🔔 Let's Rock Up! Join in our meetup for Jupyter users

Wed 28/09/2022 19:00 (KST)
311, Gangnam-daero, Seocho-gu, Seoul
Register

Robot Arm Anomaly Detection

  • Automobile
  • Anomaly detection

Challenge

Today, industrial robots perform critical operations such as assembly, painting, and welding at every automobile assembly line in an automated environment. In such environments, predictive maintenance (PdM) for robot equipments is essential for maintaining productivity. For this purpose, installation of mechanical sensors on robot equipments may be required to collect sympton measurements, which is time-consuming and costly in most cases.

Approach

We use MakinaRocks' proprietary anomaly detection algorithm based on unsupervised learning without attaching mechanical sensors to the robots. Deep neural network models customized to production environments and process scenarios detect anomalies of robot equipments before they cause critical failures.

Results

Our models can detect anomalous behavior of robot equipments (*5 days before) when the robot becomes stopped as a result of critical failure. The use of anomaly detection in industrial robots helps reduce down time and reduce costs.