Optimization the Process Indicator Right First Time Under a Brewery Context

Optimization the process indicator Right First Time under a brewery context

Marta Correia Jamal Pinto Gonc¸alves [email protected]

Instituto Superior Tecnico,´ Lisboa, Portugal November 2019

Abstract With the intent of optimizing the process indicator Right First Time (RFT) under a brewery context, the current project was developed in Super Bock group’s headquarters in Leça do Balio. The RFT indicator reflects the capacity to achieve results of the intended quality in the first attempt throughout the brewing process. Thus, the indicator promotes continuous improvement by facilitating the identification of variations in the process, which have an impact on the quality parameters. The project focuses on the improvement of the following three quality parameters– the original extract, the oxygen dissolved within the filtered beer and the bitterness of the beer during maturation. The cases were selected due to their mostly negative values on the RFT indicator. Work was conducted according to the tools and principles of continuous improvement, more specifically, the DMAIC method from Six Sigma, the 5 Whys analysis, the cause and effect diagram, and the Pareto diagram. This way, it was possible to define the primary problems in each of the parameters and formulate an hypothesis for the existing variations and defects, and, finally, to propose measures to help improve the efficiency of the brewing process. Keywords: Beer Production, Right First Time Indicator, Continuous Improvement, Quality

1. Introduction present work was developed in the production department of Super Bock company. Nowadays, it is thought that a sustainable company is the one that produces less waste, with fewer defects, consuming less resources and pre- 2. Literature Review senting shorter production times and faster pro- 2.1. Beer Production cesses. Consequently, it is suggested, the company is more efficient and competitive in the mar- Brewing is a biotechnological and enzymatic pro- ket [1]. cess [2]. Four raw materials are required for beer The process indicator Right First Time is used to production: barley, hops, water and yeast. The validate the beer quality and the efficiency of the quality of these raw materials has a decisive influ- process. It reflects the capacity to achieve results ence over the quality of beer. of the intended quality in the first attempt through- The most important process within the beer pro- out the brewing process. duction is the fermentation of sugars into alco- Thus, the RFT promotes the continuous improve- hol and carbon dioxide by the yeast. To provide ment of the process by identifying the variation and conditions for the fermentation, it is necessary to problems, enabling changes within it. The goal is produce the wort. To transform wort into beer, always to improve the results of the quality param- the sugars in the wort must be fermented by the eters. With this, the more defects pointed out by yeast. Along with this bioreaction, by-products are the indicator, the greater will be the need for re- formed, which have an effect on the taste, aroma work and therefore more waste will be produced. and other characteristic properties of beer [3]. First of all, it was necessary to identify the most The fermentation of beer is performed in cylindro- prevalent characteristics of the indicator and, con- conical vessels (CCVs). The reactions occurring sequently, in the quality of the product. These were during this process are those occurring in the main identified and prioritized. Besides that, through fermentation and those in maturation. After fer- methodologies and tools of continuous improve- mentation, maturation and lagering, the beer is ment, it was possible to particularize the problems filtered to be stabilized both biologically and col- in order to identify the causes for defects. The loidally.

1 2.2. Quality and Continuous Improvement cause and effect diagram. Quality is described as the added value delivered 3.1. Case Study I- Original extract at filtered beer to the client, and is generally regarded as excel- The first case study is based on the elimina- lence. Due to the competitiveness of the market, tion/reduction of out-of-specification original ex- companies work to achieve the quality expected tract results in the first attempt at filtered beer. by the consumer. To achieve the expected quality, The quantity of dissolved sugars present in the the process is constantly improved [4]. wort before fermentation is known as original ex- Continuous improvement is defined as an inte- tract. It is expressed in Plato Units (◦P)– quantity grated tool to affect the process or service with of sugars (g) per 100g of wort. the aim of improving the quality and productivity. It is also described as a culture, where everyone 3.1.1. Define Phase contributes to produce improvements without the The first step was to define and specify the prob- need of major investments [5]. lem. To accomplish this, the influence of products on the RFT for this characteristic was investigated; 2.3. DMAIC Methodology the products of interest were selected; the influ- The DMAIC method (Define–Measure–Analyse– ence of the filtration line and the stage of the fil- Improvement–Control) is an approach that seeks tration cycle on the out-of-specification results was to reduce variability along the process with the pur- determined. The main problem was identified as pose of minimizing the waste and achieving the ex- filtrations of product X and Y in the filtration line 2 pected quality. and prevalently at cycle ends. For that, first of all, it is necessary to define the 3.1.2. Characterisation of the current situation problem (define), than to quantify the current sta- Once the problem has been defined, the mode of tus of the project (measure), analyse the data and operation of the filtration line in question has been formulate hypotheses for the causes of the problem identified taking into account the different stages (analyse), implement solutions to minimize or elim- of the filtration cycle and also the understanding inate the causes effects (improvement), and check of the online dillution and carbonation process. In if the desired results were achieved through the addition, the measuring instruments in the line and changes made (control) [6]. their mode of operation were identified. 2.4. Right First Time Indicator 3.1.3. Measure Phase RFT is broadly used in several industries to vali- Data collection was performed and the current date the quality and efficiency of the process [7]. deviations presented by the online original wort The indicator reflects the capacity to achieve the measuring instruments (DSRn 427S and mPDS desired quality results in the first attempt. The max- 2000V3 Evaluation Unit) were quantified in rela- imum process efficiency is reached when the indi- tion to the measuring of the measuring apparatus cator is at 100 %. According to this, it is possible present in the laboratory (Alcolyzer Beer). For this to identify the problems and variations during the purpose, each sample is defined as the beer taken process easily. It has the benefit of reducing costs online each time during the filtration of product X due to reduced product reprocessing. and Y. Sampling, the process of collecting samples 3. Materials and Methods at the terminal of the filtration line using a Scan- The present study followed an approach focused dibrew tap for that purpose. Table shows the num- on identifying the quality parameters which most ber of samples that have been collected for both negatively influence the process indicator. There- products (X and Y) and their concentration range. fore, first, they were identified and then selected as Table 1: Set of samples collected at filter line 2 for product X case studies. and Y. Since, in the company, the RFT is determined by Product Concentration Number of samples the average of RFT in the fabrication (of the wort) X 10.97– 11.31◦P 26 and RFT in the filtration, the cases were selected Y 12.36–16.95◦P 11 separately, i.e., considering the characteristics that affected fabrication and filtration. At the company, the indicator regarding with fabrication is calcu- 3.1.4. Analyse Phase lated by the results in maturation. The selection Considering the characterization of the current sit- was made by the Pareto diagram. uation carried out and by the analyse made re- Then each case study was handled independently garding to the measure phase it was possible to by the DMAIC method. Also, potential causes were pointed out some of the main causes. The identi- pointed out using the 5 Why’s analysis and the fied causes have been brought together in a cause

