Double-weighted multi-source domain adaptation model for tunnel boring machine main bearing fault diagnosis at low rotating speeds

Abstract The main bearing of a tunnel boring machine (TBM) is one of its key load-bearing components, and the health condition of the main bearing significantly impacts the safety and operational efficiency of the equipment. The main bearings are often under complex working conditions such as low speeds and heavy loads, and the operating conditions change frequently, making real-time diagnosis of the main bearings a challenging research problem. To address this issue, this study proposes a novel diagnostic method based on a Double-Weighted Multi-Source Domain Adaptation (DWMDA) model. The model mainly consists of a feature fusion extractor, several domain-specific adaptation feature extractors, and a domain discriminator. First, the feature fusion extractor is used to extract the fused features from the multi-directional sensors. The kurtosis index (KI) values of the original signals are used as weights to achieve weighted fusion of multidimensional features. Then, in the domain-specific adaptation feature extractors, domain-invariant features are extracted between each source and target domain, and the classification results of each domain are output through multiple classifiers. Finally, a domain discriminator is introduced to measure the similarity between target domain data and each source domain during the testing phase, and a second weighting is applied to integrate the multi-source domain classification results. Experiments were conducted on a TBM main bearing testbed. The experimental results show that the proposed method outperforms other single-source and multi-source domain adaptation methods, successfully achieved cross-domain fault diagnosis for the TBM main bearing.