Automated visual inspection for detection, characterization, and measurement of flaws on machined metal surfaces

This work investigates the automatic inspection of machined metal surfaces through their light scattering properties. Light patterns resulting from the incidence of laser light on a surface flaw are studied to develop efficient methods for detection, characterization, and measurement of flaws on machined surfaces. The work places special emphasis on flaw depth inspection and the simplicity of the inspection system.

Experimental investigations of this study involve optical and pattern recognition techniques. Low-power laser light impinges on a machined surface normally. The scattered light is focused on a Wedge/Ring detector whose signals are processed by pattern recognition algorithms to qualify the surface.

Three theoretical light-scattering models are developed for surface flaws. These models indicate that the light scattered from a flaw surface is dependent on flaw parameters, experimental factors, and surface properties, and that the observed light intensity and flaw depth do not have a one-to-one relationship.

The major experimental results are that flaw detection based on a given depth threshold can be performed very satisfactorily if flaw width and shape do not disturb the depth inspection drastically. However, separating flaws into different classes based on depth cannot be performed so successfully. Improvement of the results generally requires finding a data feature which depends only on flaw depth or developing models which properly accommodate the effects of flaw depth, width, and shape. Experimental models are developed for construction of flaw shape and measurement of flaw dimensions. While shape construction is hindered by laser characteristics, measurement of flaw depth or width can be performed fairly satisfactorily, although it needs further improvement.

This study examines some theoretical and some experimental aspects of metal surface inspection. Although the results do not present a complete solution to the problem, they constitute a significant contribution to the state of the art in this field.