Distillery processes are ideal situations for NIR analysis since most compounds in a distillery are organic and at sufficiently high concentration. From the beginning, when grain is unloaded from the truck, to the end where product goes into a bottle, NIR analysis can provide useful information to maintain process control.
INCOMING GRAIN
Unlike fuel ethanol production where corn is 100% of the mash bill, in a Canadian whisky distillery corn is approximately 90% of the mash bill. It is the largest expense to a distillery, yet it goes largely unmonitored throughout the distilling industry. The components of importance to distillers are moisture, protein, oil, and starch, all of which can be measured by NIR spectroscopy.
Moisture content of incoming grain is of critical importance. If moisture becomes too high it can cause mechanical problems with a hammer mill and cause downtime in the distillery. Typically, a maximum specification of 15% moisture should be targeted. If moisture levels are above 15%, there is also less starch available to make alcohol. It is up to the distiller to either make concessions with the corn broker or to reject the load of grain. This type of monitoring can justify the expense of a NIR spectrometer.
Protein content of whole kernel corn is another parameter that can be analyzed, because it affects composition of DDGS, a valuable by-product at the end of distillation. For example, in 1998 a
distillery was having problems meeting the guarantee for protein content in DDGS. At first, it was thought there were issues with fermentations not finishing properly; however on investigation it was found that fermentors were finishing with a good yield. After further investigation of the NIR crude protein data, it was found that the average protein level in the incoming corn fell from 7.5–8.0% in 1997 to 7.0% in 1998 due to drought conditions. It was found that the expected crude protein content in DDGS could be roughly determined by multiplying the protein content of incoming corn by 3. A difference of 0.5–1.0% in protein content of incoming grain will therefore affect the protein content of DDGS by 1.5–3%. There is nothing a distillery can do to change the incoming crude protein levels of poor quality corn. The only option is to lower the crude protein guarantee of DDGS.
Oil content can also be monitored by NIR spectroscopy. The reason that it may be important to measure oil is that genetic varieties of corn with high oil contents exist in the commercial market. Measuring incoming loads for oil allows the distiller to prevent high oil corn varieties from entering the system.
Starch is the component with the greatest effect on overall yield because it is the source of fermentable sugar for yeast growth and ultimately alcohol production. NIR spectroscopy can allow a distiller to track the starch levels in purchased corn.
There are several advantages to using NIR spectroscopy as a primary tool for determining the quality of corn:
1. The first, and most obvious is the rapid analysis of essential components. It can take up to three days to determine the levels of starch, protein, moisture and oil in one corn sample by traditional analysis methods. 2. The distiller can track corn suppliers more
closely to prevent becoming the victim of fraud. The way samples are collected is important, because it can affect measurement results. The suggested method is to gather a composite sample from a truck with a probe that can measure the bottom, middle, and top of the load. Also, it is important to measure the front, middle, and back end of the truck.
A sample should always be collected by this consistent and statistically reliable method. Sampling with this method can also detect loads containing grain that is out of specification on the bottom that have been ‘topped up’ with good quality grain. 3. Production efficiencies can be tracked more
closely. Accurate determination of production efficiency is important, because it helps the distiller make improvements and reduce costs.
4. More crop data accumulated each year. 5. Variety selection for desired characteristics
is possible.
The calibration model Hiram Walker & Sons currently uses to monitor incoming corn was provided by Bruker Optics with the purchase of the NIR spectrometer. The expected value of each of the four major components and the predictive error of each component expressed in RMSEP are listed in Table 1.
Table 1. Expected values and predicted errors for starch, oil, protein and oil content of incoming corn expressed in RMSEP.
Component Expected result Approximate RMSEP (% as is basis) (+/- %)
Starch 60–62 1.2
Oil 2–3 0.5
Protein 7-8 0.5
Moisture 12-15 0.3
A major advantage of using an FT-NIR system is that it has high wavelength accuracy and precision, which allows calibration models to be easily transferred from instrument to instrument via the internet, without having to adjust for slope or bias on each instrument. As a precaution, the downloaded models are validated with a set of 40 corn samples that are not from the original calibration set. The predicted error of the 40 validation samples measured by the instrument in the Hiram Walker & Sons plant was roughly the same as the error in the original model, which met requirements.
