Date of Award
11-22-2024
Document Type
Dissertation
Degree Name
Doctor of Philosophy (PhD)
Department
Electrical and Computer Engineering
First Advisor
Richard Povinelli
Second Advisor
Phil Voglewede
Third Advisor
Cris Ababei
Fourth Advisor
Edwin Yaz
Fifth Advisor
George Corliss
Abstract
Surface grinding plays a pivotal role in the machining industry, constituting roughly 25% of all machining operations worldwide. Its precision and efficiency are crucial, particularly in sectors requiring high-quality surface finishes, such as aerospace and semiconductor manufacturing. For thermal damage prevention, traditional approaches to parameter selection use thresholds to exclude burn-prone parameters. However, by omitting the cost of burn, the threshold-exclusion strategy yields outcomes that fail to reflect the true costs of grinding. This dissertation introduces a novel burn cost model that transcends these limitations, offering a more nuanced and cost-effective approach to managing grinding burn. The burn cost model presented here addresses the shortcomings of traditional burn avoidance techniques by estimating the actual costs associated with grinding burn. This is achieved by integrating the economic impact of burn into the overall cost of the grinding process, allowing for a more realistic evaluation of grinding cost efficiency. This effectively balances competing objectives of throughput maximization and workpiece defect minimization. A grinding model is developed to explore the differences between the proposed cost model and traditional approaches on cost, speed, and quality of grinding. Eleven separate experimental runs are performed, and comprehensive models of grinding power, surface roughness, burn detection, and grinding ratio are developed. Machine learning techniques of linear regression, support vector machines, and physics-informed modeling are used to curate the modeling process. Parameter selection is defined by a cost function, several parameter constraints, and a two-stage design of rough and finish passes. Gradient descent and evolutionary algorithms are used to conduct the parameter search. In addition, the concept of burn risk probability is presented for estimating burn costs. This dissertation contributes to the field by challenging the conventional burn avoidance strategy based on burn threshold constraints. Results show the flaws of the burn threshold method and the advantages of the burn cost model in quantifying burn risks, enhancing robustness to parameter sensitivities, and providing value-based grinding optimization. The proposed cost framework demonstrates a more accurate reflection of the actual grinding costs. Robustness and sensitivity are also evaluated, showing a superior approach to managing and minimizing grinding costs.