Title: Developing online approaches to visually inspect quality characteristics with known tolerances

Authors: Mohammed Hussein; Safaa Labib Diab

Addresses: Mechanical Engineering Department, British University, P.O. Box 43, Al-Shorouk City 11837, Egypt ' Department of Physics, Helwan University, Helwan, Egypt

Abstract: In this work, new approaches for online 100% visual inspection are introduced. Unlike the traditional visual inspection approaches that testing the complete similarity of the inspected unit with a master one, the developed approaches test whether each quality characteristic in the product falls within its specified tolerance band. In the first approach, some statistical indices were used, e.g., correlation, to decide whether the dimensions of the inspected product are within specs. This approach is then modified, and each key dimension is treated alone to know which one (if any) falls beyond its tolerance zone. A third approach is introduced in which a feature of each quality characteristic is developed, and a NN model is used to determine whether this quality characteristic comply with its specs. The decision is released in the real time such that an out of control situation can be treated immediately.

Keywords: visual inspection; tolerance bands; neural networks; image processing; feature extraction; known tolerances; quality control; online inspection.

DOI: 10.1504/IJCENT.2015.073175

International Journal of Collaborative Enterprise, 2015 Vol.5 No.1/2, pp.19 - 31

Received: 28 Jun 2014
Accepted: 26 Aug 2014
Published online: 26 Nov 2015 *