ParticleTensorAI® (PTA for short) is an optical measurement technique for the characterization of particles and droplets in flows. It was first introduced in 2024 as an AI-assisted extension of the TSTOF technique. In conventional TSTOF instruments, the light scattering signals of individual particles are recorded and analyzed as isolated events. In contrast, ParticleTensorAI® continuously captures a large number of such individual events, thereby significantly increasing the information density per unit time.
The operating principle of classical TSTOF instruments is described in detail in [1]. Unlike the conventional TSTOF approach, ParticleTensorAI® is based on the evaluation of a continuous data stream of light scattering signals. These signals are transformed into so-called tensors, which gives rise to the name ParticleTensorAI®. The tensors (also referred to as PTAs) are analyzed using artificial intelligence, in particular by applying Convolutional Neural Networks (CNNs). From the PTAs, physical properties of the particles such as size, velocity, concentration, or refractive index can be determined. ParticleTensorAI® can also be operated using only a single detector.
The theoretical foundation of ParticleTensorAI® was established in 2024 during the development of SprayQuantAI® and was first published in [2] and presented at ICLASS 2024 in Shanghai [3]. The objective was to reduce the number of optical components while maintaining the quality of the light scattering signals, enabling reliable detection of particles with irregular geometries. The first experimental validation followed in 2025: A measurement campaign [4] demonstrated that the mean droplet size and velocity of a slurry spray can be determined from an ensemble of light scattering signals—without any triggering, filtering, or limitation of the signal bandwidth.
ParticleTensorAI® is a novel method and is under continuous development. Up-to-date information can be found on our website and on our YouTube channel. The key advantage of PTA lies in its unique signal processing approach: neither triggering nor any form of pre-processing such as filtering or bandwidth limitation is required for the analysis. Instead, the full signal diversity is exploited to statistically extract physical parameters using Convolutional Neural Networks (CNNs). As a result, ParticleTensorAI® achieves exceptional robustness and accuracy, even for complex light scattering signals.
ParticleTensorAI® can be deployed as a software update in combination with existing TSTOF instruments such as SpraySpy®. In addition, we also provide customized hardware solutions tailored to the specific requirements of your application.
References
[1] Schaefer, W., Li, L., Stegmann, P., & Terada, M. (2026). Technical report on the TSTOF measurement method: Technical basics, historical development, and comparison with other laser-based measurement methods. Photonics, 13(1), 56. https://doi.org/10.3390/photonics13010056
[2] Schaefer, W., & Li, L. (2024). Particle characterization by analyzing light scattering signals with a machine learning approach. Applied Optics, 63(29), 7701. https://doi.org/10.1364/AO.531346
[3] Schäfer, W., & Li, L. (2024). Particle characterization by analyzing light scattering signals with a machine learning approach. In Proceedings of the 16th Triennial International Conference on Liquid Atomization and Spray Systems (ICLASS 2024).
[4] Schaefer, W., Fleck, S., Haas, M., & Jakobs, T. (2025). Optical measurement method for monitoring high-mass-concentration slurry sprays: An experimental study. Photonics, 12(7), 673. https://doi.org/10.3390/photonics12070673