Research On Remaining Useful Life Prediction Method Of A Rolling Bearing Based On Improved GcForest

As one of the key components of rotating machinery,rolling bearings directly affect the operating performance of the entire equipment.Accurately predicting the degradation trend of remaining useful life(RUL)of rolling bearings can provide valuable status information and sufficient response time for equipment maintenance,which is of great significance for ensuring the normal operation of equipment and reducing maintenance costs for rolling bearings.To fully tap the deep characteristic information contained in the original vibration signal of the rolling bearing,the performance degradation index is independently constructed,and the RUL of the rolling bearing is predicted.A new method for predicting the RUL of rolling bearings is proposed based on the deep iterative feature(DIF)cascade Cat Boost(Cas Cat Boost).This method is an improved new multi-grained cascade forest(gc Forest)algorithm,and the feature extraction and the RUL prediction are respectively improved.In terms of feature extraction,the multi-grained scanning structure in the multi-grained cascade forest has a large amount of memory consumption,so it is replaced by a convolutional neural network(CNN).Iterative calculation is performed on the frequency domain features of the rolling bearing obtained by the fast Fourier transform to obtain the iterative features(IF).Using the characteristics of CNN with convolution and weight sharing,IF is further extracted its deep features and obtain DIF to build a performance degradation feature set.In terms of RUL prediction,in order to solve the problems that the prediction accuracy of a single model is not high and the computational efficiency of the cascade forest structure in multi-grained cascade forest is low,the cascade forest structure is replaced by Cas Cat Boost.Using the strong generalization ability of Cat Boost,the parallel acceleration of GPU can be realized,introduce the deterministic coefficient to cascade them,and construct Cas Cat Boost to improve the model's computing speed and representation ability.The DIF performance degradation feature set is used as the input of the model,and the difference between the decision coefficients of adjacent cascade layers is used as the cascade expansion criterion to realize the adaptive expansion of the model.The average life percentage p of the last cascade layer of the model is selected to represent the output.A linear function is used to fit the current life percentage of the bearing data to predict the future life trend at each point and obtain the RUL of the bearing.Experimental results show that the proposed method has high computational efficiency and prediction accuracy.