Machine Learning-Based Methods for Engineering System Optimization

Abstract

Optimization problems are at the core of many engineering systems, where performance, efficiency, and reliability must be improved under practical constraints. Existing studies on machine learning based optimization often emphasize algorithmic development, while providing limited discussion on how learning models support decision making under real engineering constraints. In recent years, machine learning has become an increasingly important tool for addressing optimization tasks that are difficult to solve using conventional analytical or heuristic approaches. Instead of relying on explicit system models, machine learning methods enable optimization strategies to be learned directly from data. This paper presents an overview of machine learning based methods for engineering system optimization. The discussion focuses on commonly used learning models, their roles in optimization workflows, and their application in representative engineering domains. In addition, practical challenges related to interpretability, data dependence, computational burden, and engineering deployment are examined. The aim of this review is to provide engineers and researchers with a clear understanding of how machine learning is currently used for optimization and what limitations must be considered in real engineering practice.