Why is NIR an Effective Food Testing Technique?

NIR stands for near infrared analysis, and it is a proven method for effective and efficient food testing. It has been used throughout the food and agricultural sectors since the 1970s, and is now an established global method for quantitative testing of main constituents in most types of food and agricultural products.

Perten’s Inframatic 9500 NIR Flour Module

It is a spectroscopic technique, which means it involves observing a spectrum of light, from a source, passing through a lens or prism.

NIR uses the naturally occurring electromagnetic spectrum, and the near infrared region of this spectrum is between wavelengths of 700nm and 2500nm, measured in nanometers.

This technique has clear advantages when it comes to testing foodstuffs and produce.

The Advantages of NIR for Food Testing

The NIR method for food testing provides data quickly through rapid analysis, requiring little or no sample preparation. It provides results in seconds rather than hours. NIR does not involve any chemical or consumables, which means its costs are very low, requiring minimal manpower.

The technology is very user-friendly. It is quick and easy to learn, and it does not require someone with a specialist, technical background to operate an NIR device. Personnel and temporary workers can carry out this task confidently.

Consequently, investment in NIR offers the potential for rapid payback.

How Does NIR Work?

The principal behind NIR is using light to analyse materials.

As humans, we rely on our sight to perceive colours. In a painting, for example, an artist will use different pigments of paint to create various colours and shades, which we then differentiate between visually.

We cannot, however, tell the difference between the fat or moisture content of a given item of food with the naked eye.

However, infrared light can. Just as pigments used by artists have unique colours, so these elements in foodstuffs give off a unique infrared emission.

What an NIR spectroscopic instrument does is allow us to see what the proportion of these elements are, as if we were looking at the balance of colours in a painting.

The critical thing here is not the simple fact that these compounds of foodstuffs or products contain moisture and fat, but rather how much of these things they contain.

NIR will judge the concentration of the various properties of a foodstuff through the intensity of specific infrared colours.

Transmission and Reflectance

Why do different compounds display different infrared colours?

Certain molecular bonds absorb specific infrared light wavelengths.

In transmission, you pass the infrared light through a sample and measure the amount of light the sample absorbs.

Alternatively, using reflectance, you reflect light from the sample and determine the absorbent properties from how much light reflects back.

Certain foodstuffs and types of agricultural produce are more suited to one process than the other.

For example, transmission works well with cheese, but reflectance is the preferred method for milk powder.

What are the various things NIR can measure?

Measuring Different Compounds

NIR will measure those organic compounds critical to foodstuffs, including: protein, starch, fat, sugar and many more.

These are measurable at very low concentrations, as little as 0.1%.

Because minerals and most inorganic compounds do not absorb infrared light, NIR cannot measure them. However, some minerals are bound or linked to measurable compounds, which in turn will make them measurable.

The physical properties of foodstuffs and agricultural produce such as density or texture are also non-absorbent of infrared light and therefore not normally measurable by NIR.

Again, though, there are exceptions, where the physical properties of a sample have somehow changed how its other properties absorb infrared light.

What sort of applications does NIR have in the food and agricultural sectors?

Analysing Foodstuffs for Contamination

Even in today’s climate of strict food standards and regulation, adulteration of foodstuffs can occur.

In the 21st century, food contamination is not something of the past. Notable cases of it include: spices adulterated with potentially toxic chemicals in India; melamine contamination of milk in China; and the European horsemeat scandal.

Two of the most frequently adulterated foodstuffs are alcohol and meat.

Budget varieties of alcoholic drinks, especially those sold in India and the Far East, have been found to contain toxic alcohols, industrial dyes and chemicals.

NIR can analyse alcohol for adulteration reliably and efficiently.

Adding water to meat is a controversial practice, even where regulations allow for adding a specified proportion of water.

Chicken can have 7% water added, and ham as much as 25%, but sometimes there can be other additives, such as hydrolysed protein, polyphosphates, lactose, dextrose and salt.

Again, NIR can test effectively for adulterated meat products.

It is also a highly effective method for quality testing of produce, such as flour.

NIR can test large flour samples to determine levels of moisture, protein, water absorption, starch damage and gluten strength.

How Accurate is NIR?

All analytical methods have a certain amount of variability, and some are higher than others.

Generally, the more manual a method is, the higher its variability will be.

