Optimising quality in time-variant systems: a focus on dynamics and heterogeneity
针对现代制造中随时间变化的质量特性，提出一种结合半变系数模型、高斯径向基函数插值和全时段损失优化的稳健设计方法，通过仿真和案例验证了其提升产品质量的有效性。
Traditional robust design research focuses mainly on static quality characteristics that are either time-invariant or assumed to be time-invariant for the sake of convenience. However, modern complex manufacturing environments require the consideration of dynamic quality characteristics that change over time. The goal is to optimise the design parameters so that the dynamic time-variant quality characteristic closely approximates the target profile over the entire duration. To tackle this type of quality characteristic, this paper proposes a novel robust design method that combines a semi-varying coefficient model, Gaussian Radial Basis Function interpolation (RBFI), and a time-dependent optimisation strategy that considers the full-time span loss. First, a semi-varying coefficient model incorporating both parametric and nonparametric components is constructed to explore the dynamic and complex process relationships, and a three-step iterative updating method is employed for model estimation. Second, Gaussian RBFI is used to obtain the target profile, and the optimal factor settings are determined by minimising the time-dependent loss metric between the quality characteristic and the target profile over the entire duration. The proposed method excels in capturing dynamic features and enhancing predictive accuracy. Simulation analysis and a case study validate the effectiveness of the proposed method in improving product quality.