Authors:

You-Cheng Lay

Date Created:

Not specified

Course Title:

Professor:

Not specified

About Paper:

This project aims to develop an effective method for predicting the performance of the manufacturing process and finding the factors that would affect the manufacturing outcome. The goal is to predict the performance and compare each model based on several statistical factors, including ROC, Entropy R2, AUC, RASE, and classification accuracy. The outcome is whether or not the manufacturing process has failed and what variables affect the outcome. Variable selection involved the use of stepwise selection (both backward and forward), p-value selection, BIC selection, and the contribution portion. The target variable is the defect status, which is indicated by 'YES' or '1' when the process has failed and by 'NO' or '0' when the process is in control. Models constructed include: Bootstrap Forest, K-NN, Neural Boosted Network, Support Vector Machine, Naive Bayes, and Logistic Regression. The final model is the one constructed through the bootstrapped forest, which has the highest Entropy R2 score among all the candidate models. The bootstrapped model also has the highest AUC, accuracy rate, and the lowest RASE. The variables that affect the defect status are: Maintenance hours, Quality Score, Defect Rate, Production Volume, Additive Material Cost, Energy Efficiency, and Production Cost. Keywords: Machine Learning; Manufacturing Defects; Predictive Modeling; Supervised Learning; Quality Control

Source:

Purdue University / 2025

Topics:

No topics listed

Co-authors:

You-Cheng Lay