Automated ultrasonic classification of defects in aluminium die castings

Casting is the most ancient metal-shaping technique exploited by man, and is the process of forming metal objects by melting metal and pouring into moulds[1]. There has always been the problem of discontinuities in castings such as irregularities, breaks, or gaps in the material structure. If the discontinuities are sub-surface they must be detected and identified before they can be eliminated. Automotive castings oen are subject to fluids under pressure which includes transmission fluid, engine oil and coolant. A common problem is fluid leakage through sub-surface discontinuities. Therefore, the detection of porosity and other discontinuities that cause leakage is of paramount importance in maintaining quality. Nondestructive testing (NDT) techniques (including ultrasonic, x-ray, liquid penetration, eddy current and magnetic particle testing) have previously been used in different areas of the casting industry[2]. Most of these techniques have experimented with machined castings. At present, in the automotive die casting industries, defective castings are detected off-line either through x-ray or leak testing of castings. The focus of this research work was to investigate the possibility of identifying sub-surface defects using ultrasonic inspection prior to machining the die castings. A difficulty with this technique is that a significant amount of ultrasonic energy is scattered by the rough surface of the as-cast components. Furthermore, the castings have variable internal grain size structure which leads to problems with scattering of the ultrasonic signal. Another major focus was to automate the defect classification of ultrasonic signals using various signal processing techniques. The objectives of this research program emphasized the necessity of obtaining an understanding of the casting process and use of ultrasonic inspection. Once the casting process and ultrasonic NDT methodologies are understood, then it should be possible to duplicate the expertise of a human inspector in the inspection area through an automated identification system. The experimental set-up and results obtained from ultrasonic immersion testing of the sample castings are also described in this paper. Finally, the results of an initial implementation of the inspection system are presented.