Introduction
NIR spectroscopy has been a reliable technique for distillery quality assurance at a major distiller for over twenty years, used as both a research tool and quality assurance tool. As a research tool, NIR spectroscopy can be used to evaluate raw materials, yeast, enzymes, nutritional supplements, and production parameters to optimize conditions for the production plant. As a quality assurance tool, it can be utilized to monitor and maintain control of processes. The major advantage of NIR spectroscopy is rapid analysis that requires no sample preparation or destruction of the sample. Traditional analysis of parameters of interest in distillation monitoring often take hours or even days to implement. Real-time feedback in areas of production like incoming grain monitoring, fermentation profiles, distillate analysis, dry house operations, and finished product analysis provide benefits which cannot be understated, maximizing resources and energy which subsequently leads to financial savings. The summary provided here documents the process and benefits of using NIR spectroscopy as an analytical tool throughout the entire distillation process, from incoming grain analysis to the final product being bottled. NIR spectroscopy can also be used as a screening tool to detect the present of adulterants in the final product of distilled alcoholic beverages.
Analytes
- Starch
- Oil
- Protein
- Moisture
- Ethanol
- Dextrins
- Dextrose
- Maltose
- Lactic Acid
- Glycerol
- Percent Solids in Solubles
- Fat
- Fiber
- Ash
- Methanol
- Adulteration
Summary of Published Papers, Articles, and Reference Materials
A detailed and thorough book chapter is discussed documenting the use of NIR spectroscopy at a large distillery for monitoring all points of the distillation process, including raw grain analysis, fermentation monitoring, dryhouse operations, and ethanol determination and blend analysis. Parameters of interest are thoroughly detailed as well as the benefits of using NIR spectroscopy as a real-time analytical tool in the distillery. One study used NIR spectra of stock solutions of ethanol and methanol to create calibration models for determining these parameters in commercial alcoholic beverages. Results were excellent and a similar study was conducted using stock solutions of glucose, sucrose, and fructose for fruit juice analysis. Adulteration is a major problem in the food and beverage industries and NIR spectroscopy was examined as a method for classifying different distilled beverages as well as determining the presence of water, ethanol, or methanol adulterant in them. Samples were classified at a 100% success rate and the results proved the feasibility using NIR spectroscopy as a screening tool to determine the presence of an adulterant in alcoholic beverages, allowing the classification of adulterated samples for further analysis using a traditional method like chromatography to quantify the amount of adulterant present.
Scientific References and Statistics
Livermore, Wang, and Jackson – The Alcohol Textbook, 4th Edition, Nottingham University Press, 2003, Chapter 12 – Understanding Near Infrared Spectroscopy and its Applications in the Distillery, pp. 146-170
Incoming Grain
Although it can vary by the type of whiskey, a good estimate of the mash bill for corn is around 90%. While it is the biggest expense for a distillery, corn goes largely unmonitored. For incoming grain, moisture, protein, oil, and starch are of critical importance and can all be measured by NIR spectroscopy. High moisture can cause mechanical problems with a hammer mill and thus downtime in the distillery. Grain with moisture over 15% can cause this problem as well as reduce the starch available for fermentation. Protein affects the composition of Dried Distillers Grains (DDGS), a valuable by-product at the end of distillation that is often sold as livestock feed. High oil content exists in some genetic varieties of corn and this type of corn is considered disadvantageous for distillation. Starch is the most important component for overall yield because it is the source of fermentable sugar for yeast growth and subsequent alcohol production. Besides the obvious advantages of no sample preparation and rapid analysis of these constituents in incoming grain, there are others as well. Corn from an incoming truck can be analyzed at different points to ensure that the seller did not attempt fraud by placing high quality corn at the top of the load and lower quality corn at the bottom. Other advantages include tracking production efficiencies, accumulation of crop data, and variety selection for desired characteristics. Shown below are the typical range of values and expected RMSEP (Root Mean Standard Error of Prediction) for the incoming corn calibration models used at the distillery:
| Starch | Typical Range of Values = 60-62% | RMSEP = 1.2% |
| Oil | Typical Range of Values = 2-3% | RMSEP= 0.5% |
| Protein | Typical Range of Values = 7-8% | RMSEP= 0.5% |
| Moisture | Typical Range of Values = 12-15% | RMSEP= 0.3% |
These models have been tested using validation sets and the error of prediction in the original data set was comparable to the error in the validation set, proving the accuracy and feasibility of the models for measuring these parameters.
