Reliability and availability are top priorities for any fault-tolerant machine system. To develop a robust and interruption-free system, this study incorporates features of a multiple retrial orbit space with hybrid vacation and working breakdown under a threshold admission control policy for failed units. The continuous arrival of failed units can present management challenges. To address this, multiple retrial orbit spaces have been created where failed units can wait until a server becomes available, servicing other failed units in different server states. A fixed number of failed units can join the system at a time; after that, entry is restricted until the threshold level F is reached under the F-policy. The hybrid vacation combines operational vacation (OV) and complete vacation (CV), which are also considered in this paper. The results and methods are carefully analyzed, and a sensitivity analysis is performed to assess competitiveness. System matrices are evaluated with parameter variations. To determine the cost-effectiveness of the model, a total cost function is designed, and a direct search method is used for cost evaluation. A further cost comparison study is conducted using Sine Cosine Algorithm (SCA), Grey Wolf Optimization (GWO), and Particle Swarm Optimization (PSO). The minimum optimal cost is determined using meta-heuristic optimization techniques such as SCA, GWO, and PSO and observed that PSO performed better than GWO and SCA.
Risk management in high-hazard process industries such as petrochemical facilities relies on accurate assessment of safety barrier performance under varying operational conditions. Conventional Event Tree Analysis (ETA), a widely used probabilistic risk assessment method, typically assumes fixed Probabilities of Failure on Demand (PFD), limiting its ability to capture context-dependent influences from equipment condition, human performance, and environmental factors. This paper proposes a hybrid Event Tree Analysis–Machine Learning (ETA-ML) framework in which barrier-specific PFDs are dynamically inferred from contextual operational data, rather than treated as constant values. Random Forest (RF) models are trained on realistic synthetic datasets capturing operational variability, with class imbalance associated with rare failure events mitigated using the ADASYN algorithm, enabling probabilistically calibrated, context-aware updating of barrier reliability. The framework is demonstrated on an industrial overpressure scenario involving multiple sequential safety barriers, where ML-predicted PFDs are integrated into ETA branches to update scenario frequencies. A rigorous validation protocol including stratified cross-validation, ROC/PR curves, F1, log-loss, Brier score, calibration curves, and bootstrap AUC (B = 1000), showed high discrimination (AUC-ROC 0.95–0.995) and excellent calibration (Brier = 0.05; low Expected Calibration Error). Compared with static ETA, probabilistic error was reduced by over 80
Ageing bridge infrastructure is exposed to danger regimes beyond those foreseen in its design and maintenance planning due to climate change: extreme weather events are becoming more frequent, intense, and geographically organized. Even though flood-induced scour, temperature-driven material degradation, and compound hazards evolve on time scales equal to deterioration processes, most bridge risk models treat climate, deterioration, fragility, and life-cycle assessment as weakly related or static. Fixed hazard assumptions, isolated deterioration models, single-hazard fragility curves, and deterministic life-cycle evaluations limit the ability to foresee rapid degradation and aid climate-adaptive decision-making. This study models, validates, and implements climate-change-induced bridge failure risk using data. The structure links six analytical blocks successively. CHESS-EV uses spatial extreme value theory and graph-based machine learning to downscale climate forecasts into bridge-specific hazard time series and extreme-event catalogs. The Physics Informed Graph Deterioration Network (PiGraDeN) simulates component-level deterioration trajectories using mechanistic degradation limits in a spatio-temporal graph neural architecture driven by climate threats. The projected deterioration states and several hazard intensities in a Climate-Conditioned Multi-Hazard Fragility Surface (C-MHFS) generate time-varying, deterioration-dependent failure probability across various limit states. To translate risks into actionable performance indicators, Reliability-aware Life-Cycle Assessment via Multi-Objective Bayesian Optimization (ReLCA-MOBO) assesses adaptation and maintenance options for life-cycle cost, environmental impact, and reliability under several climate scenarios. To reduce model uncertainty, HEViS calibrates risks, deterioration, fragility, and cost models against inspection records, historical damage, and realized expenditures. DeFo-RAPS optimizes network-scale inspection, retrofit, and budget allocation policies using posterior validated predictions. The proposed methodology solves major drawbacks of earlier methods by capturing non-stationary climate extremes, openly propagating uncertainty across interconnected processes, and matching forecast accuracy with decision relevance. Optimized adaption solutions minimise failure risk, life-cycle cost, and emissions while improving forecast dependability. This work offers a rigorous climate-resilient bridge management and transferable infrastructure risk assessment method under severe climatic unpredictability. Implementation of the proposed model, along with the baselines and the dataset is publically available at https://colab.research.google.com/drive/1487RPtu4jFiUW_VaoHl74Fk4VDDCpFsH .
When choosing to replace or remove vehicles, public transport infrastructure planners and managers face major environmental and economic options and constraints. Degraded operational efficiency and reduced gross revenue result in older commercial vehicles, while premature retirement results in considerable environmental impacts at the manufacturing and supply-chain levels. This paper introduces a life cycle assessment (LCA) and vehicle operational cost (VOC) analysis of public passenger buses (PPBs) on five critical routes in the city of Agartala, Tripura, India, namely Champaknagar, Teliamura, Udaipur, Khowai and Sabrum. This study models optimal operational cost degradation and gross revenue and profit margin thresholds through five operational phases (aging intervals). Results show that for all the routes, vehicle profitability decreases in all the operational phases (e.g., the Agartala–Champaknagar route is decreased from 20.44 to 15.63
