Distortion Prediction and NURBS Based Geometry Compensation for Reducing Part Errors in Additive Manufacturing

Abstract

Additive Manufacturing (AM) is a process in which a part is typically fabricated by depositing materials layer by layer. Powder Bed Fusion Additive Manufacturing Process (PBFAM) is one type of AM process which utilizes a high energy laser source to heat and fuse metal powders into a solid part. The constant cycle of heating and cooling in each layer causes thermal deformation associated with residual stresses that reduce the geometric accuracy of the build. To remedy this problem, a compensation algorithm is presented in this paper which modifies the native NURBS CAD geometry of the part to counteract the thermal distortion. An inherent strain-based fast distortion prediction model is used to predict the thermal distortion of the part. The resulting distorted FEA nodes are used to construct a NURBS based compensated geometry using nonlinear least square fitting approach using the original NURBS parameters. Compensating the native NURBS geometry of the model provides more accuracy for the part build rather than compensating STL models. Validation of the algorithm is performed using two case studies by comparing the thermal deformation of pre and post compensated NURBS geometries. The accuracy and robustness of the algorithm for achieving geometric tolerances are further assessed by comparing the flatness and cylindricity tolerances values of the part feature from the two case studies.