Editor's Choice

AI in manufacturing: a process engineer’s perspective

May 2024 Editor's Choice IT in Manufacturing

The expert will tell you what to do, the philosopher will tell you why to do it, and the engineer will get on and actually do it. As the hype around AI intensifies, the number of ‘experts’ is increasing exponentially. In contrast, the number of engineers who actually know how to implement AI technology remains small.

In past weeks, I have received a proliferation of marketing content about generative AI and how AI is transforming the way we work. Webinars and training courses are oversubscribed as budding talent worldwide recognises that AI skills are not just a passing fad, they will become fundamental to competing in the modern workplace.

With all of this information flooding my inbox, it is perhaps important to step back and ask: “What specific new engineering skills and knowledge are really necessary in order to thrive in the future environment? How should we as engineers react?”

Applied intelligence

As engineers, we are tasked with applying the right technology in a way that will add value to our organisations, and of course to society at large. This has to go beyond generating interesting pictures, getting Elon Musk to perform in the voice of Elvis Presley, and asking ChatGPT to write poetry. We have to go beyond being users of generative AI, and learn what lies under the hood, thereby unlocking the potential of AI to innovate and supercharge our business.

Where is AI innovation most rapid?

Naturally, most of the AI innovation is taking place in the tech sector. Automotive appears to be following in a very close second place.

However, according to a recent Accenture study, the process industry (specifically, chemicals) lags behind in terms of the AI Maturity Index. Accenture defines the AI maturity index as the arithmetic average between foundational and differentiation factors, the two dimensions by which they assess whether a company is an AI innovator, an AI achiever, an AI experimenter, or an AI builder.

Why is it that the chemical industry, that was once at the forefront of automation innovation in the 1970s, has seemingly now lagged and been slow on the uptake regarding AI?

Generative AI infused into business and IT systems

Microsoft recently embarked on a significant marketing campaign to explain the benefits of Copilot, which they describe as AI being ‘infused’ into the business and productivity software that we use every day. Of course, the demos were impressive, and presented by the sharpest minds. Their vision is compelling; ask Copilot to analyse the data in a spreadsheet and then to summarise the important patterns and trends. It is easy to see how generative AI can be used to analyse financial data in the ERP system to help quickly identify loss-making customers, systemic quality issues or product lines that are underperforming.

As an ordinary human, interacting with these AI agents does require a new mindset. In my experience, many people in corporate jobs barely scratch the surface of basic spreadsheet functionality, let alone have enough imagination to ask AI agents to do it for them and correctly interpret the output. This will become a challenge across the enterprise, separating out people who are unable or unwilling to embrace these new technologies in favour of others who do.

Types of AI

In my opinion, the term AI is very broad and doesn’t provide a clear definition of the underlying toolsets. There are many aspects to AI, and generative AI – where the current excitement is centered – is only one variation. Other notable AI technologies include machine learning, decision management, interactive agents and speech/image recognition. As engineers, we have to understand the underlying principles of each of these, and their differences, in order to apply the technologies correctly.

Information process flow

I am a process engineer by training and therefore I imagine a manufacturing plant to consist of a number of process flows that run in parallel. Two important and relevant flows are the material flows and information flow.

Material flows are tangible and have attributes such as composition, mass, temperature and pressure. Information flows, in contrast, are invisible and intangible. They have these attributes:

Timeliness: Information must reach the recipients within the prescribed time frame.

Accuracy: Information is said to be accurate when it represents all the facts pertaining to an issue.

Relevance: The information should be relevant to the situation or decision at hand.

Adequacy: Adequacy means information must be sufficient in quantity.

Completeness: Information is complete when there are no missing parts of the data.

Explicitness: Information should be clear and easy to understand. It should not be ambiguous or open to multiple interpretations.

Exception based: Information should highlight deviations from the standard or expected results.

Infusing AI into manufacturing essentially means infusing AI into the constant streams of information flowing through a factory. The AI technologies mentioned above each need to be applied correctly to the attributes of information flows above.

For example, AI can help summarise a random stream of IoT data so that it becomes explicit and easy to understand. This is where machine learning or generative AI tools like Copilot might, in future, have a significant role to play.

This information flow model of a plant is a conceptual framework that helps understand how AI could be applied in practical terms to a manufacturing operation where real-time data flows in information streams. However, correctly applying the appropriate tool is necessary to solve specific problems. To actually implement these technologies, engineers need to understand the underlying technology fundamentals, just as a process engineer needs to understand how a centrifugal pump works in order to specify the correct pump for an application.

