Quality improvement is a key objective in meeting reduced costs and increased productivity. It has therefore become an integral part of many organisations' business strategy. In the Non-Destructive Testing (NDT) of welds, the quality of inspection results from a welded structure is closely linked to the parameters used in the test procedure adopted. The formulation of the optimum test parameters depends on the knowledge and experience of the NDT specialist and requires inputs of several different types of expert knowledge. This experience which the human experts have is usually heuristic, judgemental, subjective or intuitive in nature. Moreover, the optimum procedures usually differ from one job to another. Hence, the application of the design of experimental techniques, such as Taguchi's approach, can be used to identify optimum conditions which are robust against unwanted disturbances in the testing environment while providing an improved degree of sensitivity. Taguchi methods (Taguchi, System of Experimental Design, Unipub Krauss, 1987; Phillip, Taguchi Techniques for Quality Engineering, McGraw-Hill, 1988) which are aimed at reducing variability and cost, can be used to improve the sensitivity of ultrasonic measurement methods through the use of the Signal-to-Noise (SN) ratio as the quality characteristic of a measuring system. Traditionally, the quality of measuring systems has been evaluated on the basis of repeatability. Another important element associated with the quality of measurement systems is ‘sensitivity’, which is the ability to perceive and discriminate between two signals or samples to be measured. Parameter design in Taguchi methodology can be used to provide the human experts with a means of optimising inspection parameters. Therefore, knowledge-based systems incorporating heuristic algorithms, as well as analytical and empirical models provided by Taguchi methods can in turn provide human experts with support to further improve their decision making. This paper will describe how parameter design can be used to increase the efficiency of NDT procedures by providing robust inspection parameters for a knowledge-based expert system, an ongoing research program to enhance industrial quality.
Read full abstract