SafeCBM: NPP Safety-constrained CBM Optimization Diagnosis

Systems in nuclear power plants (NPPs) utilizes advanced Reliability-centered Maintenance (RCM) methodologies like Time-/Condition-based maintenance (TBM , CBM) for determining the maintenance requirements of objects in its current operating context. Such techniques have been enhancing the safety of the unit, improving equipment reliability and reducing maintenance costs. However, the existing NPP RCM method limits on its abilities in standardizing Equipment/Material data classification, learning Remaining Useful Life (RUL) automation, and modernizing RCM strategies with NPP safety and reliability constraints. To fill this gap, we propose SafeCBM: NPP Safety-constrained CBM Optimization & Diagnosis, to leverage Artificial Intelligence-based process automation, machine intelligence, and computer aided decision making for NPP modernization that enhances both core operations and maintenance work activities. SafeCBM enables condition-based monitoring with dynamic-scheduled inspection for NPP components. To achieve this goal, SafeCBM defines new equipment and materials classifications based on International Standard ISO 15926 standardized information models to enhance the results of risk-based CBM optimization (described in the later proposed module). Next, an attention-based deep learning method for RUL prediction is developed. LSTM (Long Short Term Memory) neuron network is employed to learn sequential features from raw sensory data. The attention mechanism learns the importance of features and time steps by assigning them different weights. Meanwhile, a feature fusion framework combines handcrafted features (from the classic feature engineering approach) with the weighted features (from deep-learning) to boost the performance of RUL prediction. Finally, a CBM optimization model with NPP safety constraints assuming imperfect inspection is constructed to provide optimal maintenance policies. Such a policy is based on optimal maintenance cost CBM probability before a component’s lifetime failure occurs. The NPP inspection intervals can be therefore adapted with changing system conditions using dynamic NPP safety constraints. For practicability of maintenance decision, the proposed optimization model, which minimizes the expected long-term cost rate with safety constraints, also considers the influence of imperfect inspection. Business Benefits: SafeCBM achieves several impressive benefits:Conditional Monitoring/CBM Enhancement: SafeCBM can enhance the plant’s CMMS (Computerized Maintenance Management Systems) performance by minimizing overall maintenance cost, improving PLiM (Plant Life Management), and increasing process productivity, all while integrating NPP safety constraints and dynamic inspection scheduling criteria. It considers plant safety constraints and measurement errors when calculating maintenance cost optimizations and strategies. It supports CBM decision-making by suggesting predictive correction of components and accessories likely to fail in the future. Advanced Machinery Maintenance: In the fast-pace production environment, SafeCBM provides a real-time attention-based neural architecture with LSTM to enable learning from various sequential raw sensory data inputs for condition-based monitoring & analysis at the plant’s system and local levels. Monitoring the health of machines effectively enables the users to schedule repair and maintenance of the system ensuring less downtime, thereby enhancing its lifetime and production volume. Facility Management & Maintenance: SafeCBM offers these facilities management/maintenance activities in terms of reliability, availability, and serviceability. SafeCBM ensures continuous facility operability in multi-tier aspects. Its optimal CBM model is able to integrate patient and staff equipment risk and link that risk to the equipment or area being serviced. It streamlines their facility and maintenance operations. With maintenance under control, organizations are freed up to focus on first-rate customer service and improving their users’ experience.