Scholarlink

In this paper, we developed an anomaly detection algorithm for Flexible Manufacturing Systems (FMS) which is based on pneumatic actuators. Data was collected through sensors implanted on the system via the ThingWorx platform, ensuring a robust dataset encompassing both faulty and non-faulty states. The need for this research paper arises from the challenges faced in maintaining pneumatic actuator-based FMS. Traditional maintenance techniques, such as scheduled maintenance and reactive maintenance, often fail to prevent unexpected failures, leading to increased downtime and higher maintenance costs. We trained and evaluated five machine learning algorithms—Random Forest, Long Short-Term Memory (LSTM) networks, Seasonal Autoregressive Integrated Moving Average with exogenous factors (SARIMAX), k-Nearest Neighbors (KNN), and Convolutional Neural Networks (CNN)—on this dataset. The performance of these algorithms was compared based on their accuracy in detecting anomalies. The results highlight the strengths and limitations of each model, providing valuable insights into their applicability for real-time anomaly detection in pneumatic actuator-based FMS environments. This study demonstrates the potential of advanced machine learning techniques to enhance the reliability, efficiency, and longevity of modern manufacturing systems.

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

Machine Learning Based Anomaly Detection for Pneumatic Actuators Driven Flexible Manufacturing System

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Hansika Gupta,
  • Utkarsh Saini,
  • Paras Garg,
  • Nishant Rathore,
  • Gunjan Soni

摘要

In this paper, we developed an anomaly detection algorithm for Flexible Manufacturing Systems (FMS) which is based on pneumatic actuators. Data was collected through sensors implanted on the system via the ThingWorx platform, ensuring a robust dataset encompassing both faulty and non-faulty states. The need for this research paper arises from the challenges faced in maintaining pneumatic actuator-based FMS. Traditional maintenance techniques, such as scheduled maintenance and reactive maintenance, often fail to prevent unexpected failures, leading to increased downtime and higher maintenance costs. We trained and evaluated five machine learning algorithms—Random Forest, Long Short-Term Memory (LSTM) networks, Seasonal Autoregressive Integrated Moving Average with exogenous factors (SARIMAX), k-Nearest Neighbors (KNN), and Convolutional Neural Networks (CNN)—on this dataset. The performance of these algorithms was compared based on their accuracy in detecting anomalies. The results highlight the strengths and limitations of each model, providing valuable insights into their applicability for real-time anomaly detection in pneumatic actuator-based FMS environments. This study demonstrates the potential of advanced machine learning techniques to enhance the reliability, efficiency, and longevity of modern manufacturing systems.