Machine learning empowered thin film acoustic wave sensing

Abstract

Thin film-based surface acoustic wave (SAW) technology has been extensively explored for physical, chemical, and biological sensors. However, these sensors often show inferior performance for a specific sensing in complex environments, as they are affected by multiple influencing parameters and their coupling interferences. To solve these critical issues, we propose a methodology to extract critical information from the scattering parameter and combine the machine learning method to achieve multi-parameter decoupling. We used the AlScN film-based SAW device as an example in which the highly c-axis orientated and low stress AlScN film was deposited on silicon substrate. The AlScN/Si SAW device showed a Bode quality factor value of 228 and an electromechanical coupling coefficient of ∼2.3%. Two sensing parameters (i.e., ultraviolet or UV and temperature) were chosen for demonstration, and the proposed machine learning method was used to distinguish their influences. Highly precision UV sensing and temperature sensing were independently achieved without their mutual interferences. This work provides an effective solution for decoupling of multi-parameter influences and achieving anti-interference effects in thin film-based SAW sensing.

Amplitude modulated feed-forward thermal drift compensation for linear and nonlinear current sensors

“…Finally, the Op-amp-based subtraction (using Op-amp U5) and gain adjustment circuits (using Opamp U6) are used to perform the required subtraction and amplitude adjustment operation, as discussed in equations ( 15)- (18). The temperature-dependent error associated with the Hall output voltage V H (306 + ∆T) is eliminated by the term V AM (p∆T + q(∆T) 2 ), which is equivalent to equation (7) or (8). The additional term V AM (H.O.T.)…”

“…The error related to strain drift is eliminated using this special design of the fiber structure and is able to measure temperature effectively. Different random forest regression and support vector machine methods are adopted in [7] to extract multiple sensitive temperature signals and ultra-violet detection without mutual interference. Although the proposed drift elimination technique can be extended to numerous applications of measurements, this paper focuses on different types of Hall current sensors designed with temperature-sensitive materials.…”