SHRINKAGE PREDICTION USING MACHINE LEARNING FOR ADDITIVELY MANUFACTURED CERAMIC AND METALLIC COMPONENTS FOR GAS TURBINE APPLICATIONS

Additive manufacturing of metallic and ceramic components can provide several benefits for gas turbines. This technology can help to decrease time and cost required for prototyping of both operational components and assistive tooling equipment. There are currently many methodologies for metallic and ceramic material printing, and these processes typically involve a sintering phase during the manufacturing procedure. The amount of shrinkage that occurs is dependent upon the sintering oven conditions, the printed material, and the geometry of the printed component. These factors can be broken down into many subfactors leading to an equation that is too complex and with too many variables to be solved in an analytical matter. Additive manufacturing can provide the ability to rapidly manufacture geometrically specific components for turbomachinery operations. To increase the accuracy of the geometric dimensions of the final product, the sintering shrinkage must be accurately predicted. Machine learning can assist by using data-driven approaches to ensure accurate prediction of shrinkage. This ultimately will increase the dimensional accuracy of the printed components. If there is an accurate shrinkage prediction formula that accounts for the material type, sintering oven conditions, and geometric specifications, the components can be scaled up in CAD (computer aided design) software prior to 3D printing and entering the sintering oven. This predictive capability will allow the end user to create dimensionally accurate parts at a rate that has not yet been possible. In this work, a machine learning approach is developed for the prediction of shrinkage during sintering of additive manufactured components. This paper uses a combination of experimental data and artificially generated data to create a framework for implementing linear regression tools to predict sintering trajectories. An accurate knowledge of the sintering trajectory for a given material can allow for more user control over the final properties and geometrical accuracy of a given component.