Fermentation Monitoring
Fermentation monitoring is extremely complex because many factors affect the alcohol content and overall yield. Optimal efficiency can be achieved, but in practice this is difficult to do. HPLC is used by most distilleries for fermentation monitoring of sugars, acids, glycerol, and alcohol. All these parameters are interrelated and HPLC is a difficult and expensive method for monitoring. It is especially expensive to implement HPLC in a process setting. It requires highly skilled technicians, expensive accessories and volatile chemicals, sample preparation which often requires centrifuging or filtering, and a minimum feedback time of twenty minutes. The results are often not as accurate as desired for accurate monitoring. NIR monitoring can identify non-optimal conditions in fermentation almost immediately, allowing for real-time adjustments and process optimization. The benefits and potential improvements in fermentation efficiency are immense. For example, if a 200,000 L fermenter completes the fermentation process when the alcohol content reaches 9.6%, the plant will produce 19,200 liters of absolute alcohol per fermenter. If real-time feedback can provide protocols to adjust factors such as changing enzymes, process parameters, and nutritional supplements and increase alcohol yield by 1%, each fermenter will now produce 21,200 liters of absolute alcohol. The optimization produces the same amount of absolute alcohol in nine hundred five fermenters instead of one thousand, resulting in savings in raw materials, fuel, steam, labor, maintenance, and equipment. This can potentially save a large distillery millions of dollars per year. Even a 0.1% increase in alcohol yield during large-scale fermentation will result in substantial savings in resources and money for a distillery.
Ethanol content in corn mash can be correlated to NIR spectra using two different reference methods: HPLC and distillation – DMA. The distillation – DMA method is to distill 100 mL of corn mash, collect 100 mL of the distillate, and then determine percent alcohol via DMA. Models have shown that distillation – DMA provides more accurate results than HPLC as a reference method for ethanol. In contrast, HPLC is the reference method of choice for sugar analysis. During fermentation, sugar values should show a consistent decrease and if they do not, the sugar is not being properly converted to alcohol and adjustments must be made to correct the problem. Lactic acid is another important parameter to measure during fermentation. It is produced by bacteria that compete with yeast for sugar. If a high value of lactic acid is determined, bacteria are being produced and the distiller must take steps to correct the problem. Potential solutions include adjusting backset stillage rates and using antibacterial products as well as washing out if the bacteria concentration is high enough. Shown below are RMSEP values from calibration models used at the distillery for fermentation monitoring:
| Ethanol | RMSEP = 0.14% |
| Dextrins | RMSEP = 0.50% |
| Dextrose | RMSEP = 0.46% |
| Maltose | RMSEP = 0.52% |
| Lactic Acid | RMSEP = 0.11% |
| Glycerol | RMSEP = 0.07% |
Using these calibrations is an extremely valuable tool in fermentation monitoring in many ways. Fermenter troubleshooting can involve checking records for correct yeast or enzyme addition, auditing for mechanical failures, leaks in cooling coils, and problems in washer rotators. If the alcohol yield is reduced and identified in the early stages, problems can be corrected before the final yield is compromised. Fermentation efficiency and the impact of nutritional supplements such as nitrogen can be researched as well using NIR spectroscopy.
