The thesis tackles the challenge of reliably detecting defects in additive manufacturing, a bottleneck in its industrialization. The solution proposed in this thesis uses machine learning, specifically convolutional neural networks (CNNs), to correlate online monitoring data with post-process computed tomography data. By correlating both data sets, the impact of monitored anomalies on the finished sample is evaluated. Specifically, the CNNs are trained to detect defects based on the meltpool radiation measured during the printing process. The approach shows great potential for detecting pores in laser powder bed fusion parts and is adaptable to other sensor setups and defect types.
«
The thesis tackles the challenge of reliably detecting defects in additive manufacturing, a bottleneck in its industrialization. The solution proposed in this thesis uses machine learning, specifically convolutional neural networks (CNNs), to correlate online monitoring data with post-process computed tomography data. By correlating both data sets, the impact of monitored anomalies on the finished sample is evaluated. Specifically, the CNNs are trained to detect defects based on the meltpool rad...
»