The Prediction Of Clamping Deformation And The Optimization Method Of Fixture Layout For The Thin-wall Workpiece

With the development of the aerospace industry and increasing demands foraircraft, thin-wall part has become popular in recent years because it’s light weightand some other advantages. However, due to its relatively complex structure, weakstiffness and high precision requirements, it’s always easily deformed in differentdegrees when affected by the clamping force in the actual clamping process, whichmaking the machined workpiece difficult to meet the specified quality requirements.Therefore, the effective control of the clamping deformation of thin-wall part plays animportant role in our country’s aerospace industry.In the clamping process of multiple components, the deformation of thin-wallpart is not same with different layout parameters which include clamping order,clamping force, layout of locators and so on. The clamping deformation rule affectedby single layout parameters can be obtained by the finite element method, but ifconsidering the influence of multiple layout parameters, the relationship betweenlayout and deformation will not be revealed just through FEM. For this reason, aclamping deformation prediction model that based on BP neural network has beenestablished in this paper, and the layout parameters have been optimized as well, andbuild the custom-made application program for fast milling simulation platform ofbox type thin-wall part. The main research contents are as follows:1. A workpiece-fixture system was built, and the workpiece clamped by multiplecomponents was analyzed in detail to determine the friction cone constrains andunilateral contact constraints. Then, the objective function was gained by use of theminimum excess principle. The clamping deformation of thin-wall part underdifferent clamping force, location and clamping sequence was calculated throughmodel that built by using finite element software.2. The BP network was researched to decide the number of neurons betweeneach layer and the transfer function that connecting each layer, so that the predictionmodel of clamping deformation can be built. Then, the clamping deformations underdifferent layout parameters were selected as training samples of BP neural network,and the comparation between predicted value and simulation results has been made. The outcome shows that the prediction error is no more than3%.3. The optimal model based on the clamping layout plan of which the objective isminimizing the maximum clamping deformation was made and its genetic algorithmsolution techniques have been established, to optimize the clamping layout parametersof thin-wall parts. The best parameters were obtained, and they were compared withcorresponding simulated deformations to demonstrate the correctness of optimizedresults. Not only it improves the computational efficiency of clamping deformation,but also provides a basic theory for the rational design of clamping layout plan ofthin-wall parts.4. The custom-made application program for fast milling simulation platform ofbox type thin-wall parts was developed through Python language. The correspondingGUI was built by writing a script that can call library functions included in ABAQUS.The modeling required keywords were defined so that the modeling for milling can beparameterized. At the same time, the post processor for simulation results wasdeveloped, which can quick extract the needed result date.