The research objective of this award is to establish a set of experimental design, analysis, and quality control methodologies for systems with both quantitative and qualitative (QQ) responses. Systems with QQ responses are commonly encountered in various manufacturing and biomedical applications. For example, in a thermal spray coating process, the coating quality is often measured by quantitative surface roughness and a qualitative coating failure indicator. Another example can be found in an organ transplant process, where the quantitative biopsy testing scores and doctors, qualitative evaluation are used to evaluate the health condition of the organ. The new methodologies are based on a joint modeling of the QQ responses to unveil the hidden relationship between the two types of the responses, and thus are more advantageous than the traditional methods that model the two types of responses separately. These methodologies will be tested and implemented in real case studies provided by industrial collaborators. The education plan will promote undergraduate and graduate students from the underrepresented groups to obtain the rigorous quality control training in various manufacturing and biomedical applications. The research results will be used to improve the quality in real industrial practices.
If successful, this research will lead to effective and reliable modeling, experimental design, monitoring and control of the systems with QQ responses. The joint modeling method considers constrained likelihood estimation with joint variable selection. It will further accommodate various complex scenarios including functional predictors, missing values, multiple responses, and nonlinear quality-process relationship. A Bayesian optimal design of QQ responses is planned to address the experimental design needs, which can be extended to sequential designs and designs for robust systems. Based on the modeling method, the model-oriented process monitoring will detect process changes, and identify root causes from the predictors or from the system itself. The quality improvement will be fulfilled by optimizing objectives accounting for the hidden relationship between the QQ responses. These methodologies will serve a set of powerful tools for manufacturing scale-up and biomedical quality improvement. Moreover, the educational plan will provide undergraduate and graduate students with various course modules and research opportunities. Real case studies will be used in lab sessions and undergraduate capstone projects. The research will equip students with critical thinking and hands-on skills in applied statistics, manufacturing, and biomedical systems. Research results will also be disseminated to academia and industry through publications, conferences, and workshops. A website will be established to share data, case studies, and advanced tools for promoting research collaboration.
