New research integrates physics into machine learning for additive manufacturing

A new research paper explores the application of physics-informed and hybrid machine learning techniques to improve additive manufacturing processes, specifically fused filament fabrication (FFF). The study, authored by Berkcan Kapusuzoğlu, details three strategies for integrating physics knowledge into deep neural networks (DNNs) to predict part quality, such as bond quality and porosity. These methods include incorporating physics constraints into the DNN's loss function, using physics model outputs as inputs, and pre-training DNNs with physics models before fine-tuning with experimental data. The research demonstrates that combining these strategies can lead to accurate machine learning models even with limited experimental data, ensuring physically meaningful predictions. AI

IMPACT This research could lead to more accurate and efficient additive manufacturing processes by improving quality prediction with less data.

RANK_REASON The cluster contains an academic paper detailing novel research methods. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv cs.LG →

• arXiv
• Berkcan Kapusuzoğlu
• deep neural network
• fused filament fabrication

• paper
• other

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →