Automatic Alcohol and Real Extract Determination1

Michael H. Beckett2 and Vicki L. Pavelsek,3 Adolph Coors Company, Golden, CO 80401
ABSTRACT contains two catalyst coated filaments, utilizing platinum as the
catalyst. Both the density reading and the alcohol determination
The Servo-Chem automatic beer analyzer (SCABA) provides automated were then used in Tabarie's (6) or Balling's formulas (2), which are
determinations of alcohol, specific gravity, real extract, original extract, stored in the computer for the determination of the alcohol profile.
apparent and real degree of fermentation, pH, and color. The analyzer was This data can be stored for later use or can be used immediately
evaluated to determine its reliability and precision. The data obtained were depending on the computer access desired. A complete operational
compared to that obtained through the volumetric alcohol determination,
ASBC method Beer-4A. The analyzer operates by oxidizing the alcohol to description of the instrument has been published (4).
acetic acid, generating heat, which causes a change in the resistance of the
filaments, resulting in a change in the electric current in the Wheatstone EXPERIMENTAL
Bridge. Specific gravity was determined in situ with a Mettler Paar digital
density meter. The sample data were then compared to a set of standards Equipment
previously processed through the unit. All data except alcohol, specific A SCABA unit complete with Mettler Paar DMA 40
gravity, pH, and color, were calculated by means of a series of equations densitometer, colorimeter, pH meter, and automatic sampling
using the instrument's microcomputer. After analyzing replicate beer tray, was used in this study.
samples, statistical analysis of the two methods indicated no difference in A Mettler Paar DMA55 densitometer was used for the
precision (variance). The Student's ( test for individual samples generally
indicated a difference between the mean ASBC values and the mean determination of specific gravity according to ASBC methods
SCABA values for both alcohol and real extract. Twenty samples per hour (Beer-2A) (1). Additional equipment consisted of a sonic bath
could be analyzed versus 12 distillations per shift. (20° C), used for efficiently degassing the beer, and standard
laboratory equipment.
A great deal of progress has been made in the automation of the
determination of beer alcohol, and the use of the Anton Paar Reagents
digital density meter has relieved the analyst of the laborious task The cleaning and conditioning fluid was supplied by Servo
of determining specific gravity with a pycnometer. Recent work by Chem (type GR627). Ethanol (7.5% w/w) and 1% v/v of GR627
the AB Pripps Bryggerier and Servo Chem, Stockholm, Sweden, were mixed with distilled water. The standard ethanol solution was
has resulted in the development of an instrument utilizing the prepared with reagent grade Pharmco ethyl alcohol, 200 proof.
digital density meter and common physics in conjunction with
computer compilation of data (4). A similar analytical technique Methods
was reported by Sartor and Sherry in 1982 (5). Their analyses It is imperative that all samples be completely degassed and, if
indicated that the instrumental and manual methodologies were turbid, they must be centrifuged at greater than 3,000 rpm for 5
comparable. min. The beer was carefully decanted after centrifugation and then
The Servo Chem automatic beer analyzer (SCABA) unit degassed using either the shake bottle/release cap method or the
evaluated here is capable of analyzing any type of beer at any stage ultrasound bath method. In our studies, the ultrasound probe
of the brewing process. A complete alcohol profile is provided, i.e., caused excessive alcohol loss and was not incorporated into our
alcohol w/w or w/v%, real extract, original extract, apparent methods.
extract, real degree of fermentation, apparent degree of The daily alcohol standards for calibration were prepared by
fermentation, and specific gravity. In addition, the unit provides qualitatively diluting 200 proof ethanol with water to cover the
color and pH data. Alcohol, specific gravity, pH, and color are expected range of beer alcohol levels. In our case the ranges were
measured parameters, whereas the remaining parameters of the from 2.9 to 3.7% and 6.7 to 7.3% w/w. The computer was
alcohol profile are calculated values. A calculation for the programmed with a fourth degree polynomial used to determine
determination of calorie content can also be programmed into the the linearity curve after a series of ethanol mixtures was measured
unit. Twenty samples per hour can be processed. by the unit at 0.5% intervals from 0 to 7.3%. The daily standards
were automatically linearized against this curve.
