hiral-kumar/cnc-digital-twin

How hiral-kumar/cnc-digital-twin works

PC-NDT is a bearing-health digital twin: dataset-specific loaders extract vibration features and RUL labels, preprocessing prepares temporal windows, and the training entry point fits a graph-and-continuous-time model with physics constraints. Evaluation covers held-out IMS, PRONOSTIA cross-condition generalization, LSTM/GRU comparisons, and Archard validation. The graph follows documented architecture and SOURCE INDEX references; unsampled entry-point bodies limit claims about some detailed calls.

Main entry points: pc_ndt.py PC-NDT model · preprocessing.py Window preprocessing · train.py Training workflow · baselines.py Baseline comparison

Components 18

Data preparation

IMS loader src/data/ims_loader.py
PRONOSTIA loader src/data/pronostia_loader.py
Milling loader src/data/milling_loader.py
Window preprocessing src/data/preprocessing.py
Graph utilities src/data/graph_utils.py

Digital twin model

AGCRN encoder src/models/agcrn.py
Neural ODE src/models/neural_ode.py
PC-NDT model src/models/pc_ndt.py

Training and physics

Physics constraints src/physics/constraints.py
Trainer src/training/trainer.py
Training workflow scripts/train.py

Evaluation workflows

LSTM and GRU src/baselines/lstm_gru.py
Baseline comparison scripts/baselines.py
IMS evaluation scripts/evaluate.py
PRONOSTIA evaluation scripts/evaluate_pronostia.py
Archard validation scripts/validate_archard.py
Evaluation metrics src/evaluation/metrics.py

Other components

Engineer

Connections 29

Engineer → Training workflow: starts training README.md
Engineer → IMS evaluation: runs evaluation README.md
Engineer → Baseline comparison: runs comparison README.md
Engineer → PRONOSTIA evaluation: tests generalization README.md
Engineer → Archard validation: validates physics README.md
IMS loader → Window preprocessing: provides features src/data/preprocessing.py
PRONOSTIA loader → Window preprocessing: provides features src/data/preprocessing.py
Milling loader → Archard validation: supplies validation data
Training workflow → Window preprocessing: prepares windows scripts/train.py
Training workflow → PC-NDT model: trains model scripts/train.py
Training workflow → Trainer: uses trainer scripts/train.py
Training workflow → Physics constraints: configures physics scripts/train.py
PC-NDT model → AGCRN encoder: encodes sensors src/models/pc_ndt.py
PC-NDT model → Neural ODE: propagates state src/models/pc_ndt.py
PC-NDT model → Physics constraints: computes disagreement src/models/pc_ndt.py
Trainer → Physics constraints: computes physics loss src/training/trainer.py
Physics constraints → Graph utilities: uses graph operations src/physics/constraints.py
IMS evaluation → Window preprocessing: prepares inputs scripts/evaluate.py
IMS evaluation → PC-NDT model: evaluates model scripts/evaluate.py
IMS evaluation → Evaluation metrics: scores predictions scripts/evaluate.py
PRONOSTIA evaluation → PRONOSTIA loader: loads bearings scripts/evaluate_pronostia.py
PRONOSTIA evaluation → Window preprocessing: prepares inputs scripts/evaluate_pronostia.py
PRONOSTIA evaluation → PC-NDT model: evaluates model scripts/evaluate_pronostia.py
PRONOSTIA evaluation → Evaluation metrics: scores predictions scripts/evaluate_pronostia.py
Baseline comparison → IMS loader: loads IMS data scripts/baselines.py
Baseline comparison → Window preprocessing: prepares inputs scripts/baselines.py
Baseline comparison → LSTM and GRU: runs baselines scripts/baselines.py
Baseline comparison → Evaluation metrics: scores predictions scripts/baselines.py
Archard validation → Physics constraints: validates constraint scripts/validate_archard.py

More Python diagrams: circuitnet/circuitnet · idaholab/malcolm · minasaad1/pbi-cli · technologicat/pyan · tombulled/innertube

Drawn by GitDiagram on September 30, 2026 from the repository's file tree, README and source files. Read as Markdown ·