Research On Multi Source Heterogeneous Information System And Fusion State Recognition For Ultra-narrow Gap Welding

With the wide application of thick plate structure and high-strength materials in civil and national defense major equipment,the contradiction between output demand and economic benefits,production process and welding quality is becoming increasingly fierce.Ultra-narrow gap welding has become an important research direction of modern welding technology.However,the confinement arc is very sensitive in the deep narrow and time-varying groove,which is difficult to be stable and easy to produce the risk of arc climbing,resulting in the non fusion or critical fusion state of the bottom sidewall.Uniform and good fusion state is an important characterization of high-quality joint,but it is difficult to sense,and conventional post weld inspection methods also have a lot of shortcomings.The multi-source heterogeneous information in the whole welding process is rich in connotation,which is helpful to state recognition and feedback regulation.Therefore,based on its economic,non-contact and effective characteristics to identify the fusion state,it has a crucial and positive significance to promote the practical application of ultra-narrow gap welding.In this thesis,a multi-source heterogeneous information system for ultra-narrow gap welding is constructed,which combines the whole process information collection with production mechanism,aiming to provide reliable data support for knowledge acquisition,experimental research and application verification.Based on the deep analysis of the crucia signals which can effectively reflect the welding process and extract the relevant features.After the dimensionality reduction of the highdimensional feature vector,the fusion state recognition model of the bottom sidewall of ultra-narrow gap welding is established with the help of machine learning theory,so as to effectively distinguish the three typical fusion states.The main contents of this thesis are as follows.An ultra-narrow gap welding test platform is developed,a multi-source information acquisition system for welding process is designed and built,including hardware platform and software system for information acquisition.Multi-source process information such as welding voltage,arc voltage,welding current and arc sound can be accurately,real-time and synchronously obtained,which provides datadriven support for subsequent analysis.Based on Lab VIEW and Access software,combined with Lab SQL toolkit and FX3 U PLC programming port communication protocol,a set of ultra-narrow gap welding multi-source heterogeneous information system software is developed,which integrates the functions of identity verification,parameter recording,parameter preset,information monitoring interaction,acquisition and storage,waveform echo,etc.On the basis of welding test,the whole process information is collected and recorded.A classification reconstruction filtering method based on VMD,SavitzkyGolay filtering and wavelet threshold denoising is proposed to retain the effective information to the maximum extent.Through in-depth analysis of electric signal and arc acoustic signal,combined with their generation mechanism,the intelligent identification behavior of welders is simulated,the multi-source heterogeneous information is processed,and the effective features are extracted to construct highdimensional joint feature vector.Based on PSO optimized LSSVM,the fusion state recognition model of ultra-narrow gap welding bottom sidewall was established.The low dimensional principal component of the feature vector reduced by KPCA is used as the input of the recognition model.The experimental verification of thick steel plate and rail shows that the identification model has good accuracy and generalization ability,and can realize the high-precision identification of three kinds of fusion state.It provides a new basis for intelligent decision-making such as parameter adjustment,abnormal condition diagnosis and self-healing control,also lays a certain foundation for realizing intelligent online prediction and control of ultra-narrow gap welding quality.