Colgate’s ‘Aha’ Moments as Digital Twins Improve Manufacturing Results
Add bookmarkAs embedded sensors make real time tracking of data a reality, many manufacturing companies are looking to harness that information to understand what influences throughput, quality and consistency of results.
Enter the digital twin.
A digital twin is a virtual representation of a process, product or machine. Coupled with the high availability of machine data and machine learning, these virtual representations help companies form a picture on how different variables affect the outcome of a process.
At IX Network’s recent online event, Digital Twins in Manufacturing, Darren Haverkamp, Technical Director for Hill's Pet Division of Colgate, presented how the consumer products giant has been using a digital twin in their manufacturing operations in order to improve quality and reduce raw material loss. (If you’re interested, you can watch the full presentation here.)
The division has applied a digital twin at the start of their pet food manufacturing process. It all begins begins when raw materials such as rice or corn are brought into an extrusion process. After the extrusion process there is a 30 minute drying period where the materials are screened and then topicals are applied.
At that point, the material is allowed to cool down to ambient temperature. This takes an additional 20 minutes.
It is only when the product is cooled – about 50 minutes from the start of the process - that it can be sampled for quality.
“Oftentimes you find out that you have a problem and you have 50 minutes of product,” said Darren Haverkamp, Technical Director of Colgate-Palmolive. The aim of using a digital twin, he adds, is to be able to predict what influences the outcome earlier so that quality consistency can be improved.
That’s complicated with pet food manufacturing as it relies on a range of commodities such as rice and corn. Those raw materials have natural variation between suppliers and crops and can change over time. The time of the year when crops are harvested or processed, for instance, may impact on how they behave during the manufacturing process.
Equally, the machines used to process the materials change over the course of their lifecycle. Increased wear and tear on machines may influence the end results.
“We wanted to take in all those variables and really learn at a higher level from those variables so that we could predict where that product was going to be in 50 minutes,” explained Haverkamp. “We found a digital twin was a good way to do this.”
The company chose to partner with GrayMatter, a manufacturing services company that offers advanced industrial analytics.
To begin, Colgate-Palmolive first needed to identify all the potential variables that could affect the outcome. Havercamp says that they targeted extrusion first to look at how to improve the consistency of the results.
“In a start up is when you have the greatest opportunity for material loss and the greatest opportunity to improve,” he observes. The company brought began with as much data as they could get from smart sensors embedded in the machines to start training the digital twin on how historical operating parametres have affected outcomes.
“A human mind can only accept so much data upon which to make their decisions. Within the digital twin you can flow in as many variables as you can with these sensing devices that are starting to come on line within the last couple of years,” says Havercamp.
“We’re starting to take that data and push it into the models to help us predict even more the quality of our output. We’re using next generation sensors to do that.”
Sensors help them to understand how the health of the equipment – how it operates differently over time - impacts the quality of the product the machine produces.
Based on that data and machine learning, the digital twin has been able to identify the impact of different variables and can prescribe optimal operating parametres.
“Eventually what we’re learning to do is not only prescribe the adjustments to the operator but automate those decisions so that you can close the loop. Not only are you prescribing where to put your moisture settings or your horsepower settings, but you’re automating those to change those within the machine themselves and take those decisions away from the operator. And that’s where we’re at today,” says Havercamp.
The most difficult part of their digital twin implementation, Havercamp explains, had nothing to do with the technology. The cultural shift required – and ensuring that operators felt they could trust the data, that they were not overloading operators with too much information or making them feel like this was yet another application to learn – was the hardest part of the journey.
“You want to provide just the right amount of data to allow people to react. Create the right KPI reports and make those reports visible to the appropriate audience,” he says.
Additionally, you want to make sure that the application is easy and useful enough that operators don’t feel like it’s another complicated app for them to learn to use. Instead of a ‘push,’ the use of the digital twin should be a ‘pull.’
A big ‘aha’ moment was when they realized that the digital twin doesn’t need to get perfect results – it just need to get better results than historical human averages.
Another ‘aha’ was when they realized that it was critical to find the right technology partner that could help them leverage their own expertise.
“The challenge oftentimes with these types of technology is not to search for the next shiny object but to search for what fits your needs,” says Havercamp. “We’re the pet food experts. [GrayMatter are] the digital twin experts. You need to bring that expertise together to solve that problem.”
“We realized very quickly when we started into the digitization journey that we’re not an analytics company. We’re a CPG,” adds Warren Pruit, Vice President - Global Engineering Services at Colgate-Palmolive. “Instead, we partnered with the experts.”
Colgate is now looking to scale the pet nutrition digital twins into the formulation process to better predict how a particular formula will perform before even beginning manufacturing. The company is also looking to roll out digital twins in their soap extrusion models in Mexico and digital twins for energy optimization.
Interested in Learning More About this Topic?
Hear from leading technology vendors and digital twin experts at our upcoming Digital Twins in Power and Utilities event on December 2, 2021. Hear practical advice and get your questions answered by reserving your FREE spot here.