The talking shop floor – backstage
In my previous blog there was a fictitious conversation between a shop floor AI and a manager. I got some enthusiastic remarks and questions. And the most important one has been: How long will we have to wait for that?
Creating a really working intelligent assistant for such a scenario is not simply copying the current natural language command interfaces like Alexa(TM), Siri(TM), or Google(TM) but it is a holistic approach for a domain like that. In the rest of this blog I will discuss the main components that are necessary for the realization of our “intelligent” FAB:
The natural language interface
The current technologies for speech recognition and generation are quite advanced. They are already available from the big platform owners Google(TM), Microsoft(TM), IBM and Apple and from different specialist companies in the area of dictation software. On this level it is important that the underlying language base includes domain specific phrases and all products as well.
The (knowledge) domain and company model
The things that make AI systems tick are clearly the included knowledge bases: These come in different forms like ontologies, rule-bases, knowledge trees, and other structures modeling the enterprise. Clearly there is general knowledge about production, about business rules etc. But most important are the active knowledge-bases which reflect the current state of the enterprise: This means for the shop-floor the information which machine and blue-collar worker is doing what related to a production order. On the planning level the general machine utilization, running or upcoming maintenance or upgrade operations together with the regular PLM, ERP data define the semantic order and strategic models.
The (distributed) AI platform
The AI platform combines different modules which work on the domain and company models and update them continuously. Of course the language interface is one of such modules. With regards to their activity modules can be grouped according to the functionality in three major groups: knowledge extraction, knowledge integration, and reasoning components. The knowledge extraction transforms input channels of the company into meaningful information. This creates an awareness model as described in the previous section which is then integrated into the knowledge bases and the consequences of this update are analyzed. The AI platform can host a complete zoo of AI services, which range from decision management systems, rule-bases and inference engines to semantic business process agents also with machine learning elements.
The intelligent company does not have one big AI, it is more an emerging quality of different AI services, models and semantic tools which interact and create as in the dialog “talking to your shop floor” an interactive partner with all the knowledge at its disposal for the workforce.
The road to this intelligent “fab” can be built from semantic and tool bricks which are already available and need to be integrated and configured. Most steps include augmenting the existing data and control flow with semantics and processing capabilities. This creates semantic contexts which abstract from given low-level detail until they work on meaning on a symbolic level for the employees to exploit, like a blue collar specialist on the shop floor.
What does this mean for us? The first AIs we encounter will be specialists, able to analyze and “understand” a very small segment of the business. However, as integration moves on there will be more synergistic effects which when combined with control for automation will become the driver for digital transformation in production (and not only production).