OEE Analysis with Artificial Intelligence

The data flow of plant productivity is speaking, but we humans do not listen. In the future, artificial intelligence will be able to identify the signal in the noise, that is, to report the deviation from the normal state, without a production engineer having previously explicitly said what is “normal”. The algorithm makes suggestions to the human being about where to optimize.

The development path of the OEE

The OEE has taken a development in companies. The start was a tally sheet as a handwriting. The first stage of development was the use of Microsoft Excel. If you build OEE management step by step, the employees have understood it well and can argue with the key figure. But then the next steps must follow. It’s not 2018 to send an Excel list to all those responsible for operations once a week. There’s a lot more to it.
Those who use oee.cloud have taken the further development steps: The data are available online and are displayed on Andons at the machines and in offices. Besides causes of availability- and quality-losses, the causes of performance losses are also recorded. The loss detection is digital and precise, charts indicate the downtime distribution and gradients. If the system does not deliver the expected performance, a responsible person will be informed immediately by e-mail, text message or push notification.
If we are now working with the findings of the data specifically to improve, we have a management of plant productivity with methods that reflect the current state of knowledge and technology.

The OEE data stream speaks

In the OEE data stream, ie the actual system speed, the default speed and the causes for the deviation between the two data, there are a lot of statements that have not yet been used for process improvements.
Which production manager can answer the following questions:

  • Does the morning shift systematically change-over faster than the late shift?
  • Is it better to run the equipment faster with more short-stops than slower and more stable?
  • Was the OEE good for the ABC job or is it usually manufactured at a lower cost?
  • What is the best duration from the end of the change-over to reaching the ridge line of production?
  • Which products cause particularly short-term shutdowns?

And these are just 5 of many questions that can’t be answered in a production plant as of today.

Artificial intelligence in the analysis cockpit

The oee.cloud team is working on oee.ai. In the future, the production engineer will be actively informed of anomalies in the data stream in an analysis cockpit. Artificial intelligence algorithms scan the data stream for patterns and deviations.

Image: oee.ai anomaly cockpit

The data for the evaluations will continue to be collected by the proven minimally invasive sensor and tablets, so that this new level of OEE management requires no intervention in the system. If the data are available in a network-connected PLC, they can of course also be fed directly to oee.ai.
The anonymous data is collected across companies. For example, the algorithms can be trained together for all operators of bottling plants. Because we already see today: Whether mineral water, coffee, milk or bread spread is bottled, makes no difference to us in the data samples. The OEE analysis of machine tools is the next limit to be overcome.

If you are interested in keeping abreast of the current developmental stages of using artificial intelligence in OEE management, you have the opportunity to contact us at info@oee.cloud or you sign up for our newsletter below. This also applies if you want to become part of our private beta application. Let’s get in touch.