We see AI within factories as essential for further developments in quality control. Getting the needed calculating power is where fog computing comes into play.
Maillefer is closely researching the use of fog computing to augment the standard hierarchy of automation control. Such concepts show the capacity to meet future needs of complex smart manufacturing solutions. Production processes need to be continuously controlled in order to maintain product quality. Realizing systems that must balance between stringent resource consumption constraints and the risk of defective end-product in real-time is especially challenging.
We see artificial intelligence as a viable choice for developing automated quality control systems, but integrating such system with existing factory automation remains complex. This is where a progressive approach is needed.
So, what is fog computing ? Borrowing on the concept of cloud computing, fog computing platforms are positioned at ground level, close to where people actually need computing resources. In our industry, that place is within the smart cable factories.
What is a fog cluster ?
Traditionally most of big data related work has been done in data centers and centralized cloud services. However, real time control of a production process cannot rely on cloud computing. The amount of data is simply too much to transfer over distances. It needs to be processed locally. Availability, reliability, and security concerns favor the introduction of fog computing for industrial control applications.
The comparison table of factory control schemes include fog computing
In practice, this means connecting computational resources to the local factory network. The fog nodes (i.e. computer servers) provide the necessary power for running more complex algorithms in a smart environment.
Big data and machine learning require more computational power then what traditional automation systems can offer. Fog nodes support needs of multiple programs, manufacturing lines or the whole factory. Running a small local datacenter addresses concerns such as data security and ownership.
Topography scanner example
Our Topography Scanner is a smart manufacturing solution which focuses on cable surface quality. The stand-alone installation has a PLC running direct control, a server PC to crunch numbers, a database PC and an edge-computing PC with accelerator card running the AI module. That’s a lot of computing power. Now, imagine multiplying the scanner across multiple lines within a factory.
Here, it becomes more cost effective to have one central place where all these applications are shared on a local platform. Fog computing federates the computing power, rather than having multiple systems for each individual unit.
More significantly, the ability to integrate a fog computing layer with PLC and SCADA networks brings other benefits. Having a platform that can handle the combined data flow from all the individual devices offers new opportunities for data-driven intelligence.
Individual measurements are given heightened context. For example, a deviation in cable core geometry that is detected by the Topography Scanner can be cross-referenced to extrusion parameters, or all the way back to the conductor manufacturing process. Or, we can statistically track increases in surface defect frequency and correlate this with the need for maintenance actions.
All in one smart place
All of these possibilities are more readily available when the data is in the same place and is backed by supporting computing capabilities (i.e. the fog cluster). Having a centralized data warehouse opens the doors to data mining, without having to move the data outside the factory walls. By adding a layer of fog computing to the edifice, cable factories are one step closer towards data-driven process intelligence and more autonomy for the future.
Janne Harjuhahto, R&D Specialist