Another advantage of this calibration model is that reliable results can be obtained by
measuring incoming corn directly without any sample preparation such as grinding. Since the absorption coefficient in the NIR region is low, NIR light can penetrate corn kernels and whole kernels can therefore be used for model development. The NIR spectrum can be collected with the aid of an appropriate sampling device such as a rotating cup or a flow-through funnel. Depending on the available sampling devices and how a particular calibration model was developed, some instruments today still require users to grind incoming grain in order to achieve a reliable measurement.
Calibration models for other grains including rye, rye malt, barley, and barley malt are currently being developed. The components of interest include starch, oil, protein, moisture, and enzymatic activity.
FERMENTATION MONITORING
Fermentation optimization can be complex, because many parameters can affect the final alcohol content and overall yield. With the traditional analyses available today it can be time consuming for laboratory staff to keep up with the demand for fermentation analysis. Optimum efficiency can be achieved if all parameters are kept in control prior to and at fermentation, but this does not always happen as equipment failures can occur hours before being identified and corrected. NIR technology can identify non- optimum conditions in the fermentation almost immediately, allowing correction of problems. In addition, NIR analysis can aid distillers in evaluating the new enzyme and yeast nutritional supplements.
In the industry today, most distilleries use an HPLC for fermentation monitoring, including measurements of sugars, acids, glycerol, and alcohol. All these components are interrelated and it is very important to keep track of them at each stage of fermentation, but there are disadvantages to using HPLC as the primary monitoring tool:
1. The technician must be highly skilled. 2. Accessories can be quite expensive (i.e.
columns, chemicals, filters).
160 D. Livermore, Q. Wang and R.S. Jackson correct answers is at least 20 minutes. This does not include repeat samples, sample preparation time, technician training time, instrument setup time, wash outs, and calibration time. This can be frustrating to fermentor operators when quick decisions are required.
4. The sample preparation for HPLC only permits analysis of those components that are soluble in the liquid fraction. To extract the components of interest from the mash, it must be centrifuged or filtered to retain the liquid portion. This will compound the error because the solid portion that is removed has sugars, acids, and alcohols still bound to the solid particles, thus skewing results.
5. Results are not always as accurate as required.
The amount of ethanol produced in fermentation is very important, and the potential improvements in fermentation efficiency that NIR technology can support should not be underestimated. For example, if a fermentor is 200,000 liters and the average fermentation finishes when alcohol content reaches 9.6%, this means that the plant will produce 19,200 liters of absolute alcohol per fermentor. After researching different protocols to optimize
fermentations such as changing enzymes, process parameters, or nutritional supplements, the alcohol content was raised to an average of 10.6% per fermentor. This means that 21,200 liters of absolute alcohol are now made in each fermentor. If 1000 fermentors are set per year in a plant to produce a certain amount of alcohol, then one percentage point increase in fermentor alcohol content means that plant can now set only 905 instead of 1000 fermentors to produce the same volume of alcohol. This means considerable savings in raw materials, processing fuel, steam, labor, maintenance, and equipment. Also, plants strive to be energy efficient, and generating steam can be quite expensive. Overall, this can mean millions of dollars saved annually for a distillery. Figure 15 shows the results of a number of fermentation optimizations monitored by NIR spectroscopy at Hiram Walker & Sons distillery over a four- year period. NIR has proven invaluable in assisting in fermentor optimization and in saving energy costs.
It is clear from the example above that raising the percent alcohol in a fermentor by even 0.1% can increase efficiency and contribute to cost savings for a distillery. There are two reference methods that can be used to calibrate the NIR model, HPLC or distillation - DMA. The NIR model should be based on the most accurate
Hours of fermentation Ethanol (% abv) 0 15 30 45 60 75 2003 90 105 120 0 1 2 3 4 5 6 7 8 9 10 11 12 1999
Figure 15. Fermentation profiles from 1999 to 2003 during which period NIR has assisted in optimizing fermentors and thus
reference data available in order to determine when an enzyme, nutritional supplement or other change in fermentation conditions is effective.
