ICLASS 2024 – Machine Learning Spray Diagnostics

At the 16th Triennial International Conference on Liquid Atomization and Spray Systems, ICLASS 2024, in Shanghai, China, we presented a poster on machine learning spray diagnostics and particle characterization. The work shows how data-driven methods can improve optical spray measurement and support compact measurement instruments for complex atomization processes.

The poster focuses on particle characterization by analyzing light scattering signals with a machine learning approach. This method builds on the established Time-Shift Time-of-Flight, or TSTOF, measurement technique. TSTOF is also known through the commercial measurement system SpraySpy®.

Particle characterization in sprays and flows often requires detailed information about particle size and particle velocity. Classical TSTOF-based instruments use several optical signals to extract this information. These signals come from droplets or particles passing through structured light fields.

In this work, we show how machine learning can reduce the required number of signals. The model learns from the complete signal set and then uses only one signal for the final evaluation. As a result, the method can still determine important particle information such as size and velocity.

Machine learning spray diagnostics can help simplify optical measurement systems. By using only a single signal in the final evaluation, the method can reduce the number of required optical components.

This creates several practical advantages. One light source, three detectors, and parts of the related electronics and optics may no longer be necessary. Therefore, future TSTOF-based systems could become smaller, more cost-effective, and easier to integrate into laboratory and industrial environments.

The TSTOF measurement technique has a long history in optical spray diagnostics and particle measurement. It analyzes time-shifted light scattering signals from particles or droplets in a flow. The commercial device based on this method became known under the brand name SpraySpy®.

The ICLASS 2024 contribution shows how this established measurement principle can be extended with machine learning. This connection between classical optical diagnostics and AI-assisted signal evaluation forms an important step toward new systems such as SprayQuantAI®.

The presented approach supports particle characterization, droplet analysis, optical spray diagnostics, and industrial spray monitoring. It can help researchers and engineers measure sprays with reduced hardware complexity while keeping relevant information from the light scattering signal.

This makes the method interesting for atomization research, coating processes, spray development, process monitoring, and compact optical measurement instruments.