Focus on smart coatings: LLM fuels AI innovations in the coatings industry

Sarah Gibbons reports for PPCJ on how Large Language Models (LLM) is fuelling AI innovations that could transform the paint and coatings sector

From formula composition and supply chain logistics to regulatory compliance, defect identification and corporate reputation, paint and coatings industry companies are exploring the growing potential of artificial intelligence (AI).  

With AI taking many different forms, key to the paint and coatings sector’s adoption of more widespread AI options are LLMs (large language models), such as ChatGPT, GPT-4, Llama 2 and Mistral 7B: “By using LLM you can easily allow users to interact with AI in a human-readable way instead of using code, as well as allowing AI to work with vast amounts of unstructured data which may be in various different formats,” London-based AI consultant Garuka Serasinghe who works with AI consultancy Futura, told PPCJ.   

Dr Robail Yasrab, Senior Researcher in AI at the UK’s University of Cambridge, explained: “LLMs can help synthesise vast amounts of paint industry research data, identifying trends, innovations and potential improvements in paint formulations or manufacturing processes.”  

And he said by analysing and interpreting data from manufacturing pipelines, LLMs can support and identify issues that lead to defects or deviations in paint quality, thus developing more active quality control procedures and analytical maintenance strategies.  

However, he said, “the most prominent factor that could be taken care of by LLMs are the future enhancements in customer engagement” through personalised recommendations, virtual colour consultations and prompt responses to customer queries. There is also important potential for integration with virtual reality (VR) and augmented reality (AR) systems to optimise customer experience when selecting products.  

Serasinghe told PPCJ that, in the R&D phase, AI can be used to analyse many combinations of different chemicals and advise on which combination works best.  “For example, if we were to look for a light blue paint, which is mould resistant, AI can do its magic to come up with the best possible formulation for that,” he suggested.  Add in data from previous invoices and other raw material information and more detailed product viability can be established, he said.  

READ MORE:

It’s a wrap! Middle East Coatings Show welcomes record number of visitors to Cairo

Using machine learning or deep learning solutions alongside LLMs, he said can maximise data processing power: “You can submit a whole bunch of previous reports [and] data to a software solution, and then ask questions like: ‘There is a new [EU] directive on CO2 emissions on paint manufacturing, limiting CO2 emissions to 100kg for 1,000 litres of paint. Given the previous CO2 emissions from using various power sources including main electricity, solar and wind, suggest the best option to achieve those emission targets with lowest cost to the company.”  

Kurt Cagle, an AI consultant from Seattle, USA, CEO of Semantical, an enterprise data hubs and metadata management consultancy, told PPCJ that LLMs can be trained to identify which material combinations would violate environmental laws. He said significant efforts are going into building interfaces between AI systems to allow more data querying that would benefit the coatings sector.  

For example, he said, regarding logistics, asking an LLM system a consignment’s location will not work, but that system now knows where to ask for that information.  

That would be the ‘mixture of experts’ (MOE), a machine learning technique that leverages the power of collective intelligence by merging insights from various specialised models. That includes expert databases, graphs and other LLMs “and ask for a map to be generated using that data to show expected delivery dates, weak points regarding port back-ups, revised alternative routes or material suppliers”.  

Serasinghe said LLMs quickly analyse customer feedback, channelling it to the appropriate mechanism, for example complaints or refunds, adding: “Understanding natural language, it can monitor social media to prevent wildfire issues spreading which can be damaging to a company’s reputation if unchecked.”  

And he agreed with Yasrab who suggested evolution in the LLM space in the next five years will focus on integration with other forms of AI, such as computer vision and predictive analytics, “leading to more broad and multi-faceted AI applications” in the coatings industry.  

A blog from analytics platform www.datatobiz.com said last July (2023) that computer vision and AI are being used to analyse images in bulk to develop smart coatings: “By understanding the functioning of the microstructures in the materials, a manufacturer can not only create more durable and better-quality paints but can also optimise production to reduce input costs,” it said.   

READ MORE:

PPG completes US$2.7M expansion of powder coatings factory in Sumaré, Brazil

Serasinghe said such AI combinations can also identify defects and be used for predictive maintenance to keep machines working with less downtime.  

Meanwhile, Pittsburgh, USA-based AI solution provider Cognistx has launched a deep learning system using raw material and plant conditions data to help secure consistency in pH, colour and paint viscosity to reduce waste and associated costs.  

ALSO, AOM-Systems, based in Heppenheim, Germany, has launched a laser-based sensor and measurement system ‘SpraySpy’, using its integrated ‘SprayAI’ function to monitor paint droplet application.  

SprayAI algorithms detect even the smallest anomalies in spray according to set parameters, in real time, helping prevent many common paint coating defects.  

Monitoring the layer thickness enables it to detect if a minimum or maximum layer thickness is reached, which can be essential, for example in the application of a fire protection coating, said a company note.  

Also in Germany, Hirschau-based Dorfner, a raw minerals solutions provider, has partnered with generative AI platform for materials and chemicals Citrine Informatics, from California, to optimise paint formulations.  

A company note said: “Dorfner can now identify optimal ingredient combinations, refine formulations, and address customer requirements with greater precision and efficiency than ever before,” using AI algorithms and advanced data analysis based on a decade of historical data.  

Another service is the ‘Snap It Button’, from Ohio, USA-based paint producers Sherwin-Williams, which instantly turns any online image into “a personalised paint palette”, said a company note. The system lets the user select an image and instantly identify the Sherwin-Williams paint colours that correspond to the colours contained within the picture.  

The button can be downloaded onto a browser so any online image can be viewed and analysed.  

Meanwhile, French home improvement company Bricorama, based in the Paris suburb of Noisy-le-Grand, has joined forces with Dublin, Ireland-based global professional services company Accenture to launch ‘pAInt,’ a generative AI-powered shopping assistant that helps customers with their painting projects. It suggests trending styles, helps the purchase of materials and offers best practice guides on painting.  

A company note said ‘pAInt’ is a “conversational tool that customers can interact with for help with paint colours and finishes, decorating ideas from Bricorama’s video content, demonstrations, paint quantity, brushes and other accessories”.  

With research ongoing around the world, Dr Erik D Sapper, associate professor at California Polytechnic State University, predicted more areas of the paint production process aligning with diverse AI solutions soon to overhaul much of the industry’s traditional make-up.  

He told PPCJ: “Currently, automated laboratory equipment and processes require many person-hours of engineering, tinkering, programming, and troubleshooting. When human users can tell roboticised labs what to do in natural human language, the adoption and efficiency of digitalisation and roboticisation will flourish.”    

The coatings industry, it seems, is about to be front and centre of the inexorable drive to integrate AI into business.