Scholarlink

<p>Porosity defects in high-pressure die-casting (HPDC) are a significant quality concern and are challenging to diagnose due to noisy raw data, nonlinear relationships, and data imbalance. Augmenting rare faults becomes essential for machine learning (ML) results, but it can be challenging, potentially leading to the generation of unrealistic or irrelevant fault signatures. This paper considers data augmentation in an explainable AI framework for robust porosity diagnosis. Specifically, we utilize an autoencoder with a Wasserstein generative adversarial network (WGAN-GP) to augment the latent space for rare-fault data and employ the Fréchet Inception Distance (FID) metric to assess the quality of the synthetic data. Demonstrated through a real-world case study. We applied a traditional XGBoost classifier to isolate defects and achieved a 95.90% F1 score for detecting porosity, significantly outperforming the baselines on rare-fault augmentation. Shapley additive explanation (SHAP) analysis identified the most contributing parameters, including Average Slow Shot Velocity and Biscuit Size, offering interpretable insights that enabled targeted process adjustments, followed by an observed reduction in the porosity rate from 8.3 to 3%. This hybrid explainable AI fault-diagnosis approach combines advanced augmentation, robust classification, and explainability to improve the diagnosis of complex and rare manufacturing defects, thereby enhancing quality control.</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

Rare Faults Diagnosis via Latent Space Augmentation in High-pressure Die Casting

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年5月18日

查看全文
  • Amr Dahab,
  • Kevin Otto,
  • Wen Li,
  • Tansu Alpcan,
  • Karthik Sankaranarayanan

摘要

Porosity defects in high-pressure die-casting (HPDC) are a significant quality concern and are challenging to diagnose due to noisy raw data, nonlinear relationships, and data imbalance. Augmenting rare faults becomes essential for machine learning (ML) results, but it can be challenging, potentially leading to the generation of unrealistic or irrelevant fault signatures. This paper considers data augmentation in an explainable AI framework for robust porosity diagnosis. Specifically, we utilize an autoencoder with a Wasserstein generative adversarial network (WGAN-GP) to augment the latent space for rare-fault data and employ the Fréchet Inception Distance (FID) metric to assess the quality of the synthetic data. Demonstrated through a real-world case study. We applied a traditional XGBoost classifier to isolate defects and achieved a 95.90% F1 score for detecting porosity, significantly outperforming the baselines on rare-fault augmentation. Shapley additive explanation (SHAP) analysis identified the most contributing parameters, including Average Slow Shot Velocity and Biscuit Size, offering interpretable insights that enabled targeted process adjustments, followed by an observed reduction in the porosity rate from 8.3 to 3%. This hybrid explainable AI fault-diagnosis approach combines advanced augmentation, robust classification, and explainability to improve the diagnosis of complex and rare manufacturing defects, thereby enhancing quality control.