Operation of the SCABA The analyzer was conditioned for 18 min using the conditioning
The purposes of our investigation were to determine how closely solution. The alcohol standards followed the conditioning
the analyzer results compared with those of the standard ASBC solution, and 40 ml samples of degassed beer followed the
distillation method (ASBC method Beer-4A) and whether the standards. Program selection was made with the Motorola
SCABA method was as precise as the distillation method. We also computer, determining which method (ASBC or EEC) of analysis
wanted to determine the reliability of the unit. was desired. Changes can be made in the computer program to
A 40-ml degassed sample was taken from a sample vial into a accommodate special types of beer, if necessary.
peristaltic pump. The sample was then split, with the first portion Control alcohol analysis was carried out according to ASBC
going to the densitometer, colorimeter, and pH meter. The second method Beer-4A (1). Real extract determination for the control
portion was passed over a thermostatic coil and entered the was determined volumetrically (ASBC method Beer-5A).
evaporation cell where the ethanol was vaporized. The ethanol Statistical analysis was carried out using an HP85 basic statistics
vapor was then passed through a hydrophobic filter and oxidized cassette. The F ratio and Student's t test were computed manually
to acetic acid at the sensor, causing a change in the resistance (3)-
measured by the Wheatstone Bridge. The Wheatstone Bridge
RESULTS AND DISCUSSION
'Presented at the 51st Annual Meeting, Milwaukee, WI, June 1985.
2 The analytical program was broken down into three phases. The
Process Engineering and Development.
'Product Quality Control. first phase was used to compare similarities of the two methods.
Numerous data were used to evaluate the correlation coefficient
between the two methods and the differences in the average results.
©1986 American Society of Brewing Chemists, Inc. After program changes, the second phase of testing was used to
78
 ASBC Journal 79
TABLE I
Comparison of Automatic Analyzer and Distillation Method Analyses for Light Beer, Coors Beer, and Selected Specialty Beers (Phase II)*
Light Beer" Coors Beer' Specialty Beers"
Difference Differences Differences
Analysis and Sample of Standard Correlation Sample of Standard Correlation Sample of Standard Correlation
Method Average Averages Deviation Coefficient Average Averages Deviation Coefficient Average Averages Deviation Coefficient
Before blending
Alcohol
Analyzer 6.69 -0.06 0.046 0.9873 5.73 0.00 0.225 0.9964
Distillation 6.75 0.052 5.73 0.225
Real extract
Analyzer 3.24 -0.01 0.084 0.9565 5.37 0.00 0.504 0.9980
Distillation 3.25 0.085 5.37 0.398
After blending
Alcohol
Analyzer 3.30 0.01 0.090 0.9699 3.73 0.02 0.265 0.9969 4.17 0.03 0.440 0.9984
Distillation 3.29 0.084 3.71 0.246 4.14 0.428
Real Extract
Analyzer 1.55 -0.01 0.187 0.9920 3.61 -0.05 0.345 0.9966 4.24 0.02 0.500 0.9994
Distillation 1.56 0.181 3.56 0.316 4.22 0.498
"Between sample data.
""Between blending n — 12, after blending n = 22.
0
Before blending n = 18, after blending n = 37.
"After blending n = 17.

determine any differences between the means for alcohol and real TABLE II
extracting utilizing the paired two-tailed t test. For the third phase, Comparison of Automatic Analyzer and Distillation Method Analyses
overall precision was determined. Statistical analysis determined of All Beer Before and After Blending (Phase II)"
the average, standard deviation, correlation coefficient, and Analysis and Sample Difference Standard Correlation
difference between methods. The paired F ratio and paired Method Average of Averages Deviation Coefficient
Student's t tests were used to determine if there were significant Before blending
differences between the variances and means, respectively. All Alcohol
differences were tested at the 95% confidence level. Analyzer 6.11 0.509
Distillation 6.09 0.02 0.525 0.9990
Phase I Analysis Real extract
Two beer types, Coors Light and Coors Premium, were analyzed Analyzer 4.53 0.03 1.120
Distillation 4.SO l.llf 0.9998
during the first phase of testing. These two beer types were further
split into their respective alcohol categories before blending and After blending
after blending. In all, 104 individual samples were completely Alcohol
analyzed by both the control ASBC procedures and the SCABA Analyzer 3.70 0.01 0.414
Distillation 3.69 0.397 0.9985
analyzer using the SCABA program 01 (primarily ASBC
formulations). Real extract
Analyzer 3.17 0.03 1.099
The specific gravity for each method was determined by the two Distillation 3.14 1.086 0.9998
digital density meters, the DMA 40 and DMA 55. The specific
gravity data from each instrument, when subjected to the paired "Between sample data. Before blending n = 29, after blending n = 79.
