How can cognitive technology and artificial intelligence be applied in manufacturing?

From experience I used to make two kinds of engineering error. The first was to use the wrong efficiency factor. The answer was often close enough to pass scrutiny, but still wrong. The second kind of error was to get the decimal wrong. The answer had the right digits but was out by an order of magnitude. When I got involved in IT I soon realized that even the smallest error was enough to bring down an entire system. It is the same when trying to predict IT trends, you need to both get the order of magnitude right as well as get the details right, otherwise you are going to misjudge altogether.

In the field of process automation, engineers work to automate plants in order to free people from repetitive tasks. But, as we move into the age of intelligent machines these automation boundaries between human and machine are changing. The reality is that many more of the tasks currently only performed by humans will in future be performed by machines.  And many of our established engineering and business analysis techniques are ill-prepared for defining requirements in this new era of computing.

ARTIFICIAL INTELLIGENCE (“AI”) AND COGNITIVE TECHNOLOGY

Cognitive technology refers to the broad category of technologies that can perform tasks that can otherwise only be done by humans. Two simple examples are voice and image recognition. Cognitive technology is still in its relative infancy, but is expected to mature rapidly and become part of mainstream IT. While consumerization of this technology is already happening outside the enterprise space (for example personal assistants on mobile phones), developments in this space will eventually have a profound impact on the manufacturing enterprise.

Behind the scenes, artificial intelligence and cognitive technologies are already infused into everyday experiences and are in widespread use in context aware applications like personal assistants, online map services, cross selling (like Netflix or Amazon recommendations) and so on. This has the effect of setting new consumer expectations that are hard for ordinary business to fulfill, while also opening up a range of new untapped opportunities to exploit for competitive advantage.

Machine learning allows natural language to be interpreted and processed to deduce patterns and correlations. When done on scale, for example by a machine processing all the published research on a subject, these correlations can be used to make accurate predictions and augment human decision making in ways that were previously impossible. A good example is the machine diagnosis of medical conditions that guide a doctor to explore all probable likely alternative conditions before settling on a probable diagnosis.

AI IN MANUFACTURING

In the next few years, AI and cognitive technology will impact on every aspect of manufacturing.  Companies will in future need to go about defining, developing and implementing solutions that will rely on cognitive technology in some form. This is new territory for engineers, business analysts and IT professionals alike.

In the field of process automation artificial intelligence techniques are not new. Back in 1998 I remember our engineering team investigating the use of neural networks to model the behavior of an aerobic fermenter based on measured variables such as pH, temperature, historical cell growth rates and so on.  However at the time this was a very specialized field. Nowadays deep neural networks are a mainstream technique in IT and becoming more prevalent than ever before.

Mainstream IT is driving much of the innovation in machine learning

One trend to watch closely is the entry of mainstream software vendors like Microsoft and IBM into this space. These vendors have the resources to cost effectively leverage vast cloud platforms and databases of structured and unstructured information to perform data processing and analysis at scale. They can do this on powerful hardware that can be easily allocated and scaled according to real-time demand. It is significant that the established mainstream IT vendors are now making their cognitive technology available through web service API’s for everyone to access and utilize. Readily available bot frameworks and cognitive API’s for speech and image recognition can now be accessed by simply calling a web service.

Another area to watch is the evolution of innovative database techniques for unstructured data. This information can be stored in database constructs confusingly called “graphs” that make it possible for unrelated information to be correlated to detect patterns that can be utilized in ways not possible previously.

Companies are starting to use cognitive technology across the whole value chain in industries as diverse as oil and gas to automotive and in areas as diverse as R&D, plant operations, supply chain, marketing and customer service.

WHERE TO START?

There are several ways in which artificial intelligence and cognitive technology might be used in manufacturing, for example:

1. Enhance and augment products with new features/services. For example you could recognize patterns of interaction with individual customers that are used to make personalized recommendations on better use of the products. Or, in a business to business context you could provide your customers with access to your models that help them utilise your product more efficiently. An example of this might be to assist commercial farmers with AI based models that combine weather predictions, soil conditions, GIS information and supply/demand projections to help with planting and harvesting decisions.
2. Develop operational insights. For example, by monitoring key process variables and using pattern recognition techniques across all your data sources you could determine when there is an elevated risk of incidents on site, process upsets, likely logistical problems or equipment breakdowns. Or whether certain unit operations are being run sub-optimally etc.
3. Further automate and optimize your business processes. Information trapped in your ERP system can be combined with MES data and unstructured information held in a graph database to provide better real time insights into the business and manufacturing processes, with the resulting call to action served to the responsible person through a mobile device. These proactive alerts can help prioritize interventions where necessary ensuring a quick response to upset conditions.

Process control specialists, business analysts and IT professionals will in the near future need to be well versed in techniques where AI and cognitive technology are to be used.  Many of these techniques are still evolving.

The business case for these solutions will in general be motivated by the increased value as opposed to cost reduction. This will require a business mindset as well as a deep technical knowledge. Managers need to also recognize the future impact these technologies will have on the way systems are integrated and the new skills needed by your company in this emerging era of thinking machines.