Recent new finding of friction stir welding(FSW)brings a new technology for welding with physically stirring the metals in their solid states and avoiding the common fusion welding defects to produce high properties and good shape welding structures.This method has been widely used in our daily life especially in aerospace,high speed rail and other industries.However,the FSW joints often suffer from voids that affect the mechanical properties of the joints.It is necessary to select proper welding parameters,work-plate alloys and dimensions,and tool materials and geometries before welding in order to achieve void-free joints.However,even after using the proper welding conditions,voids may also form due to process instabilities and other factors.Therefore,it is essential to repair those voids using repair friction stir welding(RFSW)with adjusting welding parameters.RFSW may result in accumulation of high residual stress and distortion.Therefore,it is important to find out the proper conditions for RFSW.In this research,for the first time,both the three-dimensional numerical model and machine learning methods are employed to analyze the conditions of void formation.Two wellknown machine learning algorithms and three types of input data sets are investigated on the conditions of void formation.One hundred and eight sets of independent experimental data on void formation for the friction stir welding of three aluminum alloys,AA2024,AA2219,and AA6061,are analyzed.In this research,we combined experimental tests and simulations to find the residual stresses evolution of FSW and RFSW process.Three different methods are used to measure the residual stresses,X-ray diffraction method,the indentation strain-gauge method and the blind-hole method.Both a three-dimensional transient numerical model and a Abaqus-FEA thermomechanical model are employed to investigate the influence of RFSW process on residual stresses and distortion.The neural network-based analysis with welding parameters,specimen and tool geometries,and material properties as input predicts void formation with 83.3% accuracy.When the potential causative variables,i.e.,temperature,strain rate,torque,and maximum shear stress on the tool pin are computed from an analytical model of friction stir welding,90.0%and 93.3% accuracies of prediction are obtained using the decision tree and the neural network,respectively.When the same causative variables are computed from a rigorous threedimensional numerical model,both neural network and decision tree predict void formation with 96.6% accuracy.Among these four causative variables,the temperature and maximum shear stress show the maximum influence on void formation.The research results show that the evolution of residual stresses primary depends on the temperature field.In the FSW process,the region near the tool pin has high temperature but low residual stress,while the weld joint has low temperature and high residual stress due to the cooling down process of the material.The distribution of x-direction residual stress shows “M”shape.The residual stresses along both x and y directions are about 80 MPa.There is accumulation for both temperature and residual stress in RFSW process.However,after the work plates cooling down,the abovementioned accumulation influence disappears.With the pass number of RFSW increasing,there is no change for the x-direction residual stress,while the region for high residual stress along y-direction is increased. |