Figures 16a and b show the two calibration models for determining ethanol in corn mash. The first NIR model was built using HPLC as the reference method (Figure 16a) and the second was built using the distillation – DMA reference method (Figure 16b). The distillation – DMA method is to distill 100 mL of the corn mash, collect 100 mL of the distillate and then determine the percent alcohol via the DMA. The
value of RMSECV for the NIR models using the HPLC and distillation-DMA method is +/- 0.67% and 0.18%, respectively. This demonstrates the greater accuracy of the distillation-DMA method for the calibration of NIR models for corn mash. It is recommended that the distillation-DMA procedure be used for ethanol calibrations in corn mash instead of the HPLC method.
In contrast to alcohol, for sugar calibration models HPLC is recommended as the reference method. Even though the prediction error for sugar is roughly +/- 0.5%, it is still acceptable. The only concern to a distiller is that sugar is
Rank: 5 R2 = 96.2 RMSECV = 0.674 0 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 1 2 3 4 5 6 7 8 9 10 11 12 13 14
Reference method value
NIR predicted value
Rank: 6 R2 = 99.73 RMSECV = 0.176 0 1 2 3 4 5 6 7 8 9 10 11 12 0 1 2 3 4 5 6 7 8 9 10 11 12
Reference method value
NIR predicted value
Figure 16a. Cross validation result for ethanol content in 30 fermented corn mash samples. The reference method of analysis was
HPLC. The values of R2 and RMSECV were 0.962 and 0.67% v/v, respectively.
Figure 16b. Represents the cross validation result for ethanol content in 30 fermented corn mash samples. The reference method
162 D. Livermore, Q. Wang and R.S. Jackson being converted and is used up by the yeast. If sugar values are not decreasing then the fermentation parameters must be quickly investigated and proper actions taken to correct the problem. Figure 17 shows the carbohydrate concentration over a 140 hr period, monitored using NIR spectroscopy.
Another important parameter in corn fermentation that NIR spectroscopy can monitor is the production of lactic acid. The primary method of calibrating the NIR for lactic acid is HPLC. Lactic acid is a compound produced by
Lactobacillus sp., bacteria which compete with
yeast for sugar. The more lactic acid produced, the more bacterial contamination is present. Figure 18 illustrates the amount of lactic acid produced over time in a typical batch fermentation process. It was noted that as the backset stillage use rate increased in the fermentor, so did the starting level of lactic acid. By monitoring lactic acid, a distiller can therefore adjust backset stillage rates in order to optimize fermentations. For continuous fermentations, the lactic acid values can be used to determine when antibacterial products are required or wash out procedures should commence.
A potential application for NIR technology is optimization of fermentor cleaning. One can
optimize the amount of wash water and detergent necessary for clean fermentations by correlating the wash out procedure with the levels of lactic acid produced at the end of fermentation, thus determining which wash out procedure is most cost effective.
NIR spectroscopy is a flexible analytical tool because its results are based on the best primary methods available, distillation – DMA for ethanol, and HPLC for sugars and lactic acid. The predictive errors of the calibration models currently used to monitor corn fermentations at the Hiram Walker & Sons plant are shown in Table 2.
Table 2. RMESCV values for fermentation parameters.
Component Approximate RMESCV
Ethanol 0.14 Dextrins 0.50 Dextrose 0.46 Maltose 0.52 Lactic acid 0.11 Glycerol 0.07
Troubleshooting is one of the largest paybacks for an NIR spectrometer. Over the course of one
Figure 17. Typical reduction in carbohydrates in corn fermentations over a 140 hr period.
0.0 0 10 20 30 40 50 60 70 80 90 100 110 120 130 140 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 Hours of fermentation Dextrose Maltose Dextrins Carbohydrate (%)
year, one lab technician can measure 5600 corn mash samples with just one NIR spectrometer. There are 30 batch fermentors at the Hiram Walker & Sons facility, and all are kept full of mash until the mash is sent to the beer well. Today Hiram Walker & Sons laboratory monitors the fermentors at 12–40 hr intervals and can predict the final percent alcohol. If ethanol values are lower than expected, a course of action can be determined. Troubleshooting of fermentors involves checking records for correct enzyme or yeast addition, or auditing the plant for mechanical failures. An example of mechanical failure is an increase in lactic acid values and a decrease in ethanol in one fermentor, in which case the Butterworth washer must be closely inspected to determine if it is rotating properly. As another example, if the alcohol content is too low, check for small leaks in the cooling coils. NIR technology can assist in identifying and fixing problems in the early stage before yield is compromised.
