Rolling bearing RUL prediction method based on piecewise linear fitting HI and LSTM

Abstract

The invention discloses a rolling bearing RUL prediction method based on piecewise linear fitting HI and LSTM. The method comprises the steps: obtaining an HI curve of a plurality of rolling bearings under a preset working condition, obtaining a residual life sequence; screening out the optimal segmentation number of each rolling bearing by employing a BUP time sequence segmentation algorithm based on a clustering evaluation index and a fitting evaluation index, segmenting the HI curve, extracting a degeneration period HI curve and a degeneration period residual life sequence, and performing normalization to obtain a normalized degeneration period HI curve and a normalized degeneration period residual life sequence; and taking a long-short term memory network as an RUL prediction model, taking the normalized degeneration period HI curve as an input, taking the normalized degradation period residual life sequence as a label, training the RUL prediction model; obtaining a normalized degradation period HI curve by adopting the same method for a certain rolling bearing under a preset working condition, and inputting the normalized degradation period HI curve into the RUL prediction model to obtain a predicted residual life sequence. According to the invention, the prediction accuracy and robustness of the RUL can be effectively improved.

Description

technical field [0001] The invention belongs to the technical field of rolling bearing life prediction, and more specifically relates to a rolling bearing RUL prediction method based on piecewise linear fitting HI and LSTM. Background technique [0002] Among all the key components of rotating machinery at present, rolling bearings have always been one of the important objects of research. On the one hand, because rolling bearings are widely used, they play an irreplaceable role in the application of rotating machinery; on the other hand, rolling bearings are more prone to failure than other components. Due to the complexity and uncertainty of the service environment and working conditions of rolling bearings, the randomness of fatigue damage development and the diversity of failure modes, the maximum service life of rolling bearings has a large dispersion, and regular maintenance of rolling bearings often leads to " Undermaintenance” and “overmaintenance” issues. The RUL ...

AI Technical Summary

Problems solved by technology

At the same time, in today's era of big data, in the practical application of engineering, the existing neural network-based life prediction technology suffers from too many model input samples, ambiguous sample positioning in the degradation period, and under multiple working conditions and multiple failure modes. Problems such as the low accuracy of life prediction in the field need to be studied and solved

Images