The need for high reliability interior crack detection in jet engine turbine components requires an optimal configuration of ultrasonic test equipment. Such optimization must be based upon the specific part geometries involved and upon the crack population which is supposed to exist in the parts. Due to the high cost with respect to time and manpower of fabrication, system evaluation and destructive analysis of experimental samples, it is preferable to simulate the detection process numerically in order to identify system configurations which promise high probability of detection of cracks. This paper describes such a computer model, presents experimental verification of its accuracy and illustrates its use in the design and optimization of ultrasonic detection systems for cracks in web geometries.