Spray characterization in air-assist atomizers using flow-induced acoustic vibrations and multivariate analysis

Abstract

This feasibility study investigated a new non-intrusive approach employing acoustic chemometrics. The method includes acoustic/vibration data recording obtained utilizing two clamp-on piezoelectric accelerometers and two electret condensers-type microphones mounted on an arc. Principal Component Analysis (PCA) classification models were based on the acoustic FFT spectra from four sensors. The non-dimensional number (X) values correspond to the different breakup regimes comprising a range of air and liquid (water) flow rates in this air-assisted atomizer (one-analyte system). PCA classification model discerns the clusters belonging to similar non-dimensional number (X) values with the maximum variance in the first principal component (PC1) direction for both sensors combined. This study also assesses the utility of the acoustic chemometrics approach for predicting the flow parameter, such as Sauter mean diameter (SMD) based on the Partial Least Squares-Regression (PLS-R). The PLS-R prediction models work best for the 550 mm location with a low root mean square error of prediction (RMSEP) value of 5.443 and a high Pearson correlation coefficient (R2) value of 0.856 when validated using 50% independent data (test set validation). The comparison between the two sensor types demonstrated superior prediction performance for accelerometers for all the prediction models.