t test, indicated no variability in the instruments. If specific gravity
differences were found between the instruments, it was attributable hundredths of one percent because of the consistently higher
to poor sample preparation, i.e., failure to centrifuge all yeast out reading obtained with the SCABA unit in phase I. Phase I testing
of a fermentor or aging sample, or failure to degas the sample was repeated using 109 routine individual alcohol samples. Table I
properly. This was not a frequent occurrence, but sample shows the statistics for Coors Light, Coors Premium, and a select
preparation was observed closely. group of specialty beers. The results between sample analyses on
The phase I statistical analyses were then grouped into two specialty beers showed, in general, a statistical difference in the
areas, before blending and after blending beers. The data means between methods for individual samples. Table II is a
accumulated for each set of analyses, on a w/w% basis, included summary of all beer types before and after blending, from the
analyzer alcohol, distillation alcohol, analyzer real extract, and second phase of testing. In general, the change in the computer
distillation real extract. The statistical values, average, standard program in phase II led to a lesser difference between the means for
deviation, difference between means, and correlation coefficient the two beer groups for between-sample testing, but when the
were also calculated. individual alcohol and real extract data were subjected to the
All data from this phase of testing were subjected to the paired paired Student's / test, a significant difference was detected in the
( test. In general, for the various groups analyzed, the data means for most of the groups. Table III shows a summary of the
indicated a significant difference in mean values for each method. paired t test. Coors Light after blending and Coors Premium
The analyzer unit gave alcohol readings for the various beer before blending were the only alcohol analyses that showed no
categories from 0.02 to 0.05% w/w alcohol higher than the ASBC significant difference between the means of the two methods. In all
method. Real extract values were 0.01 to 0.02% higher for the cases, however, lowering the SCABA program by 0.03% did
analyzer unit. decrease the difference in the means between the methods.
Phase II Phase III
In phase II of this study, the SCABA computer program 01 was This data from phase II led to the decision to run a true test for
adjusted by lowering the overall alcohol reading by three- comparison of precision. The questions asked were, Do the
80 Vol. 44 No. 2
TABLE III Instrument Care, Durability, and Service
Summary of Paired t Test (a = 0.05) Automatic Analyzer vs. Distillation The care of the analyzer is quite easy. Cleaning with a solution of
Calculated t Value" 7.5% w/w ethanol and 1% v/v of GR627 (Servo Chem) is done
Degrees of Critical before and after each day's analyses and takes approximately 30
Beer Type Freedom t Value Alcohol Real Extract
min. Wine or wine-like solutions tend to harm the sensor because
Before blending of their high tannin concentrations and must not be sampled. A
Coors 18 2.101 1.099* 2.00* sensor is available for wine analyses. Samples containing free
Coors Light 12 2.179 13.95 1.47* ammonia also harm the sensor. The unit should not get wet, as may
All beers 29 2.045 4.21 6.52
Specialty 17 2.110 5.54 6.74 occur in an open production area, and should be used by a
minimum number of operators. Training is simple; it takes less
After blending than four hours to thoroughly train an operator, which includes
Coors 37 2.030 3.960 3.212
Coors Light 22 2.074 0.698* 1.893* hands-on training. The SCABA unit was operated here daily for
All beers 79 1.990 8.670 9.150 six months with no downtime.
