Smart, machine-readable Piping and Instrumentation Diagram (P&ID) data is key to unlocking innovation in the process industry, especially for emerging applications like Generative Artificial Intelligence (GenAI) - see examples of GenAI for the process industry. pyDEXPI helps enable this by making the DEXPI standard accessible and usable in Python.

pyDEXPI is an open-source tool in Python that implements the DEXPI data model. Importantly, pyDEXPI includes a Pydantic data class model of the DEXPI data model. This allows building Python applications with the DEXPI standard. pyDEXPI further provides functions to load a Proteus .xml export of DEXPI, the current DEXPI exchange format, into the data model.

pyDEXPI implements a parser to a graph representation of Piping and Instrumentation Diagrams (P&IDs) using NetworkX. This allows the user to work easily with the graph data.

Supports DEXPI version 1.3.

• DEXPI data model as Pydantic classes in Python.
• Load Proteus .xml files to a pyDEXPI instance.
• pyDEXPI toolkit to analyze and manipulate pyDEXPI models.
• Parse pyDEXPI instance to graph in NetworkX.
• Synthetic DEXPI P&ID generation for generative Artificial Intelligence (AI).

Please reference this software package as:

@InProceedings{pyDEXPI,
  author    = {Goldstein, Dominik P. and Schulze Balhorn, Lukas and Alimin, Achmad Anggawirya and Schweidtmann, Artur M.},
  booktitle = {Proceedings of the 35th European Symposium on Computer Aided Process Engineering (ESCAPE35)},
  title     = {pyDEXPI:{A} {Python} framework for piping and instrumentation diagrams using the {DEXPI} information model},
  year      = {2025},
  address   = {Ghent, Belgium},
  month     = {July},
  doi       = {https://doi.org/10.69997/sct.139043},
}

• Overview
  • Features
  • Citation
• Table of contents
• Installation
• Using pyDEXPI
  • pyDEXPI Python model
  • Proteus import
  • Serialization
  • Graph export
  • Synthetic P&ID generation
• Relevant applications and references
• Contributors
• Copyright and license
• Contact

Install the pyDEXPI package via

pip install pydexpi

or from GitHub via:

pip install git+https://github.com/process-intelligence-research/pyDEXPI

Alternatively, get the latest updates by cloning the repo and installing the editable version of the package with:

git clone https://github.com/process-intelligence-research/pyDEXPI
cd pyDEXPI
pip install .

The following section serves as a first guide to start using the package, illustrated with the DEXPI reference P&ID (data/C01V04-VER.EX01.xml © DEXPI e.V.). We recommend to study the DEXPI data model before working with the tool. For more details, see the extended documentation.

The pyDEXPI Python model is derived from the DEXPI data model and implemented using Pydantic. By using Pydantic the rules of the data model are automatically enforced, for instance, a pump cannot be added as a nozzle to a tank. Each DEXPI instance gets assigned an ID in the form of an uuid, if not specified differently by the user.

The following DEXPI data types are replaced with default Python classes:

• "String", "NullableString", "AnyURI", "NullableAnyURI" -> "str"
• "Integer", "NullableInteger", "UnsignedByte" -> "int"
• "Double" -> "float"
• "DateTime", "NullableDateTime" -> "datetime"

The Proteus serializer loads a proteus .xml file to a DEXPI model. Drawing information are currently not parsed, among others.

from pydexpi.loaders import ProteusSerializer

directory_path = "data"
filename = "C01V04-VER.EX01.xml"
my_loader = ProteusSerializer()
dexpi_model = my_loader.load(directory_path, filename)

You can load and save DEXPI Python models via the serializer. Currently pickle .pkl and json .json are offered as file formats.

For json:

from pydexpi.loaders import JsonSerializer

my_serializer = JsonSerializer()

For pickle:

from pydexpi.loaders import PickleSerializer

my_serializer = PickleSerializer()

Then:

my_serializer.save(dummy_dexpi_model, "dummy_path", "dummy_filename")
dummy_dexpi_model = my_serializer.load("dummy_path", "dummy_filename")

The graph loader exports a DEXPI model to a NetworkX graph. This is implemented as an abstract class to accommodate customized graph formats. The package includes an implementation for direct application MLGraphLoader. Here, we represent DEXPI components such as tagged plant equipment, process instrumentation functions, and piping components as nodes. The connectivity between the nodes represent piping connections or signal connections as directed edges. Data attributes of the pyDEXPI classes are added as node/edge attributes of the respective, representing node/edge. More details of the graph format are described here.

from pydexpi.loaders.ml_graph_loader import MLGraphLoader
from pydexpi.loaders import ProteusSerializer

directory_path = "data"
filename = "C01V04-VER.EX01.xml"
# load proteus
my_loader = ProteusSerializer()
dexpi_model = my_loader.load(directory_path, filename)
# export to graph
my_graph_loader = MLGraphLoader(plant_model=dexpi_model)
my_graph_loader.parse_dexpi_to_graph()
my_graph_loader.draw_process_matplotlib()

The synthetic data generation generation contains code for the creation of synthetic P&IDs. This method is based on the aggregation of P&ID module templates (or patterns) as described in the publication Toward automatic generation of control structures for process flow diagrams with large language models. An abstraction of the generation logic allows to customize the procedure underlying the module aggregation. For this, the abstract GeneratorFunction can be implemented as required. An example implementation RandomGeneratorFunction is provided. This implementation selects P&ID modules at random. P&ID modules can be used in any data representation. For this, suitable abstractions of the Pattern Class and the Connector Class need to be implemented. A pattern wraps a P&ID data structure for the generation algorithm. A connector acts as a connection interface of a P&ID module. Sample implementations of patterns and connectors are provided for pyDEXPI instances and NetworkX graphs. A UML diagram of the implementation is given here.

