Peter Seeberg and Robert Weber run the AI podcast under the label “AI in Industry.” There, the topics of artificial intelligence and machine learning are to be demystified in order to inspire industrial users and impart knowledge for everyday working life. In episode 50, our founder Jörn Steinbeck was a guest and explained an excerpt of the use of AI at oee.ai in conversation with Peter Seeberg.
At oee.ai, the use of machine learning algorithms is constantly expanding. One example of an application is the LSTM algorithm, which is used to separate a plant run from stable and unstable times via classification. The reason for the podcast was the previously published white paper together with bitkom, which is available for download here.
The following topics were discussed in the podcast from that minute on:
00:42: Part 1 — Publishing data
05:41: Part 2 — Quality of data
13:50: Main part of the 50th AI podcast — LSTM algorithm for OEE optimization
14:20: Peter explains OEE definition, his experience with a start-up
15:20: Reference to Jörn and OEE; AI in OEE process, OEE status in companies
14:42: Jörn promises data collection without an IT project, NanoMes, installation in 20 minutes
17:34: Detect stable/unstable plant operation with an algorithm
18:28: Peter: Monitoring production with a model — phase detection, detecting anomalies
22:38: Robert: The algorithm finds anomalies that you would not have seen before; special role of the domain expert; the algorithm comes with suggestions, the domain expert must confirm
24:06: Explanation of how algorithm handles data/figures; Peter talks about process-driven business models; Peter: “Data collected by an MES can be used proactively to improve OEE. ”
26:02: Robert: Jörn is a competitor to MES with oee.ai; Jörn says “... minimally invasive cloud-based NanoMes solution based on gateway”; then this is a new provider on the market
28:46: Technology, LSTM, why is LSTM so important at OEE? Jörn: “We classify as stable/unstable, use the LSTM's short-term memory, need 2-3 months of data to train, have a 'human in the loop, 'from which the algorithm learns”
31:02: Peter: Summarizes content, goes back to AutoML
34:11: Jörn about AutoML; system situation may change; algorithm must be retrained
38:57: Robert: There is a 100% funding program from the Federal Ministry of Economics for companies with less than 50 million € turnover, less than 250 employees, which can be carried out remotely in times of corona; reference to show notes; contact Jörn
40:54: Robert: The funding program focuses on small companies... that's where IT professionals and domain experts meet, Peter conducted an interview with Dr. Nabil Alsabah from bitkom in which the combination of IT and domain knowledge is well presented
The podcast is available on all popular platforms, such as
available.
More information about the podcast is available at https://kipodcast.de/ available.