Do Bayesian Neural Networks Improve Weapon System Predictive Maintenance?
This work addresses predictive maintenance for weapon systems, but it appears incremental as it applies an existing Bayesian inference method to a specific domain.
The paper tackled the problem of predicting time to failure for highly reliable weapon systems using Bayesian neural networks with interval-censored data and time-varying covariates, resulting in the development of LaplaceNN, which was benchmarked on synthetic and real datasets with metrics like ROC AUC and PR AUC.
We implement a Bayesian inference process for Neural Networks to model the time to failure of highly reliable weapon systems with interval-censored data and time-varying covariates. We analyze and benchmark our approach, LaplaceNN, on synthetic and real datasets with standard classification metrics such as Receiver Operating Characteristic (ROC) Area Under Curve (AUC) Precision-Recall (PR) AUC, and reliability curve visualizations.