Artificial intelligence approach for aerospace defect detection using single-shot multibox detector network in phased array ultrasonic

Defect detection is one of the important issues for preventive maintenance in many industries, particularly in aviation to ensure safety. Continuous safety checks during the in-service inspection guarantee the safety of an aircraft and spacecraft and defect detection are usually done by experienced engineers. Recently, the autonomous inspection system has been implemented with the support of advanced artificial intelligence (AI) technologies. In aerospace engineering, ultrasonic testing is a reliable method to examine the integrity of composite components in an aircraft. In addition, phased array probes are commonly utilized to boost the inspection process and visualize the scanning result. In this chapter, the development of an operational system using the latest AI technology for defect detection in an aircraft is demonstrated. We adopt the convolutional neural network to detect defects in the composite laminates automatically to increase the accuracy and efficiency of ultrasonic inspection. We focus on delamination which is a critical failure mode that commonly occurs in composite materials. However, the inspection system proposed in this chapter has limited performance to nonlaminated defects, such as linear cracks. We will also focus on a higher level of system development including the algorithm for defect inspection, computer programming, and training of the inspection model. The delamination defect images of ultrasonic inspection are used for the training of the AI model for illustration.