Phil Williams

Comparison of commercial near infrared transmittance and reflectance instruments for analysis of whole grains and seeds

Near infrared transmittance and reflectance instruments were compared for the determination of protein, oil, moisture and some other constituents and parameters in several grains and seeds of commerce. Both approaches were comparable in accuracy and reproducibility. The importance of optimisation of the wavelength range in whole grain analysis is demonstrated for measurements in both the NIR and visible/NlR wavelength ranges. The RPD statistic, which relates the standard error of prediction to the standard deviation of the original data, is illustrated as a method for the evaluation of calibrations. The concept of monitoring the accuracy of analysis using whole grain calibrations with ground grain calibrations is introduced.

Comparison of three whole seed near-infrared analyzers for measuring quality components of canola seed

Whole seed near-infrared (NIR) analyzers are capable of high-speed compositional analysis of oilseed commodities. This study compared the PerCon Inframatic 8144 (Perten Instruments, North America Inc., Reno, NV), the Tecator Infratec 1225 (Tecator AB, Hoganas, Sweden) and the NIR-Systems 6500 (NIR Systems, Inc., Silver Spring, MD) analyzers for measurement of oil, protein, chlorophyll and glucosinolates in intact canola seed of composite samples from the Grain Research Laboratorys (Winnipeg, Manitoba, Canada) annual Western Canada Harvest Surveys (1985–1989) for assembly of calibration and prediction sets. No significant differences were found between the three instruments for oil [standard error of prediction (SEP 0.43–0.55%)], protein (SEP 0.35–0.42%) and glucosinolates (SEP 2.4–3.8 mM/g). Neither the Tecator nor the PerCon instruments were effective for determining chlorophyll. By combining oil content and fatty acid composition data to give an estimate of the total level of each fatty acid in the sample, high correlations were obtained for total saturates, linolenic acid, and linoleic acid although the RPD (ratio of the S.E. of prediction to the S.D. of the original data) values were not high enough to enable routine use of the method to predict results.

Analysis of oilseeds for protein, oil, fiber and moisture by near-infrared reflectance spectroscopy.

Wavelength and mathematical treatments were optimized for the determination of oil, protein, moisture and crude fiber components in the ground seeds of nine oil-bearing crops [rape, flax, sunflower, safflower sesame, palm kernel, groundnut (peanut), soybean spectroscopy. Optimum wavelengths, selected for the estimation of various components, were influenced by the algorithm (math treatment) used and differed among crops. The second derivative math appeared to be better suited for the estimation of all constituents. Methods for sample preparation of all constituents. Methods for sample preparation and analytical results are discussed. The accuracy was quite satisfactory for routine quality control and evaluation purposes, and precision was equal to that of standard analyses.

Analysis of Feed Barley by near Infrared Reflectance Technology

Rapid methods for predicting feed barley quality with near infrared (NIR) reflectance spectroscopy were investigated. Reference tests for true metabolisable energy (TME), in vitro digestibility, neutral-detergent fibre, protein and kernel plumpness in feed barley is time-consuming. Near infrared technology can save considerable time by testing all of the above simultaneously, but accurate calibration of the equipment is essential. Calibration requires accurate results from chemical or physical tests and wide variance in reference data. Our calibration data sets were selected from over 800 feed barley (hulless, two- and six-rowed) samples that were grown at various locations across the Canadian Prairies in 1990 or 1991. Calibrations for a NIRSystems 6500 scanning spectrophotometer, using both whole and ground kernels, were calculated using one of three basic mathematical treatments: log (1/R) or the first or second derivative thereof. Partial Least Squares regression was applied to the best mathematical treatment and further calibrations were generated where applicable. We found correlations of 0.95 (TME), 0.98 (in vitro digestibility), 0.90 (NDF), 0.97 (protein) and 0.91 (kernel plumpness). Standard errors of prediction were 0.21, 0.97, 0.65, 0.31 and 11.5, respectively.

