Research On Modeling And Prediction Technology Of Wire Arc Additive Weld Bead Forming

With the gradual development of large-scale integration and complexity of metal structural parts required in aerospace,weapons and equipment,ships and other fields,the methods of equal and reduced materials used in traditional manufacturing gradually show the disadvantages of difficult processing,cumbersome process flow and high cost,etc.Wire arc additive manufacturing is a rapid prototyping method based on welding technology.The formed parts have the advantages of uniform microstructure,high density and reliable strength,which provides a new solution for the integrated forming of large and complex structures.However,the poor forming accuracy and low forming stability of wire arc additive restrict its further development.In order to reduce the phenomenon that the forming process cannot start normally due to the accumulated error of interlayer height,and realize the automatic and stable wire arc additive manufacturing by robot,this paper focuses on the research of weld bead forming modeling and prediction technology in arc additive.Firstly,considering the thermophysical properties of materials and the stress conditions such as surface tension,buoyancy and electromagnetic force,based on the VOF multiphase flow model,melting solidification model and energy model in the Fluent solver,the UDF port is redeveloped,the droplet transfer process parameters and source terms in the molten pool are compiled,and the three-dimensional transient numerical model of droplet transfer and molten pool flow is constructed.The three-dimensional morphology,temperature field and flow field evolution law of the molten pool and droplet transition in the single-channel wire arc additive forming are obtained through numerical simulation,and the heat and mass transfer characteristics in the molten pool are revealed,which provides a theoretical basis for arc additive forming.Then,the single-factor test method was used to analyze the influence of welding current,welding speed and wire feeding speed on the shape of the weld bead in the stable forming interval of the single bead.The general rotation combination scheme of quadratic regression is designed,and the single pass and multi-layer cladding test is carried out by using the noncontact laser on-machine measurement method,based on Design Expert fitting,a quadratic regression prediction model of process parameters,layer width and layer height was obtained.The model adaptability test was carried out by variance analysis and error calculation.The results showed that the error of the prediction model was within 6%,and the influence of the multi-parameter coupling on the weld bead morphology was obtained according to the response surface results.In order to further improve the prediction ability of the single-pass multi-layer prediction model of arc additive,a three-layer neural network structure with input process parameters and output weld bead layer width and layer height is designed based on the neural network toolbox in Matlab machine learning.After allocating training set and test set,it is used for nonlinear BP modeling prediction between wire arc additive process parameters and weld bead morphology.Aiming at the defect of large local deviation in the training and testing results of BP neural network,a genetic algorithm is added to optimize it in the learning process,and an optimized prediction model with high precision and high stability with a prediction error of less than 3%is obtained.Finally,based on the integrated test platform for wire arc addition and subtraction and onmachine measurement independently built by the research group,aiming at the best optimized neural network prediction model of weld bead morphology,two sets of typical thin straightwalled and two sets of inclined curved thin-walled structural parts were subjected to arc additive verification experiments.During the additive process,it was found that the straight-walled structural parts were formed very stably,while the inclined structural parts had a high local voltage.By calculating the relative error between the predicted value and the actual value,it is concluded that the forming dimensional errors of different types of thin straight-walled structural parts are within the accuracy range of the prediction model,while the inclined curved thin-walled structural parts are beyond the allowable error range of the prediction model.In order to explore the ability of wire arc additive forming,the additive test of the simplified design of the profile of a certain type of aviation blade was completed,and the internal non-destructive flaw detection of the thin-walled twisted blade was carried out.