Ruth H. Krywicki, PhD, MS. Minitab Insights Conference ASQ-CSSBB. Philadelphia, PA September 12-13, 2016

Transcription

1 Ruth H. Krywicki, PhD, MS ASQ-CSSBB Minitab Insights Conference Philadelphia, PA September 2-3, 206

2 Brewing beer it s more than just water, barley, hops and yeast! Brewery lab operations and challenges Where the journey started How an internal standard works FMEA and Gage R&R Check Beers, Check Standards and Control Charts Just look at the picture! Lab Results How lab results and process capability can help the Brewer to optimize his process 2

3 Historical records indicate beer has been brewed for over 7,000 years; found in hieroglyphics and stelae on pyramid walls Many styles of beer are brewed around the world by small craft breweries up to the mega-size breweries Big beer is big business.. Wall Street Journal July 2, 206/B. AB InBev s $08 billion takeover of SABMiller gets U.S. approval A large scale brewery operates 24/7 and expects to brew and package millions of cases of beer that are consistently high in quality with predictable flavor. Brewery may have multiple wort streams and many brands which adds complexity Voice Of the Customer VOC drives the business! Customer expects great taste EVERY time 3

4 4

5 Large scale brewery Quality Lab may process as many as 35,000 tests in a year operates 24/7 lean staffing Test results provide critical information to the Brewer Quality, flavor, process characteristics, etc. and help ensure yeast health and ultimately great beer every time Sophisticated instrumentation used for monitoring PM schedules and recalibrations challenge to figure out root cause for out of spec results How can test results be assured? Is the problem because the process is out of spec? Is it analyst error? Is it the instrument itself? 5

6 Recalibrated if a problem was suspected. Didn t always fix issue. HPLC-High Performance Liquid Chromatograph Anton-Paar Alcolyzer Atomic Absorption Spectrometer Head-Space Gas Chromatograph Instruments are used around the clock by several different technicians. On a PM schedule, but instrument can have problems before next service is due. UV/Vis Spectrometer 6

7 Initial use of Minitab started by doing an FMEA(Failure Modes Effects Analysis) A Quality Gate in the process also was the process bottle-neck the Vicinal Di-Ketone (VDK) check Brewer could not move the fermenter to the fast cool step [to halt fermentation] until the VDK was at an acceptable level; lab would check several times before the final result was below limit The Brewer wanted to be able to project when the fermenter would be ready based on a predictable drop in VDK Results from the lab were all over the place instead of predictably dropping with time, rechecks might be too high or too low. Brewer wanted to save time time is $$$$ 7

8 VDK is an acronym for Vicinal Di Ketone better known as 2,3 butanedione or diacetyl It occurs in fermentation along with another vicinal di-ketone known as 2,3 pentanedione At low levels, diacetyl contributes a slipperiness to the feel of the beer in the mouth. As diacetyl levels increase, it imparts a buttery or butterscotch flavor In some styles of beer (e.g. in most beers produced in the British Isles, such as India Pale Ale), the amount of diacetyl is part of the brand design. Consumer expects a certain flavor profile VDK increases during fermentation. Once fermentation is finished, the yeast will re-absorb the diacetyl back into the cell but not if yeast is stressed 8

9 A sample from the fermenter is spun down, and the top phase [supernatant] is carefully removed; a portion is incubated with sulfuric acid to force the conversion of the VDK precursors to diacetyl and pentanedione. Many steps, different people/ test The compounds of interest are volatile and when heated, the space above the liquid becomes saturated The headspace (gas phase) is then tested If we add a known amount of something similarly volatile, [hexanedione], we can quantify how much VDK is present Hexanedione is called an internal standard 9

10 0 µl 0 µl 0 µl W3 W W W W2 W3 Standards for 3 point curve 3 different concentrations of analyte of interest (W, W2, W3) Add same amount (0µl) of internal standard hexanedione to each vial Measure; generate 3-point curve [straight line] 0 µl Add 0µl internal standard to the sample; measure peak height Interpolate concentration Sample 0

11 FMEA Failure Modes Effects Analysis evaluates every step in a process (usually complex) and assesses the frequency, severity, and ability to detect the failure value from -0 is assigned Each value is multiplied to yield RPN [Risk Priority Number]; higher the RPN, the greater the impact to the process Team determined there were 34 potential failures Top hitters included operator not flushing sample tap long enough, and anything that affected the delivery of 0 µl of the hexanedione internal standard. (Six main causes that impacted consistent delivery of 0 µl of internal standard.) Minitab Gage R&R of the instrument along with control charting the internal standard helped further pin point opportunities

12 0 µl of internal standard should yield the same value every time One analyst in particular seemed to have a problem with sample range and the mean was higher as well. 2

