Advanced deep learning solution for rapid, non-destructive reservoir rock characterization
Try Live DemoGet 5 key reservoir parameters in seconds instead of days. Faster than conventional methods.
Reduce laboratory analysis costs by up to 70% while maintaining laboratory-grade accuracy.
State-of-the-art DenseNet201 CNN model trained on thousands of rock samples from Digital Rock Portal.
Analyze samples without physical alteration, preserving valuable core material.
Backend & AI
Core CNN Model
Tortuosity Calculation
Grain Segmentation
We evaluated multiple transfer learning models:
Organized team structure:
DenseNet201 showed superior performance across all parameters:
| Parameter | Best Model | RMSE | MAPE |
|---|---|---|---|
| Porosity | DenseNet201 | 0.012 | 4.2% |
| Tortuosity | DenseNet201 | 0.15 | 6.8% |
| Grain Size | DenseNet201 | 0.08 | 5.1% |
| Surface Area | DenseNet201 | 0.21 | 7.3% |
| Coordination | DenseNet201 | 0.32 | 8.5% |
Significant time reduction compared to conventional methods:
| Parameter | Conventional Method | RophysiX | Time Saved |
|---|---|---|---|
| Tortuosity | 2-3 days | 5 seconds | 99.98% |
| Coordination Number | 1-2 days | 5 seconds | 99.94% |
| Grain Size | 6-8 hours | 5 seconds | 99.98% |
Drag & drop or click to browse (2D sandstone images only)
The RophysiX model and methodology have been registered as Intellectual Property (HKI) with the Indonesian Ministry of Law and Human Rights.
Registration Number: HKI-2024-EC00202479885
Complete project documentation and technical specifications