Abstract
In response to the growing demand for autonomous adaptive control in manufacturing lines for productivity enhancement and carbon neutrality, we propose a novel methodology for autonomous control of product quality, taking into account the effects of non-measurable parameters. Using local linear regression modeling combined with temporal neighborhood data, we identified a single manufacturing parameter guided by the derived regression coefficients. Our simulation results revealed that conventional multiple regression modeling often produced undesirable control
behavior, characterized by fluctuations in product quality. To mitigate this instability, we employed semiparametric regression modeling. Notably, the semiparametric regression model succeeded in stabilizing control, through accurate selection of the control target parameter and the incorporation of an additional non-linear term that offsets time-dependent, non-measurable parameters. Our approach facilitates enhanced manufacturing control, promoting both efficiency and sustainability.
