AI company AHX.ai, a manufacturing intelligence company, has a mission to transform efficiencies in the wood-based panels manufacturing industry.
The company had an unusual beginning. Haidin Rashid Amin, CEO of the company, was actually involved in developing a new adhesive for the industry after his studies at Imperial College, London.
Mr Rashid Amin, with a background in engineering, design and innovation, was exploring the possibility of building a new type of adhesive from micro-organisms, an alternative to the bio-based and lignin/ tannin-based adhesives that have entered the sector.
In theory, the content for adhesives can be found in micro-organisms, promising unlimited production through fermentation.
It was 2017 and the fledgling project won a few grants from Dyson and Shell LiveWire. He joined an accelerator company, Deep Science Adventures, where he joined up with Adrien Hitz, who had just finished his PhD at Oxford University. Mr Hitz had carried out research on predicting the performance of steel coils in a steel manufacturing process with machine learning models.
“If you want to develop the adhesives and optimise them you need to understand the manufacturing process inside and out,” explained Mr Rashid Amin.
“Years on and thousands of experiments later we had developed an adhesive that was quite impressive in that the pressing time was close to UF and ready to commercialise it. But we needed a lot of investment.”
Ultimately, the adhesives business was transferred to Canadian biotechnology company Lallemand in 2021, which is currently conducting numerous trials with the technology.
It was the thorough research of woodbased panels production processes for the adhesive, with the accumulation of data, that directly led to a pivot in the two men’s approach and ultimately to the birth of AHX.ai.
FORMATION OF AHX.AI
“So, we understood the wood-based panels manufacturing process, having worked with CF2P in France and got access to a lot of data. We saw raw material quality changes all the time in the line, changes to the recipe and constant deviations from the actual targets, so you generally use a lot more resources to make that same product – more adhesives, more energy, and more maintenance.”
Mr Rashid Amin knew that others had investigated AI in the industry previously, with mixed results.
“We have been using AI for a very long time and knew it had a good shot of being successful. Our approach was if you can access all available data across a factory from raw material inputs to the final product you could in theory build a good quality indicator. To do that effectively at a good commercial scale you need to be incredibly transparent in terms of the quality you predict versus what actually happens.”
It’s a big characteristic of AHX.ai’s approach that it looks at AI as a continuously evolving system, which can be counterintuitive for manufactures to grasp.
“We have this radical transparent approach to whatever we predict, the customer sees and directly compares it to the ground truth – like lab results. You are able to improve the models significantly over time.”
Mr Rashid Amin said AHX.ai can predict production quality really well most of the time, but there are always some occasions where it can’t – and that is an opportunity to learn with investigating the cause of that mis-prediction.
“Maybe there was a sensor failure, or the raw material changed too dramatically, so that the model couldn’t catch up.”
On occasions where there is no known cause for mis-predictions, it gives the user a sense of caution that AI doesn’t know everything.
After quality predictions, the second AHX.ai component is cost estimations, a real opportunity to estimate the real time cost of the factory.
“You need to have a good indicator of quality, and you need to have a good estimation of how costly these processes are.”
AHX.ai has real-time intelligence apps – analytics for simulations and optimisation, a production scheduler, asset maintenance, production emission indicator, and autonomous decisions engine. It presents itself as an intelligence platform that turns live data streams quickly into predictions, 24/7.
When Mr Amin and Mr Hitz, a statistician and AI specialist, presented at the International Wood Based Panels Symposium in Hamburg in 2024. They posed questions such as:
- “What is the current quality of my product?
- “What is the current cost per cubic metre of the product I am producing?”
- “Which processing parameters can I adjust to reduce costs without compromising the quality of the product?”
They showed figures for financial savings which could be made from AHX.ai – potentially €500,000 a line annually for a production output of approximate €70-80m turnover. You can see how savings can mount up with multiple production lines.
“Over time if you really embrace AI the savings compound. If you have the quality prediction, cost estimator, recommendations and real time chat intelligence, then the savings can really go up quite exponentially.”
Mr Rashid Amin said due to production variables the savings could vary from one year to the next. “It’s more of an art than a hard science, but the clients that are paying us each month would not be paying us if they weren’t making savings.”
