Will AI Soon Be Generating Color Formulations?

Florian Streifinger about the use of artificial intelligence in the plastics industry.

Florian Streifinger wanted to know whether plastics manufacturers were already using artificial intelligence or at least testing its use, and what potential they saw in it for their work. The result surprised him.

Have you looked into ChatGPT and other AI technologies yet? I took a look at OpenAI’s much-praised wunderkind some time ago. And was impressed and disappointed in equal measure. Impressed because the intelligent helper indeed compiled very detailed and appropriate summaries to some of my questions. Even for a rather complex topic where an Internet search would have taken quite some time. ChatGPT disappointed me as, elsewhere, it blithely added incorrect and inappropriate information.

So my experience with the chatbot chimes with that of other users. The dispute prompted me to ask colorant manufacturers about the use of AI for the current trend report on coloring plastics.

I wanted to know whether they are already using such technologies or at least testing their use, and what potential they see for their work. The results surprised me. No fewer than half of the companies surveyed are already involved in this area. Some of these are smaller mid-sized companies that you might not have expected to be doing this at first glance. Their considerations for AI use are also directed at the companies‘ core competencies with color formulation and product development. Incidentally, an article also shows how color formulations can be created with AI.

One of the companies may even conceivably create trend color collections using artificial intelligence in the future. So maybe by next year you will already be reading a trend report with colors selected by AI. But for now, I wish you interesting insights with an issue that is still exclusively human-made.

16:9

Editorial Streifinger

Streifinger, Florian

“We See Our Work as Basic  Digitalization“

Christian Schiller wants to make the phrase “plastic waste is recyclable material” a reality. His tool of choice is digitalization. He is keen to help plastics recycling achieve the final breakthrough – but not everyone is on board.

Christian Schiller’s goal is to turn the phrase “plastic waste is recyclable material” into reality. His tool of choice: digitalization.

The story of Cirplus starts not in a garage, but on a sailing trip between Panama and Colombia. While at sea, founder Christian Schiller and his crew were engulfed in a huge carpet of plastic waste and algae. “At that moment, I realized I could utilize my experience in software development to help fight the plastic crisis,” says Schiller. In the interview, he discusses recycling challenges, proving himself during the Covid pandemic and the long-term goals of the “New Plastic Heroes”.

Kunststoffe: Despite the availability of modern technologies, why does plastics recycling remain so challenging?

Christian Schiller: If you look at the mechanical and chemical recycling of plastic waste, numerous plastic types still lack any commercially viable – let alone profitable – recycling processes. So I would not call any of the existing technologies “modern” or even “state-of-the-art”. Moreover, if you look at the supposedly recyclable products, you quickly realize that they are often not designed for high-quality recyclability. Compounding the issue is the core problem of the considerably opaque and complex waste and recycling streams.

Schiller: In Europe alone, more than 20 million tons of waste go unaccounted for every year, so traceability of plastic flows is the key issue. To really move the needle when it comes to recycling plastic waste and optimize recycling flows, we need to pinpoint the precise location, composition, quality and quantity of each type of plastic in real time.

Kunststoffe: How can this be changed – and how does Cirplus help?

Schiller: Cirplus paves the way to track what kinds of plastic waste and recyclates are available or traded, where and under what conditions. Our software also integrates ever deeper into the procurement process of companies to simplify the complex transaction beyond initial buyer and seller matching and make it more affordable. Building on this dataset, we are developing AI applications that can optimize waste recovery routes in real time.

Schiller: To date, more than 2500 companies have registered on the platform. As well as major packaging manufacturers and branded goods companies, we also have well-known customers from the automotive and household goods industries.

Kunststoffe: What has been the biggest success for you so far?

Schiller: DIN SPEC 91446, which we initiated and financed, is certainly one of them. It is a world-first standard for high-quality plastics recycling and digitization. In the meantime, even the VDA has taken it up and introduced it as a new guideline for enhanced sustainability for plastics in the automotive industry. Our DIN SPEC 91446 has also been included by the EU Commission when rolling out forward-looking standards to expand and improve recycling and has received several awards.

Kunststoffe: Have you encountered any obstacles or setbacks?

Schiller: In March 2020, following the Covid-19 outbreak, our lead investor withdrew a financing commitment that we had considered secure and irrevocable. Four days before the notary date! At that time, we had just EUR 15,000 in our bank account and I had no idea how we were going to pay our staff. Even worse, I had just hired two more people with the funding round in mind. It was a huge effort to keep the company alive: Everyone gave up half their salaries without compensation and I had to let the two new employees go immediately. We just remained afloat, thanks to Covid subsidies and a DBU (German Federal Environmental Foundation) support program. In September 2020, I was lucky enough to find a replacement investor in the form of Pale Blue Dot. It was a few months of harsh lessons learned that we survived with plenty of luck and toil.

