Quantitative Study Of Metal Magnetic Memory Testing Technology In Welding Defect Detection

Contains of typical petrochemical industry equipment is mostly composed of weldingstructure, and welding crack is one of the most dangerous defects in welding structure.Using nondestructive testing (NDT) technology of welding crack detection can avoidserious accident. So far, metal magnetic memory (MMM) testing technology is the onlyone which can make early detection of metal in NDT, it can realize the prediction of theaccident.In order to solve the problem of MMM technology which can not analyze thewelding defect quantitatively, this paper makes a thorough research on the relationshipbetween MMM signal and stress, the influence of the welding crack size on MMM signaland its quantitative relations, the application of the neural network in MMM testing.Which is based on the principle of MMM testing, and combine with the modern signalprocessing and neural network technology.This paper extracted the MMM signal on the surface of the specimens under differentstress state, and found that MMM signal can show the stress state effectively; This articlealso extracted the characteristic value of MMM signal, and found that the characteristicvalue can show the stress state of the specimens; It realized the identification of specimenstress state using BP neural network preliminary.This paper proposed a new way to analyze the welding crack quantitatively, which isthe combination of testing directions. Two testing route of Q345R welding plate wastested, one was along the weld direction and the other one was perpendicular to the welddirection. The effects of welding crack size (including crack length and crack depth) onMMM signal was studied. Then the quantitative relationship between welding crack sizeand MMM signal was also been established. The researches show that the MMM signal oftwo testing route both have obvious welding crack location features, but one singledirection signal can not reflect all the size information, the signal of two testing route mustbe combined.In addition, BP neural network method was studied in the quantitative study, theresults show that BP neural network can realize the quantitative evaluation of welding crack size.