With all the hype and accompanying controversy surrounding AI, it is natural – and even sensible – not to take all its promises at face value.
The manufacturing industry is generally slower to adopt new technologies, especially when it comes to heavily regulated markets such as pharmaceuticals. And while it is important to note that we are all still getting to grips with it, AI is not magic. Its adoption comes with its own challenges, but AI is here to stay. To avoid playing catch up, the manufacturing sector needs to understand and acknowledge the benefits it can bring, and to embrace it with a degree of caution.
The Manufacturer sat down with Group Chief Digital and Information Officer, Balajikasiam Sundararajan from ACG about this.
Can you give a bit of background in ACG and your role?
BS: Founded in Mumbai in 1961, ACG now serves pharmaceutical and nutraceutical companies all over the world, and touches almost every aspect of solid dosage manufacturing – from providing capsules, to manufacturing machinery and equipment, to protective packaging materials, to traceability solutions.
My current role in ACG involves; driving digital led transformation spanning transforming operations, enhancing customer experience, building connected products and services, and creating digital led business models. Information technology management comprising IT infrastructure management, IT application management, enterprise resource planning, cyber security. As well as, industrial automation initiatives covering AGVs, industrial robotics, end of line automation, end to end lights out automation initiatives.
Digital transformation of operations and increasing smart manufacturing is a huge focus area at ACG. We look at what we can do with technology, especially IoT and AI, for our existing brownfield plants. We have 20 manufacturing plants around the globe across the different business areas. We are looking at implementing these technologies, focusing on enhancing productivity, safety, quality, planning, maintenance, and supply chain operations.
How can adoption of AI benefit manufacturers and do most manufacturers realise what they are?
The landscape of AI can be confusing, and many manufacturers do not have a full understanding of what the benefits of AI are. The larger and more progressive businesses are figuring this out, but many are still questioning what AI can do for their business.
AI can be deployed across all business functions, from planning to production, but first manufacturers need to identify the right use cases that makes sense for the business.
If manufacturers approach AI partners, they will tell them what they can do from a technology standpoint, such as deploying reinforcement learning and machine learning, which is good, but it doesn’t explain how that will make sense for the business. The business needs to look at what to improve and how to get from where they are today to where they need to go using AI technology. In a lot of cases, technology is implemented, and then dropped off unless a clear pathway is set.
However, implementing AI on the manufacturing shop floor will not happen in two or three months, it can take years and will need the perseverance of the business. One of the biggest challenges will be change management. Businesses will have to look at current operators who have been doing the same work for 30-40 years and introduce a new, unfamiliar system which can be difficult to accept.
What challenges currently exist around AI deployment and adoption? Do you have any of your own use cases you can share?
At ACG, we have capsule manufacturing which uses gelatine as a raw material – it has its own natural variation because it comes from different sources. Our capsule manufacturing is very fast, producing 3,300 capsules per minute, per line from machines that range from 20 years old to brand new. Considering these factors, we wanted to know how we can produce capsules with the best yield, quality and OEE. At the time, much of this was dependent on the skill of the operators who were ensuring smooth production. However, looking at the system, we wondered if we would be able to build a system to recommend the perfect settings for operations. Considering the product, raw material, and operating conditions, could the AI system provide us with the right settings for the line to produce the perfect outcome.
We began by collecting more than 700 process and machine parameters per line to create a digital twin model. This took nine months before we were able to give the first recommendation.
The first time the model was used for production, the outcome was nowhere close to where it should be and experienced operators told us that it wasn’t going to work. However, we knew we were heading in the right direction, and it would take a while for the model to mature. We had 26 production lines in this plant, and we took help of plant operations personnel to work with us and refine the model multiple times as needed.
After the initial nine months, it then took another seven to create a prediction which was accurate and close to how an expert operator would run the machine. For this, we had to work closely with the operators on the shop floor to make it a success. Tough conversations were had on the shop floor, but it was soon realised that there were people working hard to try and make this work.
Throughout this use case, we found the fundamental need was data and we needed at least 12 months’ worth to allow the model to start providing something relevant. Even though the data was good, the model still required a human in the loop. We still have an operators look at the settings recommended by the AI system and apply the settings to the machine.
What advice would you give to manufacturers embarking on AI deployment?
It’s going to take time for things to mature, improve and deployment to quicken. But right now, if you want to implement AI applications that will result in serious operational impact, which takes a lot of data from the machine, manufacturers need to be realistic and acknowledge that this implementation is a very intense process, and it is going to take time.
Can you explain what the Global Lighthouse Network is?
The Global Lighthouse Network is a network of progressive manufacturers who use emerging listed technologies in depth and at scale, creating significant business impact.
As of today, there are 153 manufacturing sites across the globe that are part of this network, and companies apply through a comprehensive online application process.
Applications to the network are shortlisted based on four main criteria. The first is integrated use cases and how good they are. They should be significant and should impact the business in a strong manner. The second is about enablers. A company must have the right enablers within the structure, the people, the tech ecosystem and partner ecosystem so it can sustain the technology long-term. The third, and most important, is impact – financially, operationally and from an environmental sustainability point of view. And fourth, to evaluate all the tech platforms that are being used.
Once shortlisting happens, an expert team from the World Economic Forum visit the sites in person for a day. From this, a site record is created and this is presented to an expert jury panel from the industry and academia. This jury panel then votes if this company can become part of the network.
What’s next for ACG in terms of Fourth Industrial Revolution technology deployment?
New technologies are emerging at an exponential pace. I am most excited about the fusion of AI, digital twin and industrial automation. This means:
- Digital twins powered by sensor fusion, real-time data acquisition, modelling and visualization technologies help create a digital replica of the physical environment.
- Artificial Intelligence (AI) in its different forms (computer vision, machine learning, deep learning, or generative AI), helps in intelligent decision making.
- Industrial automation, whether through conventional actuator systems, automated guided vehicles, robotics, quadrupeds or humanoids, enables actions based on intelligent decisions made by AI.
The fusion of these technologies enables fully autonomous lights out factories. At ACG, we are on the journey to make this happen for our future manufacturing plants.
for more articles like, visit our Industrial Data and AI channel.
