Characterizing the Uncertainty in Near-Infrared Spectroscopic Prediction of Mixed-Oxygenate Concentrations in Gasoline: Sample-Specific Prediction Intervals

Orthogonality constrained inverse regression to improve model selectivity and analyte predictions from vibrational spectroscopic measurements

In analytical chemistry spectroscopy is attractive for high-throughput quantification, which often relies on inverse regression, like partial least squares regression. Due to a multivariate nature of spectroscopic measurements an analyte can be quantified in presence of interferences. However, if the model is not fully selective against interferences, analyte predictions may be biased. The degree of model selectivity against an interferent is defined by the inner relation between the regression vector and the pure interfering signal. If the regression vector is orthogonal to the signal, this inner relation equals zero and the model is fully selective. The degree of model selectivity largely depends on calibration data quality. Strong correlations may deteriorate calibration data resulting in poorly selective models. We show this using a fructose-maltose model system. Furthermore, we modify the NIPALS algorithm to improve model selectivity when calibration data are deteriorated. This modification is done by incorporating a projection matrix into the algorithm, which constrains regression vector estimation to the null-space of known interfering signals. This way known interfering signals are handled, while unknown signals are accounted for by latent variables. We test the modified algorithm and compare it to the conventional NIPALS algorithm using both simulated and industrial process data. The industrial process data consist of mid-infrared measurements obtained on mixtures of beta-lactoglobulin (analyte of interest), and alpha-lactalbumin and caseinoglycomacropeptide (interfering species). The root mean squared error of beta-lactoglobulin (% w/w) predictions of a test set was 0.92 and 0.33 when applying the conventional and the modified NIPALS algorithm, respectively. Our modification of the algorithm returns simpler models with improved selectivity and analyte predictions. This paper shows how known interfering signals may be utilized in a direct fashion, while benefitting from a latent variable approach. The modified algorithm can be viewed as a fusion between ordinary least squares regression and partial least squares regression and may be very useful when knowledge of some (but not all) interfering species is available.

Monitoring Process Water Quality Using Near Infrared Spectroscopy and Partial Least Squares Regression with Prediction Uncertainty Estimation

Reuse of process water in dairy ingredient production-and food processing in general-opens the possibility for sustainable water regimes. Membrane filtration processes are an attractive source of process water recovery since the technology is already utilized in the dairy industry and its use is expected to grow considerably. At Arla Foods Ingredients (AFI), permeate from a reverse osmosis polisher filtration unit is sought to be reused as process water, replacing the intake of potable water. However, as for all dairy and food producers, the process water quality must be monitored continuously to ensure food safety. In the present investigation we found urea to be the main organic compound, which potentially could represent a microbiological risk. Near infrared spectroscopy (NIRS) in combination with multivariate modeling has a long-standing reputation as a real-time measurement technology in quality assurance. Urea was quantified Using NIRS and partial least squares regression (PLS) in the concentration range 50-200 ppm (RMSEP = 12 ppm, R²= 0.88) in laboratory settings with potential for on-line application. A drawback of using NIRS together with PLS is that uncertainty estimates are seldom reported but essential to establishing real-time risk assessment. In a multivariate regression setting, sample-specific prediction errors are needed, which complicates the uncertainty estimation. We give a straightforward strategy for implementing an already developed, but seldom used, method for estimating sample-specific prediction uncertainty. We also suggest an improvement. Comparing independent reference analyses with the sample-specific prediction error estimates showed that the method worked on industrial samples when the model was appropriate and unbiased, and was simple to implement.

Selecting the optimum number of partial least squares components for the calibration of attenuated total reflectance-mid-infrared spectra of undesigned kerosene samples

Selecting the correct dimensionality is critical for obtaining partial least squares (PLS) regression models with good predictive ability. Although calibration and validation sets are best established using experimental designs, industrial laboratories cannot afford such an approach. Typically, samples are collected in an (formally) undesigned way, spread over time and their measurements are included in routine measurement processes. This makes it hard to evaluate PLS model dimensionality. In this paper, classical criteria (leave-one-out cross-validation and adjusted Wold's criterion) are compared to recently proposed alternatives (smoothed PLS-PoLiSh and a randomization test) to seek out the optimum dimensionality of PLS models. Kerosene (jet fuel) samples were measured by attenuated total reflectance-mid-IR spectrometry and their spectra where used to predict eight important properties determined using reference methods that are time-consuming and prone to analytical errors. The alternative methods were shown to give reliable dimensionality predictions when compared to external validation. By contrast, the simpler methods seemed to be largely affected by the largest changes in the modeling capabilities of the first components.

