Numerical Simulation And Control Of Welding Deformation Of Flange Plate Based On Argon Arc Welding

Welding is a common way of connecting engineering materials,which is widely used in aerospace,shipbuilding,transportation,pressure vessels,precision instruments and other industries.The welding process is accompanied by a large amount of heat input,resulting in an uneven high heat and pressure environment in and around the weld,along with a large temperature gradient.Under the condition of transient heat input,the materials expand by heat and is restricted by the low temperature area.When the internal stress is greater than the yield point,the material produces plastic deformation.After material cooling,residual plastic deformation will bring residual stress.The plastic strain and residual stress not only adversely affect the stiffness,strength,shape accuracy and installation accuracy of the material components,but also reduce the flatness and corrosion resistance of the welded joints.So it is of great significance to reduce the stress and deformation of welding.The welding method adopted in this paper is tungsten argon arc welding,and the research object is the flange ring vacuum equipment.In the assembly process,the installation accuracy is required to be high,so the research is mainly carried out on the welding deformation control problem.In addition to the elaboration of the relevant theoretical basis,the research content mainly includes four parts:(1)This chapter explores the applicability of the heat source model for argon tungsten arc welding to provide an accurate welding temperature field distribution model for subsequent research.ABAQUS was used as the platform of welding finite element simulation to simulate the process of welding filler adding.It designs temperature patrol test to obtain the temperature curves of the sampling points,and then compares with the temperature simulation results of the rotating body heat source model,uniform body heat source model and double ellipsoid heat source model.It also compares the size of weld pool and its heat affected area under test and simulation of three heat source models.(2)In order to solve the problem of time consuming and force consuming in the finite element analysis process of large welding parts,the inherent strain method was used to quickly predict the welding deformation.It studies the welding sequence of two welds of flange ring and choose the optimal welding sequence.Then it obtains the transverse and longitudinal inherent deformations by thermal elastic-plastic calculation of local characteristic structures,and converts the inherent strains into temperature loads for elastic analysis to generate the deformation results in one time.Comparing the results of inherent strain method with those of thermal elastic-plastic finite element method,it verifies that the inherent strain method can greatly reduce the calculation time and ensure the accuracy of calculation in the welding simulation of large welding parts.(3)External constraint has a great influence on welding deformation and is one of the important factors affecting welding deformation.Compared with the traditional rigid constraint,dynamic constraint can reflect the effect of external constraint force in the welding process in real time.In the case of multi-layer and multi-pass cross welding of two welds,it establishes a multi-body coupling analysis model,and simulates the dynamic constraint force effect caused by the deformation of the low-stiffness fixture due to the deformation of the welds by using the spring element.It considered the feedback of the deformation to the constraint force.The deformation of welded parts were discussed from four factors: constraint stiffness,constraint distance,initial constraint force and constraint width.(4)It can predict the welding deformation prediction of arbitrary dynamic constraint factors quickly based on the RBF neural network method.The dynamic constraint analysis samples mentioned above are used as training samples,the constraint stiffness,constraint distance,initial constraint force and constraint width are input parameters,and the component deformation is output.It calculates the center and width of hidden layer nodes by K-means clustering,and calculates the connection weight coefficient of hidden layer and output layer by violation method.The neural network structure is turned into a "white box",and the mapping relationship between the dynamic constraint factors and welding deformation is described by mathematical analytic expression.It further improves the speed and randomness of welding deformation prediction.