Cu3d-ML

In the context of ecological and economic resource utilization, the use of copper materials is becoming increasingly important due to their high recyclability. The use of additive manufacturing processes can additionally reduce lifecycle costs and the CO2 balance. An existing obstacle to the use of additive manufacturing processes in the industrial environment is the reproducible achievement of the required component quality, which results in high scrap rates.

The aim of the Cu3D-ML research project is to develop a quality control system in the additive processing of copper materials using the example of cold gas spraying. The quality control comprises the near-real-time prediction of material and component-related quality properties (e.g. electrical and thermal conductivity, hardness, grain size, microstructure composition) with the aid of an IoT infrastructure. The quality is recorded during the running process by applying machine learning methods to sensor-based process data (e.g. process gas pressure, temperature, etc.). Due to the near-real-time prediction, countermeasures in terms of process intervention or premature process termination can be taken when irregularities occur, thus preventing or minimizing scrap.

The research project is scheduled to run for 24 months and is being carried out in cooperation with the Chair of Materials Technology (LWT) and in close exchange with the members of the PA. In order to meet the quality requirements for additively manufactured components made of copper materials, a novel multivariate approach is being developed with consideration of all relevant influences on quality-related material and component properties. The properties of a component (e.g. electrical and thermal conductivity, hardness) are significantly determined by the material composition and the material state, which in additive manufacturing processes are influenced by a large number of process variables (e.g. laser power, scanning speed, residual oxygen content). These process variables are to be recorded by sensors or read out from machines. Machine learning methods will then be used to draw conclusions about the properties of additively manufactured components made of copper materials in real time on the basis of this data. In this way, if necessary, measures can be derived via process interventions that lead to compliance with the required quality requirements.

The project "Cu3d-ML" (grant number: 22076) of the Bundesvereinigung Stifterverband Metalle e.V. - Wallstr. 58/59, 10179 Berlin - is funded by the AiF within the framework of the program for the promotion of joint industrial research and development (IGF).