Machine learning method monitors non-Newtonian fluid flow in real time

Various food production, manufacturing, and healthcare applications need to control the flow rate of non-Newtonian fluids through different types of microscale channels. However, the viscosity of non-Newtonian fluids varies depending on stress, which makes these fluids more challenging to measure. Existing flow sensors tend to either disturb non-Newtonian fluids or be bulky, inaccurate, costly, slow, or difficult to integrate with existing flow systems.

Bao et al. developed a method to precisely measure and control non-Newtonian fluid flow at a microscale outlet of a fluid channel with high sensitivity and in real time. This closed-loop system is fully electronic and smaller than previous methods.

Their method combines a contactless, cuff-like flow sensor with a machine learning algorithm. When wrapped around a fluid-dispensing nozzle tip, the flow sensor, a 3-millimeter-wide copper-tape transducer, detects the triboelectricity generated by the flowing fluid with a connected coulomb meter to yield flow rates. The neural-network-based algorithm then analyzes the measured flow rates to predict future flow rates and control the flow to match desired rates. The system remained accurate despite significant noise.

Because this flow-sensing method is contactless, it measures non-Newtonian fluids without disturbing their flow and can be easily adapted to various applications that use different types of dispensing nozzles.

“Our work offers significant promise for applications in additive manufacturing, such as 3D printing, and in medicine, such as microfluidic devices,” said author Jinglei Ping. “New instrumentation, such as next-generation food dispensers, can also be developed based on this technology.”

Next, the authors plan to build on this technology to enhance the precision and resolution of 3D printing.

Source: “Neural network-enabled, all-electronic control of non-newtonian fluid flow,” by Huilu Bao, Xin Zhang, Xiaoyu Zhang, Xiao Fan, J. William Boley, and Jinglei Ping, Applied Physics Letters (2024). The article can be accessed at https://doi.org/10.1063/5.0226525 .