Animal Feed Applications

Introduction

 The animal feed industry is constantly examining new methods for the analysis of nutritional and chemical composition of products. Traditional methods are often expensive and time-consuming. They are ill-suited to measure large numbers of samples and especially for determining variation within a batch of animal feed because most methods can only measure a small portion of sample. Other disadvantages include the inability of most techniques to perform real-time, on-line measurements during the manufacturing process and being unable to determine more than one parameter from a single measurement. In recent years, NIR spectroscopy has emerged as a technique that offers many advantages over traditional analytical methods. It is fast, non-invasive, and able to determine multiple parameters of interest from a single measurement. It requires no chemical or toxic solvents and can not only measure large amounts of samples but can measure a large sample amount in one reading. This is very important when accounting for variation within a batch, as animal feed and other natural products can often show large amounts of variation.

The ability of NIR spectroscopy for use in measuring parameters in animal feed has been proven for some time. However, recent advances in technology have helped facilitate its use as a complete process analytical tool from the analysis of raw materials in the field to a final quality control check. Handheld spectrometers can be used for field measurements and to check incoming batches of raw materials before entering the manufacturing process. On-line instruments can be used as real-time process control tools to optimize the manufacturing process, resulting in improved product quality, optimization of resources, and potentially significant cost savings. It can also be used to both detect and quantify adulterants in animal feed, which is a significant problem with potential health and economic consequences. An examination and review of applications using NIR spectroscopy in the animal feed industry is presented here. Reviewed topics include the use of the technique to determine nutritive value, energy parameters, proper labeling of the finished product, and detection of adulteration in various animal feed products.

Analytes

  • In Vitro Dry Matter Digestibility (IVDMD)
  • In Vitro Organic Matter Digestibility (IVOMD)
  • Moisture
  • Crude Protein (CP)
  • Crude Fat (CF)
  • Ether Extract (EE)
  • Calcium (Ca)
  • Phosphorus (P) – Total and Phytate
  • % Meat and Bone Meal (MBM)
  • % Poultry, Pork, and Cattle Meals
  • Classification of Fish, Meat, and Soya Meal
  • Correct Labeling of Bagged Feed Based On Twenty-Four Different Feed Types
  • Comparison of Traditional NIR Spectroscopy and NIR Chemical Imaging in the Detection and Quantification of Ruminant Meal in Processed Animal Proteins
  • Soybean Meal Adulteration with Melamine, Cyanuric Acid, and Whey Powder
  • Dry Matter (DM)
  • Mineral Matter (MM)
  • Neutral Detergent Fiber (NDF)
  • Acid Detergent Fiber (ADF)
  • Ether Extract (EE)
  • Gross Energy (GE)
  • Metabolizable Energy (ME)
  • Ash

Scientific References and Statistics

Fast and Simultaneous Prediction of Animal Feed Nutritive Values Using Near Infrared Reflectance Spectroscopy

Sago starch is a starch extracted from the spongy core tissue of tropical palm stems. The tissue is often referred to as pith and is abundantly produced in Southeast Asia, especially in Malaysia and Indonesia. Animal feed typically comprises about 70% of livestock production cost and producers need efficient methods to analyze the composition of animal feed. The abundance of sago starch in Indonesia has facilitated its use as an animal feed product. Sago starch contains around 27% crude fiber. Sago residues also contain starch after extraction with an approximate composition of 57.9% starch, 23.1% cellulose, 9.2% cellulose, and 3.9% lignin. High amounts of crude fiber can cause slow and limited degradation of carbohydrates in animals. Low nitrogen and protein content can be present in sago residues, reducing the nutritional value of feed for ruminants. Microbial fermentation is one method that can be used to improve the nutritional content of animal feed from residues. However, it is crucial to properly evaluate the nutritive value of animal feed before feeding to ensure animal health and maximization of quality in livestock products. Traditional methods for evaluating the nutritive value of animal feed are often time-consuming, expensive, and require the use of expensive chemicals and reagents. They are also insufficient for measuring large amounts of samples and can only measure a small portion of an individual sample. This is especially a problem when analyzing agricultural products as there can be significant variation in nutritive values within the same batch. NIR spectroscopy offers the advantage of being fast and non-invasive while requiring little or no sample preparation and being able to determine multiple parameters of interest with a single measurement. In this study, sago residues were analyzed using NIR spectroscopy for the purpose of determining digestive organic material parameters.

