RESUMEN
Fourier transform near-infrared (FT-NIR) spectroscopy is a versatile and non-destructive analytical tool widely utilized in industries such as food, pharmaceuticals, and agriculture. While traditional FT-NIR instruments pose limitations in terms of cost and complexity, the advent of portable and affordable systems like NeoSpectra Scanners has broadened accessibility. Partial Least Squares Regression (PLSR) stands as an industry-standard method in Chemometrics for analyzing chemical compositions. This work addresses optimizing PLSR models in FT-NIR spectroscopy, focusing on enhancing accuracy and adaptability in material analysis. Unlike traditional PLSR models which often rely on grid searching a limited number of parameters, such as latent variables, the presented approach effectively expands the parameter space. A novel framework combining Bayesian search and stacking techniques is introduced to enable more customization while ensuring time and performance efficiency, along with automation in model development. Bayesian search efficiently explores hyperparameters space, enabling faster convergence to optimal model settings without exhaustive exploration. The proposed stacked model leverages learned knowledge from the top-performing PLSR models optimized through Bayesian methods, amalgamating a unified and potent body of knowledge. Bayesian-stacked models are compared with PLSR models that use grid search for a limited parameter set. Findings show a marked improvement in model performance: a 51.5% reduction in Root Mean Square Error (RMSE) for the training dataset and a 26.1% reduction for the testing dataset, alongside a 10.9% increase in the correlation coefficient square (R2) for the training dataset and a 10.4% increase for the testing dataset. Notably, Bayesian search reduces the model optimization time by approximately 90% compared with the grid search. Furthermore, when addressing instrumental variations, the models demonstrate an additional improvement, evident in the average reduction of 24.1% in the mean range of prediction. Overall, results demonstrate that the presented approach not only increases the prediction accuracy but also offers a more efficient, automated and robust solution for diverse spectroscopic applications.
RESUMEN
In this work, we study the detection of acetylene (C2H2), carbon dioxide (CO2) and water vapor (H2O) gases in the near-infrared (NIR) range using an on-chip silicon micro-electro-mechanical system (MEMS) Fourier transform infrared (FT-IR) spectrometer in the wavelength range 1300-2500 nm (4000-7692 cm(-1)). The spectrometer core engine is a scanning Michelson interferometer micro-fabricated using a deep-etching technology producing self-aligned components. The light is free-space propagating in-plane with respect to the silicon chip substrate. The moving mirror of the interferometer is driven by a relatively large stroke electrostatic comb-drive actuator corresponding to about 30 cm(-1) resolution. Multi-mode optical fibers are used to connect light between the wideband light source, the interferometer, the 10 cm gas cell, and the optical detector. A wide dynamic range of gas concentration down to 2000 parts per million (ppm) in only 10 cm length gas cell is demonstrated. Extending the wavelength range to the mid-infrared (MIR) range up to 4200 nm (2380 cm(-1)) is also experimentally demonstrated, for the first time, using a bulk micro-machined on-chip MEMS FT-IR spectrometer. The obtained results open the door for an on-chip optical gas sensor for many applications including environmental sensing and industrial process control in the NIR/MIR spectral ranges.