Automatic forgery detection for x-ray non-destructive testing of welding

Abstract

Inadvertently and intentionally incorrect welding films are very common in X-ray non-destructive testing (NDT) of welding engineering. The forgery welding films were handed in by welding workers and pretended these films were Xray photographed on the welding seams that need to be inspected, but actually, these films were X-ray photographed on other welding seams that are of high quality and had passed the X-ray imaging inspection. If some welding seams escape inspection due to forgery, these will cause many potential safety hazards. The current forgery detection method is manual inspection, which is very time-consuming and laborious. An automatic forgery detection method for X-ray non-destructive testing of welding was proposed. Three solutions were used to prevent forgery: they were the markers recognition for avoiding erroneous submission, the fingerprint extraction for identifying welding seam uniquely, and the overlap matching for preventing intentional forgery. This paper proposed a concept of welding seam fingerprint. The welding seam features were extracted based on an improved SPP-net deep learning model. Those deep learning features of welding seam were called as the welding seam fingerprint. The proposed method can automatically accomplish the forgery detection, which can greatly reduce the workload of manual detection and make the welding inspectors focus their attention on the welding defects inspection rather than on the forgery prevention.

Acoustic Non-Destructive Testing Technology in Concrete Bridge Inspection and Pile Foundation Detection

“…After treatment, acoustic nondestructive testing was performed again. In the latter inspection, it was found that the concrete strength of all parts of the bridge floor reached the design standards of this project [9] .…”

NDT Method for Weld Defects Based on FMPVit Transformer Model

“…Modern welding testing methods can be divided into destructive testing (DT) and nondestructive testing (NDT) methods, depending on whether the test method is destructive. NDT is highly efficient and safe; it is the current mainstream welding defect detection method [3]. In welding NDT, most methods use images as the input, and the convolutional neural networks (CNNs) are employed.…”