Detection of deviations in high productivity cycle times

Participants

Etxetar

Universidad Politécnica de Madrid

Aingura insights

Transformation Transformation
Delivery Delivery
Data Acquisition Data Acquisition
Preprocessing Preprocessing
Transportation and storage Transportation and storage

Sectors

Manufacturing Manufacturing

Applications

Maintenance Maintenance
Condition Monitoring Condition Monitoring

impact

  • Detect deviations in the process parameters below 15% of the cycle time, commonly invisible to the deployment technicians.
  • Increased productivity by an average of 6 hours a day.

Develop AI Technologies

  • Data intelligence.
  • Machine Learning.
  • Pattern analysis, prediction and intelligent classification.

objectives

  • To increase in machine availability in high production systems in order to maintain or exceed 95%.
  • To search for inefficiencies in the operating pattern that represent losses of time and prevent compliance with line cycle times.
  • Reduction of machine deployment times by detecting imperceptible deviations.

results

  • Dataset size: 12GB
  • Number of variables: 22
  • Tiempo de muestreo: 100ms
  • Development of an algorithm for the detection of deviations based on a hybrid system between Kernel Density Estimation (KDE) and Hidden Markov Models (HMM) for training and detection of deviations based on machine data acquired at 100 ms.
  • Offline data analysis system that has allowed to find deviations below 2 seconds, as shown in graphics bellow.
  • The inefficiency detected is related to a CNC code line that unexpectedly moved the drilling references in the B axis. This error would have been practically impossible to detect with the naked eye.
Machine cycle data for each servo.
Machine cycle data for each servo.
 Differences between cycles
 Differences between cycles

Contact us