The code below demonstrates the synthetic data generation algorithm with the RandomGeneratorFunction and the pyDEXPI/Graph patterns.

import os
from pydexpi.syndata import SyntheticPIDGenerator, PatternDistribution
from pydexpi.syndata.generator_function import RandomGeneratorFunction

# Load distributions
the_path = "./data/dexpi_sample_patterns"
# the_path = "./data/graph_sample_patterns" # Alternatively for graph patterns
pattern_distr_names = [name for name in os.listdir(the_path) if os.path.isdir(os.path.join(the_path, name))]
distributions = [PatternDistribution.load(the_path, name) for name in pattern_distr_names]
distribution_dict = {distribution.name: distribution for distribution in distributions}

generator_function = RandomGeneratorFunction(distribution_range=distribution_dict)
the_generator = SyntheticPIDGenerator(generator_function, max_steps=5)
syn_pattern = the_generator.generate_pattern("New pattern label")

Vision of GenAI for process engineering

• Schweidtmann, A. M. (2024). Generative artificial intelligence in chemical engineering. Nature Chemical Engineering, 1(3), 193-193. https://doi.org/10.1038/s44286-024-00041-5

Chat interface for P&IDs using Large Language Models (LLMs)

• Alimin, A. A., Goldstein, D. P., Balhorn, L. S., & Schweidtmann, A. M. (2025). Talking like piping and instrumentation diagrams (p&ids). Proceedings of the 35th European Symposium on Computer Aided Process Engineering (ESCAPE35), Ghent, Belgium. https://doi.org/10.69997/sct.159477

Error correction of P&IDs

• Balhorn, L. S., Seijsener, N., Dao, K., Kim, M., Goldstein, D. P., Driessen, G. H., & Schweidtmann, A. M. (2025). Rule-based autocorrection of Piping and Instrumentation Diagrams (P&IDs) on graphs. Proceedings of the 35th European Symposium on Computer Aided Process Engineering (ESCAPE35), Ghent, Belgium. https://doi.org/10.69997/sct.150968
• Balhorn, L. S., Caballero, M., & Schweidtmann, A. M. (2024). Toward autocorrection of chemical process flowsheets using large language models. In Computer Aided Chemical Engineering (Vol. 53, pp. 3109-3114). Elsevier. https://doi.org/10.1016/B978-0-443-28824-1.50519-6

Process development

• Vogel, G., Balhorn, L. S., & Schweidtmann, A. M. (2023). Learning from flowsheets: A generative transformer model for autocompletion of flowsheets. Computers & Chemical Engineering, 171, 108162. https://doi.org/10.1016/j.compchemeng.2023.108162
• Balhorn, L. S., Hirtreiter, E., Luderer, L., & Schweidtmann, A. M. (2023). Data augmentation for machine learning of chemical process flowsheets. In Computer Aided Chemical Engineering (Vol. 52, pp. 2011-2016). Elsevier. https://doi.org/10.1016/B978-0-443-15274-0.50320-6
• Hirtreiter, E., Schulze Balhorn, L., & Schweidtmann, A. M. (2024). Toward automatic generation of control structures for process flow diagrams with large language models. AIChE Journal, 70(1), e18259. https://doi.org/10.1002/aic.18259
• Balhorn, L. S., Degens, K., & Schweidtmann, A. M. (2025). Graph-to-SFILES: Control structure prediction from process topologies using generative artificial intelligence. Computers & Chemical Engineering, 109121. https://doi.org/10.1016/j.compchemeng.2025.109121

Digitization of paper P&IDs to smartP&IDs

• Theisen, M. F., Flores, K. N., Balhorn, L. S., & Schweidtmann, A. M. (2023). Digitization of chemical process flow diagrams using deep convolutional neural networks. Digital chemical engineering, 6, 100072. https://doi.org/10.1016/j.dche.2022.100072
• Digitization Companion

	Dominik P. Goldstein
	Lukas Schulze Balhorn
	Achmad Anggawirya Alimin
	Artur M. Schweidtmann

This software is released under the OSI-approved GNU Affero General Public License (AGPL-3.0) license (see license file file for details). We believe in open collaboration and knowledge sharing, and encourage use by students, researchers, open-source contributors, and industry. You are free to use, modify, and distribute the software under the given license terms. This is a copyleft license, which means that any software based on pyDEXPI, or any modified version thereof, must be published under the same open-source license.

Commercial or Proprietary Use?

If you would like to:

• Use this software in a proprietary or closed-source product,
• Use it in a way that is not compatible with AGPL copyleft obligations, or
• Obtain dedicated support or feature extensions,

We’re happy to discuss a commercial or custom license on a case-by-case basis.
Please reach out to a.schweidtmann@tudelft.nl for more information.

Copyright (C) 2025 Artur Schweidtmann.

📧 Contact

🌐 PI research