Twenty-five sago residues which were fermented by a standard process were procured for the study. Samples were scanned using an FT-NIR spectrometer from 1000 nm to 2500 nm at 0.2 nm intervals. Twenty scans were collected per reading and averaged into one spectrum. Standard laboratory methods were performed on the samples for two digestive parameters: in vitro dry matter digestibility (IVDMD) and in vitro organic matter digestibility (IVOMD). Both are evaluations of the nutritional value of ruminant feed by measuring the portion of matter that is digested by animals at a specific level of feed intake. Various pre-processing methods were performed on the spectral data before chemometric modeling. Partial Least Squares (PLS) calibration models were created using the NIR spectra and reference values for IVDMD and IVOMD. Results are shown below.

IVDMDR2 = 0.865RMSEC = 1.20
IVOMDR2 = 0.810RMSEC = 1.71

The results of this study showed that measuring digestibility parameters using NIR spectroscopy and calibration models is a feasible method of analysis.  The results shown were the best obtained and the Standard Normal Variate (SNV) pre-processing method was used on the spectral data.  Results could likely be improved using a larger sample set which extends the range of values of the digestibility parameters as the range for both IVDMD and IVOMD was limited for the samples.  Further study and calibration work would be necessary before using this method in a real-time setting.

Rapid and Simultaneous Determination of Feed Nutritive Values by Means of Near Infrared Spectroscopy | Tropical Animal Science Journal (ipb.ac.id)

Nutritional Evaluation of Commercial Broiler Feeds by Using Near Infrared Reflectance Spectroscopy

Broiler farming is one of the fastest growing industries in Bangladesh and production of broiler feeds is dependent on the composition of compound feeds. Proper nutritional evaluation of feed ingredients is essential to achieve a balance between providing adequate nutrition to the animals and maximizing cost efficiency. Standard methods for evaluating nutritional parameters in broiler feeds are often time-consuming, expensive, and require the use of expensive chemicals and reagents. They are also insufficient for measuring large amounts of samples and can only measure a small portion of an individual sample. NIR spectroscopy offers the advantage of being fast and non-invasive while requiring little or no sample preparation and being able to determine multiple parameters of interest with a single measurement. In this study, locally available broiler feeds in Bangladesh were evaluated using NIR spectroscopy to determine the feasibility of using this method to measure nutritive value in a fast and efficient manner.

Over five hundred broiler feed samples were collected from different commercial feed mills and feed markets over four consecutive years for the study. Samples were ground before scanning to obtain homogenous particle size. The ground samples were scanned using a FT-NIR spectrometer from 700 nm to 2400 nm at 8 nm resolution. Thirty-two scans were collected per reading and averaged into a single spectrum. Standard AOAC procedures were used to obtain reference values for moisture, crude protein (CP), crude fat (CF), ether extract (EE), calcium (Ca), and potassium (P). Various pre-processing methods were performed on the spectral data before chemometric modeling. Partial Least Squares (PLS) calibration models were created using the NIR spectra and the reference values for parameters of interest. It must be noted that all the reference tests were not performed on all the samples and each model contains a different amount of samples. Results are shown below.

ParameterSamplesRangeR2RMSECV
Moisture54310.90%-13.11%0.8430.843
CP35213.12%-22.96%0.9520.371
CF2151.78%-7.20%0.8180.406
EE3743.09%-11.35%0.9550.390
Ca2300.63%-1.27%0.7500.066
P2410.38%-0.83%0.9470.031

Cross validation was used to evaluate the predictive capability of the models and the results proved the feasibility of using NIR spectroscopy and calibration models to evaluate nutritive values in broiler feeds. Results were excellent for CP, EE, and P which showed correlation coefficients around 0.95 and low predictive error. The range of these samples were large and likely contributed to the good results. In the case of moisture and EE, correlation coefficients were lower. The correlation for moisture was especially low considering that water bonds are highly absorbing of NIR light and usually shows very high correlation in calibration models. It is unclear whether the samples were stored for a period of time before reference tests were performed and if this is the case, it is likely that the moisture values may have changed over time. There was also an uneven distribution of moisture values in the samples and a more even distribution may improve results. Correlation for Ca was low at 0.75 and non-organic minerals cannot be directly measured using NIR spectroscopy. However, in some cases a change in the mineral content may affect the organic bonds in the molecule and an indirect correlation is possible (such as the P model in this study). Such models must be carefully validated to ensure that the correlation is valid. Overall, the results of this study proved the feasibility of using NIR spectroscopy and calibration models to determine nutritive value of broiler feeds.

https://www.researchgate.net/publication/329738144_Nutritional_Evaluation_of_Commercial_Broiler_Feeds_by_Using_Near_Infrared_Reflectance_Spectroscopy

Near Infrared Spectroscopy for Enforcement of European Legislation Concerning the Use of Animal By-Products in Animal Feeds

