Using Phased Array Ultrasound to Localize Probes During the Inspection of Welds

In this paper, an image processing-based localization system is developed for remote nondestructive evaluation of welds within industrial assets. Manual ultrasonic inspection of large-scale structures is often repetitive, time-consuming, and benefits greatly from robotic support, however, these robotic systems are often fixed to a single purpose, lack self-awareness of their surrounding environment, and can be limited to simple geometry. For the inspection of welds, which are often carried out using phased array ultrasonic testing, there is a reliance on the use of surface features for automated tracking such as the laser profiling of a weld cap. For the inspection of more complex geometry such as non-linear or saddle welds a more positionally sensitive method is required. The proposed system utilizes information already available to a nondestructive inspector in the form of live phased array ultrasonic images to estimate the location of the weld using non-surface, volumetric data. Data is captured using a 64-element, 10 MHz phased array probe mounted to the end-effector of a small robotic manipulator which increases the scope of applications due to its heightened flexibility when compared to on-the-market alternatives. Morphological operations are applied to the ultrasonic data to reduce the noise apparent from regions of parent material and promote the data reflected from grain boundaries within the weld material. Through a series of image processing techniques, it is possible to predict the position of a weld under inspection with an absolute mean positional error of 0.8 mm. From this study, the localization system is to be embedded within a remote system for extensive data acquisition of welds on large structures.