AI Prediction using IR data for Feedback Control in NiTi Laser Powder Bed Fusion

Laser powder bed fusion (L-PBF) is an additive manufacturing technique commonly used in the biomedical and aerospace sectors. In the L-PBF process, a thin layer of powder is melted at specific targeted areas of a build area to form a 2D geometry, a new layer of powder is spread and micro-welded to the previous layer; the cycle repeats until a 3D completed part is formed.

Porosity defects can commonly occur in parts produced with L-PBF and have a significant impact on the component’s mechanical properties. Rapid cooling rates experienced during the process can also lead to high residual stresses, resulting in cracking and delamination during printing. Therefore, in-situ monitoring is often employed to monitor the process. The development of closed-loop feedback control systems for L-PBF is of great importance to actively prevent and correct these defects in real-time.

In this study, in-situ infra-red pyrometers were used to monitor the melt pool temperature during the L-PBF process whilst NiTi samples were printed. A four-level full factorial design of experiments was conducted with varying laser power, scanning speed, hatch spacing and spot size. Machine learning models were developed using IR in-situ sensor data for predicting the thermal information for the next layer, based on the prior thermal history and process parameters. Prior models have been developed which relate the thermal field to produced part quality. By predicting the thermal field of future layers using past in-situ thermal data and process parameters, optimal parameters for achieving the best part quality can be identified.