Prognostics With Variational Autoencoder by Generative Adversarial Learning

Prognostics predicts the future performance progression and remaining useful life (RUL) of in-service systems based on historical and contemporary data. One of the challenges in prognostics is the development of methods that are capable of handling real-world uncertainties that typically lead to inaccurate predictions. To alleviate the impacts of uncertainties and to achieve accurate degradation trajectory and RUL predictions, a novel sequence-to-sequence predictive model is proposed based on a variational autoencoder that is trained with generative adversarial networks. A long short-term memory network and a Gaussian mixture model are utilized as building blocks so that the model is capable of providing probabilistic predictions. Correlative and monotonic metrics are applied to identify sensitive features in the degradation progress, in order to reduce the uncertainty induced from raw data. Then, the selected features are concatenated with one-hot health state indicators as training data for the model to learn end of life without the need for prior knowledge of failure thresholds. Performance of the proposed model is validated by health monitoring data collected from real-world aeroengines, wind turbines, and lithium-ion batteries. The results demonstrate that significant performance improvement can be achieved in long-term degradation progress and RUL prediction tasks.

Automated summary

This paper proposes a novel sequence-to-sequence predictive model based on a variational autoencoder for predicting future performance progression and remaining useful life of systems. The model, trained with generative adversarial networks, is capable of handling uncertainties and providing probabilistic predictions. Validation using real-world health monitoring data shows significant performance improvement in long-term degradation progress and RUL prediction tasks.