Research On Weld Defect Detection Technology Of LPG Spherical Tank Based On Ultrasonic Phased Array

Liquefied petroleum gas(LPG)is widely used in the chemical industry,and the quality of LPG spherical tanks is very important to production safety and personnel safety.The weld seam of the spherical tank is squeezed by the internal high-pressure gas,and the stress is relatively high.At the same time,it will suffer from the corrosion of H2 S and other gases in the natural gas,resulting in defects in the welding seam,which seriously affects production and life.Usually,when inspecting weld defects in spherical tanks,scaffolding needs to be set up inside,then the welds are polished,and finally the inspectors hold the inspection equipment to inspect the welds.After the inspection is completed,the inspector needs to check the inspection result and determine whether it is a defect based on experience.It also needs to qualitatively determine the defect.The defect determination is affected by experience,and it will be artificially missed and misjudged.On the basis of relevant domestic and foreign researches,this thesis takes the realization of LPG spherical tank weld defect recognition and classification as the design goal,and proposes a defect detection technology based on ultrasonic phased array.The thesis focuses on the study of key technologies such as feature energy extraction of defect echo signals and defect recognition based on neural networks.The main research contents are as follows:First of all,to solve the problem that the defect echo signal is difficult to characterize,a feature energy extraction method based on wavelet packet transform is proposed.The Q345 R steel welding test plate was made experimentally,and five defects of porosity,incomplete penetration,cracks,slag inclusion,and lack of fusion were prefabricated in the weld;according to the A-scan signals of the five defects obtained,the db18 wavelet base was selected as the optimal wavelet Based on the signal wavelet packet decomposition,the defect energy features are extracted.The five types of defects have little difference in high-frequency feature energy ratios,but their low-band energy values are quite different.According to the results,for the pore(3,0)node,the energy ratio is higher than 86%,for the crack defect(3,0)node,the energy ratio is about 79%,and the slag defect(3,0)node,the energy ratio of the node is about 62%,and the energy ratio of the non-fusion defect and the non-penetration defect node is not much different.Thirdly,a neural network model is designed for the recognition and classification of defects.The characteristic energy ratios of the five defects are used as the input parameters of the neural network.240 sets of data are used for network training,and there are 60 sets of test data.The final BP neural network identification and classification accuracy rate is 86.67%;then the deepneural network is used for identification and classification,Its classification accuracy reaches90%,and its accuracy rate is high,which can meet the requirements of practical applications.Finally,the development of the phased array detection system is completed.Ultrasonic phased array equipment selects CTS-PA22 S phased array module,based on the.NET platform,uses C++ language to carry out the secondary development of phased array software to improve the function of the detection system.According to research and analysis,ultrasonic phased array inspection can accurately detect weld defects of LPG spherical tanks.Using d B18 wavelet and deep neural network technology,five types of defects can be accurately extracted and identified and classified to achieve the expected results.LPG spherical tank weld defects can be accurately identified and classified.