Progressive measurement and monitoring for multi-resolution data in surface manufacturing considering spatial and cross correlations

Controlling variations in part surface shapes is critical to high-precision manufacturing. To estimate the surface variations, a manufacturing plant usually employs a number of multi-resolution metrology systems to measure surface flatness and roughness with limited information about surface shape. Conventional research establishes surface models by considering spatial correlation; however, the prediction accuracy is restricted by the measurement range, speed, and resolution of metrology systems. In addition, existing monitoring approaches do not locate abnormal variations and lead to high rates of false alarms or misdetections. This article proposes a new methodology for efficiently measuring and monitoring surface variations by fusing in-plant multi-resolution measurements and process information. The fusion is achieved by considering cross-correlations among the measured data and manufacturing process variables along with spatial correlations. Suchcross-correlations are induced by cutting force dynamics and can be used to reduce the amount of measurements or improve prediction precision. Under a Bayesian framework, the prediction model is combined with measurements on incoming parts to progressively make inferences on surface shapes. Based on the inference, a new monitoring scheme is proposed for jointly detecting and locating defective areas without significantly increasing false alarms. A case study demonstrates the effectiveness of the method.