COSMINO makes Production Data AI-ready
By susanne.gutermuth@cosmino.de In Artificial intelligence opens up new possibilities for manufacturing companies: Detect errors faster, analyze causes automatically and make informed decisions - directly during ongoing production, without wasting time or manual intervention. AI is becoming the key technology for quality, efficiency and reaction speed in production. But for AI to be effective, it needs structured, networked and comprehensible data. This is exactly where COSMINO AG comes in. As a specialist in PDA, MDA and MES solutions, COSMINO is currently developing AI functions that are integrated directly into existing systems - with the aim of making everyday production more intelligent and responsive. The Cosmino Digital Twin forms the central database for this.COSMINO makes correlations in your data usable for AI
Production data - such as downtimes or process values - is automatically read from the Cosmino Digital Twin and transferred to a graph database. There, an intelligent network maps the relationships between data: e.g. which job is running in which work center with which tool. This structured link makes data machine-readable and forms the basis for AI-supported queries, for example: „Why does this machine stop so often?“ or „Which orders had quality deviations?“Cosmino's own chatbot uses large language models (LLMs) such as ChatGPT or Mistral: it automatically generates suitable Cypher queries from a user question and delivers a comprehensible answer - without the external LLM having direct access to operational data. Sensitive content remains protected, even if LLMs from the cloud are used. AI finds patterns and correlations in the neural network that are often overlooked in purely manual analysis - for example, recurring sources of error, hidden causes of rejects or critical combinations of order and machine. This generates knowledge that can be used specifically to optimize processes, product quality and system availability.
Anomaly Detection in real-time
In other projects with various partners (including KnowGrADe with Fraunhofer IGCV and other project partners), AI-based neural networks are being set up and trained to detect deviations in the production process not only in retrospect, but at the moment they occur - before rejects are produced or a machine breaks down. This not only saves costs, but also protects processes, deadlines and quality.The large amount of data required for this comes from the Cosmino machine and operating data acquisition system via the digital twins. If anomalies are found in the data, the Cosmino software can react in real time: from alarms, to additional checks initiated in the data acquisition dialog, to automated intervention in the machine control system.
From reaction to prevention
Cosmino software already helps manufacturing companies to react quickly and in a targeted manner to faults and deviations. With the new AI-based solutions, the next level is now emerging: problems are not only detected more quickly - they can be specifically avoided before downtime, rework or rejects occur.Would you also like to make your production data usable for AI and analyze it? Let's find out together what potential lies in your data - and how artificial intelligence can turn it into usable information for your production.
Fig.: System architecture with integration of AI solutions. Data is made available for analysis in near real time. Identified anomalies and the like are then reported to the Cosmino software, also in near real time.