[Data] Federated Learning for Self-Supervised Acoustic Intelligence in LPBF: Physics-Guided Learnable-QSincNet for Cross-Domain Representation with Causally Explainable Spectral Insights
Abstract
Airborne acoustic emission (AE) forms a rich component of Laser Powder Bed Fusion (LPBF) process emissions and offers a high-value channel for in-situ process monitoring; however, leveraging this information for industrial deployment remains challenging due to cross-domain variability in process conditions, strict data-governance constraints, and the absence of physics-guided representation learning. This work introduces a federated learning framework for self-supervised acoustic intelligence in LPBF, enabling distributed model training across heterogeneous data domains without sharing raw sensor signals. At the core of the method is a Physics-Guided Learnable-Q SincNet, a learnable front-end that adaptively discovers defect-sensitive spectral bands while enforcing physically meaningful frequency parametrization. Combined with a joint inter–intra self-supervised objective, the framework learns cross-domain representations that remain robust across diverse LPBF datasets reflecting variations in defect regimes, melt-pool dynamics, and spectral characteristics. To ensure interpretability, we develop a causally explainable spectral analysis pipeline, revealing how specific learned frequency bands contribute to identifying lack-of-fusion, conduction-mode, and keyhole regimes. The resulting causal spectral pathways highlight physically interpretable links between AE signatures and melt-pool stability, offering insight into the acoustic mechanisms underlying defect formation.…
Read the paper
Related studies
- Active Learning during Federated Learning for Object Detection
- Fairness-aware federated learning
- Decentralized Federated Learning
- Building a Personally Identifiable Information Recognizer in a Privacy Preserved Manner Using Automated Annotation and Federated Learning
- Preservation of Patient Level Privacy: Federated Classification and Calibration Models
- Efficient explainable federated learning
Go deeper in Evidence Hub
Evidence Hub adds a study data sheet (design, sample size, evidence level, key numbers), an AI summary and a chat grounded in the paper's text.