Abstract
Structural damage is a significant factor that threatens the service life and operation safety. At present, many technical methods can only achieve qualitative damage detection, without providing specific information. Quantitative damage evaluation is more challenging and significant than qualitative damage detection of structures. Pipe-like damage is a common occurrence, caused by factors such as corrosion, welding, and excavation, etc. Ultrasonic detection is a common technique which is very suitable for pipe detection. In this paper, a quantitative evaluation technology based on ultrasonic detection and deep learning is proposed. The 3 dimensional structure is converted into 2 dimensional plate. The plate is discretized into pixels that represent the spatial structure in 2 dimensions. Then, the ultrasonic signal is mapped to 2 dimensional pixels through deep learning to realize the imaging of the damage of the pipe. This method is commonly referred to as 2.5 dimensional imaging because it is not a direct 3D imaging method but can be converted into a 3D structure. Numerical experiments were conducted on typical pipe-like structures, and the results demonstrate that physical information of damage, including location, size and shape, can be accurately detected. Although this imaging method has only been verified by simulation, and it is expected to be explored and applied in future studies to more complex structures, such as aircraft wings, trusses, and beams, for the purpose of detecting and imaging damage.
