The elevator is a vital apparatus in everyday life, and precise fault identification is critical for guaranteeing its safe operation. This paper offers an elevator bearing fault diagnosis approach utilizing MHO-BPNet, as current methods frequently exhibit low accuracy rates. The main aspects of this method are: first, redundant and noisy features are removed using Mean Influence Value (MIV) feature dimensionality reduction method. Second, the Hippopotamus Optimization (HO) algorithm is introduced to optimize the initial weights and thresholds of the Backpropagation Neural Networks (BPNNs) in order to avoid local optimal solutions and gradient vanishing problems. Finally, the MHO-BPNet model is experimentally verified with two datasets to achieve more than 96.5% accuracy in both cases and accurately identify the fault states of the elevator.
