Intelligence Analysis And Prediction Of Joints Quality For Laser Welding Of Automotive High Strength Steels

Laser welding of high-strength steel is one of the important ways to realize lighter weight of automobiles.The application of high-strength steel can reduce the thickness of car body steel plates,reduce the body quality,and ensure the strength performance of the car body to achieve energy-saving and emission reduction effects.In this thesis,the laser welding of dual-phase steel DP600 is taken as the research object,and the performance of the joint can only be predicted as the research objective.The artificial neural network and cluster analysis method are used to study the correlation between process parameters and formability and joint performance to achieve welding.The size of the seam area and the intelligent prediction of the performance of the welded joints provide intelligent analysis methods for the process parameters of the dual-phase steel DP600 laser welding for optimum joint performance.The main conclusions of this paper are as follows:Using BP neural network to establish the nonlinear prediction model based on laser welding power,welding speed and defocus amount as input variables,and using the width of the upper and lower surfaces of the weld zone and the width of the narrowest part as output variables,the maximum relative error of the verification result is 8.94.%.According to the model fitting welding process parameters and the weld zone morphology,it was found that the overall width of the weld increases with the increase of the defocus amount;decreases with the increase of the welding speed;with the increase of the laser power,the width of the upper surface changes less,the most Narrow and lower surface widths gradually increase.Using the width data of the upper and lower surfaces and the narrowest part of the weld zone as three variables,the K-means algorithm was used to divide the data into 4 categories,and a classification rule set was established.The width of the upper surface was between 1.06 mm and 1.19 mm and the narrowest The width at 0.74 mm~0.91 mm can be divided into the first type,the width of the upper surface is 0.84 mm~1.06 mm and the width of the narrowest part is 0.62 mm~0.74 mm and the width of the bottom surface is 0.65 mm~0.81 mm In the class,the width of the upper surface is in the range of 0.79 mm to 0.91 mm and the width in the narrowest portion is in the range of 0.74 mm to 0.85 mm.It can be divided into the third category.The width at the narrowest part is 0.48 mm to 0.67 mm and the width of the lower surface is 0 mm to 0.65 mm.Points to the fourth category.Performance evaluation of various types of welds shows that the second type of weld has a better profile and the performance is closest to the parent material;the first type of weld zone has a larger width and poorer elongation at break;the third type of weld and There is a dent in the base metal joint;the fourth type of heat input is small,and the weld seam is in a completely welded state.Combining artificial neural network and cluster analysis,a neural network-cluster analysis model was established based on laser welding process parameters as input variables,with weld types and properties as output variables,and welding experiment verification was carried out.The measured topography and weld performance of the weld zone are consistent with the prediction results,which verifies the rationality of the model prediction.