Study On The Fusion Detection Method Of Steel Welding Seam Defects Based On Ultrasound Communication Characteristics

With the widespread application of welding components in the field of mechanical manufacturing industry,the impact of welding defects on welding component performance is increasingly valued.Therefore,the detection requirements of welds have also been continuously improved.Among them,manual ultrasound testing is the main detection method of welding parts.However,traditional ultrasound detection identification and welding defects have problems such as low efficiency of manual classification and inaccurate classification of defect types,and the detection methods of different ultrasound probes have the characteristics of tendency and uncertainty.This article proposes a method of fusion detection method based on the fusion of poly source decision-making based on ultrasonic communication characteristics.The main research content is as follows:(1)Using COMSOL to analyze the spread of ultrasonic waves in the defect medium,establish a simulation model,and set the parameters of the ultrasonic detection system.Through the analysis of the simulation results,it is found that the deficiency waves of 5types of defects have different characteristics in terms of waveform amplitude,reflection strength,and reflection form.It provides the theoretical basis for collecting welded defect waveforms,ensuring the objectivity of the data set.Provide possibilities for realizing intelligent identification.(2)In order to realize the automation of welded ultrasound recognition and image classification,combined with multi-scale structure and attention mechanism,a improved resnet18 network model(Multi-SCALE-SE-Res Net18)was proposed.Add the Inception module to the residual network model,and use different standard convolution nuclear structure to extract defective waveform image characteristics to increase the diversity and richness of the characteristics.At the same time,the attention mechanism is added to optimize the right value.This method makes full use of the characteristic information of the defective waveform,which effectively improves the recognition rate of defects.(3)Considering that the defect information obtained by a single probe is increasing and one-sided,it proposes a method of combining cloud models and D-S evidence theory to solve the problem of strong subjectivity of basic probability value matrix in the traditional D-S evidence theory.Different defects extracted from multiple probes are expressed one by one through the number of cloud ginseng,and the defect information is combined to optimize the defect information in the theoretical model of the D-S evidence and obtains more effective information from it.Finally Fusion,overcome the negative effects caused by random factors,and improve the reliability and accuracy of identification on defect type.This article realizes the automatic classification of welded defect ultrasound images,and provides a new research idea for the decision-making method of multi-probe recognition defective types,and finally achieves a good classification results and providing assistance for the development of artificial intelligence in non-destructive testing.