Kunststoffe: You have been online for several years now. How has the industry’s attitude toward Cirplus and recycling changed during this time?

Schiller: Where do we start? Basically, we have succeeded in raising awareness of the crucial need for digitalization in closing the plastics loop and even if many value-adding participants are still struggling to implement this insight, more and more market players now view us as a competitive force.

Schiller: They give us additional hurdles. Our approach is not to replace companies, but quite the opposite: leveraging digitalization to help them do better business.

Kunststoffe: But surely that doesn’t apply to everyone?

Schiller: No! Overall, industry, science and politics appreciate our work. We can also do our bit to ensure recycling is no longer generally frowned upon or seen as an excuse. We show young people how to get involved in a sustainable plastics industry and make the transition from protesting to taking action. For me, the fact that we can inspire the next generation of plastics engineers, mechanical engineers, managers and investors makes all the hard work worthwhile.

Kunststoffe: Finally, where do you envision your company in five years and what are the next steps to get closer to these goals?

Schiller: In five years, our vision is that every company, no matter where they are in the world, will see Cirplus as the first and key contact point when it comes to using, procuring or distributing plastic recyclates and waste.

Hauptartikel

Circular Economy: Using Digitization as Crucial Catalyst

<ul><li>Name: Cirplus GmbH</li><li>Founded: 2018 in Hamburg, Germany</li><li>Number of employees: 11</li><li>Total customers: > 2000 companies</li><li>Homepage: <a href="https://www.cirplus.com/en" target="_blank">www.cirplus.com</a></li></ul>

Read the German version of the article in our magazine Kunststoffe or at <a href="https://www.kunststoffe.de/a/interview/wir-sehen-uns-als-grundlagendigitalisier-4774271" target="_blank">www.kunststoffe.de</a>

Digitale Transformation

Wirtschaft

Innovationen und neue Technologie

Recycling

Umweltmanagement

Start up

Unternehmen und Unternehmer

Ehrhardt, Melanie

Kunststoff-Recycling

Kunststoffverarbeitung: Verfahren

Kunststoff-Aufbereitung

Kunststoff-Rezyklat

Kunststoff Werkstoffe

Energy Consumption Reduced, Processes Optimized

Reducing energy costs is crucial for the competitiveness of many plastics processing companies. Fried uses an energy management system from Econ Solutions for this purpose – but not only for this: thanks to the connection with ProSeS, it also provides insights into the production processes.

Reducing energy costs is crucial to competitiveness for many plastics processors. Fried uses an energy management system from Econ for this purpose.

With over 40 years‘ experience in plastics technology and coating, Fried develops individual technical precision parts and assemblies made of plastic in long and short runs for many global market leaders. In so doing, Fried places great importance on a sustainable and environmentally friendly approach; since July 2021, the company has been producing in a CO2-neutral manner.

Energy management makes an important contribution to this. In 2014, Fried received ISO 50001 certification and has since step-by-step expanded the energy management system used for this purpose. Today, the company records the energy data of its entire production with 142 measuring points.

 The Econ Sens3 energy and power measuring device from Econ Solutions is used for this purpose. It can be installed quickly and in a space-saving manner and then measures the key electrical energy parameters. Since the measuring device comes with all common interfaces, additional measurement data can also be integrated and forwarded to higher-level systems. All data flows into the Econ4 energy management software, and gas and water consumption are added manually on a monthly basis.

Integration of ProSES for More Insights and Process Understanding

The software merges the measurement data at system level with production and process data from various company systems and creates analyses and reports from them. Fried has connected the MDE system ProSES with Econ4. This enables the plastics specialist to determine order-related energy variables and provides a fact-based foundation for more accurate calculations. “The higher energy prices are, the more important it is to take them into account appropriately in quotations. This was not possible with the consumption measurements from the supplier and connected load values from the machinery manufacturers but now we can quantify the energy costs fairly accurately based on the measurement data,” explains Dr. Andreas Tewald, Energy Management Officer and Head of Work Preparation, Facility Management, Maintenance/

Repair at Fried. He adds, “There is another advantage of integrating ProSeS into Econ: the overview of our processes has expanded so that we can understand them even better, for example by comparing different processes on the individual machines.”

Key Performance Indicator “Energy Input per Processed Quantity of Pellets”

Weekly pellet consumption is also entered manually into the energy management software. This provides the energy input per processed quantity of plastic pellets. “A key parameter for our calculations,” says Tewald.

Likewise, the system can also be used to provide other key figures that are relevant for the particular company. What these are depends on various factors, such as the production process and the machines and systems used.

In addition to the key figures, Econ4 software also provides the load profile as well as consumption, cost, and CO2 reports in almost any time resolution, plus special and management reports. Among other things, these show the main consumers, the energy flow in the company or correlations between several data points.

