High Definition Metrology based Process Control: Measurement System Analysis and Process Monitoring.

Abstract

Process control in high precision machining necessitates high-definition metrology (HDM) systems that provide fine resolution data needed to characterize surface shape. HDM data is critical for the evaluation of process surface variation, as it reveals local surface patterns that are undetectable using low definition metrology (LDM) systems. Monitoring of the part-to-part variation of these patterns identified by HDM enables the detection of abnormal surface variation and the degradation of process conditions. HDM systems present many opportunities for surface variation reduction. However, there are challenges to using HDM data for process control. Conventional HDM systems are inefficient and may take a long time to measure a part, such that sufficient samples cannot be obtained for process control purposes. In addition, conventional monitoring methods are difficult to implement due to the high density of data. A new study uncovered significant cross-correlations between part surface height and process variables in an automotive engine milling process. This dissertation aims to apply new insights gained from HDM to develop algorithms and methods for surface variation control, specifically:

- Surface modeling through fusion of process variables and HDM data: An improved surface model is developed by incorporating process and multi-resolution data through spatial and cross-correlation to increase prediction accuracy and reduce the amount of HDM measurements necessary for process control. - Measurement system analysis for HDM using: A method to effectively estimate the gage capability for HDM systems is proposed. - Surface variation monitoring using HDM data: A sequential monitoring framework is developed to monitor surface variations as reflected by HDM data. Based on the surface data-process fusion model, a progressive monitoring algorithm under a Bayesian framework is developed to monitor surface variations when limited HDM measurements are available. - Multistage modeling and monitoring of HDM Data: A morphing-based approach is proposed to model process multistage interdependence. A new multistage monitoring procedure is developed based on the morphing model. The research presented in this dissertation will aid in transforming quality control practices from dimensional variation reduction to surface shape variation control. The proposed HDM data monitoring algorithms can be extended to other high precision manufacturing processes.