Off-campus Michigan Tech users: To download campus access theses or dissertations, please use the following button to log in with your Michigan Tech ID and password: log in to proxy server
Non-Michigan Tech users: Please talk to your librarian about requesting this thesis or dissertation through interlibrary loan.
Date of Award
2025
Document Type
Campus Access Dissertation
Degree Name
Doctor of Philosophy in Mechanical Engineering-Engineering Mechanics (PhD)
Administrative Home Department
Department of Mechanical and Aerospace Engineering
Advisor 1
Jeremy Worm
Committee Member 1
Jason Blough
Committee Member 2
Darrell Robinette
Committee Member 3
Gowtham S.
Committee Member 4
Scott Wagner
Abstract
Automotive warranty problems can be expensive ($20M per manufacturer or more annually) and can cost customer loyalty if not corrected quickly. Quickly determining a root cause and corrective action for automotive warranty concerns is imperative if warranty costs and failures are to be minimized. This becomes exceedingly difficult if the problem is intermittent or has not been experienced before. Dealerships diagnose problems every day but are often unable to find a root cause. When they are unable to find a root cause this claim is referred to as Trouble Not Found (TNF) and the dealer bills the automotive manufacturer for the time they spent trying to find the cause. This project reviewed automotive warranty analysis, specifically trouble not found (TNF), and found a way to predict root cause using statistical pattern recognition and numerous years of warranty data sets. Research into other industries was done to determine how they corrected their own TNF warranty reports, as well as other industries already using some statistical pattern recognition. Utilizing multiple methods (Doc2Vec, K-Means clustering and anomaly detection) research was done to determine the best method for solving TNF warranty. Finally, an algorithm was determined to be used for TNF warranty to assist in root cause determination and corrective actions. Testing with this algorithm was done utilizing a known issue and an unknown issue for proof of concept. The algorithm was determined after iterative testing and adjustments where each test run provided insight that guided the next steps. The process of refining with each test run led to substantial improvements. The fine-tuning process ensures that the algorithm evolves to handle challenges more effectively, improving its performance with each cycle. Promising success, showing expectations for 90% successful prediction rate, was accomplished by a combination of these techniques. These predictions will direct the dealership maintenance team to the part and/or software that holds the root cause.
Recommended Citation
Hand, Jody J., "DIAGNOSTIC LABOR OPERATION REIMAGINED THROUGH MACHINE LEARNING AND STATISTICAL PATTERN RECOGNITION", Campus Access Dissertation, Michigan Technological University, 2025.