Strict legislation was first passed by the EU in 2000 with further legislation in subsequent years that bans the use of animal origin meals in compound feeds and intra-species recycling. The use of animal-by-products such as meat and bone meals (MBM) play a crucial role in feed manufacturing, but the safe and healthy use of them in animal feeds is important to prevent the spread of Bovine Spongiform Encephalopathy (BSE). The legislation banning the use of by-products in animal feeds will only be lifted if analytical methods that are validated that can not only detect the presence of MBM, but also identify the animal species origin of the meal. NIR spectroscopy is a proven method for measuring parameters of interest in animal feed, especially for detecting and quantifying adulterated samples. It offers the advantages of being fast and non-invasive while requiring little or no sample preparation. It can measure both large amounts of samples and large portions of each individual measurement within the same batch. The method is proven for both quantitative and qualitative analysis of animal protein by-products and the feed mixtures containing them.

Two separate R&D projects are examined here that demonstrated the ability of NIR spectroscopy to help the enforcement of EU legislation governing the use of by-products in animal feed. Three types of animal feed were used for the study: compound feeds (CFs), animal protein by-products (APBPs), and animal fat by-products (AFBPs). For CF analysis, two separate sets of samples were used: one thousand five ground samples and five hundred and twenty-three unground samples. Samples were scanned using an NIR spectrometer from 400 nm to 2498 nm at 2 nm intervals. Reference data for the inclusion of meat and bone meal were either provided by the feed manufacturer or determined from optical microscopy. The NIR spectra and reference values for %MBM were used to create a Partial Least Squares (PLS) calibration model for both sets of data. Results are shown below.

Ground Samples
Range of %MBM = 0.00 – 34.85R2 = 0.97SECV = 0.94
Unground Samples
Range of %MBM = 0.00 – 32.55R2 = 0.98SECV = 0.80

A separate set of samples was used to validate both sets of models and showed good prediction results.  The second part of the study examined two hundred eighty samples of APBPs from different providers.  The following types of meals were included: pure poultry, pure pork, pure cattle, cattle-poultry mixture, cattle-pork mixture, poultry-pork mixture, blood, fish, hydrolysed feather meal, feather meal, and cattle-sheep-pork & poultry mixture.  All samples were scanned using the same parameters as the CF samples and visual examination of the NIR spectra showed clear and visible differences in the wavelength range from 1680 nm to 1760 nm, indicating that NIR spectroscopy could be used to differentiate between different types of APBPs.  PLS regression models were created for the following types of meal in the samples: poultry, pork, and cattle.

Poultry MealRange = 0% -100%R2 = 0.94SECV = 8.73
Pork MealRange = 0% -100%R2 = 0.93SECV = 8.12
Cattle MealRange = 0% -26.6%R2 = 0.81SECV = 3.5

The PLS models for the three types of meal showed good results and proved the feasibility of using NIR spectra and calibration models to quantify the amount of a particular type of meal in APBPs.  In the case of cattle meal, the range of values was smaller and more samples with a high percentage of cattle meal should improve results.  The third portion of the study scanned AFBP samples to determine if visual spectral examination could distinguish between poultry and pork fat samples.  The sample size was limited but a calibration model correctly determined whether a sample was poultry, pork, or a mixture for thirty-seven out of thirty-nine samples.  Some differences were observed around 1710 nm and 1725 nm, likely due to differences in oleic and linoleic acids content as well as fat insaturation degree.  The results of this study could lay the groundwork for using NIR spectroscopy as a method for determining the amount and type of MBM in animal feed, allowing for less restriction in current legislation if the type of MBM can be determined to be a safe one. 

(PDF) Near infrared spectroscopy for enforcement of European legislation concerning the use of animal by-products in animal feeds (researchgate.net)

Usefulness of Near Infrared Reflectance (NIR) Spectroscopy and Chemometrics to Discriminate Between Fishmeal, Meat Meal, and Soya Meal Samples

Qualitative and quantitative quality analytical control is essential in food manufacturing for the assessment of raw materials, finished products, and by-products as well as for optimization of the manufacturing process itself.  While nutrient composition is important, there are also other types of analysis that can relate to product processing history or geographical origin.  Examples of this include fresh vs. frozen meat and assessing the origin of olive oil as there are strict regulations regarding the origin of Italian olive oil.  Along similar lines, adulteration detection is of utmost importance as high value products are subject to adulteration.  Intentional adulteration from mixture of a lower quality product can be profitable and difficult to detect from visual examination.  Selling of an inferior product that was subject to environmental exposure (such as excess rain or sun) can also constitute a form of adulteration.  One product that is subject to adulteration is fish meal.  It has a higher protein content and market value than other types of meal, such as meat meal and soya meal.  Traditional methods for detecting adulteration are often time-consuming, expensive, and require the use of expensive chemicals and reagents.  They are also insufficient for measuring large amounts of samples and can only measure a small portion of an individual sample.  This is especially a problem when analyzing agricultural products as there can be significant variation within the same batch.  NIR spectroscopy offers the advantage of being fast and non-invasive while requiring little or no sample preparation. In this study, NIR spectroscopy was examined for the purpose of detecting adulteration in fish meal. 

