Operation condition assessment for elevators under limited High-Quality label and Unbalanced data using feature alignment encoder GAN

As an important part of equipment prognostics and health management (PHM), condition evaluation plays a vital role in ensuring the safe operation of the elevator. However, in the actual condition assessment process, there are problems of insufficient label data and imbalance in data categories. So, it can bring significant resistance to the task of elevators’ condition assessment. To address this issue, this paper proposes a novel feature alignment encoder generative adversarial network with auxiliary classifier (AC-FAEGAN) for data augmentation (DA) and condition assessment. Firstly, the raw data is mapped by the encoder into hidden variables obeying a Gaussian distribution, and the generator is designed to implement data enhancement based on feature alignment. The improved discriminator and generator approximate the distribution of real data in an adversarial manner. The auxiliary classifier is trained with the generative model simultaneously. It functions in two ways, the first is to generate losses to constrain the generator, encouraging it to produce data that is closer to the original data. The second is to screen the generated intermediate data for fine-tuning to improve its generalization ability. Finally, the effectiveness and feasibility of the proposed method are verified by the elevator operation data. Meanwhile, different comparative and ablation experiments were carried out to validate the advancement and necessity of the proposed methodology. Compared to the state-of-the-art (SOTA) models, the proposed method shows optimal results.