On this basis, Fried has already implemented numerous efficiency measures and significantly reduced its energy consumption. For example, the analyses made it obvious how energy could be saved in the automatic coating process by optimizing control parameter settings – and without any loss of quality. The situation was similar for the compressed air system with frequency-controlled compressors. “We didn’t even know before that we could reduce our energy consumption here. But Econ not only shows the starting points for efficiency measures but also allows us to precisely track and document the effects,” explains Tewald.

A look at the reports also reveals findings that are not primarily related to energy efficiency. This was the case at Fried when the energy consumption of the coating shop rose disproportionately. The search for the cause led to a dirty filter. The company now uses this fact to replace the filters as needed based on the measured values. “This means we no longer have to adhere to the specified maintenance cycles, which makes maintenance much easier and more cost-effective,” says Tewald. Twice a year, Fried uses the analyses to determine the overhead costs for the production machines and adjust them if necessary.

To minimize expensive peak loads, Fried uses Econ Peak peak load optimization. It collects the quarter-hourly values of the production loads from the energy supplier‘s meter. If the threshold value defined by Tewald is exceeded, the load management system automatically switches off the loads that were previously classified as unimportant for a short time. In addition, the energy management officer receives a notification and a visual display appears in production.

If the threshold is always exceeded at the same time, it is worth checking whether the production sequence can be adjusted. This was the case at Fried. A peak load was always reached when starting up the machines and systems after the production-free weekend. Now the machines are switched on with a time delay and the peak load is avoided.

However, determining the peak load was not a simple matter. This was because Fried operates a photovoltaic system and feeds the electricity generated into the grid within the meters but “behind” the peak load calculator. During production hours, the company purchases the electricity again; on weekends, it is sold. This electricity had to be included in peak load control, otherwise the determined value would always be too low by this amount. Since the peak load control can only handle a limited number of input signals, a signal summer was used. “This was not easy in terms of measurement, but Econ provided us with excellent support,” says Tewald.

Energy Efficiency Is Part of the Decarbonization Strategy

Since July 2021, Fried has been using only certified green electricity to cover the electricity requirements that cannot be met by its own PV system on the plant building. This is supplemented by eco-gas, so that all production processes from raw material receipt to shipping are 100 % CO2-neutral.

The company offsets the CO2 emissions of the natural gas with certificates based on the Life Cycle Assessment concept of the German Federal Environment Agency, which are secured through Green Book management by TÜV Süd.

The load curves make anomalous consumption patterns immediately visible.

The Econ Sens3 energy and power meter can be installed quickly and without disruption, providing immediate transparency on energy consumption.

Econ Peak enables optimized dynamic load and charge management and supports a wide range of electricity meters, wallboxes, as well as charging stations, power generation and storage systems.

Econ Optimizes Energy Management at Fried Kunststofftechnik

Rolf Wagner is authorized signatory and head of sales at Econ Solutions GmbH.

Econ Solutions offers solutions for effective operational energy management. The open plug & play software and hardware are flexibly scalable and can work on their own or integrate existing hardware or systems from other manufacturers. Services include integration and customizing, system sizing and planning, commissioning, software maintenance, and training and workshops. More than 600 companies rely on Econ software, including BASF, Continental, ebm-papst, Trumpf Werkzeugmaschinen, Unilever, and ZF TRW.

Read the German version of the article in our magazine Kunststoffe or at <a href="https://www.kunststoffe.de/a/fachartikel/energieverbrauch-reduziert-prozesse-opti-4774263" target="_blank">www.kunststoffe.de</a>

Kunststoffverarbeitung

Dienstleistungen

Lohnfertigung

Energiemanagement

Energieeffizienz

Consumers’ color preferences change over time. At the same time, the choice of colors and colorants is influenced by the materials used. This is currently the case, for example, due to the increased use of recycled materials. The Kunststoffe international editorial team asked colorant manufacturers about current color trends. We also wanted to know what role the carbon footprint plays in their products and whether they already use artificial intelligence.

Colorant manufacturers inform about current color trends and the role of the carbon footprint in their products.

The relationship between customer preferences and the color of products is an ambivalent one. On the one hand, color choices influence customers’ perceptions of products, deciding whether they are seen as valuable, exciting and sustainable. On the other hand, customers’ lifestyles and desires create color trends, which are then reflected in the appearance of products. Ultimately, manufacturers want their products to address and serve consumers’ needs and preferences. Jessica Knoch, marketing manager at masterbatch manufacturer Lifocolor (Fig. 1), puts it very well: “Colors always influence and express our attitudes, emotions and actions.”

According to her, global crises are having a particularly strong impact on trend colors. The war in Ukraine and other global conflicts have created a longing for peace and harmony. The Covid 19 pandemic is also still having an impact. That’s why, according to Knoch, soft, subtle and neutral colors and mid-tones are particularly in demand (Fig. 2).