Seven samples of fish meal, ten samples of meat meal, and fifteen samples of soya meal were procured from different industrial manufacturing plants for the study.  All samples were scanned using an NIR spectrometer from 1100 nm to 2500 nm at 2 nm intervals.  Various pre-processing methods were applied to the spectral data before chemometric analysis.  Principle Component Analysis (PCA) was performed to analyze differences in the spectra.  A Partial Least Squares-Discriminant Analysis (PLS-DA) model was created to use the NIR spectra to classify the type of meal sample.  In a PLS-DA model, arbitrary values are assigned to different sample types for classification purposes.  In this case, fish meal was given a value of 1, meat meal a value of 2, and soya meal a value of 3.  Samples were given a cut-off value of 0.5 and the NIR spectra were used with the model to choose the sample type.  Results are shown below.

PLS-DA Model
R2 = 0.92RMSECV = 0.20
Correct Classification
Fish Meal85.7%
Meat Meal100.0%
Soya Meal100.0%

The results of the study were excellent and proved the feasibility of using NIR spectroscopy and a PLS-DA calibration model to determine meal type.  In the cases of meat meal and soya meal, validation samples were correctly classified every single time.  Some samples of fish meal were classified as meat meal but protein analysis of these samples showed that the protein content was abnormally low for fish meal and showed a protein content that is closer to an average meat meal sample.  The fact that the classification is determined by protein content, which is the most important ingredient in meal used for animal, bodes well for use of this type of spectroscopic analysis in a practical setting.  Further analysis should include mixtures of samples to quantitatively determine the percentage of different meals for the determination and detection of adulteration.

(PDF) Usefulness of near infrared reflectance (NIR) spectroscopy and chemometrics to discriminate between fishmeal, meat meal and soya meal samples (researchgate.net)

Multivariate Near-Infrared Reflection Spectroscopy Strategies for Ensuring Correct Labeling at Feed Bagging in the Animal Feed Industry

Correct labeling of compound feed is a key concern in the animal feed industry due to traceability and safety issues.  A requirement of feed labels is the “purpose statement”, which conveys the intended purpose to the user.  It must communicate the species and physiological status for which the feed is manufactured.  At the labeling and packaging stages of feed manufacturing, there are potential mechanical and human errors that can lead to incorrect labeling.  There is a need for control methods that ensure that animal feed products are properly labeled.  NIR spectroscopy offers the advantages of being fast and non-invasive while requiring little or no sample preparation.  In this study, NIR spectroscopy was examined for the purpose of classifying twenty-four different feed types using different linear and non-linear multivariate algorithms. 

A factory provided 17,306 NIR spectra of feed compound samples in their original physical final market form that were collected over four years.  Reference data regarding ingredient composition and feed type were also provided.  Spectra were divided into separate groups.  The first division was between granules and meals.  Granules undergo a process of heating, steaming, and pelleting while meals are simply mixtures of different ground ingredients.  Within meals, there were three groups corresponding to different species: porcine, ovine, and bovine.  Porcine meals have six types of feeds, ovine have two types, and bovine have three types.  Granules have the same three groupings for species with four types of feed for porcine and ovine and five types for bovine.  Types were based on differences in the age or physiological status of the animal.  In practice, there are more types of feed that were not included due to low availability. 

The NIR spectra were analyzed using two linear methods and one non-linear method.  The first linear method is Soft Independent Modeling of Class Analogy (SIMCA).  The second linear method is Partial Least Squares (PLS) with two different approaches to classification (PLSD and PLS-DA).  Support Vector Machines (SVM) is the non-linear method used.  Within the database structure, two separate strategies were used to evaluate the methods.  The first was the development of a model composed of the nine classification models corresponding to the structure of the data.  The nine classification models were using each of the three multivariate methods with no spectra pre-processing, first derivative & Standard Normal Variate (SNV) processing, and second derivative & SNV processing.   The best classification results using this strategy were obtained using SVM with a classification error of 3.96%.  The second strategy developed one unique model that discriminates among the twenty-four different classes of feed.  SVM also showed the best results of the three multivariate methods using this strategy with a classification error of 2.31%.  Among the two linear methods, SIMCA showed the best results with a classification error of 8.47% using the first strategy and 4.05% classification error.  The results of this study showed the ability of NIR spectroscopy to make acceptable classification of feed types using multivariate algorithms and a NIR spectra database.

