A Multi-Level Fault Diagnosis Continual Learning Method Using AIS-Based Hyper-Shell T Cell and Hypercube B Cell

In recent years, the continual learning method is one of the critical research frontiers in artificial intelligence fault diagnosis. This paper proposes an artificial immune system (AIS) based continual learning technique with hyper-shell T cells and hypercube B cells for multi-level fault diagnosis. According to different engineering requirements, multi-level fault diagnosis is proposed and a cascade AIS architecture of multi-level fault diagnosis is established. The generation, evolution, and detection of detectors determine the application effect of the algorithm. Firstly, self-space is constructed by cell cloning to detect the existence of fault. Secondly, generates hyper-shell T cells by searching boundary cells to determine the severity of damage. Lastly, activate hypercube B cells to identify the categories of failures. Investigations on the Paderborn University bearing data set demonstrate the effectiveness.