NIR techniques are calibrated to mimic reference methods, and consequently they will inherit the reference method’s variability.

Generally, NIR has an accuracy of 1.5 times the error of the reference method it is using.

It is also worth noting that accuracy is different from precision.

Precision can repeat the same result several times. Accuracy, on the other hand, is about achieving the correct result.

 The main factor contributing to the effectiveness of NIR as a technique for testing foodstuffs is its calibration.

Why Calibration Matters in NIR

For the NIR device to be able to read different concentrations, and to see the difference between them, you have to first train it.

If you were tasting the strength of something through its concentration, your eyes might indicate this first by showing how dark or light the specific thing was.

The eye is calibrated to predict the taste on the tongue.

Using a stroboscopic NIR instrument accurately, means first calibrating it by comparing its initial infrared data with corresponding reference analysis data.

This is a mathematical equation, upon which the NIR will base its subsequent readings for a particular material or compound.

This sounds complex, but in fact, the instrument stores this calibration in its memory.

Therefore, when analysing a previously unknown sample, the device will take its initial infrared absorption reading, use it to create the equation, then calibrate itself accordingly.

NIR Calibration Techniques

NIR instruments will have pre-calibrated factory settings, but to use them effectively it helps to have a basic understanding of calibration techniques.

As mentioned earlier, NIR is a secondary form of analysis. This means it has to have something to refer to, on which it bases its readings and results.

This is the reference method which provides measurements, and, along with sample readings from the NIR instrument, it becomes the calibration model.

It is expressed as a mathematical equation, or algorithm.

There are different calibration techniques to produce this algorithm.

Multi Linear Regression (MLR)

This is the most basic calibration technique, suitable for only the simplest of NIR applications. Its success relies on there being a very strong correlation between the parameter of what is going to be measured and a few, specific wavelengths of infrared light.

It could apply in situations where you were measuring moisture variations where all other elements remained stable.

Partial Least Squares (PLS)

This more advanced calibration technique is suitable for most situations, including times where there is a high degree of diversity and variation.

As such, it is the most common calibration technique in NIR. Using it, it is possible to develop good calibration with as few as 100 samples. It can also provide reasonable predictions for those samples that lie a little outside the range that the calibration was originally based on.

PLS will not always give an optimal performance, however, if you are seeking one calibration to provide a very wide variation of samples.

One way of overcoming this is with the local calibration technique.

Local Calibration

In local, there is no pre-defined calibration model. Instead, the technique draws on a database of samples with spectra and reference values.

Every time you analyse an unknown sample, local with evaluate its infrared signal, locate a similar sample from the database, then calculate a PLS calibration based on this similar sample.

This is a useful calibration technique for situations where there is a great deal of diversity, so long as there are enough samples in the database to match up with the field samples being taken.

It is more demanding on data and input though, requiring lots of calibration samples. It can prove challenging to document the different calibration models used for specific analysis.

Artificial Neural Networks (ANN)

ANN relies on slightly different mathematics to the other calibration techniques. It allows for one calibration model to cover very different sample types but still provide accurate results.

It can also handle variations in instrument hardware, temperature or in the reference methods themselves.

Its major drawback is that it requires a very high number of samples in the database to be effective, at least 1,000. In fact, it can require as many as 4,500 to be effective.

It also requires that you include all possible future variations in your calibration model for it to deliver results accurately.

Honigs Regression (HR)

This is a unique calibration technique from Perten Instruments, developed to combine the benefits of PLS calibration with the adaptability of Local and ANN techniques to handle large sample variations.

HR combines a PLS model with a sample database containing spectra and reference values.

This allows for adjustment of the PLS calibration prediction based on the database samples.

Honigs Regression will therefore perform well where one calibration model is required to cover different types of samples. And it means you can update it easily by adding samples to its database to cover new variations.

NIR Spectroscopy for Fast, Accurate Food Testing

Large scale analysis of foodstuffs, including grain, has been possible since the early 1970s, using NIR technology.

This form of testing has remained essential because of its proven effectiveness, and the ongoing development and refinement of NIR instruments.

Calibre Control provides cutting edge testing equipment, including advanced NIR instruments, such as the Perten IM9500 Wholegrain NIR Analyser and the Perten IM9520 Four Analyser.

For more information, please get in touch, using our online contact form. Alternatively, you can call us on +44 (0) 1925 860 401, or email info@calibrecontrol.com