A data-efficient sequential learning framework for melt pool defect classification in Laser Powder Bed Fusion

Abstract

Ensuring the quality and reliability of Metal Additive Manufacturing (MAM) components is crucial, especially in the Laser Powder Bed Fusion (L-PBF) process, where melt pool defects such as keyhole, balling, and lack of fusion can significantly compromise structural integrity. This study presents SL-RF+ (Sequentially Learned Random Forest with Enhanced Sampling), a novel Sequential Learning (SL) framework for melt pool defect classification designed to maximize data efficiency and model accuracy in data-scarce environments. SL-RF+ utilizes an RF classifier combined with Least Confidence Sampling (LCS) and Sobol sequence-based synthetic sampling to iteratively select the most informative samples, refining the model’s decision boundaries with minimal labeled data. Results demonstrate that SL-RF+ achieves an accuracy of 83.3%, outperforming the traditional RF model (78.8%) with significantly fewer labeled samples in melt pool defect classification. Moreover, SL-RF+ improves precision (83.1%), recall (76.9%), and F1-score (78.9%), surpassing the baseline model in all key performance metrics. Notably, SL-RF+ achieves competitive classification performance with fewer than 150 sequentially added samples, whereas the traditional RF model requires all 275 labeled samples to reach similar accuracy levels. By prioritizing high-uncertainty regions in the process parameter space, this framework efficiently captures complex defect patterns, ultimately achieving superior classification performance without the need for extensive labeled datasets. While this study utilizes pre-existing experimental data, SL-RF+ shows strong potential for real-world applications in pure sequential learning settings, where data is acquired and labeled incrementally, mitigating the high costs and time constraints of sample acquisition.