An efficient and generalized parameter optimization methodology for Gough-Stewart platform

Parameter optimization of Gough-Stewart platforms (GSPs) is crucial for enhancing their kinematic and dynamic performance. Traditional approaches, which involve solving complex equations and performing comprehensive workspace analyses, are often computationally intensive and mathematically complex. To overcome these challenges, this paper proposes a novel deep learning-driven framework, termed OMSP-DLPM, for efficient and generalizable parameter optimization of GSPs. The core innovation lies in developing a performance prediction model (DLPM) that accurately maps design parameters to performance metrics, thereby bypassing traditional, computationally expensive simulations. The methodology integrates four key components: theoretical analysis to identify critical performance determinants, an adaptive non-uniform sampling strategy based on sensitivity analysis, construction and training of a deep learning model with customized loss functions, and validation across multiple GSP configurations. Experimental results demonstrate that the DLPM achieves exceptional prediction accuracy, with a mean absolute percentage error of 0.39% and a coefficient of determination of 0.9996. In optimization trials, the proposed method limited performance metric errors to within 5%, maintained deviations in optimal design parameters below 7%, and reduced computation time to under 27 min per case. This research provides a practical framework for efficient parameter optimization of Gough-Stewart platforms.