A multiscale Machine Learning based Software for the Simulation of Catalytic Processes

Project Information

CATALYSE

Grant agreement ID: 101112677

DOI

10.3030/101112677

Project closed

EC signature date 10 March 2023

Start date 1 June 2023

End date 30 November 2024

Funded under

• European Research Council (ERC)

Total cost

No data

EU contribution

€ 150 000,00

Coordinated by

POLITECNICO DI MILANO
Italy

Periodic Reporting for period 1 - CATALYSE (A multiscale Machine Learning based Software for the Simulation of Catalytic Processes)

Reporting period: 2023-06-01 to 2024-11-30

Summary of the context and overall objectives of the project

In this project, we have further developed and optimized towards commercialization MultiCAT, an innovative multi-scale, physics-guided, machine-learning-based surrogate modeling framework for catalytic reactor simulations. MultiCAT is designed to bridge the gap between first-principles-based multiscale modeling and computational efficiency, enabling the simulation of catalytic reactors from the atomic to the process scale while significantly reducing computational costs. First-principles models offer high accuracy but are computationally demanding, making them impractical for large-scale reactor and plant-level simulations. On the other hand, empirical models are computationally efficient but lack predictive accuracy. MultiCAT overcomes these limitations by integrating first-principles modeling with machine learning, providing fast and accurate reactor simulations to support process development, optimization, monitoring, and control.

The framework includes several key features. The physics engine is a modular system that efficiently simulates detailed atomically resolved chemistries in catalytic reactors. The surrogate chemistry module automatically generates ready-to-use kinetic model surrogates, adaptable to different resolution levels from microkinetics to macrokinetics. The reactor digital twin technology allows for the creation of highly accurate yet computationally efficient models. MultiCAT represents a paradigm shift in computational catalysis by making high-fidelity multiscale modeling faster, more accessible, and seamlessly integrated into industrial workflows. The software features an intuitive graphical user interface that removes technical barriers to multiscale modeling, along with advanced tools for integrating and interpreting kinetic and spectroscopic data. Additionally, its digital twin capabilities enable real-time process monitoring and control.

MultiCAT is the only software capable of simulating an entire reactor across all relevant scales, from atomic interactions to full-scale process operations. Leveraging machine learning and physics-based modeling approaches, it extends multiscale modeling accuracy and resolution from the design phase to real-time industrial applications. Its accessibility makes it suitable even for companies without specialized expertise in kinetics and catalysis.
Based on a market study and business feasibility analysis, this project has led to the creation of a spin-off company from Politecnico di Milano for the commercialization of MultiCAT.

Work performed from the beginning of the project to the end of the period covered by the report and main results achieved so far

The project focused on upgrading the alpha version of MultiCAT through code optimization and the development of a graphical user interface (GUI). A key technical effort was beta testing and software optimization, aimed at bringing the methodologies developed in the ERC Stg SHAPE project to a commercial-ready multiscale software package. This involved optimizing the existing reactor models and refining the active learning approach to improve computational efficiency. The first-principles-based reactor models were numerically optimized by implementing advanced macroscopic reactor models, crucial for achieving physically-based machine learning representations. These optimizations enhanced computational performance and introduced more efficient numerical methods for solving discretized systems. The beta version of the software was developed alongside the GUI, improving accessibility and usability. The machine learning (ML) models were automated through a physically-based active learning workflow, iteratively refining the training dataset based on predictive performance.
The GUI development played a central role in making MultiCAT more intuitive and user-friendly. It supports all major features, including the automatic generation of ML surrogates for kinetic models with minimal user input. Additionally, a dedicated interface was developed to manage simulations via the computational backend engine, enabling visualization and post-processing of results.
Extensive beta testing was performed to validate both numerical algorithms and GUI functionalities. Various test cases were used to assess the software's performance, drawing from literature sources and industry collaborations. Beyond the technical aspects, the project also investigated opportunities for knowledge transfer and exploitation. With the support of specialized external advisors, a market analysis was conducted, identifying potential competitive advantages and refining the business model for commercialization. One of the key achievements was the successful accreditation of a spin-off company at Politecnico di Milano, aimed at commercializing MultiCAT (MCE- Multiscale Catalysis and Engineering s.r.l.).

Progress beyond the state of the art and expected potential impact (including the socio-economic impact and the wider societal implications of the project so far)

MultiCAT provides a comprehensive suite of functionalities designed to optimize the analysis, design, and control of chemical reactions. Leveraging a unique multi-scale, machine-learning, and physics-guided computational framework, it delivers an integrated software environment for chemical and catalytic reaction engineering. In its final release, MultiCAT encompasses chemical reactor modeling, kinetic and spectroscopic data analysis, and digital twin creation, enabling seamless integration with process simulators for optimization and control purposes.
Through extensive applied research in reaction engineering and direct engagement with potential users—including chemical industries, process engineering firms, and academic institutions—the team has identified a strong market demand for such tools. By offering an efficient and high-accuracy approach to process optimization, the software contributes to enhancing product quality, improving operational efficiency, and reducing costs. It is particularly valuable for industries managing complex chemical reactions, such as petrochemicals, pharmaceuticals, polymers, and materials science.
To support commercialization efforts of the spin-off, initial discussions have begun with possible venture capitals (VC) fund specializing in knowledge-based start-ups. To sustain future growth and investment, the team is exploring funding opportunities from public funding schemes, including regional (Lombardia), national (FISA, Italian Government), and European (EIC Transition) programs.
For MultiCAT to reach full market potential, several key areas require further development and strategic support. These include ongoing research and technical enhancements. Additionally, demonstration projects in industrial environments are essential to validate its performance and increase user adoption. We have already started presentations of the tool in industries and research lab potentially interested (Casale, Clariant, Maire, Floatech, IMDEA Materials, …) in the use of the software, exploring potential industry partnerships.
At the conclusion of the project, we have successfully achieved the core technical objectives, delivering an optimized beta version with enhanced numerical models, machine-learning-driven active learning frameworks, and a user-friendly GUI. The software now provides a powerful, efficient, and accessible tool for next-generation catalytic reactor modeling, offering highly scalable solutions for both research and industrial applications. Moving forward, securing investment and further industry collaboration will be instrumental in ensuring continued development, widespread adoption, and long-term commercial success.