I strongly believe that we are only at the beginning of understanding the practical value of AI and its applications. Those who dismiss AI in manufacturing as mere hype are mistaken this time. There are many use cases. The issue is the scarcity of new skills to bring these ideas to reality.

Fasten your seatbelts, hold onto your hats

According to the same Accenture study mentioned above, the current AI transformation process will likely take less time to disrupt industry than digital transformation. It seems that when we are only just getting to grips with digital transformation, things are about to get interesting again. AI is moving quickly and the stakes are higher than ever. Now is perhaps a good time to seek out training opportunities to better prepare you as an engineer for the next five years.

About Gavin Halse

Gavin Halse.
Gavin Halse.

Gavin Halse is a chemical process engineer who has been involved in the manufacturing sector since mid-1980. He founded a software business in 1999 which grew to develop specialised applications for mining, energy and process manufacturing in several countries. Gavin is most interested in the effective use of IT in industrial environments and now consults part time to manufacturing and software companies around the effective use of IT to achieve business results.

For more information contact Gavin Halse, Absolute Perspectives, +27 83 274 7180, [email protected], https://www.linkedin.com/in/gavinhalse/

Share this article:
Share via emailShare via LinkedInPrint this page

Further reading:

PC-based control for waterjets
Beckhoff Automation Editor's Choice
Control systems are subject to very specific demands from every form of production technology, including waterjet cutting. IGEMS from Sweden set itself the goal of taking this process to a new level in terms of precision and speed, and achieved it with the help of PC-based control.

Case History 193: The big dipper.
Michael Brown Control Engineering Editor's Choice
This article gives an example of an important flow control system for flare gas in a petrochemical refinery. The control was suffering from serious problems, often cycling severely, and at other times the flow process variable (PV) would suddenly move quite a lot away from the setpoint (SP).

Edge computing: Introducing AI into the factory
Editor's Choice IT in Manufacturing
As AI evolves, it is evident that the most powerful models will be cloud-based, and hosted in data centres that are beyond the control of the average business. The practical application of AI in manufacturing control and automation will only be possible if some of the computing workloads can be brought onto the plant, inside the firewall and inside the plant network.

36 years of innovation and success
SAM Systems Automation & Management Editor's Choice System Integration & Control Systems Design
Systems Automation & Management was established in 1988 at a time when there were no other systems integrators (SIs) in the process business. SA Instrumentation & Control’s editor caught up with managing director, Claudio Agostinetto to find out more about how this thriving company has prospered over the last 36 years.

The world’s greatest model railway
Horne Technologies Editor's Choice Motion Control & Drives
Located in Hamburg’s traditional warehouse district, Speicherstadt features the largest model railway in the world, and is one of the most exciting tourist attractions in Germany.

Loop signature 23: Tuning part 1.
Michael Brown Control Engineering Editor's Choice
This is the first of several articles dealing with the subject of tuning. I have found that many people think that optimisation consists solely of tuning. I would stress once again that tuning is the last thing one should do when optimising regulatory controls.

Plastics meets packaging for consistent and efficient process control
Beckhoff Automation Editor's Choice
PC- based and EtherCAT-based control and drive technology from Beckhoff represent a universal solution that transcends industry and application boundaries. This standardised and scalable automation platform offers numerous advantages. Industry experts delve into how machine builders and end users in the plastics and packaging industry can capitalise on these advantages.

Continuous corrosion resistance
ifm - South Africa Editor's Choice Sensors & Transducers
The polypropylene version of ifm’s LDL400 conductivity sensor is based on the proven LDL200 inductive conductivity sensor. Its material properties make it the ideal choice for applications in which metallic sensors tend to corrode.

Control architecture leads to faster, easier product development for refrigeration
Opto Africa Automation Editor's Choice IT in Manufacturing
What’s the secret to providing superior service and staying competitive in a changing market? You might learn something from ALTA Refrigeration’s experience. Over ten years, it transformed itself from a custom engineering services company into a scalable industrial equipment manufacturer, using an edge-oriented control architecture to manage a growing installed base.

Step into the visual factory
Turck Banner Southern Africa Editor's Choice Electrical Power & Protection
At Banner, the visual factory comprises three key applications for lighting and indication in industrial settings. These applications include the ability to help machines and workstations quickly communicate their status to people nearby, to use light to guide workers to perform certain tasks such as part picking, and to provide illumination for work areas and tasks.