Statistical Process Monitoring Based on Region Growing for Fused Deposition Modeling with Image Data

Fused Deposition Modeling (FDM) is a highly prevalent additive manufacturing technique. Product quality is of utmost importance for the success of FDM manufacturing process. Effective monitoring of the printing process enables timely detection of product defects and facilitates significant quality improvement by providing timely corrective actions. The advancement of machine vision is driving the development of statistical process monitoring (SPM) with images data in FDM process. Given that out-of-control FDM process may appear due to parameter drifts, a SPM method is proposed to monitor the status of the process. This method is for grayscale images of each printing layer, which are continuously captured by real-time industrial camera systems. Regions of interest (ROIs) are first determined by region growing, so as to locate the different positions of the unfilled regions. The generalized likelihood ratio (GLR) is introduced to establish statistical variables and find out the change point where the parameter shifts in the process, so as to realize the monitoring of the ROIs of each image. The probability of alarm in a specified period (PASP) and the cumulative PASP are used to determine the control limit and measure the monitoring performance of control chart. Both simulation and case study results demonstrate the proposed method exhibits the effectiveness in detecting faults in time and the precision in estimating their locations.