Smart Machines & Factories
Evolution towards the smart factory
Published:  16 May, 2017

By the time Industry 4.0 peaks, the functions of the machines producing the goods will no longer be clearly defined. Instead the goods themselves will tell the machine how they want to look at the end of the day. Frank Maier, CTO Lenze Group reports.

We are taking leaps and bounds towards this scenario ̶ the upside-down factory as many would see it ̶ is fast becoming a reality and there are already specific approaches to achieving maximum efficiency even for a batch size of 1. Basically, you need smart technical systems with inherent self-organisation and self-optimisation potential. These cyber physical systems (CPS) are no longer regarded as Utopian on today's production lines.

So why all the hype about the so-called brave new industrial world? The answer has to do with striving towards greater efficiency throughout the product life cycle. Self-learning systems can make a valuable contribution to minimising resource consumption by linking existing information with physical processes and components. Within this complete network, the production processes can be automated to the extent that they adapt flexibly to prevailing – and continuously changing - factory conditions like a human organism.

This approach becomes tangible when self-learning systems work out machine module performance and speed needs from the specific production requirements at the time and adapt accordingly. Possible results might be, for instance, conveying belts or storage and retrieval units that operate less dynamically in quiet times, yet with enhanced motion optimisation. The benefits of automated adjustment to actual conditions trip off the tongue: energy saving, less wear and tear, longer maintenance intervals, fewer costs and less ecologically damaging emissions.

When operating at low capacity, when it is irrelevant whether a travel request, for instance, is completed in two minutes instead of 20 seconds, the focus can be shifted from productivity towards efficiency. Self-learning systems are therefore extensively networked to deduce the right measures from over-arching trends, economic impact or seasonal peaks. It is beyond question that this holistic automation approach has effects on engineering. If all physical components, their relationships to one another and connections to the outside world can be transposed into a virtual world, then it stands to reason that real system control programs can already be created on virtual systems. This degree of virtualisation increasingly calls for reliable models that provide a true image of real-life practice, complete with regularities, facts and coherencies.

The whole is supported by open systems that can be seamlessly linked together without interface issues. The final product will be a homogeneous entity in the shape of a cyber-physical networked system with de-central intelligence that can self-optimise, detect faults or even initiate predictive maintenance. At the moment, we have a host of working partial solutions, which however are still far from constituting a smart factory. The evolution is well underway, but there is still a lot of ground to be covered.