Multivariate near-infrared reflection spectroscopy strategies for ensuring correct labeling at feed bagging in the animal feed industry – PubMed (nih.gov)

Detection and Quantification of Ruminant Meal in Processed Animal Proteins: A Comparative Study of Near Infrared Spectroscopy and Near Infrared Chemical Imaging

In January 2001, a suspension was implemented in all EU countries on the use of processed animal proteins (PAP) in feed for any animals farmed for production of food.  There are some exceptions, such as the use of fish meal for non-ruminants.  Ten years later in 2011, there was a growing concern about the protein deficit in the compound feed manufacturing sector and portions of the ban were relaxed.  Careful examination of the situation made it clear that the total ban was impractical and from an adulteration standpoint, it was necessary to only determine whether significant amounts of an adulterant were added to animal feed products and not trace amounts.  Adding trace amounts of an adulterant will make little economic impact on a product and it makes little sense for those attempting to make money from adulterating products to do so.  While light microscopy is the only method authorized for official inspections for adulteration, it is impractical to use on large amounts of samples.  A need exists for fast, non-invasive methods that can test on a large scale to determine the presence of adulteration in PAP animal feed.  In this study, the methods of NIR spectroscopy and NIR chemical imaging were used to determine the feasibility of using them for detecting and quantifying ruminant meat meal in PAP animal feed. 

A set of one hundred and twenty-six fish meal samples were procured for the study.  Samples were adulterated with controlled amount of ruminant meat meal ranging from 0.25% to 16%.  For the NIR spectroscopy portion of the study, all samples were scanned from 400 nm to 2498 nm at 2 nm intervals.  Thirty-two scans were collected per reading and averaged into one spectrum.  For the NIR chemical imaging of the sample, samples were analyzed with a single scan using a chemical in the absence of external light from 1000 nm to 1700 nm at 10 nm intervals.  Various pre-processing methods were performed on both sets of spectral data before chemometric analysis.  The first type of analysis was qualitative to determine if either method could accurately detect the presence of adulteration in the samples.  Eighty-eight additional samples were procured as a validation set.  Partial Least Squares Discriminant Analysis (PLS-DA) models were created to classify whether or not a sample contained an adulterant based on the spectral data.  The validation results showed that the imaging method accurately classified the presence of an adulterant or not with over 98% accuracy, while the traditional NIR method predicted with 86% accuracy.  The next portion of the study used both sets of spectral data and the reference values of % adulteration to create PLS models that quantify the amount of adulteration present.  Results are shown below.

Traditional NIR Method
R2 = 0.97RMSEP = 1.69%
NIR Chemical Imaging
R2 = 0.84RMSEP = 2.17%

The results for quantitative analysis of the adulteration were better using the traditional NIR method and independent predictions performed on the validation set confirmed that this was the case.  NIR imaging was shown to be more effective for qualitative analysis and traditional NIR spectroscopy was more effective for quantitative analysis. Therefore, imaging is better suited to species discrimination and identification, while the traditional method is better suited to the quantification of a meal adulterant in a given species of feed.  It must be noted that the traditional method still showed decent results for the qualitative analysis but the imaging method showed very poor results in the predictions for the quantitative measurement of the adulterant.  This is likely due to poorer spectral quality from imaging compared to traditional NIR, an issue that should improve in the future as the technology of NIR cameras advances.  Both methods should play an important role in the future of animal feed analysis and adulteration detection.

Detection and Quantification of Ruminant Meal in Processed Animal Proteins: A Comparative Study of near Infrared Spectroscopy and near Infrared Chemical Imaging (optica.org)

NIR Fingerprint Screening for Early Control of Non-Conformity at Feed Mills

A number of high profile incidents related to health have increased the awareness of human food and animal feed safety.  One such incident is the recall of pet food by several pet food manufacturers in North America after a number of dogs and cats died after consuming contaminated pet food.  The FDA reported finding melamine in the pet food and in samples of wheat gluten imported from China.  Melamine shows as regular crude protein in traditional testing methods.  Similarly, milk and infant formula contaminated with melamine were discovered in China the following year after more than three thousand people were taken ill and six infants died from kidney damage.  Soybean meal is the feed remaining after the solvent extraction of oil from soybean flakes.  It consists of more than 36% protein and 30% carbohydrate and is an important source of dietary fiber, vitamins, and minerals.  Approximately 90% of soybean seeds produced globally are used as animal feed, with an amount exceeding two hundred and five million tons. 