The use of NIR spectroscopy for fermentation analysis allows the distillery to rapidly perform research on fermentation supplements. One example is experimenting with various forms of nitrogen. Nitrogen is the limiting substrate for yeast growth in fermentations. Figure 19 shows NIR results from nitrogen supplement experiments. It was determined that nitrogen at a specific level decreased fermentation time (increased rate of fermentation) by 30 hrs. A
Figure 18. Production of lactic acid during the course of corn fermentations.
0.40 0.50 0.60 0.70 0.80 0.90 1.00 1.10 0 10 20 30 40 50 60 70 80 90 100 110 120 130 140 Hours of fermentation Lactic acid (%)
fermentor with a nitrogen supplement added finished in 50 hrs while the control fermentors finished in 80 hrs. A downside of the nitrogen supplement, however, was that if the control fermentors were given enough time, they would finish with more alcohol than the nitrogen-fed fermentors. This is because the nitrogen induced yeast cell growth from the carbohydrate source instead of alcohol production. In the end, it was not economical for this distillery to use nitrogen as a supplement, since 80 hrs of fermentation was acceptable. However, if production capacity ever needed to be increased, or if the gravity of fermentation was increased, then a nitrogen supplement could be an asset to this operation. The NIR analysis allowed the distillery get analytical results quickly, utilize personnel to their maximum potential, and optimized production.
NIR is an asset in determining the efficiency of fermentations. For example, incoming corn was monitored by NIR spectroscopy and found to contain 62.6% starch (as is). This means that there were 626 kg of starch per metric tonne of corn. The 626 kg of starch should produce 696 kg of dextrose if it were completely converted. The theoretical ethanol yield from yeast is 51.1% ethanol; therefore, the theoretical yield from 1000 kg of this grain is 451 liters. If scales are in place to measure the amount of corn that enters the fermentor, then the efficiency of fermentations can be calculated. If a 200,000
164 D. Livermore, Q. Wang and R.S. Jackson
liter fermentor contains 50 metric tonnes of corn and finishes at 10.0% alcohol, then final yield is 400 liters of absolute alcohol per tonne of grain. But theoretical yield is 451; therefore, this fermentation was 89% efficient.
The efficiency of the distillation column can also be determined. If the above fermentor contained 20,000 liters of ethanol, and the tank that recovered the ethanol contained 20,500 liters at 96.0% ethanol, then the distillation efficiency was 98.4%.
Calibration models for rye mash, barley mash, and molasses have also been developed at Hiram Walker & Sons distillery. Optimization studies, efficiency determinations, and troubleshooting currently are carried out for these types of fermentations. A future calibration model also possible using NIR technology is determination of the percent solids in mash prior to exiting the pre-liquefaction tank or prior to fermentation. This can help quickly identify if there are milling problems or belt scale problems.
DRYHOUSE OPERATIONS
There are several applications for NIR analysis in the dryhouse. The first application is the determination of the percent solids in solubles,
or the condensed syrup discharged from the evaporators. After the third effect of a falling film evaporator, the desired level of solids is around 40%. If the solids level gets much higher, it can cause fouling and eventual plugging of the discharge lines. It is very expensive to clean pipes plugged with solid material; therefore it is highly desirable to have a quick and easy method to determine the percent solids. NIR has the ability to measure the percent moisture of the syrup in a Pyrex beaker at very warm temperatures with the integrating sphere accessory. After the percent moisture is determined, simply subtract this value from 100 to get the percent solids. At Hiram Walker & Sons, the RMSECV for this calibration model is +/- 0.8%.
The most important application for NIR analysis in the dryhouse area is DDGS analysis. After the rotary driers, most distilleries will guarantee to their customers minimum levels for protein and fat; while there are maximum levels for moisture, fiber, and ash. Protein levels for DDGS are probably the most important guarantee for the end customer. The most accurate wet chemical method for protein is the Kjeldahl method. Figure 20 illustrates the cross validation for a calibration model using the Kjeldahl method as the reference method.
Moisture analysis is also important for the
0 1 2 3 4 5 6 7 8 9 10 11 0 10 20 30 40 50 60 70 80 Hours of fermentation Ethanol (%) Nitrogen Supplement Control
quality of DDGS. To get optimum results, the most accurate wet chemical procedure for moisture analysis is the Bidwell-Sterling method, as compared to the oven moisture method. The