Influence of Water on Prediction of Composition and Quality Factors: The Aquaphotomics of Low Moisture Agricultural Materials:

The main constituents of agricultural materials are starch, protein, oil (lipid), cellulose and minerals. These constituents all exist in the molecular state and are constantly in motion, but are essentially functionally static in the raw materials from which foods and feeds are made. It is only when the molecules are brought together by an external medium that brings them into contact with each other in a way in which they can interact more freely, that the molecules are able to create new products. The most important medium is water. Water has a profound influence on the behaviour and the spectra of agricultural commodities. The paper will demonstrate that during the development of near infrared (NIR) calibration models, using several different approaches, the selection of wavelengths in areas where water is known to absorb in the NIR region, is widespread among calibrations for prediction of composition, and functionality in systems in which the moisture content is not high (up to about 15% moisture), and water is not the dominant constituent. Over 200 references have been made to assignments for water in some form, between 700–2200 nm. The influence of water on NIR calibration models for prediction of composition and functionality of materials of even fairly low moisture content is convincing. Evidence is presented in terms of partial least squares loadings for NIR calibration models developed on these materials. The relevance of this is discussed with respect to the possibility of interaction with other absorbers at the same, or adjacent wavelengths. The concept is presented that studies of changes in the intensity and positions of the absorbances of water, and those of other major constituents during development of NIR calibration models, can provide new information on the interaction among these constituents that result in differences in the structure and functionality of the materials, and on the efficiency of NIR analysis.

Prediction of wheat bread-baking functionality in whole kernels, using near infrared reflectance spectroscopy

Efficiency of prediction of wheat strength by “wet chemistry” prediction tests is summarised. Inter-relationships among these tests, and recognised criteria of wheat strength (physico-chemical properties and baking data) are reviewed. Calibrations for predicting both “prediction-type” and data from recognised test procedures have been developed for use with whole wheat. The relative efficiencies of near infrared and established wet chemistry methods for the prediction of wheat strength for use in plant-breeding are discussed. For screening large numbers of lines for several strength parameters, near infrared methods are more practicable.

Tutorial: Items to be included in a report on a near infrared spectroscopy project:

There are nearly 40 items that should ideally be reported when an NIR (near infrared) spectroscopy project is completed, either as a report or as a scientific paper. However, in our reading of the extensive literature, many of the papers presented or published report no more than 6–10 of these. The purpose of this tutorial is to indicate all of the items and the reasons for reporting them. Most of the items that need to be reported are important for anyone who seeks to duplicate the type of application and methods reported in a peer-reviewed journal article for their own work. Practically, all of the items are significant to any worker if the eventual objective of their work is to extend it to the level of industrial application. The tutorial will summarize these items, and give some explanation for their inclusion. The tutorial should be useful to potential authors, as well as to reviewers.

Feasibility study on the evaluation of the dry rubber content of field and concentrated latex of Para rubber by diffuse reflectance near infrared spectroscopy

The analysis of the dry rubber content (DRC) of Para rubber latex, including field latex and concentrated latex, using near infrared spectroscopy was conducted using a Fourier transform near infrared (FT-NIR) spectrometer in diffuse reflection mode over the wavenumber range of 4000–10,000 cm−1. The proposed method is useful for industrial purposes. The best model was developed using the partial least square regression (PLSR) from the spectra, which were pretreated using the 2nd derivative method, where the correlation (r2), standard error of prediction (SEP) and bias were 0.997, 0.3398% and −0.0239%, respectively. The ratio of standard deviation (SD) to SEP of the reference data in the prediction sample set (RPD) was 18.18 and the ratio of the range to the SEP of the prediction set (REP) was 74.4. The model was validated using a new batch of samples and the prediction performance was good with an r2 of 0.999, a SEP of 0.3898% and a bias of −0.0008%. Therefore, the NIR spectroscopy technique can be used as an accurate and rapid method for estimating the DRC of Para rubber latex for both field and concentrated latex.

Observations on the use, in prediction of functionality in cereals, of weights derived during development of partial least squares regression

Partial least squares regression is a multivariate method commonly used for the development of near infrared calibrations for prediction of composition and functionality in wheat and other cereals. Modern software enables storage of the “weights” and “loadings” realised during equation development. These can be used in the interpretation of the factors affecting the development of the calibration equations. The weights indicate areas of wavelength where variance in the optical signal has been used in the development of the calibration equation. This paper gives examples of the use of weights in interpretation of calibrations for the prediction of composition and functionality factors.

What does the raw material have to say

Very poor Not recommended Poor Not recommended Fair Rough scr~ening Reasonable Screening Good Quality control Very good Quality assurance Excellent Any application Superior As good as reference This short feature draws attention to the importance of the relationship of the standard deviation (SD) of the reference data for analysis of any material to the SEP (Standard Error of Performance or Prediction) achieved, using a new NIA or NIT calibration. The chief objective of NIA/NIT analysis for composition or functionality is to achieve results as accurate as the reference method, but much quicker and cheaper. success has been measured in terms of the coefficients of correlation, or determination , and the SEP, which is the standard deviation of differences between reference and NIA methods. ·e.g. in plant-breeding programmes.