In-situ porosity monitoring in laser powder bed fusion through acoustic signal and interpretable multi-sensor fusion

Porosity is a harmful defect in alloys produced by the laser powder bed fusion (LPBF) process. Acoustic signals from the laser-material interaction can capture complex events such as pore formation. However, the defect-signal correlation mechanism and porosity monitoring in complex LPBF processes influenced by non-laser energy parameters have received limited attention. Existing research rarely considers the potential impact of sensor configuration on monitoring, and it lacks interpretable acoustic signal processing methods and multi-sensor fusion monitoring approaches. This paper proposed an interpretable multi-sensor acoustic monitoring framework. First, a multi-source acoustic monitoring system was developed, tailored to the LPBF acoustic characteristics. The complementarity and effectiveness of 20 kHz-40 kHz resonant multi-sensor monitoring were validated from signal and model perspectives. Next, the physics-informed frequency-scale texture image method was proposed to enhance the expression of acoustic defect features. Finally, an interpretable multi-sensor fusion model was established for in-situ porosity monitoring in the LPBF process. It embedded prior knowledge of sensor performance and acoustic defect features. Results show the method achieves 99.22 % accuracy in identifying LPBF porosity defects with eight different quantization labels. Visualization and interpretability analysis confirm that the model achieves reliable multi-sensor fusion and extracts the physically reasonable global or local key spectrum features from the acoustic signals. With a monitoring response time of 1.69 ms and interpretable multi-sensor fusion monitoring capabilities, this method offers significant potential for industrial deployment.