AHX.ai says the cost of poor-quality manufacturing can range from 5-30% of gross sales for manufacturing, while static scheduling can cause the loss of up to 15% of revenue, and the cost of unplanned downtime can cost between 5-20% of production capacity.
“So, you need to have a good cost estimator and a good quality indicator. Once you’ve got these two you can start to build a recommendation system on top. That is quite critical. If the quality is too high, you may need to speed up the line or reduce the glue loading, but you don’t exactly know by how much the recipes needs to change. With a good AI, you can build a very good recommendation system, but it is technically a very difficult problem.”
He encourages producers to embrace AI to help staff become more data-driven, so they rely on data to make decisions. The information generated compounds the longer the approach is used.
Mr Rashid Amin also said companies begin to attract the right sort of people into their companies. “You attract the right type of people because you do not bring people in that say they’ve done a certain approach in another factory and that’s how were going to do this here.”
AHX.ai’s recommendation system tells the operator what they should be doing, but it isn’t yet able to generate recommendations in all eventualities. Out of context issues can be a problem, for instance.
“But if you have this radical loop of transparency and you get your people in your company acquainted with AI then the savings are crazy.”
Operators can work with their intuition and also with the indicators generated by AI. By working with each other, you make much better decisions on the production line and it is more transparent.
“The operator can make a decision, and we directly monitor it – what was the decision, what was the recommendation and what was the output. This way a company learns.”
Mr Rashid Amin emphasises that more data does not mean more intelligence. Sometimes having very good small data steps that are well structured and informative of the process is better than large datasets.
“Our approach to AI in the manufacturing sector, is you cannot build these large AI models that are trained on all available data. What you want to do is build AI models specifically to that company, which makes that company more competitive than their competitors.
DIFFERENT APPROACHES TO AI IN THE WOOD-BASED PANELS INDUSTRY
Mr Rashid Amin recognises three different approaches to AI in the wood-based panels industry.
The first is the panel manufacturers wanting to build their own in-house systems. This self-reliance approach may typically see panel manufacturers hire some specialists from a local university and build their own AI.
The second is the machine manufacturers, who develop their own AI on top of the systems they already have in place in the market – essentially adding value to what they already do.
AHX.ai sees itself in the third camp. “We work as a third-party company that is impartial to this industry.
“Each approach has its own pros and cons. In the future we believe there will be many different solutions and whichever one makes the best decision shall win.”
Mr Rashid Amin says AHX.ai is not overly promotional in the space but has connected with a lot of stakeholders and values trust.
“I am a true believer in open markets, but you get companies who feel they need to protect what they have. If you embrace our super transparent approach, then it forces you to innovate.”
Currently, AHX.ai is working with several European wood-based panels producers, including a major German company.
“We have managed to establish ourselves against all odds in a very conservative industry and the reason why we do this is because we know the process well, so we have credibility. It is an industry that needs to embrace technology faster.”
AHX.ai’s proof of concept (POC) work in adjacent industries, such as gypsum and steel, sees a lot faster response from firms.
“From first meeting, to POC, to commercial it is a lot faster,” said Mr Rashid Amin.
“The process in the wood-based panels industry is very large and complex. In Europe you have to build trust. It is a very traditional, family-owned industry, and we’re trying to help them over a long period of time. A lot of things need to align to build these systems. If you meet the right owner, it can move really quickly. If you start working with public companies then that is the other extreme in terms of approvals.
“There will be a turning point over the next five years, with AI, data integration, access, cloud connectivity, where you see an influx of technology all of a sudden digitalising everything in the span of two to three years. But everyone is trying to learn now, and we want to learn with them and prepare for those days.”
Mr Rashid Amin credited former European Panel Federation managing director Clive Pinnington with showing interest in his work with adhesive research, remembering how the latter visited him at Imperial College in the early days, recommending he join last year’s Symposium.
He summarised by saying the industry needed to be able to move faster to adopt technologies.
“A lot of things are coming together for AI – from government policy to overseas competition. The last bit is the human aspect, people in the organisations are slowly getting more acquainted with this. It is going in the right direction.”
