An Evidential Real-Time Multi-Mode Fault Diagnosis Approach Based on Broad Learning System

Su et al. proposed a dilated convolution deep belief network-dynamic multi-layer perceptron (DCDBN-DMLP) Fault diagnosis plays a crucial role in ensuring the efficiency, for recognizing bearing faults under varying operating conditions, stability, and reliability of industrial processes, making which uses dilated convolution deep belief network, it a focal point in both academic research and industrial multi-layer domain adaptation, and pseudo label technology applications [1, 2]. However, with the development of integrated, to address distribution discrepancies between source and target scaled, and complex systems, the challenges posed domains [8]. Li et al. proposed the modified auxiliary by fault diagnosis in industrial processes are becoming increasingly classifier GAN (MACGAN) as a novel supervised fault demanding. Recent advances in computer and diagnosis model for limited data in rotational machinery sensor technologies have simplified the data acquisition process [9]. Moreover, Hanachi et al. proposed a hybrid diagnostic and given rise to significant developments in data-driven framework combining a data-driven multi-mode fault parameter methods for fault diagnosis [3]. Practical industrial processes estimation scheme with a fault propagation model to often involve multiple operating modes, which give diagnose hidden incipient faults in gas turbine engine components rise to non-Gaussian, multi-modal, and center-drifting data [10]. However, deep learning methods depend on a features. These characteristics pose a challenge for research large amount of feature data from different operating conditions, into fault diagnosis in industrial production [4]. There are which is often difficult to obtain in practical engineering.