Dryhouse Operations
There are several applications for using NIR spectroscopy as an analytical tool in the dryhouse. Determining percent solids in solubles in the condensed syrup discharged from evaporators is important because a solids level above 40% can cause fouling and eventual plugging of the discharge lines. NIR can accurately measure the moisture percent of the condensed syrup and subsequently the percent solids. This method is used at the distillery with good accuracy.
| Percent Solids in Solubles | RMSEP = 0.8% |
The most important application for NIR during dryhouse operations is analysis of DDGS. Most distilleries provide guaranteed specifications for the constituents of interest when selling DDGS: minimum levels of protein and fat and maximum levels of moisture, fiber, and ash. Protein is especially important as DDGS is often sold as a high protein livestock feed that increases efficiency and lowers the risk of subacute acidosis in beef cattle. The current reference methods for measuring these constituents are time-consuming and sometimes expensive as well as requiring the use of hazardous chemicals: Kjeldahl for protein, Bidwell-Sterling for moisture, extraction units for fat and fiber, and oven burning for ash. Shown below are RMSEP for these parameters in the calibration models used at the distillery:
| Protein | RMSEP = 0.46% |
| Moisture | RMSEP = 0.35% |
| Fat | Fat RMSEP = 0.35% |
| Fiber | RMSEP = 0.49% |
| Ash | RMSEP = 0.13% |
There are other potential applications in the dryhouse that could benefit the distillation and fermentation process. One example is monitoring lactic acid in the backset stillage that is recycled to the fermenters. This would help control the amount of backset stillage in the mash bill and thus help optimize alcohol production. Residual starch in DDGS indicates unfermented sugar during the fermentation process. Monitoring this in the dryhouse would indicate that there are problems with alcohol yield during the fermentation process.
Blending Opportunities and Ethanol Determination
There are strict protocols for the strength of alcohol in whiskeys and liquors. Most whiskeys have a target specification of 40% alcohol by volume. The allowed deviation from this is dependent on government regulations but it generally varies from +/- 0.15% to 0.2%. One issue with alcohol blending is that suspended solids can obscure the final percent alcohol when measured by a DMA. The sample must either be distilled or oven dried to determine the solids by weight to compensate for solids in the calculation. The longer the blending product stays in the tank, the more money it costs the company. Percent ethanol in liquors has been calibrated to NIR spectra using DMA as the reference method with good results. Temperature effects on this model have been studied as well and samples collected at different temperatures proved that the model could measure accurately even if there is a failure in the temperature control system. Titratable acidity has also been measured in high proof alcohol with good accuracy.
| Ethanol | R² = 0.999 | RMSEP = 0.038% |
| Titratable Acidity | R² = 0.9952 | RMSEP = 0.106% |
The RMSEP for ethanol is small enough to make a measurement within the tolerance error for government alcohol regulations. At the time of publication, it was unclear whether or not the distillery discussed in this book chapter summary has actually implemented in-line measurements to determine the proof of spirit. Possible advantages of using in-line alcohol measurement for spirit proof include determining alcohol strength in the dilution process immediately before bottling, requiring less tank space and reducing use of resources and costs.
Partial Least Squares-Near Infrared Spectroscopic Determination of Ethanol in Distilled Alcoholic Beverages – Debebe, Temesgen, Redi-Abshiro, Chandravanshi, Chemical Society of Ethiopia 2017, 32(2), 201-209
NIR spectroscopy was examined as a method for determining ethanol and methanol in stock solutions as well as examination using the stock solution model to determine ethanol content in alcoholic beverages. NIR spectroscopy is a proven method for measuring ethanol in distilled beverages as well as determining the presence of methanol. However, the authors’ literature survey determined that there was no known study for quantifying both ethanol and methanol simultaneously using regression models. Twenty-four samples were prepared in distilled water ranging from 2% to 15% ethanol and 0.1% to 1.0% methanol. NIR spectra were collected from 1660 nm to 1720 nm at 2 nm intervals. Reference values were determined using GC and Partial Least Squares (PLS) regression models were created for ethanol and methanol. Six commercial alcoholic beverages were procured for ethanol model validation. The NIR spectra of the six commercial beverages were used with the ethanol PLS model for predictions and these predictions were compared with the reference GC analysis performed on the six samples. GC was also performed to determine the presence of methanol in the six commercial samples.