 These convey a sense of well-being, peace and harmony with oneself and the environment. Appropriate colors are, for example, warm gray, brown and beige tones, but also strong mineral rust red and certain blue tones. For more striking accents, she says vivid colors that stand for communication and dynamism and generate energy are ideal. “Bright orange and mid-blue, intense purple and pink, and vitalizing turquoise will definitely provide incentives at points-of-sale,” she says.

Sami T. K. Palanisami, Senior Global Marketing & Sales Manager for Non-Polyolefin Colors & Additives at US additives manufacturer Milliken, makes a similar prediction. For him, the current color trends embody a desire for cooperation and creativity. Every year, the company puts together a trend collection, the so-called ColorDirection collection, of colors that are particularly in demand. This time, it also includes strong shades, such as a corresponding pink, orange and red.

Recycled Materials Drive Demand for Lighter Colors

For Philippe Lazerme, Head of Marketing Plastics at Heubach (Fig. 3), color trends clearly depend on the field of application. In consumer goods, he sees more colorful products on the rise. In the automotive sector, he says, interest in orange is very high. Orange is used in electric vehicles as a warning color for high-performance components. The strong demand for it has already been unbroken for several years. Lazerme also sees no great change in the color scheme for plastic packaging. There, the colors of the respective brands continue to predominate. However, with one important change: “We are noticing that new packaging is being colored with lighter tones. This has to do with design for recycling. Lighter colors are considered easier to recycle,” he explains.

However, because many plastic products have been darkly colored in the past, many of the recyclates produced also have a dark hue. In addition, mixing plastic waste of different colors often results in recyclates that are gray in color. To recolor them, dyes and colorants are needed to cover dark shades and grays. “Recyclates are already loaded with color due to the conglomerate of more or less colored and reprocessed products from the collection. They must therefore be overcolored. Since the recyclates bring with them different inherent colors – from lighter to darker, gray to colored qualities – colorants with a color intensity as high as possible are required,” says Marco Meixner, head of research and development at Lifocolor, describing the problem (Fig. 4).

Less Color Variety with Recycled Plastics

This is also confirmed by Felix Finke, CEO of Karl Finke. According to him, in some cases the colorants also have to be dosed higher to cover up the inherent color of the recyclates. “Coloring a recycled material is more demanding than coloring virgin material. We need pigments that can cover the often grayish hue of the recyclate, and we dose the masterbatch higher accordingly,” he explains. This also has an impact on color selection, according to him. Customers request less brilliant colors, for example.

The use of recyclates can also mean that product manufacturers have to cut back on color selection, as some shades cannot be realized. “Not every desired shade can always be set exactly because of the recyclate type and inherent color. The manufacturer may have to compromise on the color shade,” says Meixner from Lifocolor.

However, colorant manufacturers have adapted to this and in some cases developed special solutions. Heubach, for example, offers masterbatch manufacturers a free file with calibration colors that can be used to achieve specific color settings quickly and easily, even with dark recycled materials. At Karl Finke, one of the methods used is layer combinations. The functionally sensitive layers are made from virgin material, the others from recycled materials. For example, the outer layer, which is relevant for the appearance, consists of virgin material, while the rest of the product is made from recycled waste. This ensures a high degree of design freedom, which is not restricted by the inherent color of the recyclate, and at the same time allows the use of a high proportion of reprocessed waste plastics. In addition, higher colorant concentrations and intensities are not required for coloring.

The concept can also be applied to additional layers. If, for example, packaged products must not come into contact with recyclates, the packaging can be implemented in a three-layer variant, with a recyclate layer between two made of virgin material. Karl Finke developed a coextrusion process for this some time ago and offers a special compound that is said to be suitable for two- and three-layer combinations. “In this way, our customers can use recycled goods and thus meet EU sustainability guidelines, while at the same time choosing one of numerous color shades for the 20 percent decorative layer,” says managing director Finke, summarizing the advantages.

With the increasing recycling of plastics, the stability of colorants and their impact on the recycling process are also coming more to the fore. During both processing and reprocessing, not only the plastics but also the colorants are subject to thermal and mechanical stresses. Stability plays a role here primarily for two reasons: if it is too low, the colorants can decompose and release harmful components. In addition, stable colorants ensure a more uniform coloration of the recyclates, since their color tone does not change as a result of decomposition. They can therefore be recolored more easily and with lower colorant concentrations.

According to Stefano Bartolucci, Market Segment Manager Plastics of Lanxess’ Inorganic Pigments Business Unit (Fig. 5), inorganic pigments such as iron oxides have an advantage over organic ones because they are more color-stable. According to Bartolucci, the post-coloring process is also less stressful for the plastic in the case of iron oxides (Fig. 6). Another argument in their favor is the frequent use of recycled plastic waste: “Recyclates are often used for outdoor applications. Here, too, iron oxides can score points, since they are quasi inherently UV-stable. Further addition of UV stabilizers can thus be kept to a minimum,” he explains.

