Machining process monitoring

Participants

Etxetar

Universidad Politécnica de Madrid

Aingura insights

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

Sectors

Manufacturing Manufacturing

Applications

Condition Monitoring Condition Monitoring

impact

  • Machine monitoring using machine learning techniques appropriate for application in dynamic environments, where failures are not very common and are related to degradation.
  • Monitor in real time the status of the bearings with respect to the generated reference.

Develop AI Technologies

  • Data intelligence.
  • Machines/people interfaces.
  • Smart sensors.
  • Machine learning.
  • Pattern analysis, prediction and intelligent classification.

objectives

  • To develop machine monitoring system that allows failure detection in critical elements, in this case, the spindle header.
  • To develop first steps for the detection of anomalies in highly dynamic processes.

results

  • Dataset size: 1GB / day
  • Number of variables: 10 - 110
  • Sampling time: 50 - 100ms
  • It has been possible to develop data acquisition systems that ensure their quality by extracting them from production systems such as machine tools.
  • Once this quality was assured, the first machine patterns were developed.
  • These patterns have been examined and compared as a reference to detect possible deviations in relation to the reference.
  • This system allows monitoring the state of the front bearing of the spindle head. Its failure can cause the machine and production line to stop.
Spindle head working pattern
Spindle head working pattern

Publications

Author: J. Diaz-Rozo, C. Bielza, y P. Larrañaga

Title: Machine learning-based CPS for clustering high throughput machining cycle conditions

Publication: Procedia Manufacturing, vol. 10, pp. 997-1008

Date: January 2017

More information

Contact us