Scholarlink

<p>Additive Manufacturing (AM), specifically Laser-Powder Bed Fusion (L-PBF), is increasingly recognized for its transformative role in modern manufacturing. Despite its potential, achieving consistent product quality and real-time anomaly detection during the L-PBF process remains a critical challenge. To address this, we propose an ontology-driven framework that systematically encapsulates the key parameters, materials, and governing physical principles of L-PBF. This framework integrates with a Physics-Informed Neural Network (PINN) to create the O-PINN model, which harmonizes domain-specific knowledge with data-driven learning. The O-PINN model accurately predicts melt pool dynamics while enhancing interpretability by revealing the underlying physics of the process to provide a critical step toward deeper insights into L-PBF mechanisms. In addition to its predictive capabilities, the O-PINN framework provides robust real-time anomaly detection by continuously monitoring melt pool behavior and identifying deviations from established norms that indicate potential defects. This capability significantly improves quality control, optimizes process parameters, and reduces production costs, ultimately enhancing the reliability and consistency of AM-produced components.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Ontology-driven Physics Informed Neural Network (O-PINN) of interpretable melt pool behavior for real-time process anomalies’ prognosis/diagnosis

  • 期刊论文

  • 发表日期：

    2026年1月7日

查看全文
  • Byeong-Min Roh,
  • Xinyi Xiao

摘要

Additive Manufacturing (AM), specifically Laser-Powder Bed Fusion (L-PBF), is increasingly recognized for its transformative role in modern manufacturing. Despite its potential, achieving consistent product quality and real-time anomaly detection during the L-PBF process remains a critical challenge. To address this, we propose an ontology-driven framework that systematically encapsulates the key parameters, materials, and governing physical principles of L-PBF. This framework integrates with a Physics-Informed Neural Network (PINN) to create the O-PINN model, which harmonizes domain-specific knowledge with data-driven learning. The O-PINN model accurately predicts melt pool dynamics while enhancing interpretability by revealing the underlying physics of the process to provide a critical step toward deeper insights into L-PBF mechanisms. In addition to its predictive capabilities, the O-PINN framework provides robust real-time anomaly detection by continuously monitoring melt pool behavior and identifying deviations from established norms that indicate potential defects. This capability significantly improves quality control, optimizes process parameters, and reduces production costs, ultimately enhancing the reliability and consistency of AM-produced components.