Remaining useful life (RUL) prediction is a critical technology for preventing unexpected failures and reducing maintenance costs in modern industrial systems. Traditional model-based approaches, however, are constrained by the need for explicit mathe...
Remaining useful life (RUL) prediction is a critical technology for preventing unexpected failures and reducing maintenance costs in modern industrial systems. Traditional model-based approaches, however, are constrained by the need for explicit mathematical modeling of degradation mechanisms, whereas data-driven methods often require large-scale datasets and suffer from limited interpretability. To address these limitations, this study proposes a Weibull Variational Autoencoder (WVAE). The WVAE is designed to learn probabilistic characteristics grounded in the Weibull distribution from historical failure data and to predict failure times accordingly. A composite loss function that integrates mean squared error (MSE) with negative log-likelihood is employed to jointly ensure predictive accuracy and distributional fidelity, while Monte Carlo Dropout–based inference is used to quantify predictive uncertainty and generate confidence intervals. Simulated datasets incorporating three representative failure types—infant mortality, random, and wear-out—together with multiple noise levels were constructed to reflect diverse system characteristics and realistic industrial conditions. Evaluation results show that, compared with several benchmark models, the WVAE produces predictions that remain statistically consistent with historical failure distributions and maintain stable performance even under high noise conditions. Moreover, by directly estimating the parameters of the Weibull distribution, the WVAE provides statistically interpretable predictions that can be readily trusted by domain experts.