However, Heubach sees its own organic pigments also suitable for several recycling cycles. According to Philippe Lazerme of the pigment manufacturer, the stability of colorants is also becoming more important, which is why internal studies have been carried out on this. “We tested ten of our organic pigments and analyzed them chromatographically after up to ten extrusion processes. These pigments show no changes in the spectrum. They are therefore suitable for coloring packaging that is recycled,” he reports.

Customers Demand Product  Carbon Footprint

The manufacturers agree that the environmental impact of products is becoming increasingly important to customers. The pigment and masterbatch manufacturers surveyed unanimously report that CO2 emissions are increasingly being asked about. “Our customers are becoming more and more sensitive to this issue,” says Dirk Schöning, Sales Director at AF-Color, for example. The so-called Product Carbon Footprint (PCF) has established itself as the most relevant parameter. “We are seeing increasing interest from our customers and a rapidly growing need for reliable CO2 data, especially product carbon footprint data,” says Dr. Thomas Lindner, Vice President Sustainability at Heubach (Fig. 7). This is also confirmed by Marco Meixner from Lifocolor: “The requests for CO2 data are constantly increasing.” Stefano Bartolucci of Lanxess also assumes that “the demand for detailed life cycle assessment data across the entire supply chain will continue to grow.”

What Progress Have Manufacturers Made on the Carbon Footprint?

According to Lindner, the background to this is increasing legal regulations, but also the greater consumer interest in such information. “In our opinion, the growing demand is due on the one hand to increasing regulation, but also to the significantly growing awareness on the part of consumers. Data on the environmental impact of products is playing an increasingly important role in purchasing decisions,” he summarizes. According to Marco Meixner, the required or desired compensation of CO2 emissions for products is also relevant for some companies.

To meet these customer requirements, colorant manufacturers are currently recording the carbon footprint of their products. The extent to which they are doing this varies greatly. Some already offer detailed and certified data for some of their products, while others are initially limiting themselves to new products or are planning to implement this in the future. In many cases, however, the aim is not only to provide customers with information, but also to use the data to reduce their own CO2 emissions. “The PCF makes sustainability measurable and provides us with reliable data for future emission savings,” explains Lindner, for example.

In addition to the perennial topic of sustainability, artificial intelligence (AI) has also established itself as an important topic in society and business in recent months. The hype was triggered by the chatbot ChatGPT presented by the US company OpenAI, which vividly demonstrated the current possibilities of AI technologies. There are also already ideas on how AI technologies can be used for coloring plastics. The Kunststoff-Zentrum Leipzig, Germany, for example, has developed a neural network in a project to improve color formulation. Envalior is also already using an AI tool for this purpose (see article on p. 16 in this issue).

Among colorant manufacturers, opinions on the use of AI differ. Half of the companies surveyed do not see the technologies as relevant to them or have not yet looked into them. Others recognize great potential for certain areas of application. The aforementioned creation of color formulations is frequently mentioned. “AI offers enormously great development potential in generating recipes for color development– especially where it is necessary, for example, to minimize metamerism effects,” explains Dirk Schöning of AF Color. Lifocolor’s Head of Development Meixner agrees: “We can well imagine that AI will also be adopted by masterbatchers. There are already software applications that assist with recipe calculations.”

And he can also imagine other areas of application, such as in trend color research or for product development. The latter is currently being examined at Karl Finke. According to CEO Felix Finke, the company already uses AI technologies in market analysis and internal process control, and is “currently evaluating the use of AI-supported optimization in the area of product development.”

In addition to the great opportunities offered by AI, there has also been a lot of talk recently about risks. Frequently, for example, the question arose as to whether the technology will replace humans. At least for the plastic color sector, Dirk Schöning of AF-Color gives the all-clear: “In the end – fortunately – the eye of the colorist is indispensable when it comes to fine-tuning.”

Fig. 1. Jessica Knoch, Lifocolor: “A bright orange and medium blue, an intense purple and pink shade, and vitalizing turquoise will definitely provide incentives at the point of sale.“

Fig. 2. At Lifocolor, soft and subtle colors are seen to be particularly on the rise in the near term.

Fig. 3. Philippe Lazerme, Heubach: “New packaging is being colored with lighter tones. This has to do with design for recycling.”

Fig. 4. Marco Meixner, Lifocolor: “With recyclates, it may not be possible to set every desired shade. The manufacturer may have to compromise on the color shade.”

Fig. 5. Stefano Bartolucci, Lanxess: “Recyclates are often used for exterior applications. Here, iron oxides score points because they are quasi-inherently UV-stable.”

