The LSTM network model is established by selecting appropriate parameters to predict the oil temperature of gearbox. In order to verify the advantages of this method, it is compared with neural network (BP), SVM, random forest (RF) and ridge regression (RR). All methods adopt the same preprocessing, and train, test and verify the model on the same data set. The figure shows the prediction curves of several different methods for gearbox model test under normal state. MAPE and sdape of each method were calculated.
It can be seen from the figure that LSTM network model can better fit the oil temperature of gearbox in healthy state; it can be seen from table 3 that compared with the other four models, the error indexes of MAPE and sdape obtained by LSTM model are the smallest, which are 0.52 and 0.57 respectively, which also shows that LSTM method has strong fitting ability of oil temperature modeling and is more suitable for the prediction model of gearbox oil temperature.