Study On The Shrinkage During Selective Laser Sintering Based On The Data Mining

Selective laser sintering (SLS) is a three-dimensional forming rapid prototypingmethod by sintering the powder material layer by layer with laser. As the technologydevelops, the selective laser sintering process has become more mature and specific.Inorder to promote a broader application of the technology in the high-precisionmanufacturing such as aerospace, the accuracy of this technology becomes an importantresearch focus; On the other hand,with the more and more extensive applications ofcomputer technology in the field of manufacturing, how to combine the traditional craftmethods with computer technology to achieve efficient and accurate accuracy controllingand realize the intelligent systematic research on process is also an important research area.A systematic study on the factors that affects the accuracy of selective laser sintering iscarried out by being combined with the partial results of the data mining field. Byrespectively using the artificial neural networks(ANN) and the support vectorregression(SVR),the complex non-liner relationship between the selective laser sinteringprocess and the shrinkage is established with these two data mining methods and realizesthe efficient&accurate analysis and prediction on the shrinkage.By combining theestablished non-liner model with the sintering theory,a study on the process is also taken.The factors affect forming accuracy and the corresponding solutions are researchedfrom the aspects of the preliminary data processing,forming process and thepost-processing. Preliminary data processing error mainly includes the fitting error of theSTL file and the staircase effect size error results from slicing.The equipment positioningaccuracy during the forming process, such as the motion accuracy of the powder spreadingsystem, the positioning accuracy of the laser scanning system and the temperaturecontrolling accuracy of the heating system also has an important effect on the accuracy oreven affects the formability.The way the process parameters such as preheatingtemperature, laser power, scanning speed, laser radius compensation parameters, scan mode affect the forming accuracy is also discussed. Finally an introduction to thepost-processing materials, parameters that may affect accuracy is given.An improved BP neural network is established to establish the non-linear modelbetween the selective laser sintering process parameters and the accuracy. Orthogonalexperimental design method is utilized to design the training data with the purpose ofreducing the data redundancy. By referring to the existing neural network structuraldesigning experience, the optimized structure parameters are chosen through multiplecomparison experiments.The prediction results show that the established model hasexcellent generalization performance with an average prediction error of4.53%.Due to the lack of a unified theory guiding the design of the neural network, and theempirical model is built on the basis of empirical risk minimization, the generalizationcapability of neural network can’t be guaranteed. Another data mining method namedsupport vector regression,which is based on the statistical learning theory, is used toestablish the nonlinear modeling between process parameters and shrinkage.Consideringthat the support vector regression involves the optimization of the structure parameters,the uniform design initialized particles swarm optimization algorithm is adapted. Theeventually established model has an average prediction error of3.95%,reflecting theaccurate relationship between the process parameters and shrinkage.Finally,the established support vector regression model is used to investigate theselective laser sintering process.Analysis of how each factor as well as the interaction ofdifferent factors affect the shrinkage is performed.The quick and easy model greatlysimplifies the research on the process.By combining the predictive ability of the modeland the optimization algorithm, the intelligent selection of process parameters based onthe actual need of forming is aviable,which supports the intelligent application andpromotes the use of selective laser sintering.