A comprehensive framework for computationally efficient system-level design optimization of machine tools

Abstract

Mass reduction of machine tool components is a crucial task that can improve performance, accuracy, and energy efficiency. System-level optimization, where multiple components are simultaneously optimized, is noted in the literature as a challenging necessity for complete performance improvement of machine tools. Existing methods focus on optimizing the machine tool components individually, neglecting the critical effects of simultaneous modification on the machine performance and thorough exploration of the design space. To the authors’ knowledge, for the first time in the literature, this paper presents a comprehensive framework for system-level machine tool design optimization considering the most significant multi-objective performance indicators for the machining process. Static and dynamic stiffness, thermal and dynamic stability, and fatigue life are evaluated as performance indicators using a multi-objective finite element response set that includes coupled thermal-structural, modal, and frequency response analyses. A minimal parameter set approach is proposed which uses the linear guide joints to minimize the number of design variables, addressing the challenge of increased computational cost in system-level modeling. Machine responses during optimization iterations are predicted by a machine learning model trained on the machine tool's multi-objective finite element response set, achieving higher accuracy than commonly used polynomial-based methods. A constraint relaxation method is proposed that permits limited degradation relative to the base design, yielding designs that substantially outperform those obtained from unconstrained optimization while avoiding over-constraining. Up to 20 % mass reduction is achieved across the machine tool components while the performance indicators are either improved or maintained with negligible degradation.