Evaluation of a Neuro-fuzzy Machine Vision System for Intelligent Inspection

This project is in partnership with AUTO21. Van Rob is an automotive parts manufacturer of metal stampings with plants in the Toronto area. The objective of this internship is to evaluate the performance of a machine vision inspection system in an industrial environment. The evaluation of the intelligent neuro-fuzzy inspection algorithm used by the machine vision system known as QVision is the subject of the intern’s Master’s thesis. The original QVision was developed by a previous Master’s student. QVision has been installed on one of Van Rob’s Manufacturing cells. The basic methodology is for the intern to test the performance of QVision as the cell operates. Different techniques will be studied to correct for problems revealed during the testing process. The criteria for success is for QVision to operate with less than 2% false negatives (rejection of a good part) and zero false positives (acceptance of a bad part). If the test is successful, Van Rob will have a production proven machine vision inspection system that can be applied to other areas of production. A working machine vision inspection system will reduce the number of false negatives and consequently reduce scrap rates. Furthermore, by eliminating the number of false positives, Van Rob will avoid economic penalties associated with containment action by their customers.

Faculty Supervisor:

Dr. Brian Surgenor

Student:

Brandon Miles

Partner:

Aurora Corporate Specialist

Discipline:

Engineering

Sector:

Automotive and transportation

University:

Queen's University