Stabilising machine performance using hidden data

Do you think have a machine with no problems? If so, you may already be collecting production and machine data but not using it for stabilising machine performance. A quick scan of it can assess the usability and quality of the data and to start using the information it holds. This will help you define the first or next step towards a data enabled organization.

 Check for your machines

How does this scan work? By logging and analysing high frequency signals from sensors and actuators and linking them to the machine process and control program. By doing this together with the machine operators and engineers, gives you detailed insights for spotting deviations and inconsistencies, such as sensor misalignment, misconfiguration, and worn-out parts. It can reduce micro machine stoppages that develop over time and help find the causes of intermittent and indistinct problems.

Stabilising machine performance

With constant monitoring it is possible to detect process and quality anomalies and to discover trend drifts for predictive maintenance. Your machines change over time and more input data for model training becomes available. A regular update of the models is necessary to keep and enhance the accuracy of predicting.

OMRON’s Sysmac AI Controller to work

OMRON uses a machine controller with a Sysmac library for artificial intelligence. This allows them to support with a proof-of-concept and help during the implementation of the AI-Controller on a production machine. They also offer AI-as-a-service for taking care of the complete implementation, updates, upgrades, and maintenance. Even AI systems need regular attention to improve and adapt to the detected changes in machine behaviour and production process.

Case study

A recent project was to improve the performance of a pin stitcher machine on the NX assembly line at Omron’s Netherlands factory. The machine stitches pins into plastic cases and contains several motors and sensors that generate more than 50 signals in parallel. The goal of this project was to monitor all the signals concurrently and discover abnormal situations. The AI Controller captured signals every two milliseconds and stores them for data analytics and anomaly detection.

A regular issue was with the bending and reel feeding motors when the machine did not move the reel correctly.  When this caused a stoppage, the maintenance activity needed took almost one hour. Anomaly detection now allows the early discovery of faults and sending of alarms to the engineers and maintenance teams. In the case of the pin stitcher machine, it triggers the alarm before a major problem occurs. This avoids the unplanned stopping of the machine by ensuring timely maintenance.

Besides, humans cannot always detect the subtle changes in the machine, for example if it slows down a little. Overtime, the machine could produce 15% less per day. Monitoring the behaviour of all signals using anomaly detection models helps the operator identify the root causes. Moreover, this information allows adjusting the machine to keep it performing at a high level.

To help you stabilise machine performance by exploit your industrial data, download the white paper for more details.

 

 

Subscribe
Notify of
guest
0 Comments
Inline Feedbacks
View all comments