Integration of Risk-Based Self-Maintenance Activities: Study Case for Crane Distributed System

Abstract

In today's practice, the coordination of maintenance activities is steadily improved with the use of models. Among them, simulation-based models are attractive because they offer flexibility and can absorb system complexity with ease. Although we avoid state enumeration by using simulation-based models, they can sometimes become computationally demanding. To obtain efficient results, simulation-based models are often implemented on a digital platform. However, there is always concern about whether the implemented model performs adequately and delivers the expected results. In this paper, to address this concern, we propose to perform model calibration. The subject of the paper is calibration through comparison of a risk-based self-maintenance layer, which was created to be executed online without human intervention. The layer is intended for self-assessment of a critical system of overhead cranes operating in hazardous conditions. The case study of the paper is a critical system of overhead cranes operating at a continuous rate, which are part of a hot rolling mill system in a steel plant. The online self-maintenance layer is a nonlinear stochastic optimization model with bounded constraints that aims to evaluate a system-level risk indicator. The estimation of the risk indicator is based on Monte Carlo simulations. Given the features of the model, to evaluate the accuracy and robustness of the layer, only free derivative algorithms, Genetic Algorithm (GA), Particle Swarm Optimization (PSO), and Nelder-Mead Modified (NMm) in our case, are used in the comparison. The presented work shows in practice common challenges when facing complex nondeterministic polynomial (NP) problems with stochastic functions.