Predictive-Inspection Based Process Control in End Milling Operations

Abstract

In the manufacturing industry, product quality control is often treated as a completely different problem than process diagnostics. Diagnostics methods are used to quickly identify faults during the process, whereas product quality is assessed at the final inspection stage, after the process is completed. The large amount of data that is collected to enable the on-line diagnostics, however, can be used to predict the product quality before the part is measured at the final inspection stage. This predicted quality can be used as feedback to a control system that improves the process quality by adjusting the process inputs. In this paper we propose a predictive inspection based process control solution for a manufacturing process. A manufacturing process is modelled as an input-output system with the machine settings as inputs and two kinds of outputs: diagnostic data and quality. This model is obtained from off-line experiments using a combination of process inputs and external disturbances. In the online implementation the predictive process model is updated with the diagnostic data collected during runtime and quality is improved by using a two-loop control strategy, on a part-to-part or run-to-run (R2R) basis. The proposed approach is applied to and developed for an end milling operation and simulation and experimental results are presented.