Adulteration of diesel/biodiesel blends by vegetable oil as determined by Fourier transform (FT) near infrared spectrometry and FT-Raman spectroscopy

In this work it has been shown that the routine ASTM methods (ASTM 4052, ASTM D 445, ASTM D 4737, ASTM D 93, and ASTM D 86) recommended by the ANP (the Brazilian National Agency for Petroleum, Natural Gas and Biofuels) to determine the quality of diesel/biodiesel blends are not suitable to prevent the adulteration of B2 or B5 blends with vegetable oils. Considering the previous and actual problems with fuel adulterations in Brazil, we have investigated the application of vibrational spectroscopy (Fourier transform (FT) near infrared spectrometry and FT-Raman) to identify adulterations of B2 and B5 blends with vegetable oils. Partial least square regression (PLS), principal component regression (PCR), and artificial neural network (ANN) calibration models were designed and their relative performances were evaluated by external validation using the F-test. The PCR, PLS, and ANN calibration models based on the Fourier transform (FT) near infrared spectrometry and FT-Raman spectroscopy were designed using 120 samples. Other 62 samples were used in the validation and external validation, for a total of 182 samples. The results have shown that among the designed calibration models, the ANN/FT-Raman presented the best accuracy (0.028%, w/w) for samples used in the external validation.

Spectral range optimization for the near-infrared quantitative analysis of petrochemical and petroleum products: naphtha and gasoline

The proper selection of the spectral range in partial least squares (PLS) calibration is critical when highly overlapping spectra from compositionally complex samples are used, such as naphtha and gasoline. In particular, the relevant spectral information related to a given property is frequently localized in a narrow range, and the most selective region may be difficult to locate. We have presented the importance of range optimization in near-infrared (NIR) spectroscopy for the analyses of petrochemical and petroleum products that are generally highly complex in composition. For this purpose, the determination of a detailed compositional analysis (so called PIONA) and the distillation temperature of naphtha were evaluated. In the same fashion, the research octane number (RON) and Reid vapor pressure (RVP) were selected for gasoline. By optimizing the range using moving window (MW) PLS, the overall calibration performance was improved by finding the optimal spectral range for each property. In particular, for a detailed compositional analysis of naphtha, it was effective to search for localized spectral information in a relatively narrow range with fewer factors.

Moment combined partial least squares (MC-PLS) as an improved quantitative calibration method: application to the analyses of petroleum and petrochemical products

A new quantitative calibration algorithm, called "Moment Combined Partial Least Squares (MC-PLS)", which combines the moment of spectrum and conventional PLS was proposed. Its calibration performance was evaluated for the analyses of three import petroleum and petrochemical products: gasoline, naphtha and polyol samples. The selected properties for these products included the research octane number (RON) and Reid vapor pressure (RVP) for gasoline, the distillation temperature at 10% (D 10%) for naphtha and the hydroxyl (OH) number for polyol. The major concept presented here used the moment to find the closest spectrum of a sample in a given dataset, and generate the difference spectrum and the corresponding difference in the property. These difference spectra and property differences were then used for PLS calibration. The moment has been employed in spectroscopic fields as a simple and effective "spectral feature characteristic" using just a few scalar values (moments). MC-PLS showed improved prediction performance over PLS for each case. In MC-PLS, the difference spectra generated using the moments were used as explained; therefore, additional detail in spectral variations can be utilized for calibrations. Additionally, the difference in the property was employed as reference data, so that its variation range was smaller when compared with that of the original property. Consequently, the MC-PLS performance could be better since the feature-enhanced spectra were used to model a narrower range of property variations. In the case of the D 10% prediction for naphtha, a non-linear prediction pattern that occurred in conventional PLS was effectively corrected using the MC-PLS method.