Chemistry Industry Solution

Acceleration of Custom-Made Formulations via Federated Machine Learning

Advanced custom-made prepolymer and polymer formulations are increasingly replacing metal in vehicle applications. Material providers are therefore asked to offer high-performing, modular formulations that meet the needs of the vehicle OEMs. The Apheris Platform for federated and privacy-preserving data science enables material providers to train and employ powerful AI models for accelerated development of superior formulations.

Image alt
Our awesome features

Situation

Vehicle OEMs require specialty materials with complex property-combinations (e.g., minimal weight materials maintaining high strength and stiffness). Two material providers are highly specialized on different materials required for advanced polyurethane synthesis (e.g., organic surfactants and low molecular weight polyols) and possess large amounts of data about the respective material class. They operate in a similar product space and it is known that their datasets are complementary. Both providers want to leverage their data for accelerating the discovery of novel formulations with defined properties. This requires novel AI models that reliably predict formulation-properties and generalize well across different chemical compound classes.

Problem

The two providers’ datasets are material-focused and highly biased. AI models that are trained on each dataset individually do not generalize well and are of limited predictive value. Training the model on both datasets would significantly enhance model performance, however the material providers cannot openly share their data because it is highly sensitive and contains intellectual property (e.g., test data and recipe information of their portfolios).

Apheris solution

The Apheris Platform for federated and privacy preserving data science allows training ML models on the datasets of both material providers without revealing sensitive information. The complementary expertise can be leveraged to generate ML models that generalize well and offer reliable formulation prediction. Both providers benefit from an accelerated formulation discovery process within a strong partnership, while maintaining their independence as no sensitive information is revealed to the other party.

Advantages of using Apheris

Discover new polymer candidates

New models are more robust and generalize better to new formulations leading to the discovery of entirely novel polymer candidates

Increase project performance

Requests from OEMs can be fulfilled faster and more effectively. The AI-based project planning tools lead to significant increases in both performance and success rate

Develop new partnerships

The specialty chemical manufacturers increase their offering portfolios and develop new partnerships with challenging OEM requirements

Our awesome features