Physics-guided denoiser network for enhanced additive manufacturing data quality

2025-07-31

AI Summary

High sensor data noise and poor physical interpretability in additive manufacturing hinder real-time process control and defect diagnosis. To address this, we propose a physics-guided denoising framework that innovatively integrates a physics-informed neural network (PINN) surrogate model, energy-based modeling, and Fisher score regularization—explicitly embedding multi-physics constraints (e.g., heat transfer, mass conservation) into the denoising process to ensure physical consistency of solutions. Unlike purely data-driven approaches, our method circumvents generalization bottlenecks by leveraging domain knowledge, thereby enhancing noise robustness and cross-condition transferability. Experiments demonstrate superior performance over state-of-the-art denoising networks across diverse noise levels. Furthermore, the framework is successfully deployed for real-time analysis of thermal radiation data in laser powder bed fusion (LPBF), enabling high-accuracy defect prediction and data-informed process parameter optimization.

Technology Category

• Computer Vision: Low Level & Physics-based Vision
• Knowledge Representation and Reasoning: Diagnosis and Abductive Reasoning
• Machine Learning: Calibration & Uncertainty Quantification

Application Category

• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Systems and Infrastructure for Web, Mobile and WoT: Applied ML and AI for Web-based mobile applications
• User Modeling, Personalization and Recommendation: On-Device user modeling, personalization, and recommendation

Abstract

Modern engineering systems are increasingly equipped with sensors for real-time monitoring and decision-making. However, the data collected by these sensors is often noisy and difficult to interpret, limiting its utility for control and diagnostics. In this work, we propose a physics-informed denoising framework that integrates energy-based model and Fisher score regularization to jointly reduce data noise and enforce physical consistency with a physics-based model. The approach is first validated on benchmark problems, including the simple harmonic oscillator, Burgers' equation, and Laplace's equation, across varying noise levels. We then apply the denoising framework to real thermal emission data from laser powder bed fusion (LPBF) additive manufacturing experiments, using a trained Physics-Informed Neural Network (PINN) surrogate model of the LPBF process to guide denoising. Results show that the proposed method outperforms baseline neural network denoisers, effectively reducing noise under a range of LPBF processing conditions. This physics-guided denoising strategy enables robust, real-time interpretation of low-cost sensor data, facilitating predictive control and improved defect mitigation in additive manufacturing.

Problem

• Reducing noise in sensor data for real-time monitoring
• Enforcing physical consistency with physics-based models
• Improving additive manufacturing data quality and defect mitigation

Innovation

• Physics-guided denoiser network for data quality
• Energy-based model with Fisher score regularization
• PINN surrogate model for real-time denoising