NEAR–INFRARED SPECTROSCOPY AS A TOOL FOR REAL–TIME DETERMINATION OF BOD5 FOR SINGLE–SOURCE SAMPLES

ABSTRACT. Near–infrared/visible spectroscopy was used to develop predictive models for rapid measurement of five–day biochemical oxygen demand (BOD5), bypassing the traditional five–day incubation time. Spectral absorption measurements and standard method BOD5 tests of 158 wastewater samples were completed. The transmission spectral measurements were performed on the samples after filtering out particles larger than 140 m. Spectral data from 400 to 1900 nm were collected using silicon and lead sulfide detectors at a 1.0–nm resolution using a 1.0–mm transmission pathlength. Partial least squares (PLS) and principal component regression (PCR) models were developed through the regression of spectral data against traditional BOD5 measurements. The PLS model was shown to have an error of approximately 4.7 percentage points greater than standard methods and a standard error of prediction (SEP) of 27.6 mg/L for actual BOD5 values ranging from 35 to 275 mg/L. The samples for this study were colleted from the Penn State Wastewater Treatment Plant. It is expected that the NIR technique demonstrated in this study for measuring BOD5 will also work for multi–site sample sources, returning results with higher variation. If the additional variation is not acceptable, then it would still be possible to reap the benefits of on–line sensing and control by developing source–specific calibrations