The traditional method for determining protein content in feed products is the Kjeldahl method, which is based on the decomposition of nitrogen in organic samples by using concentrated acid, followed by a distillation and titration.  This method is not only time-consuming and requires the use of toxic chemicals and solvents, it is unable to distinguish the inexpensive and potentially toxic melamine which can be added as an adulterant.  There is a need for a fast, effective method to detect melamine adulteration in animal feed.  In this study, a procedure was developed using NIR spectroscopy and chemometrics to characterize soybean meal and to detect the presence of unusual ingredients such as melamine, cyanuric acid, and whey powder. 

The procedure was validated at the laboratory level and later adapted for application at the Cargill Animal Nutrition feed mill.  The first step in the study was to create Partial Least Squares (PLS) regression models for fat and protein using a historical data set and then use those models to predict and characterize unknown samples.  In total, eight thousand and ten samples were used to create the calibration models.

ProteinR2 = 0.88RMSECV = 0.68%
FatR2 = 0.98RMSECV = 0.25%

After these regression models were constructed, they were used with the NIR spectra of unknown samples to both predict fat and protein values and to extract the spectral residuals from those samples.  The spectral F ratio for unknown samples is traditionally used for detecting samples that are not in compliance with the calibration set in prediction, often referred to as prediction outliers.  In this case, the technique is used with the PLS models to not characterize based on prediction values for fat and protein, but rather to check whether the spectral residuals of unknown samples fall within the limits defined by the spectral residual of the calibration set obtained using different PLS models.  Limits were defined by the values of the 99.7th percentile calculated for known pure soybean samples. 

The study was first validated at the laboratory level by using sixty-five samples of soybean meal. Samples were both contaminated and uncontaminated.  A percentage of melamine, cyanuric acid, or both were added to contaminated samples as well as whey powder ranging from 0.5% to 5% by weight.  Using the technique defined by the spectral residual and limits, all pure samples were characterized correctly.  Contaminated samples were characterized correctly as well except for three samples with a very low quantity of cyanuric acid (between 0.13% and 0.53%).  Even for these three samples, the results were very close to the limit defined by the soybean samples, giving a good indication of the threshold of detection for this technique.  Further tests were later carried out at a feed mill based on the actual contamination of twenty-five metric tons of soybean meal with whey power at the feed mill entrance.  The tests were extensive and varied the method of unloading the samples as well as the additional of extra whey powder as samples were moved along a transporter.  Test results were so encouraging that an online NIR device was installed opposite the sampling device that pulls samples before they are loaded into the feed mill.  Further testing should include additional types of contaminates used to adulterate soybean meal.

NIR fingerprint screening for early control of non-conformity at feed mills – PubMed (nih.gov)

Using Near Infrared Spectroscopy to Predict Metabolizable Energy of Corn for Pigs

Corn is considered food of well-defined chemical composition on paper tables but in reality, a number of factors can significantly alter chemical composition and nutritional value.  These include soil fertility, genetic variety of cultivars, planting conditions, processing, and storage.  Traditional testing methods for determining the nutritive and energy values of corn are often expensive, time-consuming, and require the use of toxic chemical and reagents.  These methods are also only able to determine a single parameter at a time and can only testing a small sample portion, which is significant because large variation can exist within the same batch of sample.  NIR spectroscopy offers the advantages of being fast and non-invasive while requiring little or no sample preparation.  Multiple parameters of interest can be determined from a single measurement and the technique can not only measure large amounts of samples, it can measure a large amount of sample with each reading, helping to account for variation within batches.  In this study, NIR spectroscopy was examined as a method for determining the chemical composition and energy values of different corn varieties to predict the metabolizable energy (ME) in corn fed to pigs. 

Ninety-nine corn samples of different varieties from various growing areas in Brazil were procured for the study.  Eighty-nine samples were used for model calibration and the remaining test as a validation set.  Samples were ground and traditional reference tests were performed for the following parameters: dry matter (DM), mineral matter (MM), neutral detergent fiber (NDF), acid detergent fiber (ADF), ether extract (EE), and gross energy (GE).  A portion of each sample was saved for scanning with a NIR spectrometer.  Samples were scanned from 400 nm to 2500 nm.  Various pre-processing techniques were applied to the spectral data before chemometric modeling.  Partial Least Squares (PLS) regression models were created using the NIR spectra and reference values.  Results are shown below.

