<p>With the continuous expansion of natural gas pipeline networks and the increasing complexity of equipment in transmission stations, reliability analysis has become a key challenge for ensuring pipeline safety and enhancing energy supply efficiency. Bayesian neural networks (BNNs), known for their ability to handle uncertainties, have emerged as a powerful tool for reliability analysis, particularly in life prediction applications. This study first conducts reliability analysis on natural gas transmission station equipment, fitting the life distribution and calculating distribution parameters to serve as the prior distribution for the BNN. To enhance computational efficiency, the variational inference method is introduced, optimizing both efficiency and inference accuracy. The optimized BNN is then applied to predict equipment lifespan and compared with models such as Long Short-Term Memory (LSTM), Autoregressive Integrated Moving Average (ARIMA), and the bathtub curve. Experimental results demonstrate the superior performance of the BNN in equipment life prediction. By modeling the trend of equipment reliability over time, this research highlights the method’s effectiveness in real-world industrial applications, validating its potential for equipment maintenance decision-making and health management in industrial applications.</p>