ParticleTensorAI® analyses time-resolved light-scattering data using an AI-based approach. The specifications below describe the current reference parameters for signal acquisition and Particle Tensor processing. The available outputs depend on the measurement configuration and the AI model validated for the application.
Signal and Tensor Data
| Parameter | Current PTA reference value |
|---|---|
| Evaluation approach | AI-based signal-sequence analysis |
| Signal sampling interval | Approximately 6.4 ns |
| Signal buffer duration | 100 ms |
| Particle Tensor dimensions | 3 × 256 × 256 |
| Optical detector channels | 1, 2, or 4, depending on the system configuration |
The tensor dimensions describe the data structure supplied for AI analysis. The precise meaning and arrangement of its dimensions depend on the selected tensor representation and measurement task.
ParticleTensorAI® System Configurations
| Version | Optical probe | Control unit | Detector channels |
|---|---|---|---|
| PTA1 | LSS1 | ZEON | 1 |
| PTA2 | LSS2 | IMEA | 2 |
| PTA4 | LSS4 | CLEON | 4 |
All three versions use the ParticleTensorAI® evaluation approach. They differ in their optical detector-channel configuration and associated measurement hardware.
AI-Based Evaluation
ParticleTensorAI® uses AI models to classify or estimate properties defined for a measurement task. Model outputs depend on the recorded data, the reference information used during model development, and validation under relevant measurement conditions.
Possible research objectives include particle or droplet size, concentration, material-related classification, and refractive-index-related characterization. These are application-dependent objectives and are not guaranteed outputs for every version or measurement setup.
Notes on Specification and Validation
- The sampling interval and buffer duration describe signal acquisition parameters.
- The Particle Tensor dimensions describe a data representation for AI processing.
- Detector-channel count varies between PTA1, PTA2, and PTA4.
- Measurement outputs require a model developed and validated for the relevant application.
Research Reference
A published study demonstrates one ParticleTensorAI® implementation using detector signals represented as RGB images and convolutional neural networks for classification. The reported validation results apply to the tested coal-slurry spray conditions.