DMR2 = 0.99SECV = 0.39%
MMR2 = 0.30SECV = 0.12%
NDFR2 = 0.15SECV = 0.84%
ADFR2 = 0.40SECV = 0.52%
EER2 = 0.83SECV = 0.24%
CPR2 = 0.94SECV = 0.39
GER2 = 0.86SECV = 19.62

Results were good for DM, EE, CP, and CP with good correlation and prediction validation from the validation samples.  MM results were poor because minerals do not directly absorb NIR light.  While indirect correlation to measure mineral content is possible if the mineral concentration affects organic bonds, this does not appear to be the case in this model.  In the case of NDF and ADF, it was determined that the low correlation can be attributed to the difficulty of accurately measuring the different fractions that comprise the fiber content, thus creating reference error.  Both fiber measurements have accurately been correlated to NIR spectra in other studies.  Despite the low correlation, the validation predictions for MM, ADF, and NDF were reasonable.  From these results, different prediction equations for ME were used to estimate the metabolizable energy values from the NIR spectra.  Results were decent but more accurate calibrations and further study would be necessary before applying the results of this application study in a real-time setting. 

SciELO – Brazil – Using near infrared spectroscopy to predict metabolizable energy of corn for pigs Using near infrared spectroscopy to predict metabolizable energy of corn for pigs

Evaluation of Near-Infrared Reflectance Spectroscopy (NIRS) Techniques for Total and Phytate Phosphorus of Common Poultry Feed Ingredients

Total and phytate phosphorus are important nutritional parameters in poultry feed because excess phosphorus is polluting in the excrement of poultry.  Phytate phosphorus is the organic phosphorus bound in the form of phytic acid in plants.  Traditional methods for determining phosphorus content in poultry feed are time-consuming, expensive, and require the use of toxic chemicals and solvents.  NIR spectroscopy offers the advantages of being fast and non-invasive while requiring little or no sample preparation.  In this study, the feasibility of using NIR spectroscopy for the estimation of total and phytate phosphorus in poultry feed was examined.  

Various types of poultry feed from different feed mills in the United States and Canada were procured for the study.  The sample types are as follows: one hundred thirty-three corn, one hundred fourteen soybean meal, eighty-nine corn DDGS, ninety-five bakery by-product meal, twenty-two wheat, thirty-one wheat middlings, twenty-one canola, and fifteen wheat shorts samples.  All samples were ground and a portion of each sample was used to determine total and phytate phosphorus using traditional testing methods.  The remaining portion of each sample was scanned using an NIR spectrometer from 400 nm to 2498 nm.  Sixteen scans were collected per reading and averaged into one spectrum.  Various pre-processing algorithms were applied to the spectral data before chemometric modeling.  The NIR spectra was used along with the reference values for both phosphorus parameters to create Partial Least Squares (PLS) calibration models for each sample type.  Results are shown below.

Total Phosphorus
R2 Range0.03 to 0.85SECV Range 0.03 to 0.14
Phytate Phosphorus
R2 Range0.64 to 0.89SECV Range 0.01 to 0.04

As expected, the results for phytate phosphorus were much better than those for total phosphorus.  It is known that only organic molecules absorb light in the NIR region of the spectrum and the only way that minerals can be measured is by an indirect correlation from mineral concentration affecting organic molecules.  In the case of phytate phosphorus, the phosphate functional groups of organic phytate affect C-H and O-H bonds, allowing for an indirect correlation.  It can be concluded that the correlation for phytate phosphorus is more directly related to its molecular structure, while any correlation for total phosphorus is more related to correlations between those molecules and organics that reflect in the NIR region.  Interestingly, the correlation for total phosphorus in corn was very low despite having the largest sample set.  Correlation for both types of phosphorus in DDGS was much higher than that for corn.  While the results of this study were mixed, it is true that the total phosphorus can be estimated from phytate phosphorus and the results for the phytate phosphorus were considered sufficient for screening purposes of many of the sample types.  It is recommended that further study be conducted before using any of these calibrations in a practical setting.

Evaluation of near-infrared reflectance spectroscopy (NIRS) techniques for total and phytate phosphorus of common poultry feed ingredients – PubMed (nih.gov)

The Use of Near Infrared Spectroscopy (NIRS) to Predict the Chemical Composition of Feed Samples Used in Ostrich Total Mixed Rations

Properly determining the nutritive value of animal feed is essential for farmers.  It is crucial in maintaining both animal health and good economics.  There can be considerable variation in the nutritive value of compound feed products due to the wide range of raw materials and by-products that are used in their composition.  Traditional analytical methods are often expensive, time-consuming, and require the use of toxic chemicals and solvents.  They are also only able to measure a single parameter at a time and are insufficient for both measuring large amounts of samples and determining variation within a given batch because they can only measure a small portion of sample.  NIR spectroscopy offers the advantages of being fast and non-invasive while requiring little or no sample preparation.  It can also determine multiple parameters with a single measurement while being able to measure both large amounts of samples and a large portion of sample within a given batch.  While NIR spectroscopy is a proven method for measuring nutritive values in individual raw materials, using the technique to measure Total Mixed Rations (TMR) is more difficult because of variation created within the spectra from different raw materials being used.  In this study, the feasibility of using NIR spectroscopy to determine the nutritive value of ostrich feed was examined. 

