Machine Learning Based Classification of Atomization from a Rotating Bell Nozzle

Spray painting in the automotive manufacturing industry is a tough technical challenge. Due to a poor paint job, approximately 25% of cars coming off the assembly line must be reworked or scrapped. Previous research has shown that the droplet size distribution of the paint applied to the vehicle is important for the final coating finish and colour. Paint droplets are formed in an atomization process, and changes to this process can cause considerable differences in the coating finish and quality. In this project, experiments using an automotive paint spray robot will be done to collect atomization data of the paint as it leaves the nozzle. A machine learning based model will be developed to detect when the atomization changes, which can be used to flag defects on the assembly line. This project will further validate Mazlite’s technology in optimizing the automotive paint spray process and accelerate the commercialization of its product line.

Faculty Supervisor:

Zheng Liu

Student:

Partner:

Mazlite Inc.

Discipline:

Engineering

Sector:

Manufacturing

University:

The University of British Columbia - Okanagan

Program: