Sandia National Labs Datathon

During the Sandia National Laboratories Data Challenge, I analyzed over 1,500 additive manufacturing observations to identify key drivers of part scrap and nonconformity. I engineered process-level features from geometric specifications such as lip diameter, height, and thickness measurements, creating binary “in-spec” indicators to quantify deviation from manufacturing tolerances. Using R, I conducted exploratory data analysis to understand patterns in the data and identify potential relationships between process variables and part quality. I then applied statistical and machine learning models, including logistic regression, ANOVA, and Random Forest classification, to evaluate these relationships. Through this analysis, I determined which process and dimensional variables most strongly influenced the probability of a part being scrapped. I visualized spatial and temporal defect patterns through heatmaps, week-by-week scatter plots, and specification limit overlays, uncovering that lip thickness variability (particularly LT3 and LT4) and layout configuration were the most significant predictors of scrap. The project concluded with actionable insights linking geometric precision to material type and layout design, supporting data-driven process optimization for additive manufacturing quality control.