Industry 4.0 applications generate large amounts of data. Collecting data is good, analyzing this data is better—and it's best to implement the findings. With oee.ai, we have overcome the hurdle of simple and flexible recording. Visually processing the data so that a person can recognize patterns has also been solved. Generating push notifications when the OEE is below Y% for more than X minutes also works. But there is much more to artificial intelligence...
Artificial intelligence as an overarching force and machine learning as a specific sub-area are developing rapidly. The method is efficient: By continuously processing large amounts of data (big data) based on statistical analyses, the algorithm learns from experience — more, faster and better — than humans can. Then let the computer do the work.
The use case of OEE analysis and optimization is predestined for the pattern analysis of large amounts of data that humans no longer have an overview of:
- Basic disturbance frequency depending on the shift
- Set-up times depending on the products to be set up
- Losses of availability depending on the time
It is relatively easy for experienced plant operators to carry out a graphical analysis with human intelligence and good data preparation. Here are three examples from different companies:
Example 1: A very unstable product. After setting it up, it only takes effort to reach the ridge line and before it reaches power, the lot is already done. The OEE for this order was 31.6%.
Example 2: A generally well-running product had five major faults during the run time from 9:30 a.m. to 6:00 p.m.
Example 3: Between 1:00 and 4:00 in the morning, the prepared product ran significantly worse than the other products on the machine.
However, artificial intelligence searches for these patterns and exceptions around the clock and can prioritize the significance of individual events in the overall context — and graphically present them to humans in an understandable way.
A “anomaly cockpit” shows the special features of the data stream. This algorithm learns independently and therefore becomes more and more reliable in classification over time. People evaluate whether it was an event, which was an exception and is therefore not being followed up further (e.g. components from the replacement supplier were of insufficient quality), or whether it was marked for optimization.
This is the OEE improvement process of the future: Number-driven, data-driven and fact-driven by an algorithm, evaluated by humans and implemented with their creative solutions.
The technology to create a “flashy cockpit” is available. Google has at the end of 2015 TensorFlow, its machine learning software, released under an open-source license. Since then, the tool has become the market leader in this field of application. The TensorFlow framework combined with the programming language python, which, for example, with the Scikit-learn A framework for machine learning is a powerful programming environment for solving these issues.
Together with our academic university environment, we research and develop OEE analysis through artificial intelligence — algorithm-based OEE optimization. If you are interested, please contact us: info@oee.ai or +49 (0) 241/401 842 75.