Optimization of Additively Manufactured and Lattice-Structured Hip Implants Using the Linear Regression Algorithm from the Scikit-Learn Library

Abstract

As the name implies, patient-specific latticed hip implants vary in design depending on the properties required by the patient to serve as a valid suitable organ. Unit cells are typically built based on a 3D design of beams, and the properties of unit cells change depending on their geometries, which, in turn, are defined by two main parameters: beam length and beam thickness. Due to the continuous increase in the complexity of the unit cells’ designs and their reactions against different loads, the call for machine learning techniques is inevitable to help explore the parameters of the unit cells that can build lattice structures with specific desirable properties. In this study, a machine learning technique is used to predict the best defining parameters (length and thickness) to create a latticed design with a set of required properties (mainly porosity). The data (porosity, mass, and latticed area) from the properties of three unit-cell types, applied to the latticed part of a hip implant design, were collected based on the random length and thickness for three unit-cell types. Using the linear regression algorithm (a supervised machine learning method) from the scikit-learn library, a machine learning model was developed to predict the value of the porosity for the lattice structures based on the length and thickness as input data. The number of samples needed to generate an accurate result for each type of unit cell is also discussed.