Perfect Colors with Digital Assistant CoatinoTM

At home, in the car or while shopping, voice-controlled digital assistants have become part of our daily lives. In the future, they will also simplify laboratory work. Specifically, a pioneering digital assistant designed to help scientists in the paints and coatings sector is called Coatino™. Virtual assistants bearing names like Siri and Alexa are part of the family in one out of every five American homes. Boasting infinite patience, they accept every voice command given and, if everything works properly, play the desired music, provide weather forecasts or update the user's calendar. Voice-controlled digital assistants promise to help us in our daily lives. Their use is spreading rapidly. In 2016, they were found in only 1% of homes in America, but just two years later this figure rose to 2%. In Germany, digital assistants are currently used in 13% of homes. If everything develops according to the wishes of Dr. Gaetano Blanda (Senior Vice President & General Manager, Evonik Resource Efficiency Business Line Coating Additives), this will only be the beginning. Dr. Blanda wants to transform voice-controlled digital assistants from simple helpers in daily life into chemistry experts. He wants to use these experts in laboratories—places where comprehensive specialist knowledge is needed and a technical language is spoken. Digital laboratory assistant: Thanks to modern methods in high data science and artificial intelligence learning fields, Coatino™ always recommends the most suitable additive from a very wide range of options and enables laboratory staff to work in a goal-oriented manner to achieve success faster. Source: Evonik

As Senior Vice President & General Manager of Evonik's Coating Additives Business Line, Dr. Blanda is aware of the great needs and demands that new laboratory assistants must meet regarding scientific expertise and language skills. His team is expert in formulations in the paints and coatings industry and related additives.

In order to fully meet customers' demands and wishes regarding color, gloss and durability, experts in laboratories need to create complex mixtures by adding the correct additives. There are thousands of possibilities—actually, far more than the human brain can handle. Depending on these possibilities, experts spend a lot of time searching through notes and data sheets at hand. For precisely this reason, a digital assistant would help users directly research and adjust raw materials in the laboratory. The assistant's name is Coatino™. The idea for Coatino™ was born during a strategy meeting of the Coating Additives department. Business Development Manager and Project Manager Dr. Oliver Kröhl says, "We talked about new ways that the business could be developed. Innovations are no longer limited to just the final product or processes. Instead, you need to demonstrate your abilities by finding solutions in the form of new services and business models." Researchers focused on the daily challenges faced by laboratory experts working with coating and paint formulations. Soon they decided that scientists would be able to use a voice-controlled digital formulation assistant. This idea excited the team greatly. But was this really what the experts needed in the laboratory?

From a Paint Can to a Prototype

The scientists decided to simply start working on the idea. They painted an empty paint can in the company's colors. Then they placed it in the laboratory and filmed a conversation between a colleague and the paint can. In the video, the user asked the paint can for a suitable water-based foam inhibitor recommendation for wood coating. The paint can responded by providing product options to the laboratory worker and placed a sample order. Kröhl says, "At that point, a colleague standing behind the wall answered the questions. Although this was a bit unplanned, we wanted to test our idea concretely with customers and get feedback quickly." The video was shared with a certain number of customers and the team conducted conversations with customers based on specific questions. The idea generated great interest. Customer validation of the idea encouraged the developers to enter uncharted territory. Kröhl says, "We are paints and coatings experts, not voice-controlled assistant experts. For this reason, we knew there was a possibility the project might fail. But we and our customers thought there was a huge potential. We were willing to take the risk." Their goal was to develop a prototype assistant in time for the European Coatings Show, the world's most important trade fair for the paints and coatings sector. This was not an easy task because traditional speech recognition systems could not use technical terms and vocabulary. Kröhl says, "Simply put, normal assistants don't understand our language." For example, when you tell them about dispersion, rheology or silicone resin, they quickly reach their limits and at best can only provide general information.

"Traditional assistants can't understand us." Oliver Kröhl, Project Manager of Coatino™

Kröhl says, "To prepare a coating formula, assistants need to be able to do much more. If they don't know the properties of the components and how they interact, they won't be helpful in the laboratory."

A Test Track for Coatings

For testing paint formulas, the high-throughput experimentation and testing unit within the Paints and Coatings Additives Business Line provides an important data repository for the industry-specific digital assistant. Located at Evonik's facility on Goldschmidtstrasse in Essen, this unit determines raw material dosages, formulates them to create coatings and describes the properties of the final coating product. This entire process is carried out fully automatically according to a strictly defined program that is always reproducible. As a result, the unit can prepare an average of 120 sample formulas in 24 hours. The results are always visible and reproducible. If this data is linked with Coatino™, customers can be provided with daily updated information about their personal coating formulations. Test track for paints: With a fully automated paint formula testing system, Evonik reduces the time needed to find the right formula. In the system, more than 100 samples can be tested within 24 hours. Source: Evonik

Prevention of Foam, Scratches and Streaks

Paints, varnishes, lacquers and other coatings consist of four basic components: solvents, binders, pigments and additives. For example, a solvent keeps a wall paint in liquid form. After the paint is applied, the solvent evaporates and allows the paint to dry. Pigments give the paint the desired color. Binders are used to ensure the paint adheres to the wall and remains effective. This component is colorless and binds the paint to the surface. Additives make up the smallest percentage of a formula. Although their share is less than 5%, they play a key role. Additives eliminate foam created when paint is applied. They prevent the paint's pigments from clumping. They make coatings thixotropic, meaning they make them easier to apply, while preventing streaks from forming as they dry on vertical surfaces. Other additives make coatings more resistant to scratches. Different components in a formula affect each other's effects depending on the mixture. The number of combinations is very large. In the development of a coating formula, if only 10 curing materials, 10 binders, 10 pigments and 10 additives are considered, what this numbers mean is 10,000 different combinations. Moreover, this calculation does not even take into account variations in the proportions of the components used. Blanda says, "Customers have very specific ideas about the properties a product should have when it is completed." To develop a voice-controlled, functional assistant for the coating sector, scientists initially began structuring all available information and entering it into a massive database. In the next step, they enabled retrieving this information using a voice control function.

