Misalignment fault identification of a multi-span rotor system enabled by triboelectric nanogenerators

Misalignment is widely manifested in a complex, multi-span rotor system, and its fault identification is highly preferred for health condition monitoring, which is, however, challenged by troublesome sensor installation and complex transmission path. Here, following a self-sensing solution, we design and fabricate a triboelectric nanogenerator (TENG) multi-span rotor system with the aid of freestanding-mode TENG bearings as self-aware measurement points. A single-span rotor system is first tested, showing that the triboelectric current signals under different health conditions lack universality for time and frequency domain analyses, which calls for the attendance of deep learning. By comparing the diagnostic accuracy of four deep learning models based on convolutional network structures, the ResNet 18 is determined to mine the feature patterns from the triboelectric current signals. Our TENG multi-span rotor system is demonstrated via single and three measurement point experiments, finding that the marriage between triboelectric current signals and ResNet 18 is able to reach a high diagnostic accuracy at 90.28% for misalignment fault, which can be further enhanced by an ensemble learning to reach 95.76% regardless of measurement points positioning. This is the first demonstration of TENGs for misalignment fault identification in a multi-span rotor system, paving the way for the health condition monitoring of a complex rotor system out of the limitation from troublesome sensor installation and complex transmission path.