"* — No statistical difference in means between methods. Summary
The phase I analysis was utilized to check the closeness of the
results from the two methods for a wide variety of alcohol levels.
TABLE iv The SCABA continually produced results 0.02-0.05% higher than
Variability Analysis for Automatic Analysis vs. Distillation" the ASBC distillation method. Phase II lowered the alcohol in the
Alcohol Real Extract SCABA alcohol program by 0.03% to align it more closely with the
Analysis distillation results. Paired t tests indicated that there was a
Analyzer Distillation Analyzer Distillation
difference in the means of the two methods.
X 3.114 3.115- 1.419 1.405 The third phase of testing was used to answer the questions, Do
SD(n~ I) 0.0063 0.0065 0.0047 0.0190 the two methods have different variability? and, Does one method
0.000040 0.000042 0.000023 0.00036 vary less than the other? The two-tailed paired F ratio replicate
0.321 < 0.952 ^3. 12 0.0638 < 2.96 sample test showed that the alcohol methods do not vary whereas
F ratio two-tailed one-tailed the real extract methods do. The one-tailed F ratio test showed
'Determined on 14 replicate analyses, n - 1. that the SCABA real extract had better precision than the
distillation real extract.
The routine check service showed that the SCABA was reliable.
methods of analysis differ in variability? and, Is the variability of The unit has replaced full-time use of the refractive index
SCABA less than that of distillation? A two-tailed paired F ratio methodology in the Quality Control laboratory. SCABA units will
analysis was run on 14 replicate samples for each method. One beer soon be used in production to replace the labor intensive and
type was degassed and split into 28 aliquots for this analysis. uncertain refractive index method. The SCABA unit is capable of
Table IV shows a summary of the F ratio analysis on 14 replicate processing more than 100 samples per 8-hr shift and provides a
samples. The F ratio was between the 95% confidence limits for complete beer distillation profile. A program is also available for
both the percentage alcohol and real extract analyses. There was basic wort gravity analysis. The SCABA unit, if installed in a
no difference in precision between the two methods of analysis for protected production environment, can precisely calculate the
alcohol determination. For 13 degrees of freedom (n — 1), alcohol profile and also provide phase change information for
F= 3.12, while l/F= 0.321. As the calculated caustic or acid phase changes.
F calc = var'ance method A, ACKNOWLEDGMENTS
variance method B
then, A coordinated effort by the Process Engineering and Development,
Fcalc. = 0.000040/0.000042 = 0.952. Research and Development, Quality Assurance, and Quality Control
Departments was necessary to rapidly obtain the large amount of data
required for this study.
The calculated Fcalc. is not larger than the Fand is not smaller
than \j F, therefore, the two methods do not differ in precision. LITERATURE CITED
For the real extract, the F ratio showed that there was a
significant difference in precision between the methods. The 1. American Society of Brewing Chemists. Methods of Analysis, 7th ed.
F ratio for the real extract analysis indicated that the variability Beer-2A, 4A, and 5A. The Society: St. Paul, MN.
was less for the analyzer than for the distillation real extract 2. Balling, C. J. N. Die Attenduationslehre: Die Bierbrauerei I. Prague,
determination. This is shown by a one-tailed F ratio analysis. A 1865.
paired t test also showed that there was no difference in the mean 3. Natrella, M. G. Experimental Statistics. Hdbk. 91. National Bureau of
values for the alcohol methods, whereas there was a difference for Standards: Washington, DC, 1966.
4. Pripps Korduner, H., and Westilius, R. Eur. Brew. Conv. Proc. Congr.
the real extract determination. 18th, Copenhagen, 1981, pp. 615-622.
A continual check service is conducted weekly on individual 5. Sartor, D. J., and Sherry, N. B. Tech. Q. Master Brew. Assoc. Am.
samples in the Quality Control laboratory. The SCABA alcohol 20(4):160, 1983.
and real extract averages are compared to the values obtained by 6. Tabarie, Z. Anal. Chem. Vol. 50, 1912.
the distillation method. To date, the difference in the averages has
ranged from 0.00 to 0.04%, with an average difference of 0.01%. [Received June 24, 1985. Accepted January 18, 1986.]