Stock Solutions:
| Ethanol | R² = 0.999 | RMSEP= 0.06% w/w |
| Methanol | R² = 0.929 | RMSEP= 0.08% w/w |
Commercial Sample Ethanol Predictions (v/v):
| Sample 1 | PLS-NIR= 48.90 | GC= 47.40 | Reported= 48.7 |
| Sample 2 | PLS-NIR= 39.96 | GC= 40.89 | Reported= 39.1 |
| Sample 3 | PLS-NIR= 40.43 | GC= 41.84 | Reported= 39.4 |
| Sample 4 | PLS-NIR= 49.39 | GC= 48.29 | Reported= 48.2 |
| Sample 5 | PLS-NIR= 37.19 | GC= 38.61 | Reported= Not Reported |
| Sample 6 | PLS-NIR= 48.61 | GC= 47.97 | Reported= 48.8 |
Prediction results showed good comparison between the PLS-NIR predictions and the GC reference tests as well as the reported ethanol content on the label of the commercial samples. Further validation was conducted by spiking three of the commercial samples with additional ethanol and the results were comparable with the predictions shown above. No methanol was determined to be present in any of the samples, both by the PLS-NIR and GC methods. The results here show promise for using calibration models created from stock solutions to universally measure ethanol and methanol in distilled alcoholic beverages. A similar study was conducted using calibration models made from NIR spectra of stock solutions of glucose, sucrose and fructose to measure sugar in fruit juices with comparable results.
https://www.ajol.info/index.php/bcse/article/view/163100
Fruit Juice Study:
Rapid Analysis of Sugars in Fruit Juices by FT-NIR Spectroscopy – Rodriguez-Saona, Fry, McLaughlin, Calvey, Carbohydrate Research 336 (2001) 63-74
https://www.sciencedirect.com/science/article/abs/pii/S0008621501002440
Classification of Distilled Alcoholic Beverages and Verification of Adulteration by Near Infrared Spectroscopy – Pontes, Santos, Araujo, et al. Food Research International 39 (2006) 182-189
NIR spectroscopy was examined as a method for classifying alcoholic beverage samples (whiskey, brandy, rum, and vodka) as well as verification of adulteration in the samples. Sixty-nine total samples were used for the study. NIR spectra of the pure samples was collected first. Various samples were then adulterated with 5% and 10% v/v of water, ethanol, or methanol and scanned as well. Spectra were collected from 1100 nm to 2500 nm using and sixteen scans were collected and averaged for each spectrum. Principle Component Analysis (PCA) and Soft Independent Modeling of Class Analogies (SIMCA) classification algorithms were used to determine pattern recognition and to characterize each group. The classification models were able to successfully classify at a 100% rate, both for determination of the type of pure sample as well as the presence of an adulterant in any given sample. The results here can be used as a screening tool to determine the presence of an adulterant in alcoholic beverages and choosing samples which show adulteration for quantitative analysis using a traditional reference method like chromatography.
https://www.sciencedirect.com/science/article/pii/S0963996905001638
Conclusion
NIR spectroscopy shows great potential as an analytical tool in every stage of the distillation process. Parameters of interest such as moisture, ethanol, sugars, starch, protein, oil, fat, ash, and fiber are all proven constituents that can be monitored quickly, easily, and reliably using NIR spectroscopy. Other potential applications include simple ethanol analysis from stock solution models, testing for methanol contamination, and discovering the presence of adulterants in alcoholic beverages. NIR spectroscopy can be used for real-time feedback allowing optimization and troubleshooting of the entire process and will replace older, slower, and more expensive techniques. Hardware, software, and calibration model advancements in NIR spectrometers are continuously evolving, as are new research projects and applications. This evolution makes the idea of using NIR spectrometers more appealing as not only analytical and research tools for the distilling industry, but for many other types of food and beverage industries as well.