Document Type

Conference Proceeding

Publication Date



Stamping process, Design of experiments, Generalized Linear Model, Processimprovement


This paper presents an approach used for improving the visual quality of body panels produced in an automated press shop of a car manufacturer. To solve the complexproblem of panels’ visual defects,the optimal values of process parameters based on Generalized Linear Models (GLZ) were identified and deployed. The project focused on parts exhibiting excessive rejection rates. To deal effectively with the case in which the distribution function representing the number of defects in the panels was found to be binomial with presence of some covariates in the model, the research approach focused on development of Log it Models.

After identifying the potential causes of defects and selecting the most significant process factors from typical factor sets varying from 20 to 40 for each part, the levels of factors, number of replications and appropriate methods for running the experiments regarding to practical limitations and shop floor scheduling were established. Finally, the optimal values for process variables minimizing number of defects and used effectively to set up action plans for improving the panels’ quality were identified.Implementing the plans, the process capability was increased by 0.25 for those non-satisfactory parts and production costs fell between 3-7% among them.