Fig. 6. Inorganic pigments such as iron oxides are said to be particularly suitable for recyclates because of their high color stability.

Fig. 7. Dr. Thomas Lindner, Heubach: “The PCF makes sustainability measurable and provides us with reliable data for future emission savings.”

KUint5-23_Farbe-Trendbericht - Hauptartikel

Color Trends for Plastic Products

Masterbatches

Additive

Farbstoffe und Pigmente

Farbkonzentrate (Batches)

Farbstoffe

Farben, flüssig

It is a true art to create color formulations for high-performance materials and engineering plastics, while meeting all the requirements for mechanical properties and the necessary temperature resistance. Envalior is taking the next step for color development: With the help of a large treasure trove of historical data, the AI tool Lucidiris is designed to significantly shorten development times for color formulations.

How the AI-based formulation developer Lucidiris reduces development times for color formulations.

It is quite a challenge to develop the right color for engineering materials, such as high-performance polyamide and polyester based compounds. There is a limited amount of high temperature resistant colorants available and at the same time, effects on material performance must be minimized. That can be a puzzle since many operable colorants are of crystalline inorganic nature, which can adversely affect the glass fibers and therefore lower the mechanical properties.

At Envalior, combining formerly DSM Engineering Materials with Lanxess High Performance Materials, experts develop color types based on a new experimental approach. However, it is difficult to utilize the experience of all colleagues, and the results of thousands of historical experiments performed.

The solution was the development of the AI tool Lucidiris. This AI based recipe builder is fueled with all historical data and combines it with prior knowledge, physical models, and advanced algorithms. The software is now used by Envalior and its customers for various color development requests. This collaboration with AI supports the reduction of the number of experimental iterations required and shorts lead times (Fig. 1).

A color development request from a customer starts with the specifications of color values for a selected grade with material properties suitable for their application. These color specifications can be expressed as L*, a*, b* values under a certain light source, providing a color standard or a reference target is handed over. In the latter case, the color, opaqueness, and surface quality values are determined and translated to the base material.

After the final intake discussion, the color specialists start to design experiments. Next to their experience and reasoning, they have access to color test plates ranging from tens to hundreds depending on the required grade. Often, the color of a test plate that appears closest to the customer color target is used as a starting point. If such a recipe does not meet the target on mechanical properties, the recipe needs to be designed from scratch. In several iterations, the recipe is altered to minimize the color difference with the target color values. This is mostly done by changing the concentration of one colorant at a time.

For every experiment, the recipe is compounded and injection molded into standardized plates, and finally tested on color. When the color values are too far off, the development cycle repeats until the L*, a*, b* color values are within specifications.

In color development, there are some other challenges that can arise before customer acceptance, and some are specific for high performance materials. When the color target is met, several mechanical tests are done on specimens of the developed colored compound. If the mechanical properties are deviating too much from the specifications, the color development must be performed all over again. This can happen when inorganic crystalline colorants are present that can damage reinforcement glass fibers and subsequently affect the properties.

Organic colorants are often no viable alternative since they can have stability issues when processed at the required high temperatures. Another difficulty can be that the color of the target plate and the developed plate are perfectly matched with the applied light source, while they are very different when tested under other light conditions. This phenomenon is called metamerism and is in most cases not acceptable for the customer, which implicates additional development iterations. For all these reasons, the overall color development process can be time-consuming, expensive, and can have a long lead time. It can even result in failing to meet the multi-dimensional specifications window.

Archived Data Is Considered  “the New Gold“ for AI

As a material producer, Envalior owns a huge amount of data – thousands of recipes with measured color characteristics. For many of those recipes the material properties have been tested as well. Data is an important raw material to feed AI. Together with the know-how of material and color experts, it was possible to develop the virtual assistant from this, which predicts achievable colors and suggests recipes, while taking the material properties, such as tensile strength, tensile modulus, elongation of break, Charpy notched, and reflow into account.

This novel digital concept makes use of several combinations of advanced machine learning algorithms. However, it all starts with collecting all historical data in a structured and machine-readable form, according to FAIR standards (Findable-Accessible-Interoperable-Reusable) [1]. All available data including full reflection curves are incorporated and cleansed with outlier detection methods. A new data workflow has been designed to incorporate every new experiment into the database and used to build and update the models. This data structure by itself already brought useful insights allowing to visualize correlations with a random forest model.

Quantitative models such as neural networks (NN) and Gaussian processes (GP) have also been used to make predictions about color properties when developing new recipes [2, 3]. The different models have their specific strengths for outlier detection, precision, overall accuracy prediction, calculation speed, and the ability to deal with full reflection curves to account for the metamerism effects.

