An Explainable Reconfiguration-Based Optimization Algorithm for Industrial and Reliability-Redundancy Allocation Problems

TL;DR

The paper tackles large-scale constrained optimization in industrial and reliability-redundancy contexts, introducing AI-AEFA, an intelligent parameter reconfiguration-based metaheuristic built on AEFA. It integrates a log-sigmoid-based adaptive Coulomb's constant with chaotic mapping to enhance exploration and convergence, and applies a parameter-free constraint-handling technique. A key advance is the SHAP-based explainability, which quantifies the influence of Coulomb's constant, charge, acceleration, and electrostatic force on decisions. Across 28 CEC 2017 constrained benchmarks, 15 industrial problems, and 7 RRA problems, AI-AEFA demonstrates superior feasibility, convergence speed, and computational efficiency, while offering actionable interpretability for real-world deployment and future multi-objective extensions.

Abstract

Industrial and reliability optimization problems often involve complex constraints and require efficient, interpretable solutions. This paper presents AI-AEFA, an advanced parameter reconfiguration-based metaheuristic algorithm designed to address large-scale industrial and reliability-redundancy allocation problems. AI-AEFA enhances search space exploration and convergence efficiency through a novel log-sigmoid-based parameter adaptation and chaotic mapping mechanism. The algorithm is validated across twenty-eight IEEE CEC 2017 constrained benchmark problems, fifteen large-scale industrial optimization problems, and seven reliability-redundancy allocation problems, consistently outperforming state-of-the-art optimization techniques in terms of feasibility, computational efficiency, and convergence speed. The additional key contribution of this work is the integration of SHAP (Shapley Additive Explanations) to enhance the interpretability of AI-AEFA, providing insights into the impact of key parameters such as Coulomb's constant, charge, acceleration, and electrostatic force. This explainability feature enables a deeper understanding of decision-making within the AI-AEFA framework during the optimization processes. The findings confirm AI-AEFA as a robust, scalable, and interpretable optimization tool with significant real-world applications.

Figures (12)

• Figure 1: Interaction of agents in AEFA.
• Figure 2: The plot of the original and proposed Coulomb's constant
• Figure 3: Procedure for embedding the chaotic map into the proposed Coulomb's constant as follows: Figs: A $\rightarrow$ B$\rightarrow$ C $\rightarrow$ D.
• Figure 4: Flowchart of the proposed algorithm.
• Figure 5: Schematic procedure diagram of sorting agent in the proposed algorithm.