Simulation And Precision Control Of Titanium Alloy Laser Bending Process

As one of advanced manufacturing technologies, laser bending, with the advantages of high flexibility and non-mold, has broad application prospects. This thesis has researched TC4 sheets laser bending with finite element simulation, machining experiments of sheet, neural network and particle swarm optimization:(1) Finite element model simulation of titanium alloy laser bending has been carried out based on ANSYS. According to the result research, at the time of laser heating, the upper surface of the sample is under compression stress because of the tensile stress, while the lower surface is under tensile stress. So the sample bends opposite the laser source. During the cooling phase, because of the decrease of temperature, the upper surface contracts, while the tensile stress of lower surface decreases. The sample bends to the laser source when the stress of upper surface is higher than the one of lower surface. By simulating the condition of five scans, the result shows that the peak temperature of each scan is same, so the cooling time is reasonable. The relationship between bending angle and scanning number is linear and the error between simulation results and actual machining is 0.6mm.(2) The experiments of different power, speed and scanning numbers based on existing fiber laser are used to search the process variables ’effect on bending angle. The research result indicates that under the condition that other technology parameters are invariable, the sample’s bending angle increases with the increase of scanning numbers and laser power. With the increase of the scanning speed and line energy density, at first, the bending angle increases, and then decreases. The biggest bending angle can be obtained with 6.7J/mm2 of line energy density when the numbers of scanning and other conditions are constant.(3) Radial basis function(RBF) neural network is developed to express the relationship between laser power, scanning velocity, scanning number and bending angle. The samples which were obtained in experiments of of TC4 sheets laser forming are used to train the neural network to get the accurate map relation between the process variables and bending angle.(4) The particle swarm optimization(PSO) algorithm is used to get the batter parameters of RBF network. When the optimized neural network is use to forecast the test samples, the error reduces to 1% which was 4.7% before optimizing. Based on the neural network, PSO algorithm is used to optimize the process variables. The experimental angles formed with optimized process variables are coincident well with the expected ones and the error is within 5%. The result shows that the method is practicable and effective.