A game theory enhanced domain adaptation network for mechanical fault diagnosis

Abstract

Transfer learning (TL) technology have been successfully applied to address the domain adaptation (DA) problem in machinery fault diagnosis. However, partial DA problem is more suitable for industrial applications, where the target data only covers a subset of the source classes, which makes it difficult to know where to transfer the target data. To overcome this problem, a novel game theory enhanced domain adaptation network (GT-DAN) is proposed in this paper. Based on different metrics, including Maximum mean discrepancy (MMD), Jensen-Shannon (JS) divergence and Wasserstein distance, three attention matrices are constructed to describe the distribution discrepancies between the source domain and the target domain, and the optimal coordination between these attention matrices is achieved by the combination weighting based on game theory to generate the optimal probability weights, which can guide to filter out the irrelevant source examples in domain adaptation. Two experiments show that the proposed GT-DAN is superior to the existing methods in the partial DA diagnosis performance.

Remaining Useful Life Prediction of Rolling Bearings Based on ECA-CAE and Autoformer

“…However, it requires monitoring of the equipment operation and collecting valid failure data or performance degradation data. Data-driven approaches generally include those based on statistically driven models and reliability functions, as well as those based on machine learning and deep learning models are the main directions of current life prediction research [11][12][13][14]. For rolling bearings, the main prediction steps include: (1) extracting failure characteristics and degradation curves using signal processing or machine learning; (2) constructing health indicators using deep learning or degradation function models to characterize life thresholds; (3) using the trained deep learning degradation models for life prediction or model fitting the obtained degradation curves to finally obtain the RUL [15][16][17].…”

Fault Diagnosis of Planetary Gearbox Based on Dynamic Simulation and Partial Transfer Learning

“…Li et al [ 10 ] proposed a class-weighted adversarial neural network that encourages positive transfers of shared classes and ignores source class outliers through class-weighting strategies. Sun et al [ 11 ] suggested a game theory-enhanced domain adaptation network to solve partial domain adaptation problems. The network constructs three attention matrices using maximum mean discrepancy, Jensen-Shannon divergence, and Wasserstein distance and generates the best probability weight through the combination of game theory weights, thereby filtering out irrelevant source domain samples and improving mechanical fault diagnosis performance.…”