KG4DT

Local Deployment

GraphDB

Download Graphdb Free Version
Put the downloaded zip file under the graphdb/dist folder

Docker

git clone https://github.com/sustainable-processes/KG4DT.git
cd KG4DT
docker-compose up

Find deployed KG4DT and example cases at http://127.0.0.1:5000/index!

Features

Metagraph

Inspired by the pioneering schema OntoCAPE, OntoModel and OntoProcess as ontologies are designed to bridge the gap between model and process knowledge for developing physical model-based chemical process digital twins.

Agents

Multiple functional agents are developed to harness the knowledge graph for model assembly, model calibration, rule inference, database access, AI model invocation, and LLM utilisation.

• Available databases: PubChem, ChemSpider, and Wikipedia
• Example solubility prediction model from the RMG Group
• ChatGPT integrated

Cases

Prerequisite: KG4DT is locally deployed successfully.

• Bottom-up case: annular microreactor
  • Model Description
  • Model Structure
  • Model Application
    • select dushman, click Import
    • select simulation, click Run to simulate the pristine model
    • select calibration and ktm, set min to 0.001 and max to 0.01, click Run to calibrate the model
    • select I3− to check the result
• Top-down case: Taylor-Couette reactor
  • Model Knowledge Graph
    • select top-down mode
    • select esterification, click Import
    • click Infer to proceed rule inference
    • click SPARQL of the Rule card to check SPARQL codes
  • Model Calibration and Comparison
    • select esterification base case, click Import
    • click Run Model Exploration to proceed parallel model calibration
• Amidation reaction optimisation
  • Model ChatGPT
    • input your own OpenAI API Key. Find information here
    • click Go to get the answer