In addition, a way was found to indicate a set of achievable colors, starting from a certain point in the color space. This concept can be used for virgin material, but also for recyclate-based polymers (Fig. 2). The model successfully shortens the iterative development process because the predictive model improves itself with each new experimental result (Fig. 3). These functions are used for all new color developments.

A prescriptive model provides suggestions for color recipes – based on the target color values. For this purpose, a combination of so-called evolutionary algorithms such as survival-of-the-fittest algorithms and other optimization models was used (Fig. 4) [4]. They provide a flexible equilibrium between using the current design space and exploring beyond it. The user can control the design by assigning different weights to certain parameters, thus influencing the AI‘s recipe suggestions.

The AI based color recipe builder engine developed is a complex concept. In order to enable the users to operate the system as easily as possible, an expedient user interface was developed. The first and most important feature was the option for users to select a certain material grade and build a color recipe to predict its properties. Based on feedback from previous users, Envalior has added functionality to the Lucidiris user interface, improving the user experience.

Prediction sessions can be stored and made accessible to the full color development team, and it has been made easy to incorporate results of new experimental runs to improve predictions for the next development cycle. Furthermore, it is possible to work backwards starting from target values to find nearest historical experiments easily and to suggest recipes. Next to the color itself, the effect on the mechanical properties of a given colorant cocktail are predicted as well. The user interface also shows the prediction accuracy, which makes it easier to choose recipe suggestions, which improved the effectivity of the AI tool.

Since its introduction in 2022, Lucidiris has been used consistently by Envalior for color development requests. This showed that the number of repetitions in color development and thus the development time could be reduced. These successes are a great motivation for the development team to continue along this path. Envalior is currently working on expanding the tool so that it can also support customers with non-color-related product developments.

The use of the AI-based formulation concept is possible for all applications in which formulations are developed. However, the prerequisite is to have sufficient experimental data and, if possible, a translation of raw material and application know-how into design rules. On this basis, combinations of FAIR data management, automated data workflows and machine learning algorithms lead to product development cycles (design-build-test-learn) to be fundamentally changed and development times to be shortened.

[1] Wilkinson, M., Dumontier, M., Aalbersberg, I. et al.(2016). The FAIR Guiding Principles for Scientific Data Management and Stewardship. Sci Data 3, 160018

[2] Hardesty, Larry (2017). „Explained: Neural Networks“. MIT News Office. Abruf vom 2. Juni 2022.

[3] Rasmussen, C.E.; Williams, C.K.I (2006). Gaussian Processes for Machine Learning. MIT Press. ISBN 978–0–262–18253–9.

[4] Slowik, Adam und Kwasnicka, Halina. (2020). Evolutionary Algorithms and Their Applications to Engineering Problems. Neural Computing and Applications. 32. 12363–12379.

Fig. 1. The AI-based recipe developer Lucidiris is permanently enriched with new data and thus delivers very accurate solutions.

Fig. 2. Representation of Gaussian processes in one dimension: The prediction accuracy improves with each additional measured data point. Model improvement occurs in a multidimensional space that takes into account L*, a* and b* color values, many color concentrations and mechanical properties.

Fig. 3. Representation of a predicted envelope of colors from specific starting color values for L*, a*, b*: In practice, the envelope curve is a three-dimensional volume that also includes the L* axis.

Fig. 4. Modeling workflow: In each new generation, mutations are evaluated against the predictive model until it has optimized a recipe to the virtual target values.

Hauptartikel Envalior

Plastics Technology: Finding the Right Shade with AI

<ul><li>Erwin Houben leads digitalization in R&D at Envalior.</li><li>Angelika Schmidt leads the global product and color development team at Envalior.</li><li>Chunxia Sun is global technical product manager color at Envalior.</li><li>Willem Godlieb is a data scientist specialized in machine learning and modeling at Envalior.</li><li>Arno Wilbers is a data scientist specialized in machine learning and statistics at Envalior.</li><li>Jelle de Jong is a data scientist specialized in modeling and deployment at Envalior.</li><li>Brady Bai leads the color development team at Envalior.</li><li>Youfu Chen is a color expert at Envalior.</li><li>Wang Wei is a color expert and compounding lab supervisor at Envalior.</li></ul>

More informations at <a href="http://www.envalior.com" target="_blank">www.envalior.com</a>

The authors would like to thank Sioux Technologies for their support in the enhanced algorithm evaluations.

Read the German version of the article in our magazine Kunststoffe or at <a href="https://www.kunststoffe.de/a/fachartikel/mit-ki-den-richtigen-farbton-treffen-4888907" target="_blank">www.kunststoffe.de</a>

Künstliche Intelligenz für die Qualitätssicherung

Industrie 4.0

Industrie 4.0 für die Qualitätssicherung

Simulation

Software

Metal-effect pigments make for attractive designs, but they might influence  detection when it comes to recycling. NIR sensors are unable to reliably sort plastics containing them. Eckart has now made sorting easier by developing a special silver pigment based on a combination of two conventional pigment technologies.

