Study of geometrical defects of free-form surface machined using neural network

Abstract

The manufacture of total hip arthroplasty (THA) requires the control of the quality of free form surfaces. In fact, the polyethylene insert is deformed to fit the overall geometry of the femoral part, which has an impact on the quality of the contact. In this paper, we propose a method for evaluating the defects of complex forms. The originality of the approach is the use of artificial intelligence to position the cloud of measured points, obtained with a three-dimensional measuring machine equipped with a contactless sensor, with regard to the 3D CAD model of the THA. The artificial intelligence algorithm used is based on neural networks that are trained using a virtual positioning realized with 3D CAD software. Finally, the difference between the positioned point cloud and the CAD model allows us to evaluate the shape defect of the measured THA surface. We found that the error of the proposed method is at the vicinity of micron scale.