Date of Award

6-2019

Degree Name

Doctor of Philosophy

Department

Industrial and Entrepreneurial Engineering and Engineering Management

First Advisor

Dr. Lee Wells

Second Advisor

Dr. Azim Houshyar

Third Advisor

Dr. Steven Butt

Fourth Advisor

Dr. Sam Ramrattan

Keywords

chemically bonded sand, statistical process control, quality control, classification, principal component analysis, qualification framework

Abstract

The goal of this dissertation is to establish a new quality control framework that combines a statistical process control (SPC) approach to casting quality for chemically bonded sand systems used in foundries. Foundries in the United States use the American Foundry Society standardized sand testing to monitor chemically bonded sand systems. These standardized tests are inefficient for two reasons. Firstly, standard tests are based on mechanical, physical, chemical and thermal properties of a sand system that do not consider interaction between these properties, but sand casting processes are inherently thermo-mechanical, thermo-chemical and thermo-physical. Secondly, these tests can only detect large shifts in sand systems due to higher variability in their measured data. Because of this, the application of SPC using the results of standardized tests cannot detect subtle changes in sand systems.

In contrast, disc-shaped specimen tests developed at Western Michigan University collect data of independent as well as coupled thermal, mechanical, and physical properties of sand systems. In addition, these tests have shown their ability in reducing specimen-to-specimen and test-to-test variability. Therefore, in this research, these disc-shaped specimen tests are used to develop a SPC method for detecting a change in sand system by performing feature engineering on the data sets acquired from these tests.

The proposed approach to accomplish the goal is divided into three objectives. First objective is to determine that the disc-shaped specimen tests can differentiate changes in sand systems. To develop an effective SPC method, it is important to research the mentioned objective of this phase. In this research, several different sand systems are considered that represent changes in sand systems. Disc-shaped specimen tests are performed on disc-shaped specimens of these sand systems and test data are collected using current practice. Collected data are analyzed for their classification capability.

Second objective is to perform feature engineering on the data set acquired from disc-shaped specimen tests to detect a shift in sand systems. This step focuses on implementing SPC using feature/s of these test results to detect any shift in a sand system. Thermal distortion test which is one of the important disc-shaped specimen tests, produces time-series data for multiple variables. SPC method using these time-series data was considered in this phase and results suggests that SPC using principal component analysis can help better discriminate changes in sand systems.

Third objective of this research is to combine SPC method to casting quality. In this phase, a quality control framework is proposed that can be used to verify the casting quality. After detecting a change in sand system, foundry need to verify effect of that change on casting quality because not all process shifts reduce casting quality passed an acceptable limit. The quality control framework, in this phase, uses surface defect casting trial that uses disc-shaped specimens to validate casting quality and “qualify” a chemically bonded sand system. A sand system is deemed qualified if casting trials are free of a specific defect under given process parameters.

Access Setting

Dissertation-Open Access

Download

Share

COinS