KEYWORDS: Data modeling, Machine learning, Education and training, Resonators, Silicon, Temperature metrology, Microrings, Microresonators, Bioalcohols, Water
An approach to measuring chemical concentrations using a micro-ring resonator (MRR) is proposed which is robust to thermo-optic noise and spectral shifts caused by temperature changes. The method uses a modified ResNet50 with varied kernel size and achieved a mean-square error (MSE) of 4.548E-4, and performance is compared to other machine learning methods including VGG20 and XGBoost. The model was trained to read the transmission spectra of a slotted MRR etched into heavily doped silicon and output the concentrations of chemicals in the surrounding analyte. The chemicals tested on were water, ethanol, methanol, and propanol, with concentrations ranging from 0-100%, with a dataset containing . This occurs over the mid-infrared wavelengths and within the temperature range of 290-310 K. Transfer learning was also utilized to retrain the models on several other datasets, consisting of 528 transmissions each. These datasets operated over different temperature ranges (310-320), and the other with a different set of chemicals, (water, ethanol, methanol and butanol). Similar results were achieved, with both networks achieving similar MSE. We then perform the same process on another design with the same chemicals, also operating over the infrared range, demonstrating the robustness of the method. All datasets used in the study were obtained through simulation, although we hope to test on real data.
KEYWORDS: Signal to noise ratio, Resonators, Neural networks, Microrings, Microresonators, Education and training, Machine learning, Data modeling, Refractive index, Mixtures
A new approach for determining the concentration composition of a multi-element media using a micro-ring resonator (MRR) is proposed which allows for noise removal as well as moderately higher average accuracy. This method uses two neural networks, namely a convolutional neural network (CNN) and a deep neural network (DNN). The CNN differentiates the transmission spectrum from the noise. This spectrum is used to obtain selected features before being fed into the DNN, which determines the concentration of each chemical in the analyte. Both models are trained to work with a silicon on-insulator ring resonator operating between the infrared wavelengths of λ=1.46 μm to λ=1.6μm on mixtures of water, ethanol, methanol, and propanol by using simulation data obtained from finite difference eigenmode, although the same approach can be used with other designs and chemical combinations. The CNN was trained using the MRR transmission spectra superimposed with white Gaussian noise as well as Poisson noise to mimic various noise sources, while the DNN underwent training on the extracted features. Average Root-Mean-Square Error was for a range of concentrations from 0.0357-75% is 5.531%.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
INSTITUTIONAL Select your institution to access the SPIE Digital Library.
PERSONAL Sign in with your SPIE account to access your personal subscriptions or to use specific features such as save to my library, sign up for alerts, save searches, etc.