| In this paper,the shape of single-pass bead-on-plate weld and the surface smoothness of single layer multi-pass bead-on-plate weld with 1.2mm CN645 ACW wire were taken as the research objects,and the optimization method of welding process parameters applied in high efficiency and high quality automatic welding production were studied.Firstly,based on the large self-developed prototype equipment of arc fuse augmentation,single-pass bead-on-plate welding experiments of CN645 ACW welding wire were completed.The shape of welds(weld width,weld height)under different welding parameters were collected,and the BP neural network of single-pass weld shape was established.The generalization ability of this model was used to expand the data of weld shape.Based on the expanded weld shape data,the effects of different welding parameters on the weld shape were analyzed,and the process parameters of single layer multi-pass bead-on-plate welding experiments were obtained by inverse solution.Based on this,single layer multi-pass bead-on-plate welding experiments were carried out.The surface smoothness of multi-pass welds under different weld shapes and weld spacing were collected,and the BP neural network training of the surface roughness of weld was carried out.The experimental materials,equipment and conditions were identical with those of single-pass bead-on-plate welding experiments.According to the BP neural network of multi-pass welds surface smoothness,the influence of weld shapes and weld spacing on weld surface smoothness were analyzed,and the response surface model of weld surface smoothness was established to obtain the global optimal solution.Finally,combined with the improved mountain climbing algorithm,the optimization model of process parameters was built,and the optimization software of process parameters was compiled.The main contents and conclusions of this paper are as follows:(1)64 groups of single-pass bead-on-plate welding experiments with welding voltage of 24V-30 V,wire feeding speed of 7000mm/min-13000mm/min and welding speed of 500mm/min-1100mm/min were completed.The data of weld shape were collected.The variation range of weld width is 3.47mm-10.53 mm,and the variation range of weld height is 1.25mm-4.06 mm.(2)According to results of single-pass bead-on-plate welding experiments,the BP neural networks with different hidden layer number and node number were trained,and the BP neural networks with 2 layers and 16 nodes was determined.The influence of different welding parameters on weld shape was analyzed.The width of weld increases with the increase of welding voltage,the height of weld decreases with the increase of welding voltage.The width and height of weld increase with the increase of wire feeding speed.The width and height of weld decrease with the increase of welding speed.(3)Single layer multi-pass bead-on-plate welding experiments with weld width ranging from 6mm to 10 mm,weld height ranging from 2.5mm to 3.5mm and weld spacing ranging from 0.6 to 0.9 times of weld width were completed.The BP neural network for multi-pass welds surface smoothness with 1 layer and 9 nodes was determined.The influence of different factors on the surface smoothness of weld was analyzed.The surface smoothness increases with the increase of weld width,weld height and weld spacing.The response surface model of weld surface smoothness was established,and the optimal welding parameters with welding voltage of 30.3V,wire feeding speed of 12600 mm/min,welding speed of 950 mm/min and weld spacing of 5.35 mm were obtained.(4)The advantages and disadvantages of climbing algorithm were analyzed,and the improved climbing algorithm was improved according to actual needs.By combining the improved climbing algorithm and the previous two BP neural networks,the optimization model and software of process parameters were constructed.The software includes two functions: the prediction of weld shape and the optimization of process parameters. |
PDF Full Text Request