The effect of flow rate in acoustic chemometrics on liquid flow: Transfer of calibration models

摘要: Abstract Previous studies on acoustic chemometrics liquid flow have demonstrated that rate, accelerometer location and temperature affect the passive spectra prediction results. Changes in rate result spectral variations, causing resultant local calibration model to perform poorly predicting new samples measured at other rates. Developing good robust models can be done using several approaches. Global methods were discussed a previous study flow. In present paper preprocessing techniques such as multiplicative scatter correction (MSC) piecewise (PMSC), variable selection interval partial least squares (iPLS) regression powered (PPLS) are studied. The validated two experimental data sets from measurements of through an orifice plate. Acoustic registered four different locations. liquids two-component mixtures sucrose water, three-component ethanol, water. MSC resulted improvement performance (preprocessed) Sucrose ethanol improved terms bias correlation coefficients respectively. Absolute values for range 0.84–2.57 wt.% preprocessed by compared 1.17–22.38 wt.% uncorrected located plate highest studied rates Correlation 0.80–0.97 0.76–0.97 Limited systematic was observed water mixtures. PMSC slightly mixture MSC. iPLS indicated some intervals which less affected fluctuations. Regression these instead full lower errors sucrose, models. PPLS frequency (peak position) matrix derived did not determine any peaks Effect positions important chemical composition difficult establish. However, shift with increasing could observed.