Metal-effect pigments make for attractive designs, but they cause problems with recycling. These can be resolved with a special silver pigment.

Near-infrared spectroscopy (NIR) is extensively used in recycling plants to uniquely identify polymers (Fig. 1). Among its many advantages are remote high-speed measurement, deep penetration of the NIR radiation and high sensitivity. With NIR spectroscopy, recycling plants are able to identify and automatically sort different polymers, so that they can be separated by resin type.

Success in the sorting of plastics waste is determined by the color, surface composition and shape of the plastics. These can influence the intensity of the measured spectra. Optical NIR sorters, for instance, have been unable to reliably sort black plastics. The reason is that the oft-used carbon black pigments absorb a large portion of the light, thereby rendering identification impossible. Solutions for black-pigmented plastics have since become available on the market.

Optical NIR sorters are also fairly unreliable when it comes to sorting plastics containing metal-effect pigments. These pigments act as microscopic mirrors and scattering centers for the NIR radiation, and that may lead to attenuation of the reflected NIR signal. In the case of conventional metal-effect pigments, that signal may experience interference. As a result, the polymer may not be correctly recognized and thus uniquely identified.

Metallic Effect Bestows an Aura  of Exclusivity

Metallic effects, however, play a major role in the finishing of products, especially packaging. Surfaces with a metallic appearance convey the impression of high value and exclusivity. Golden and silver elements expand the respective color space considerably; they can accentuate and unlock interesting new design options for packaging that appeals to the consumers’ senses and entices them to buy. In the plastics sector, packaging is therefore tinted  with metal-effect pigments that generate a silver or golden shimmer along with mirror-like and chrome effects, and create intense light-dark transitions. Combinations of metallic pigments with transparent, colored pigments and dyes offer a vast range of design possibilities.

To ensure that the desired metallic brilliance gets past the point of sale and also that the metallic-colored packaging can be recycled at the end of its life, Eckart has devised an alternative to the commercial metal-effect pigments usually employed for plastics.

A Pigment Development Brings  NIR Detectability

By combining selected metallic pigments with the recently developed Symic 604 series of pigments, the manufacturer has succeeded in producing very attractive silver shades for plastics that offer high levels of NIR detectability. Called NIR Silver, this new product range has been designed to meet the detectability and automated-sorting requirements of recycling plants (Fig. 2). The pigments reflect the specific wavelengths in the fingerprint range that are needed for correct polymer identification. This is often not possible with conventional silver pigments (Fig. 3). NIR Silver does not give rise to any signal interference. The characteristic reflection curve of the polymer (its footprint) is reproduced clearly. This ensures that the polymer is uniquely identified.

The pigment’s composition combines the best of two traditional pigment worlds: metal-effect and pearlescent pigments. The metallic look and flop of an aluminum pigment has been merged with the functional properties and radiation transmission of a pearlescent pigment.

A Silver Pigment for All Standard Polymers

NIR Silver is available in three degrees of fineness. As a result, different looks are achievable: from an elegant, texture-less silver-gray to a very industrial steel-gray to a shimmering silver-sparkle effect (Fig. 4). The pigments are easy to incorporate into all standard polymers. In particular, polyethylene terephthalate (PET), low and high density polyethylene (LDPE and HDPE), polypropylene (PP) and polystyrene (PS), which are prime recycling candidates, readily lend themselves to pigmentation with NIR Silver. The pigments are authorized for use in the food-packaging sector. They meet all regulatory requirements on product safety in key regions. Custom coloring in a very wide range of shades and effects is accessible through combinations with both organic and inorganic colored pigments. The degree of gloss and hiding power can be customized as well, thereby providing extensive formulation freedom.

Particularly when used in combination with post-consumer recyclate (PCR), NIR Silver represents a good opportunity for the packaging industry to reduce its products’ ecological footprint and to boost their recyclability. By making recycling easier, NIR Silver is helping to realize a modern circular economy that makes waste available again as secondary raw materials and enables raw materials to be used for as long as possible.

Plastic products with a metal-effect finish are attractive. However, the effect pigments make recycling difficult.

Fig. 1. NIR spectroscopy is widely used in recycling plants to sort plastics waste.

Fig. 2. The NIR Silver range of metal effect pigments developed by Eckart can be employed in various plastics. The characteristic fingerprints of the various polymers remain detectable.

Fig. 3. Conventional silver pigments interfere with the plastic’s specific fingerprint and prevent it from being recognized by sorting systems.  In contrast, NIR Silver causes very little interference and this ensures good detectability.

Fig. 4. Eckart offers different pigment preparations in the product series that enable a very wide range of silver designs to be implemented.