SymRegOpt: An Explainable Symbolic Regression-Based Framework for Adaptive Optimization of Dry Turning Process Parameters

Abstract

Dry turning has emerged as a sustainable and environmentally conscious machining strategy due to the complete elimination of cutting fluids from the manufacturing process. However, the simultaneous optimization of machining responses such as crater wear, face wear, and surface roughness remains challenging because of the highly nonlinear interactions among cutting speed, feed rate, and depth of cut. This paper presents Sym Reg Opt, a symbolic regressionbased multi-objective optimization framework developed for modelling and optimization of dry turning parameters. A factorial experimental design was conducted considering cutting speed, feed rate, and depth of cut as input variables, while crater wear, face wear, and surface roughness were considered as output responses. The input parameters were normalized using statistical standardization to improve symbolic search stability. Symbolic regression implemented in Julia was employed to generate interpretable analytical expressions correlating machining parameters with the output responses. The generated equations were subsequently integrated with the NSGA-II evolutionary optimization algorithm to identify Pareto-optimal machining conditions. The proposed framework enables transparent and optimization-ready analytical modelling without relying on blackbox learning systems. Unlike conventional black-box machine learning approaches, the proposed framework generates explicit analytical equations that are directly interpretable, optimization-ready, and physically meaningful. The results demonstrate that SymRegOpt success-fully captures the nonlinear machining behavior of dry turning while simultaneously supporting sustainable and interpretable manufacturing intelligence, using Evolutionary Computation.