I recently hosted a Coffee With Digital Trailblazers on AI-era transformation and the advantages and challenges of Industry 4.0. Most of the recordings from the Coffee Hour are available to StarCIO Digital Trailblazer Community members, but we had some issues with this recording. So, why not a blog post?

Our discussion covered manufacturing use cases for AI/ML, Industry 4.0 versus 5.0, and the business value and challenges from these types of investments.

AI and ML use cases in manufacturing

“When you think about Industry 4.0, Industry 5.0, and smart manufacturing, you have a lot of moving parts,” says Joanne Friedman, PhD, CEO of Connektedminds and Principal, Smart Manufacturing. “ People can make mistakes, processes can be inaccurate, the potential for cyber-attack, and even raw materials going into a machine may not be exactly within the expected tolerance. Add to that new devices such as  Autonomous Mobile Robots (AMR), humanoid robots, cobots, and a plethora of other technologies. We get into things that are more physics-based AI, than GenAI to help workers identify root cause of a failure quickly and not only figure out who was on the line in what shift but if it was part of a changeover, [a different product being made] or the need for maintenance.”

Experts on the Coffee Hour shared their top AI and ML use cases in manufacturing, from the basic to the large-scale smart manufacturing.

John Patrick Luethe shared this basic example that we can all relate to whenever we dive into a broken system and try to identify the root cause.

1. Issue root cause – When something breaks, if you know what part went wrong, there’s a repair manual that allows you to fix it immediately. But many times, the real effort is spent in the diagnostics of determining which parts went wrong and how it failed. Where I’ve seen machine learning to be incredibly valuable is sending the fault codes off into the system, having the system review them, and just telling the repair person what part needs troubleshooting.

Martin Davis, CIO and managing partner at DUNELM Associates, shared three examples of quality, production, and predictive maintenance.

2. Identify defects faster– A manufacturer of electric bus batteries wanted to understand the optimum speed of the production line of the circuits and the batteries related to this. These batteries are absolutely massive, and the number of cells within them is huge. If you get to the end of the production line with one defective cell, that could lead to either scrapping the whole battery or disassembling it to try and find which cell is defective. The ability to use machine learning to detect a problem early in the cycle as individual cells are created has a massive saving to that company, to their productivity, and to the long life of their batteries.

3. Optimize production – Continuous process type lines, such as in a fry company, and looking at the golden batch aiming to identify the set of settings that give you the optimal result. The combination of the speed of the line, the temperature of the fryers, the amount of steam applied when you’re peeling, and many other parameters go to make the optimal output while making the best use of the raw material coming in, which is obviously variable. With all these variables changing very rapidly, how do you understand, for a certain type of input, what gives you the best output?

4. Smart predictive maintenance – Paper machines with massive rollers; a bearing might cost you $50K. If that bearing fails and takes out other parts of that machine, it could be $500M to repair, and downtime could cost $150K an hour. If you are able to predict when that bearing is starting to fail – maybe it’s starting to create more heat in the bearing, or it’s starting to have more vibration. If you can start to look at the patterns of behavior from multiple sensors and predict when a bearing is starting to fail, then you can plan maintenance and replace it before it has a catastrophic failure.

Joe Puglisi, growth strategist and fractional CIO at 10Xnewco,  shared another quality example.

Joanne Friedman expands on the quality challenge:

 6. Smart manufacturing – A large factory has many moving parts of technology and equipment. You want to get to the root cause of a failure very quickly because the costs are huge—$2.37M per hour times the number of hours to repair. Today, we can use machine learning and AI to look at the contextual and simulated environments through a digital twin. Number one, did a person make a mistake? Did the raw materials coming into the process contribute to a failure? Today, we can leverage all the different variables involved in the different stages of the manufacturing process equipment, and all the data that comes out of this analysis can help companies be anticipatory, not just predictive.

Industry 5.0 defined and contrasted with Industry 4.0

I asked Joanne to explain the differences between Industry 4.0 and 5.0 briefly.

“Time to data, time to decision, and time to overall value are the three pillars underpinning Industry 4.0,” says Joanne. “Industry 5.0 is not only human-centric and human in the loop but looks to add human intelligence and institutional knowledge and look at the best way to accomplish the goal as opposed to just automating processes.”

Joanne explains where Industry 5.0 provides long-term business value. She says, “OEE metrics, that is, operational efficiency and effectiveness, become the bridge between how companies can gain cost savings and growth – and growth includes humans, how we retrain our workforce, how we look at creating what’s called regenerative businesses and using corporate resiliency as your north star in doing the transformation.”

Benefits of Industry 5.0 and smart manufacturing

Joanne also shared two clear benefits of Industry 5.0 and smart manufacturing.

• “We’re beginning to see the emergence of what was not traditionally considered knowledge workers, becoming knowledge workers. The cost of labor and the shortage of labor are pushing a lot of manufacturing companies to look at robots and cobots as a better way to do the mundane or the not-so-interesting roles in a factory. How can they upskill that workforce that used to work on  factory lines in new ways? Taking advantage of that institutional knowledge of someone working in production for 20 or 25 years is incredibly valuable. I’m hearing about programs that are taking the workforce off the line and putting them into new roles where they’re being taught how to be the source of data for machine learning. It’s the machine that’s learning from the human, not the human learning from the machine.”
• “One of the things that was missing or very hard to create under the Industry 4.0 paradigm as opposed to smart manufacturing (a subset) is the necessary feedback loops between the customer or the customer’s customer and the earliest point of ideation in design and engineering. That’s where I think this new level of knowledge working is really going to come into play because it’s not just about automating the processes to get more products out the door. It’s also about improving the design of those processes of those products, making them more sustainable, and lowering the overall energy usage of the factory.”

Challenges of AI in manufacturing

We also covered some challenges in applying AI and ML in manufacturing. “Sometimes deep learning cannot explain why a decision is made based on some inputs, so interpretability has been an issue,” says Moutushi Dey, PhD. and innovative tech engineering leader. “We may solve this with more interaction between the manufacturing and IT industry so that experts can go in a feedback loop, share their knowledge, and learn from each other.”

Moutushi adds that a second problem is that manufacturing and AI are very energy-intensive and require technical skills. “What happens to the very small farms and small manufacturers? We have to make AI accessible to everybody.”

This week, all these amazing Experts participated in a new episode covering green technology and platforms. We covered the challenges of creating business cases and how to scale green initiatives. You can access the recording on the StarCIO Digital Trailblazer Community.