13 Height Pentane Height Butane Individual Value Plot of Height Butane The 3 standards (W, W2, W3) were then injected into the GC as unknowns Peak heights were graphed by 0000 Minitab for individual value plots of the runs 5000 W W2 Standard W3 Individual Value Plot of Height Pentane The GC can be reasonably eliminated as a source of the variation by looking at the individual peak heights value plots i.e. recovery off the column of the standards made 0000 for the 3-point curve. (One bad prep outlier.) 0 W W2 Standard W3 3

14 Moving Range Individual Value Height Hexane Individual Value Plot of Height Hexane Look at the individual value plot of the hexanedione internal standard peak heights 3000 Notice how they are all over the 2000 place? Variation within each 000 W W2 Standard W3 standard should be consistent I-MR Chart of Height Hexane When looking at a control chart of the UC L=5967 data you can see that there is quite a bit 3500 _ X=3408 of data spread 2000 LC L=0849 This implies that the introduction of the internal standard is where much of the test error is incorporated for the W, W2, & W3 standards as well as the process samples A pr -A pr 3-A pr 3-A pr 4-A pr 4-A pr 4-A pr 4-A pr 5-A pr 5-A pr 5-A pr 6-A pr Date 5-A pr 6-A pr Date 7-A pr 7-A pr 8-A pr 8-A pr 9-A pr 9-A pr 0-A pr 0-A pr -A pr -A pr UC L=344 MR=962 4 LC L=0

15 Moving Range Individual Value I-MR Chart of IS Ht Counts // Downward shift caused by degradation 3/7/200 3/3/200 3/8/200 3/25/200 3/30/200 4/4/200 Sample Pull T ime 4/9/200 4/3/200 4/8/200 UC L=2292 _ X=0494 LC L=8695 UC L=220 Prior to using Minitab, the lab would have recalibrated the instrument and assumed calibration was the problem MR=676 LC L=0 Recalibration may take 8-0 hours 3//200 3/7/200 3/3/200 3/8/200 3/25/200 3/30/200 4/4/200 Sample Pull T ime 4/9/200 4/3/200 4/8/200 Control limits established after instrument was recalibrated. Data points from each test were added and trends observed to figure out where the problem was Graph of individual values of internal standard height counts shows three distinct populations of data The first shift was caused by using an old vial of Hexanedione that deteriorated [over-punctured]. It is likely the second shift was caused by the same thing. Note the dip. Loss of internal standard affects the interpolation of results; false high VDK results for samples tested 5

16 Moving Range Individual Value Internal Standard Height Outliers removed from natural control limit calculations UCL= _ X= LCL= Data points omitted from the calculations:, 43, 64, 66 UCL=925 MR=589 LCL=0 Typical limit ranges before investigation were ~0,000; Reduced to <,000 Further investigation showed that there was poor control over the number of internal standard vials being used. Since the internal standard affected ALL results, this was critical. New program put in place along with a better size syringe for delivery of the internal standard. Control charting showed big improvement. Spurious outliers investigated quickly. 6

17 FMEA is used for complicated or multi-step processes where it s not easy to figure out what is contributing to the problem Gage R&R used if good indication that problem appears to be related to people/instrument interaction Lab did Gage R&R [w/o FMEA] on an instrument used to measure an analyte from a new process introduced to the brewery Instrument already in use to measure something else Master Brewer felt certain the problem was instrument related not process related No historical performance for the new process; no analyte trends 7

18 Gage study involved 3 operators [lab techs], 7 different sample source-types, n=3 replicates Could clearly see that variation was in source samples not the lab techs or the instrument Further investigation showed that the Brewer could adjust the process to reduce the variability in the analyte concentration Later introduced a control chart to monitor analyte trends 8

19 Lab implemented a Check Standard program for Atomic Absorption Spectrometer [ AA ] AA measures metals several of which are critical components of yeast vitality Poor yeast health creates off-flavors in beer that are not part of brand design; also results in slower fermentation Check standard is something purchased that has a Certificate of Analysis Control-charting a Check Standard ensures the AA is operating correctly but more importantly reveals trends that might not otherwise be recognized when burner head is dirty, or lamp is going out 9

