Aection – Automatic EHDA Control and Mode Recognition.

Electrohydrodynamic Atomization (EHDA), also known as Electrospray (ES), is a technology which uses strong electric fields to manipulate liquid atomization. Among many other areas, electrospray is currently used as an important tool for biomedical applications (droplet encapsulation), water technology (thermal desalination and metal recovery) and material sciences (nanofibers and nano spheres fabrication, metal recovery, selective membranes and batteries). A complete review about the particularities of this technology and its applications was recently published in a special edition of the Journal of Aerosol Sciences [1]. Even though EHDA is already applied in many different industrial processes, there are not many controlling tools commercially available which can be used to remotely operate the system as well as identify some spray characteristics, e.g. droplet size, operational mode, droplet production ratio. The AECTion project proposes the development of an innovative controlling system based on the electrospray current, signal processing & control and artificial intelligence to build a non-visual tool to control and characterize EHDA processes.

Eindrapportage

This project was conducted in collaboration with Gilbert, a Dutch medical technology company developing the next generation of soft mist inhalers based on electrodynamic atomization (EHDA). While EHDA is widely applied in various industries, there is a lack of commercially available tools capable of remotely controlling and characterizing the process, such as determining droplet size, operational mode, and production rate. The AECTion project addressed this gap by developing an innovative control system based on electrospray current monitoring, signal processing, and artificial intelligence, enabling non-visual, real-time control and characterization of EHDA processes. The project successfully demonstrated and validated a real-time closed-loop control system capable of detecting and stabilizing electrospray operation - specifically in the cone-jet mode, known for its high stability and uniform droplet formation - within a multi-nozzle configuration. The optimal machine learning model - an Extreme Learning Machine (ELM) trained on key statistical features (mean current, standard deviation, and mean voltage) - achieved a response time of 1.75 seconds, ensuring stable electrospray operation. These results demonstrate that combining advanced measurement strategies with computationally efficient machine learning models can significantly improve both the stability and control performance of electrospray systems in real-world applications.