رضا آذین

The imperative to mitigate global climate change has accelerated the deployment of Carbon Capture and Storage (CCS) technologies within the oil and gas sector. However, the operational complexity and high maintenance costs of CCS infrastructure remain significant barriers to widespread adoption. This study proposes a novel Closed-Loop Deep Reliability Framework that integrates data-driven prognostics with dynamic strategic planning. Bridging the gap between classical reliability engineering and deep learning, we utilise a Bi-Directional LSTM (Bi-LSTM) architecture to predict the degradation of critical rotating assets under non-linear sensor dynamics. The framework introduces an adaptive “Safety Margin” inspection policy and a risk-aware Fuzzy-AHP resource allocation mechanism. Validated on the gold-standard NASA C-MAPSS benchmark (FD001), the proposed prognostic model achieves a Concordance Index (CIndex) of 0.863 and an RMSE of 14.95, significantly outperforming traditional stochastic models. More significantly, operational simulations reveal that this predictive capability translates into a massive 83.9% reduction in unexpected failures and a 15.0% decrease in total maintenance costs. Furthermore, the dynamic allocation mechanism demonstrates strategic agility by automatically reducing exposure to high-sensitivity EOR technologies by 8.06% during critical risk periods, thereby preventing cascading failures while optimising Operational Expenditure (OPEX). By dynamically shifting resources to robust alternatives (e.g., Geological Storage) during high-risk periods, this framework offers policymakers a resilient roadmap to achieve decarbonisation targets.