Aiming at the problems of complex features,difficult extraction and low recognition accuracy of EECU excitation signals,this paper designs a multi-label EECU excitation signal classification method based on Convolutional Neural Network(EESC-MLCNN).Firstly,the working principle of the excitation signal is analyzed,and the hierarchical features are extracted from the original data.Secondly,by stacking multiple nonlinear transformation layers,Sigmoid activation output,and Binary Cross-Entropy loss function,a new CNN network is constructed for multi-label classification of signal features.Finally,the combination of real vehicle data and simulation data is used to make up for the lack of training data sets,and EESC-MLCNN is used for multi-label feature recognition and classification.In the experimental study of EECU excitation signal detection,the proposed EESC-MLCNN method is compared with MLKNN,MLDT and other algorithms.The experimental results show that the accuracy of the comparison algorithm based on the crankshaft excitation signal data set is in the range of 73.8%～87%,and the average recognition accuracy of the EESC-MLCNN method proposed in this paper is 88.3%;the accuracy of the algorithm based on the camshaft excitation signal data set comparison is in the range of 68.3%-87%,the average recognition accuracy of the EESC-MLCNN method proposed in this paper is 91.4%.Compared with the comparison algorithms,EESC-MLCNN has improved.2.To solve the problem that the synchronization characteristics of crankshaft and camshaft signals in excitation signals are not obvious,which leads to the difficulty of detection,this paper proposes a synchronization signal classification method based on 1DCNN network model(SSC-1DCNN).Firstly,this paper proposes a synchronization signal identification algorithm based on theoretical-practical error detection.By analyzing the characteristic parameters of crankshaft and camshaft signals,the theoretical synchronous offset value of signals is obtained.According to the type of sensor,the actual synchronous offset value of the signal is obtained by using the peak second-order differential identification algorithm.If the difference between the two offsets meets a certain error range,it is a synchronous signal.Then,automatically label the excitation signal data with synchronization offset,and put it into SSC-1DCNN network model for training and adaptively extract synchronization features.Finally,the optimal model trained by the network is used to verify the excitation signal test set.The SSC-1DCNN method proposed in this paper is compared and evaluated with LTSM,GRU and random forest RF algorithms.The experimental results show that the average recognition accuracy of LTSM is 96%,that of GRU is 97%,that of RF algorithm is 99%,and that of EESC-MLCNN method proposed in this paper is 100%.Compared with the above models,the network has stronger recognition and classification ability.3.Combining the above two algorithms,this paper develops a visual system of EECU excitation signal classification based on QT framework.The system mainly includes the functions of multi-label excitation signal identification and classification,synchronous discrimination of crankshaft and camshaft signals,etc.The experimental results show that the system developed in this paper can accurately extract the characteristics of EECU excitation signals and judge whether crankshaft and camshaft signals are synchronous,which is convenient for users to identify the characteristics of excitation signals and synchronous fault information.