| Air v-bending is an important sheet metal forming process,and the set bending angle is obtained by controlling the stroke of the punch by the bending machine.Because the plate spring-back after the mold is unloaded,the bending angle after the springback is the bending Angle required by the product.Therefore,the amount of spring-back is a process parameter that must be considered in the design of the Air v-bending process.The key of sheet metal free bending process is how to establish an accurate spring-back prediction model to determine the spring-back amount,and how to accurately compensate the spring-back amount through the punch stroke control,so as to reduce The Times of bending trial and achieve the required bending Angle precision accurately and quickly.Based on this,this article focuses on the spring-back problem of the Air v-bending process of sheet metal,through the combination of finite element simulation and bending experiment,analyzes the influencing factors of bending spring-back,and uses machine learning algorithms to establish bending spring-back prediction Model and bending trial bending correction model,is used to guide the free bending precision sheet metal rapid prototyping.The main research results are as follows:(1)Based on the ABAQUS finite element software platform,the free bending model is established,and the explicit and implicit combination method is used to complete the bending and spring-back process of the plate.Comprehensive consideration of finite element model parameter selection and experimental design methods,integrated into the ISIGHT platform to realize parametric modeling,and provide a large number of samples for the sensitivity analysis of rebound factors and the establishment of machine learning models.(2)Average impact value based on artificial neural network method for Air v-bending spring-back factors sensitivity analysis,select sensitive degree larger factors,provide the basis for subsequent model input parameters selection,and discuss the sensitive factors influence on the spring-back,concluded that the plate thickness,elastic modulus,hardening exponent,punch stroke is inversely proportional to the plate of spring-back amount;The hardening coefficient,yield strength and slot width of die are directly proportional to the spring-back of plate.(3)Using machine learning artificial neural network technology as the research means,the spring-back prediction model under different factors is established,and the genetic algorithm is introduced to optimize the initial weight and bias of the network,so as to avoid falling into local optimum in the network training.Subsequently,GA-BPNN rebound Angle prediction model and punch stroke prediction model under 13 factors and 7 factors are obtained.The verified model can meet the requirements within a certain accuracy.The GA-BPNN rebound Angle prediction model and punch stroke prediction model under 7factors are suitable for engineering applications.(4)Based on the GA-BPNN punch stroke prediction model under 7 factors,two research methods,artificial neural network and dimensional analysis method,are used to establish the punch stroke correction model respectively.After comparison,the neural network model can play a more accurate correction role.The GA-BPNN punch stroke prediction model and punch stroke correction model under 7 factors were tested by an application example.The results show that the error between the final forming Angle and the target Angle can be kept within 0.5° by using one prediction model and two correction models.Finally,based on Matlab platform,the interactive interface of Air v-bending spring-back prediction is designed to facilitate the use of the two models.By selecting different models and inputting the required parameter values,it can be used for punch stroke prediction and correction. |
PDF Full Text Request