20 Individual Value Moving Range Purchased Metals Check Standard UCL=0.25 _ X= LCL= /02/2 0:05 04/8/ 2 6:55 05/06/2 09:48 05/22/2 0:46 06/05/2 9:27 06/24/2 07:27 07/6/2 02:09 08/09/2 05:26 09/2/2 :36 0/03/2 0:9 0/29/2 7:04 Sample Pull Time UCL= MR=3.3 LCL=0 04/02/2 0:05 04/8/ 2 6:55 05/06/2 09:48 05/22/2 0:46 06/05/2 9:27 06/24/2 07:27 07/6/2 02:09 08/09/2 05:26 09/2/2 :36 0/03/2 0:9 0/29/2 7:04 Sample Pull Time Supplier guaranteed 00 +/- 0 units of metal. Prior to using Minitab to control chart instrument performance, the lab assumed that everything worked ok if a check of the standard fell somewhere between 90 and 0 units. Example above shows multiple data points at lower control limit. Investigation found the lamp was going bad. A new one had to be ordered took some time. Result kept source lamps in inventory and reacted to shifts without having to wait for critical supplies to be delivered. 20

21 Check Beer program implemented to observe performance of alcolyzers Check beer is a sample from production same matrix, same prep conditions as routine samples. Alcolyzer [Anton Paar] measures alcohol, real-extract, calories, and specific gravity of beer and ensures packaged product meets label claim requirements Used to ensure in-line analysis for making up tanks was correct OEM indicates instrument repeatability is +/-0.02 % w/w alcohol ( ROH ) Lab was responsible for 4 instruments 2 in the Quality Lab, and 2 in the Brewing Lab (remote location) As long as all four instruments operated with min-max range of 0.03 from each other on the same check beer, they were considered to be operating correctly 2

22 Before Control-Chart Program After Implementing Control Charts Max difference does not address individual performance for bias or drift. One instrument may be off consistently by 0.03 Limits set +/- 2 sigma for each instrument; Lab react to shifts. Instrument consistently performs +/- 0.0 w/w% alcohol Check Beer program developed: analysis done daily to ensure instrument performance 22

23 Moving Range Individual Value I-MR Chart of Rslt (Numeric) UCL= _ X= LCL= Observation UCL= MR= LCL= Observation Qualification of a new Paar unit showed an interesting saw-tooth pattern. Before control-charting, this would not have been recognized Note that the control chart violated test 4 4 points in a row alternating up and down not likely to be analyst related Investigation showed that the pressure/flow settings were not optimized on installation 23

24 Observed lab trends can be used to create control limits for process Happy Yeast = Great Tasting Beer Fermentation step produces lots more yeast as well as beer Some yeast is wasted; while a portion is recovered in a yeast brink and reused in a subsequent fermentation Each recovery is called a generation If yeast is too old or another factor impacts process, off-flavors and/or negative quality attributes can be created Fermenter Yeast Brink Fermenter Monitoring a few key analytes helps Brewer determine yeast health Waste Yeast 24

25 Moving Range Individual Value 20 5 Brand B Analyte Results for EOF UCL= _ X= LCL=2.52 0/02/0 04:29 0/5/0 23:07 0/3/0 6:0 02/25/0 5:48 03/08/0 5:29 03/5/0 6:39 03/30/0 20:57 04/0/0 7:2 04/9/0 :3 04/29/0 6:5 05/08/0 03: Sample Pull Time 2 8 UCL= MR= LCL=0 0/02/0 04:29 0/5/0 23:07 0/3/0 6:0 02/25/0 5:48 03/08/0 5:29 03/5/0 6:39 03/30/0 20:57 04/0/0 7:2 04/9/0 :3 04/29/0 6:5 05/08/0 03: Sample pull time Sample Pull Time Analyte is a lagging indicator of yeast health Investigation revealed two outlier fermenters did not get enough yeast added Yeast from these fermenters not re-used Capability study shows process is centered well below the upper specification limit Generation mix and yeast health co-factors are right Brewer acts on control chart not spec! Happy yeast!!!! 25

26 Prior to using Minitab in the Quality Lab, technicians relied on re-calibration of equipment to solve problems which was not always effective The journey for continuous improvement in the lab started with using an FMEA and Gage R&R help pinpoint source of variability in test results Control charting key test results provides assurance to the Brewer that lab process is in control and that data are believable. No more gut feel that instruments seem to be ok Trends for key analytes in process helps Brewer assess yeast health before problems start assurance of a predictable process that produces beer consistent with brand design 26

27 The Practical Brewer Master Brewers Association of the Americas 999. (J. T. McCabe Editor) The Brewers Handbook The Complete Guide to Brewing Beer; 999. Ted Goldammer. ISBN: How to Brew Everything You Need to Know to Brew Beer Right the First Time; John J. Palmer. ISBN-3: Diacetyl in Fermented Foods and Beverages; Takashi Inoue. ISBN: ASBC Beer-25. Diacetyl Gas Chromatographic Method: (method to analyze VDK) ASQ Six Sigma Tools: Minitab: Assistant; Help; StatGuide (available with user license) The photographs and clip art in this presentation are from publicly available sources. The views expressed herein are strictly those of the presenter and not MillerCoors. The identification of specific process analytes or brands have been intentionally removed. 27