It is difficult to inspect for defects in austenitic welds ultrasonically due to complicated material properties inside the weld. Weld microstructures typically lead to weld stiffnesses that are both anisotropic and inhomogeneous, so that ultrasonic waves tend to deviate and scatter. A weld performance map is commonly used to describe how the material properties vary throughout the weld, and this idea has been applied to wave propagation models. In this work, we developed a non-destructive method to measure this map using ultrasonic arrays. A material model (previously published by others) with a small number of parameters has been applied to describe the weld performance map. It uses the information of the welding procedure and rules for crystalline growth to predict the orientations, therefore it has a good physical foundation. An inverse model has then been developed to measure the weld performance map based on the matching of predictions by the ray tracing method to selected experimental array measurements. The process is validated by both finite element models and experiments. The results have been applied to correct array images to compensate for deviations of the ultrasonic rays.