| Weld defects mainly include slag inclusion, pores, non-fusion, and lack of penetration.These defects would impair welds’ bearing capability and cause stress concentration, which then results in cracks, and finally leads to weld fracture and accidents. The normal nondestructive testing methods cannot detect early-stage stress concentration. In contrast, new metal magnetic memory technology can detect not only test pieces’ macroscopic defects, but also the stress concentration areas and early impairments in them, so it can be used to prevent equipment from potential safety hazards caused by early impairments. However,due to different loading conditions and welded materials, along with the interference of non-defective factors, such as noise and lift-off effect during measurement, surface deposit and support frames of measured objects, etc.magnetic memory data is featured by dispersion and uncertainty, thus bringing difficulties to identification of the degree of weld defects. To solve this problem, this paper first introduced D-S evidence theory in identifying the types of weld defects, and based on that, further carried out quantitative assessment on weld defect levels by using the combination of entropy and D-S theory.First, it carried out the weld fatigue test, designed the welded test pieces made of Q235 B,and prefabricated the four defect types: slag inclusion, pores, non-fusion, and lack of penetration. By comparing the magnetic memory signal of these different defect types, it found that: the magnetic memory signal curve of the test piece with slag inclusion changed gently; there were little fluctuations on the magnetic memory signal curve of the test piece with pores, but it changed gently on the whole; the magnetic memory signal curve of the non-fusion test piece changed violently and several obvious wave crests and troughs,representing the signal’s reciprocating changes, could be seen; the magnetic memory signal curve of the test piece lack of penetration changed more sharply than the other three, and the entire magnetic memory signal curve presented large jumps, with prominent wave crests and troughs. Besides, the characteristic values of the magnetic memory signal of the different defect types, such as peak-to-peak value, gradient value, limit coefficient of gradient condition, regional energy value, and change rate of strength, overlapped with each other to a great extent, and it was hard to identify the defect types. For this purpose, D-S evidence theory was used to combine the five kinds of characteristic values together and form theevidence body to perform integrated probability judgment. In this way, the difficulty in identifying weld defect types, caused by overlapped magnetic memory characteristic values was solved and the identification accuracy was improved.In order to further realize quantitative assessment on weld defect levels, by reference to the evaluation criterion of X-ray detection, this paper introduced four kinds of entropy,namely singular spectrum entropy, power spectrum entropy, and characteristic spectrum entropy and relative spectrum entropy of wavelet space state, which can reflect the characteristics of magnetic memory signal at the defects, and established the magnetic memory quantitative model with 4-level weld defects. Through magnetic memory tests for detecting weld fatigue damage, the four kinds of entropy and the critical characteristic values of entropy bands, corresponding to different levels of weld defects, were extracted. Then the mathematical expectations of the entropy bands were obtained by means of mathematical statistics, and thus the nearness formula, based on the expectations of entropy bands, was proposed. Then, the basic probability assignment function was created by combining with D-S evidence theory and taking entropy nearness as the evidence body. Information fusion of basic probability was performed in accordance with D-S evidence theory’s evidence combination principle, and the identification results of weld defect levels were finally output according to D-S diagnostic rules. In this way, the approach of assessing weld defect levels was established based on the combination of entropy and D-S evidence theory, serving as a new idea and basis of methods for the quantitative assessment on weld defect levels in practical projects. |
