Scholarlink

<p>The increasing demand for quality and efficiency in the automotive industry has driven the adoption of advanced monitoring and predictive technologies in sheet metal forming processes. This study proposes a low-cost, real-time methodology for failure prediction using artificial intelligence (AI) and sensor data collected during Nakazima tests. An experimental setup was developed with sensors for temperature, electric current, vibration, force, and strain, integrated into a hydraulic press. The collected data were analyzed using statistical methods and machine learning models, including Local Outlier Factor, Isolation Forest, and K-Means. The results show that temperature is the most relevant variable for early failure detection, showing the strongest correlation with deformation and the highest contribution across the evaluated machine learning models and indicator analysis. Among the models tested, Local Outlier Factor and Isolation Forest achieved over 80% accuracy, while K-Means performed poorly. The adherence indicators confirmed the limited relevance of current and vibration data, while temperature, force, and strain demonstrated high potential for industrial applications. The methodology proved effective and scalable, offering a viable solution for real-time quality control in metal forming environments.</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

An approach of artificial inteligence use in real time processing to predict failure during sheet metal forming

  • 期刊论文

  • 发表日期：

    2026年6月24日

查看全文
  • Carlos Alberto Vicari,
  • Chetan P Nikhare,
  • Pablo Deivid Valle,
  • Paulo Victor Prestes Marcondes

摘要

The increasing demand for quality and efficiency in the automotive industry has driven the adoption of advanced monitoring and predictive technologies in sheet metal forming processes. This study proposes a low-cost, real-time methodology for failure prediction using artificial intelligence (AI) and sensor data collected during Nakazima tests. An experimental setup was developed with sensors for temperature, electric current, vibration, force, and strain, integrated into a hydraulic press. The collected data were analyzed using statistical methods and machine learning models, including Local Outlier Factor, Isolation Forest, and K-Means. The results show that temperature is the most relevant variable for early failure detection, showing the strongest correlation with deformation and the highest contribution across the evaluated machine learning models and indicator analysis. Among the models tested, Local Outlier Factor and Isolation Forest achieved over 80% accuracy, while K-Means performed poorly. The adherence indicators confirmed the limited relevance of current and vibration data, while temperature, force, and strain demonstrated high potential for industrial applications. The methodology proved effective and scalable, offering a viable solution for real-time quality control in metal forming environments.