Probabilistic measure-based risk evaluation models for complex engineering operations under uncertainty

Abstract

In the modern world, complex engineering processes become more and more exposed to uncertainty due to equipment deterioration, varying operation regimes, changing environmental conditions, maintenance considerations, and part failures. Classical deterministic risk assessment techniques may poorly characterize such variability; classical probabilistic risk analysis may be focused on average values but offer no guarantees for protecting systems from adverse situations. This work proposes a risk analysis methodology based on probabilistic measures and minimax optimization. The proposed approach considers uncertain engineering variables as random variables with specified probability distributions, constructs realistic scenarios of system operation for the purpose of risk analysis, and uses various risk measures, namely, expected loss, variance based risk, Value-at-Risk, Conditional Value-at-Risk, and reliability-based measures, to define the composite probabilistic risk function. The minimax method is next used to find out the engineering choices that ensure minimum possible risks in the maximum possible scenario within an uncertainty set. The numerical implementation for a particular example of an industrial engineering process shows how effective the proposed approach can be in lowering worst-case risk while increasing reliability and robustness of the decision-making process. The comparison with deterministic, traditional probabilistic, VaR, and CVaR approaches confirms that the proposed approach can offer greater protection under unfavorable scenarios. It is also revealed by sensitivity and ablation studies that probabilistic quantification, dependency modeling, uncertainty propagation, and minimax optimization are all crucial for robust engineering decision-making.