Research On Arc Dynamic Behaviors And Penetration Recognition In K-TIG Welding Based On Multi-source Information

Keyhole tungsten inert gas(K-TIG)welding,characterized by the backside keyhole during the deep penetration welding process,has capacity of one-side welded with dual-side formation on welding the middle thick plates without preparing the grooves and filling the wires.A free burning arc with high stiffness and penetration ability is obtained through using higher welding current and enhancing tungsten cooling on the basis of traditional TIG welding.Thus,the K-TIG welding has the advantages of high efficiency,good welding quality,and clean process.So far,the research direction of the K-TIG welding is to intelligentize the welding process and optimize the welding technique.To intelligentize the welding process is to track the weld seam and control the weld penetration based on the multi-source information in the welding process.To optimize the welding technique is to reduce the penetrated threshold welding current and widen the penetrated process window by enhancing the penetration ability of the arc through exerting the magnetic field,etc.However,on one hand,most of the researchers focused on tracking the weld seam and obtaining a good result,while the accuracy of recognizing the penetration states were still low.The source of information was single,and the in-depth exploration of multi-source information in the welding process was not sufficient.On the other hand,the research on dynamic behaviors of arc-weld pool-keyhole system remained blank in the research direction of process optimization.The arc pressure,which is the key factor to achieve the deep penetration welding,has not been accurately measured and in-depth analyzed.Therefore,the research object of this paper is the deep penetration process of K-TIG welding.The penetration process and the dynamic behaviors of the arc-weld pool-keyhole system are analyzed by improving the experimental techniques and measurement devices.The inner relation between the arc pressure and welding parameters is revealed.The different penetration states are accurately recognized through in-depth analysis of the multi-source information in the welding process.Firstly,a K-TIG welding system embedded with the multi-source information sensing system is established,which is able to collect the arc sound information,the arc electrical information,the vision information,and the arc pressure information in the real-time welding process.The sandwich experiment method is introduced to explore the real-time penetration process.The high temperature resistant arc pressure measurement device is invented,which is capable of measuring the K-TIG arc pressure accurately in the standard welding conditions.Then,according to the physics connotation of the multi-source information during the welding process,the signal preprocessing and characteristics extraction methods are proposed to improve the signal-to-noise ratio(SNR)and represent the physics properties of the welding process,respectively.Based on the established sandwich experimental system,it is found that the full length of the keyhole channel increases slowly during the penetrating process and it becomes stable after the plates penetrated,while the shape of the keyhole channel varies with the realtime welding condition.Based on the arc pressure measurement system,it is found that the arc pressure is approximately normally distributed.When an axial magnetic field(AMF)is exerted,the arc shape appears like a gong,with the top of the arc constricted to the torch while the bottom spreads in the radial direction.The inner relationships between the arc pressure and the welding current,AMF strength,and the arc pressure effective distribution diameter are revealed.Next,during the research on penetration process and arc penetration ability of the K-TIG welding,a single particle description model is proposed to analyze the dynamic arc behaviors.It is found that the arc emitting area decreases with the exertion of the AMF,which leads to an increase in the arc force.According to the proposed single particle description model,the energy carried by the charged particles decreases when the AMF strength becomes too large,resulting in the low arc pressure and weak penetration ability.Finally,the penetration recognition method is studied based on the multi-source information during the welding process.It is found that the feature frequency of the arc voltage signal has a clear physical meaning,related to the weld pool oscillation frequency.The application of the feature frequency on recognizing the penetration states is clarified.An ECOC-SVMGSCV penetration recognition model based on the arc sound signal and the arc voltage signal is designed.The feature evaluation method based on Spearman monotonic correlation and permutation importance is proposed.It is found that the arc voltage signal is more important than the arc sound signal with respect to the task of penetration classification.However,the arc sound signal and the arc voltage signal have a certain complementarity.The proposed model has a high accuracy of 99.50% for the recognition of different penetration states.