Process Intelligence Research · TU DelftTransforming chemical engineering with artificial intelligence

Research themes

We develop AI methods that build on chemical engineering knowledge. Our work is organized in seven themes.

• Generative AI for engineering diagrams

  Digitization, autocompletion and autocorrection of P&IDs and PFDs, Smart P&IDs based on the DEXPI standard, and ChatP&ID.

  • Digitization
  • Autocompletion
  • Autocorrection
  • ChatP&ID
  Read more →
• AI for process safety

  AI support for hazard and operability (HAZOP) studies, built on machine-readable engineering diagrams.

  Read more →
• Hybrid and physics-constrained machine learning

  Models that combine mechanistic knowledge with machine learning and respect physical and operational constraints.

  • ENFORCE
  • Physics-informed neural networks
  Read more →
• Optimization with embedded machine learning

  Deterministic global optimization of problems that contain trained neural networks, Gaussian processes or KANs.

  • MeLOn
  • reluMIP
  • KANs
  Read more →
• Multi-agent and LLM systems for process design

  Teams of collaborating LLM agents with specialized knowledge and tools for chemical engineering workflows.

  Read more →
• Graph machine learning

  Graph neural networks for molecular property prediction and topology-aware soft sensors for process plants.

  • Spray drying soft sensors
  Read more →
• Reinforcement learning for process synthesis

  Agents that learn to design chemical process flowsheets by interacting with process simulators.

  Read more →
• All research projects →

Highlights

• 2023

  NWO Veni grant

  Awarded to Artur M. Schweidtmann for research on reinforcement learning for chemical process design.

• 2024

  IChemE Senior Moulton Medal

  For the group's paper on digitizing process flow diagrams with deep convolutional neural networks.

• 2024

  AIChE W. David Smith, Jr. Graduate Publication Award

  Awarded to Artur M. Schweidtmann.

• 2025

  Open Science Award

  Faculty of Applied Sciences, TU Delft, awarded to the pyDEXPI team.

• Prototype

  DigiCo prototype

  Our research on P&ID digitization is being developed into prototype software for Smart P&IDs.

• 2026

  Marie Skłodowska-Curie fellowship

  The group hosts an MSCA Postdoctoral Fellow (project CoAUTHOR).

• Editorial

  Subject Editor, ChERD

  Artur M. Schweidtmann is Subject Editor of Chemical Engineering Research and Design.

• Since 2025

  DEXPI Scientific Advisory Board

  Artur M. Schweidtmann is Vice Chair of the Scientific Advisory Board of the DEXPI Initiative.

Industry partners

We have collaborated with industry partners including Shell, Danone, Fluor, Siemens and Linde.

• Shell
• Danone
• Fluor
• Siemens
• Linde

Open-source software

We release our methods as open-source software so that others can use and build on them.

• pyDEXPI

  Python implementation of the DEXPI data model for machine-readable P&IDs.

• SFILES 2.0

  Text-based representation of process flowsheets and P&IDs.

• ENFORCE

  Neural networks that satisfy nonlinear equality and inequality constraints.

• reluMIP

  Mixed-integer formulations of trained ReLU neural networks for optimization.

• MeLOn

  Machine learning models for deterministic global optimization.

• DEXPI standard

  pyDEXPI builds on DEXPI, the open data exchange standard for P&IDs.