Could machine-driven algorithms revolutionise the future of coating design?
Could machine-driven algorithms revolutionise the future of coating design? Jake Holmes finds out…
“Those familiar with the mechanical manufacturing sector will know of the great demand for tough metal components,” says Professor Yi Qin from the department of design, manufacturing and engineering management at the University of Strathclyde. According to him, this remains true when it comes to materials for complex engines. For these types of materials, coatings able to prevent the effects of corrosion, oxidations or other potentially damaging factors are crucial.
Qin and his department are responsible for the development of a new computer model which predicts coatings for specific applications for exact circumstances. The programme aims to remove the need for trial-and-error testing of coatings for critical applications.
Hot forging dyes have gained popularity for engine moulding and metal components, but if a coating is removed from this process the product life will be significantly shorter, he says. “Adding coatings is not just a mechanical need,” outlines Qin, “but also a business need, as shorter product lives will result in more expensive repairs and more downtime, leading to less productivity.”
Coatings in themselves are not necessarily a one-stop-shop answer to these concerns, however. Some coatings can contain toxic properties making them unsuitable for certain applications. Qin outlines the term “Green coating” to highlight coatings that do not have these toxic qualities and are safe for human interaction.
Green coating not only has application benefits but also environmental, as these coatings do not add to greenhouse gas emissions. These types of coatings are more sustainable and have the added benefit of being cheaper than counter coating.
CODING COATINGS
Qin’s research was made possible through the University of Strathclyde receiving a £600,000 fund from the European Union’s Horizon Europe fund. Being part-funded by the EU allows Qin and his team to access researchers across the continent to assist with research and progress the project faster.
“We can also gather all the experts we need from different countries,” Qin says. “Of course, another plus point is through the network, people exchange experience and also learn from each other.”
The team’s computer programme uses a material design approach, creating a design based on a given element. The design determines the properties of coatings. When the intended coating’s scale is completed, the programme will calculate and adjust its structure accordingly, whether for micro or nanoscale applications, or the largest available scale.
At the selected scale, the software can predict protein performance. When processing the coating, the model takes into account the design material to ensure its suitability for the material being applied. The programme can make further predictions on coating layers and identify suitable materials as substitutes after conducting initial coating analysis and making predictions.
This process can create a sample coating. The programme can make characterisations of this coating, providing the model with an additional dataset to work with. While real-time data is being created, users search the dataset for different resources.
DATA BEHIND DESIGN
The model uses a massive data set to come to its conclusions. “All our effort goes into the data,” Qin says. “But of course, we need to update the data continuously, because that is what’s important. We use AI and machine learning techniques, and we have some experts in this project to focus on the data quality and also the creative side for the data management and then system, to ensure that there is enough data for us to use for discoveries.”
First, a demonstrator is selected to help create forging tools for forging dyes and cuttings tools. This is typically done for hard-coating proteins, as these tend to be more challenging. Coatings for such materials and tools must be able to withstand extreme conditions such as 900°C heat and dry cutting.
Qin explains: “Then, we go back to look at the product we’re predicting the coating for. Then comes design, then we think about the design alloys or coating, and even improve the coating processes, and then optimise the coating layers and how they are combined with sub materials.”
The ultimate feature of any predictive computer program is if it works. With a programme that has too many variable outcomes to test them all, there needs to be a robust way of ensuring the algorithm gives reliable results and does not make costly or potentially dangerous mistakes.
“To make sure the model is accurate enough to give results that are reliable, we have a lot of modelling work, but there is also the parallel development,” Qin says. “This will have the validation processes which produce samples, numbers, validation, and maps, to tell us to verify models or improve models. There’s a lot of iterations developed and designed for this project.”
AI BEATS TRIAL-AND-ERROR
Despite Qin saying AI does not work as fast as people may expect, it is still significantly faster than using trial-and-error testing. The computer programme can significantly cut down lead times as it can assess and predict coating needs much faster than any other method, leading to reduced lead times and higher levels of productivity.
Another benefit of not having to test coatings through trial and error is less waste. Using the model can result in significantly reduced raw material waste, as the computer can tell the user the exact measurements which are required for making coatings. This is of massive benefit to industry, as this reduces the costs of coatings and allows companies to produce more coatings with the same quantity of raw materials.
Safety may also be an unforeseen positive side effect of the programme, as workers are no longer at risk of being exposed to toxic coatings. The programme can now predict what coating combinations are toxic, enabling workers to properly prepare with protective gear or select an alternative coating.
Sustainability is at the forefront of everyone’s mind. Being able to predict toxic substances and therefore avoid producing them will help to reduce greenhouse gas emissions and achieve a net zero status more quickly in the coatings industry.
