Microstructures And Mechanical Properties Of Friction Stir Welded Marine Aluminum Alloy And Welding Data Mining

Marine equipment manufacturing is the key to China’s coastal defense industry,and lightweight and large-scale manufacturing has become the mainstream trend in ship manufacturing.Therefore,aluminum alloy is widely used in the manufacture of ships and offshore landing crafts due to its low density,high specific strength and good low temperature performance.As a solid-state welding process,low heat input of Friction Stir Welding(FSW)can avoid welding defect often caused by fusion welding.In recent years,FSW has developed rapidly in aluminum alloy shipbuilding.However,there are problems with aluminum alloy robot FSW,such as long welding process development cycle,high trial and error cost,and low efficiency.Therefore,this article conducts research on robot FSW process and joint organization performance,a machine learning model based on BP neural network was established to achieve prediction of joint mechanical properties and artificial intelligence optimization of welding process.For the 3mm thick 2195-T8 Al-Li alloy,process experiments were carried out by the single-factor test and the effects of the robotic FSW process on the mechanical properties and weld formation were studied.The results showed that the tensile strength of the joint increased and then decreased with the rotational speed increased from 1200 ～ 2000 rpm.Tensile strength decreased with increasing welding speed within 100 ～ 250mm/min.Tensile strength increased and then decreased with increasing welding pressure within 2 ～ 3.5k N.Microhardness distribution curves are mainly "W" and "U" shaped,the location of fracture shifts from SZ to BM and the fracture behavior changes from brittle fracture to toughness fracture.The influence of joint microstructure was studied.The results show that at a process parameter of 1800rpm-100mm/min-3k N,the grains in the BM are slatted,the average grain size is 11.54μm,the percentage of large angle grain boundaries is 75.8%.TMAZ is affected by heat input and plastic deformation,comprehensively.Part of grain undergo dynamic recrystallisation with an average grain size of 7.36μm,the overall shape of the grain is elongated and oriented by stirring.SZ is subjected to the highest heat input,the morphology of the grain is equiaxed with an average size of 3.34 μm,the small angle grain boundaries account for 18.6%,the weave is dominated by shear weave B.Process data of 2195-T8 aluminum alloy robot was collected,the joint mechanical properties prediction model based on BP neural network was established,the intelligent optimization of FSW joint properties prediction and welding process were achieved,and obtained the best process window.The results show that the optimal process which obtained by model is 1810rpm-105mm/min-3k N,corresponding to a predicted tensile strength of 415 MPa,which reaches 82% of the BM,with an error of 3.71% compared to the experimental value.The significant degree of influence of the process parameters on the mechanical properties of the joints was obtained by the data mining.The results indicate that for yield strength,the welding speed has the greatest impact,followed by the rotational speed and welding pressure,respectively.For tensile strength,rotational speed has the greatest impact,followed by welding speed and welding pressure,respectively.For the elongation,the influence of rotational speed is the greatest.