Working for Global Use

For example, if you ask the assistant "Which additive is suitable for printer ink?", the system needs to understand every word. Among other things, Coatino™ had to learn that "additive" describes paint components in a specific category. In the next step, the assistant needs to access available data, search through it, create appropriate connections and display the data in possible relevant results. To do this, it first breaks down the sequence of sounds into its smallest components and performs a data search based on characteristic features. Another challenge for the assistant is that Coatino™ not only needs to understand plain, straightforward German names, but also other forms of words. Scientists also want to ensure that the speaker's dialect or accent does not prevent the result. The ultimate goal is to make Coatino™ understand the pronunciation of customers worldwide. To this are added different speech rates and voice pitches of speakers as well as the special context of speech. Kröhl says, "The testing process is very challenging. After we finally succeeded in a trial with a colleague in Shanghai, we failed with our colleagues in Essen." For approximately two years, Coatino™ has been developed and worked on together with Evonik's Paint Additives department and an external development company based in Berlin. When the prototype assistant was presented at ECS, it passed the first important development test. Coatino™ has a lot to say. When asked about suitable additives, it not only presents a product list, but also prioritizes the products. Blanda says, "Coatino™ can tell me which additive would be most suitable for my formula and requirements. In this way, it can make well-founded recommendations to me." When a user finds the product they are looking for, they can give Coatino™ a voice command to place a sample order, retrieve the relevant technical documentation directly via email, or arrange a meeting with an assigned specialist. Kröhl says, "For us, customer-focused digital solutions enable people to communicate more efficiently with each other about innovative solutions."

New Formulas from the Database

The Coatino™ prototype was ready just in time for the opening of the European Coatings Show. Blanda says, "We immediately presented it to a select group of our customers." Test users conveyed their requests through a microphone to a tablet instead of a paint can. "The feedback was even better than we hoped. We were able to win some of our customers as the first users to test the assistant." These customers will pass their experiences on to Blanda and his team. Blanda says, "We wanted to involve customers at an early stage of such a project. Such a project can only be implemented if customers also think the project will be useful to them." Researchers plan to make Coatino™ available for the entire coating sector by 2020. However, for the system to be developed further, there appears to be no final end in sight on the horizon. Kröhl says, "When you use digital assistants, you constantly find ideas for new features." For example, Coatino™ will probably not only provide existing formulas, but could also suggest its own new mixtures. Scientists and technicians will be able to directly test these mixtures in the laboratory and develop them for their own use. Blanda says, "One day our Coatino™ will truly be an artificial intelligence entity. But until then, we have a long way to go." [gallery ids="eyJ1cmwiOiJodHRwczpcL1wvd3d3LnR1cmtjaGVtLm5ldFwvd3AtY29udGVudFwvdXBsb2Fkc1wvMjAxOVwvMDhcL2dhZXRhbm9fLmpwZyIsInRpdGxlIjoiZ2FldGFub18iLCJjYXB0aW9uIjoiRHIuIEdhZXRhbm8gQmxhbmRhPGJyPlNlbmlvciBWaWNlIFByZXNpZGVudCAmIEdlbmVyYWwgTWFuYWdlcjxicj5Fdm9uaWsgUmVzb3VyY2UgRWZmaWNpZW5jeSBCdXNpbmVzcyBMaW5lIENvYXRpbmcgQWRkaXRpdmVzIiwiYWx0IjoiIiwiZGVzY3JpcHRpb24iOiIifQ==,eyJ1cmwiOiJodHRwczpcL1wvd3d3LnR1cmtjaGVtLm5ldFwvd3AtY29udGVudFwvdXBsb2Fkc1wvMjAxOVwvMDhcL29saXZlcl8uanBnIiwidGl0bGUiOiJvbGl2ZXJfIiwiY2FwdGlvbiI6IkRyLiBPbGl2ZXIgS3JvZWhsPGJyPlZpY2UgUHJlc2lkZW50IFN0cmF0ZWdpYyBCdXNpbmVzcyBEZXZlbG9wbWVudDxicj5Fdm9uaWsgUmVzb3VyY2UgRWZmaWNpZW5jeSBCdXNpbmVzcyBMaW5lIENvYXRpbmcgQWRkaXRpdmVzIiwiYWx0IjoiIiwiZGVzY3JpcHRpb24iOiIifQ==,eyJ1cmwiOiJodHRwczpcL1wvd3d3LnR1cmtjaGVtLm5ldFwvd3AtY29udGVudFwvdXBsb2Fkc1wvMjAxOVwvMDhcL2R1eWd1Xy5qcGciLCJ0aXRsZSI6ImR1eWd1XyIsImNhcHRpb24iOiJEdXlndSBcdTAwZDZ6Z1x1MDBmY24gPGJyPkJ1c2luZXNzIE1hbmFnZXIgIDxicj5Fdm9uaWsgUmVzb3VyY2UgRWZmaWNpZW5jeSBCdXNpbmVzcyBMaW5lIENvYXRpbmcgQWRkaXRpdmVzIiwiYWx0IjoiIiwiZGVzY3JpcHRpb24iOiIifQ=="]