Course Project Work 

This study aims to solve the data synchronization problem between two in-process monitoring sensors used in the Laser Powder Bed Fusion (LBPF) process: the Field-Programmable Gate Array (FPGA) sensor and the Melt Pool Monitoring (MPM) camera. The study investigates the statistical relationships among five variables: FPGA timestamp, initial time-shift, layer index, part index, and time-shift adjustment. The study uses regression analysis methods, such as multiple linear regression with interaction terms, LASSO-penalized model selection, and step-wise model selection. The study also employs bootstrapping for uncertainty quantification, as the normality assumption of the residuals is not valid for the fitted models. The main findings of the study are: (1) layer index and part index have statistically significant main and interaction effects on initial time-shift, meaning that initial time-shift varies from layer to layer and part to part; (2) FPGA timestamp, initial time-shift, layer index, and part index have statistically significant effects on time-shift adjustment, meaning that these variables can be used to predict the time-shift adjustment; (3) the best model for predicting the time-shift adjustment is the one that uses FPGA timestamp, and layer index as predictors, with the highest prediction accuracy. The main implications and recommendations of the study are: (1) A lower bound on initial time-shift can substantially reduce the computational burden of data synchronization between the sensors, as it can help skip frames from the beginning of the MPM camera video; (2) a confidence interval on the time-shift adjustment prediction-error can inform the accommodations for accurate synchronization of FPGA and MPM data.