Synthetic Integrity Digital Twin (PIML + PINN Framework)
Overview
This project generates a fully synthetic refinery-style digital twin including:
3D equipment visualization (tanks, vessels, columns, pumps, exchangers)
Routed piping topology
Corrosion and chemical attributes
RBI-style integrity metadata
100 synthetic SCC alternative scenarios
Optional Physics-Informed Machine Learning (PIML / PINN) demo
All data is synthetic and anonymized.
This repository is intended for:
Research
Digital twin prototyping
Integrity analytics architecture design
AI experimentation for corrosion and cracking
It is NOT a certified engineering calculator and does not replace API 579 or ASME FFS assessments.
Engineering Context
Conceptually aligned with:
API 570 (Piping Inspection)
API 571 (Damage Mechanisms)
API 580 / 581 (RBI)
API 579-1 / ASME FFS-1 (Fitness-for-Service)
ASME B31.3 (Process Piping)
These references are conceptual only.
Features
- Synthetic Equipment Model
Tagged equipment
Design and operating conditions
Material specification
Insulation and coating metadata
- Piping Attributes
Service type
Chlorides
pH
H2S
Corrosion rate (MPY and mm/y)
SCC susceptibility score (0–1)
- 3D Interactive Visualization
Generates: synthetic_refinery_unit_3d.html
Open in browser to rotate, zoom and inspect.
- PIML / PINN Demo
Optional physics-informed neural network to:
Predict SCC susceptibility
Apply Paris-type crack growth constraint
Demonstrate hybrid physics + ML architecture
Installation
pip install -r requirements.txt
Optional (for ML): pip install torch
Run
python synthetic_unit_piml_pinn.py
Outputs
data/equipment_anonymized.tsv
data/piping_attributes_anonymized.tsv
outputs/synthetic_refinery_unit_3d.html
Future Extensions
Real P&ID parser integration
Crack growth time simulation
Remaining life forecasting
Monte Carlo reliability modeling
Integration with Power BI dashboards
Risk prioritization engine