Four hundred and seventy-nine samples of ostrich TMR were procured for the study.  All samples were scanned using a NIR spectrometer from 1100 nm to 2500 nm at 2 nm intervals.  Traditional reference methods were performed on a portion of each sample for the following parameters of interest: dry matter (DM), ash, crude protein (CP), ether extract (EE), crude fiber (CF), acid detergent fiber (ADF), neutral detergent fiber (NDF), gross energy (GE), calcium (Ca), and phosphorus (P). Various pre-processing algorithms were applied to the spectral data before chemometric modeling.  Partial Least Squares (PLS) calibration models were created to correlate the NIR spectra to the parameters of interest.  Results are shown below.

DMR2 = 0.77 SEC = 0.63
AshR2 = 0.87SEC = 0.79
CPR2 = 0.96SEC = 0.77
EER2 = 0.93SEC = 0.37
CFR2 = 0.95SEC = 1.38
ADFR2 = 0.95SEC = 1.71
NDFR2 = 0.94SEC = 2.85
GER2 = 0.87SEC = 0.22
CaR2 = 0.75SEC = 0.42
PR2 = 0.74SEC = 0.09

Ninety-four of the samples were chosen for cross-validation and the predicted results from the calibration models were compared with the reference values.  Good correlation and predicted results were obtained for the CP, EE, CF, ADF, NDF, and GE of the ostrich TMRs.  Interestingly, the results were less accurate for DM and ash despite the fact that both of these parameters are proven applications using NIR spectroscopy.  Water is especially absorbing of light in the NIR portion of the spectrum.  There are two possible reasons for the lower correlation for DM and ash.  One is potential error in the reference method.  The second is that the mixing of the rations may affect the NIR spectra in a way that makes these models more difficult to fit the data to changes in the DM and ash values.  Closer examination of the model statistics would be necessary to see if this is a possibility.  Lower correlation and worse predicted results were expected for Ca and P as they are not organic molecules and do not absorb in the NIR spectrum.  While indirect correlation of non-organic molecules is possible, this can only occur if the minerals affect the organics enough to create a correlation.  Overall, this study proved that NIR spectroscopy can be used a successful technique to measure chemical components of interest in ostrich TMR.

 (PDF) The use of near infrared spectroscopy (NIRS) to predict the chemical composition of feed samples used in ostrich total mixed ration (researchgate.net)

Application of NIR Technology in the Animal Food Industry

In order to provide animals with adequate nutrition, proper analytical techniques are important to meet requirements for nutritional parameters.  Constant monitoring during animal feed production is necessary but the results obtained from using traditional tests are often received too late to be of use during the actual manufacturing process.  Fast and accurate methods are needed to optimize the manufacturing process and one proven method for such analysis is NIR spectroscopy.  Real-time analysis can result in optimization of the manufacturing process, improving quality and potentially achieving significant savings.  In this study, the potential of NIR spectroscopy for use as a technique to measure standard nutritional parameters in soybean cakes used for animal feed was demonstrated. 

Fourteen samples of soybean cake were procured for the study.  All samples were scanned using an NIR spectrometer from 950 nm to 1650 nm.  Samples were scanned in their whole form and no grinding was required.  Standard reference tests were performed on each sample for the following parameters:  moisture, crude protein, crude fat, and crude fiber.  Various pre-processing algorithms were performed on the NIR spectra before chemometric modeling.  The NIR spectra were used with the reference values to create Partial Least Squares (PLS) regression models correlating the spectra to the parameters of interest.  Results are shown below.

MoistureR2 = 0.9783RMSEP = 0.2442
Crude ProteinR2 = 0.9905RMSEP = 0.0517
Crude FatR2 = 0.9884RMSEP = 0.0972
Crude FiberR2 = 0.9351RMSEP = 0.0433

While the sample set was very limited in this study, the results were excellent and proved the feasibility of using NIR spectra and calibration models for determining basic nutritional parameters of interest in soybean cake.  It must be noted that the results were obtained using the whole and untreated sample portion of the soybean cake, which bodes well for further development of these calibrations for use in a real-time manufacturing setting.  More work would be required by adding additional samples to ensure the calibration models are robust enough for practical use.

(PDF) Application of NIR technology in the animal food industry (researchgate.net)