Abstract:In the nuclear industry, high system reliability requirements are essential since in-service failure can result in undesirable consequences in terms of costs or safety. However, the current approach to maintaining systems and components is costly and known to involve overly conservative periodic maintenance activities. It is, therefore, appropriate to develop monitoring, detection, and predictive tools to enable operators to create optimal maintenance strategies. These strategies can vary from the substitution of an item to its repair, intending to avoid unexpected consequences. The repair can restore the item's functionality to an as-good-as-new condition (perfect repair) or sometimes can keep some accumulated degradation and change the item's degradation rate (imperfect or partial repair). Current techniques and models that can perform prognostics with extraordinary accuracy are often designed on the assumption that following maintenance, the item is restored to an as-good-as-new condition. When these models are used to predict items that follow imperfect repairs, the predictions are likely to be inaccurate. Therefore, the present work focuses on the condition-based prognostics of items, considering and handling the criticalities that arise after the items undergo different kinds of repairs. The proposed solution involves a data-driven framework that employs Left-Right Gaussian Hidden Markov Models (LR-GHMMs). These models can intrinsically manage accumulated degradation. The idea is to train different LR-GHMMs, each specific to a degradation path, and then combine them to cover possible intermediate paths. The effectiveness of the approach is tested in two case studies. In the first one, we consider simple artificial sequences that are useful to explain the method's capabilities. In the second case study, we consider semi-simulated nuclear data describing the degradation transients of a condenser that undergoes fouling. The framework is trained with data collected from items that start without accumulated degradation. The test data represent either new items or items that undergo imperfect repairs. The results demonstrate an attractive elasticity of the framework in adapting to nonstandard degradation behaviors. In addition, the applications provide interpretable and highly accurate outputs.

Advancing Predictive Maintenance with PHM-ML Modeling: Optimal Covariate Weight Estimation and State Band Definition under Multi-Condition Scenarios

Adopting New Machine Learning Approaches on Cox's Partial Likelihood Parameter Estimation for Predictive Maintenance Decisions

The Study of Large Equipments Condition Maintenance Policy Based on Kalman Filtering

Condition-based recursive estimation of proportional hazards model, for equipment remaining useful life prediction

PHM Techniques for Condition-Based Maintenance Based on Hybrid System Model Representation

Combining Similarity Measures and Left-Right Hidden Markov Models for Prognostics of Items Subjected to Perfect and Imperfect Maintenance

Strategies for overcoming data scarcity, imbalance, and feature selection challenges in machine learning models for predictive maintenance

Prognostics and Health Management for Maintenance-Dependent Processes

A metric for assessing and optimizing data-driven prognostic algorithms for predictive maintenance

Condition-based Maintenance Optimization for Multi-component Systems Considering Prognostic Information and Degraded Working Efficiency

SHM-informed life-cycle intelligent maintenance of fatigue-sensitive detail using Bayesian forecasting and Markov decision process

Health status assessment and remaining useful life prediction of aero-engine based on BiGRU and MMoE

Optimized Task Assignment and Predictive Maintenance for Industrial Machines using Markov Decision Process

Advanced ML for predictive maintenance: a case study on remaining useful life prediction and reliability enhancement

A New Methodological Framework for Optimizing Predictive Maintenance Using Machine Learning Combined with Product Quality Parameters

Hidden Markov Models and their Application for Predicting Failure Events

A Dynamic Condition-Based Maintenance Model Using Inverse Gaussian Process

Prescriptive maintenance with causal machine learning

Knowledge Informed Machine Learning using a Weibull-based Loss Function

Health Status-Based Predictive Maintenance Decision-Making via LSTM and Markov Decision Process

A hybrid prognostic & health management framework across multi-level engineering systems with scalable convolution neural networks and adjustable functional regression models