Joint allocation of measurement points and controllable tooling machines in multistage manufacturing processes

Abstract

Stream of variations (SoV) modeling of multistage manufacturing process has been studied for the past 15 years and has been used for identification of root causes of manufacturing errors, characterization and optimal allocation of measurements, process-oriented tolerance allocation, fixture design, and operation sequence optimization. Most recently, it was used for optimal in-process adjustments of programmable, controllable tooling (controllable fixtures, CNC machines) in order to enable autonomous minimization of errors in dimensional product quality. However, due to the time and resources needed to take the measurements and the high cost of controllable tooling, it is plausible to strategically position such measurements and controllable devices across a manufacturing system in a way that the ability to mitigate quality problems is maximized. In this article, a distributed stochastic feed-forward control method is devised to optimally (in the least square sense) reduce the variations in dimensional workpiece quality with a limited number of controllable tooling components and measurements distributed across a multistage manufacturing process. Based on this, a reactive tabu search algorithm is proposed to enable joint optimal allocation of measurement points a controllable tooling devices. Theoretical results are evaluated and demonstrate using the SoV model of an actual industrial process for automotive cylinder head machining.