2 and effect diagram. At this point, since common Table 2: Set of samples collected. causes were found for other products, the prob- Type of tank Number of samples lem was extended to products other X and Y. The 3000 hL 3 causes were discussed for further analysis. 3000E hL 4 1000E hL 17 3.1.5. Improving phase Finally, solutions were prioritized taking into ac- 3.2.3. Analyse Phase count the causes identified. Not all the proposed The data collected was analysed and root-causes solutions were carried out during the internship pe- were identified by the use of 5 Why’s tool. riod but are thought to contribute to improvement of the current situation. 3.2.4. Improving Phase The possible causes were identified and prioritized 3.2. Case Study II- dissolved oxygen at filtered beer for future improvement tests. Then three different tests were performed to reduce or eliminate three 3.2.1. Define Phase of the causes pointed out by 5 Why’s too.

First, the influence of filtration lines on out-of- 3.2.5. Control-Validation specification results was investigated. The topic The success of the tests carried out was discussed was then discussed with the technicians in order to taking into account the effectiveness of the problem identify the causes. Two main causes were pointed resolution and the associated costs. (a) filtration starts in filtration line 1 and (b) the final beer stored in the tank (CCVs) during unloading to 3.3. Case Study III– bitterness of maturing beer the filter. From the same data the two causes were The bitter character of the beer is acquired by the confirmed and the problem has been prioritized – addition of hops during the boiling process. The in- high dissolved oxygen due to final beer in the tank. soluble α–acids are converted during wort boiling into soluble iso–α–acids (isomerisation). From the 3.2.2. Measure Phase end of boiling to the end of fermentation (maturation) the theoretical bitterness loss is around 25- To define the increment of oxygen during the dis- 30% [3]. The bitterness is expressed in units of charge of beer from the tank to the filter some vari- bitterness corresponding to mg iso–α–acids/ L of ables were chosen and data was collected. As beer. when the beer ends (in the tank) the piping between the tank and the filtration matrix (valve block 3.3.1. Define Phase that directs the beer to the centrifuge), is still full of Firstly, the products with the greatest impact on the beer and a final push with water or beer is provided indicator were determined on the basis of the indi- to further process the beer present on the line. cator results for the characteristic of interest. Prod- The variables were oxygen during the discharge uct X has been selected for furthers studies. The of beer from the tank; volume of beer in the incomplete isomerisation and the loss bitterness tank; oxygen measured at the final push, type of units during fermentation by foaming and by the tank (fermentation – 3000 hL– or storage– 1000E adherence of bitter substances to yeast cell walls and 3000E hL). The type of tank was selected as influences the bitterness result after the fermenta- variable since the tanks 1000E and 3000E hL do tion. not have any counter-pressure of carbon dioxide Since an anti-foam agent is used in the brewery, it during the process of filtration, what is thought was decided to focus the study on the loss of bit- to be an effective factor for the incorporation of terness by foaming. oxygen. The measurement of dissolved oxygen was carried out by the device – Haffmans-CO2/O2 3.3.2. Measure and Analyse Phase GEHALTEMETER,c-DGM (measuring range 0- From the data provided by the company, the aver- 2000 ppb; accuracy ± 1 ppb). age and standard deviation of the bitterness values A sample is defined as the set of fractions collected for periods with and without the addition of anti- during unloading and final pushing for a given tank, foaming agent were determined. From the determi- considering the variables under study. Sampling is nation made the loss of bitterness was calculated described as taking samples from the filtration ma- for both periods. The results considered for this de- trix using a Scandibrew tap. Twenty-four samples termination are based on the samples taken from referring to product X were collected and for subse- the process control (cold wort and maturated beer). quent analyses the samples were divided into the Table 3 shows the number of results considered for three groups of tanks (Table 4). both periods.

3 Table 3: Number of results considered at measure phase. 3.3.6. Problem Characterization Cold wort Matured beer From the tests performed it was possible to iden- With Anti-foam 102 167 tify a new situation so the problem in question was Without Anti-foam 43 77 narrowed down– Loss of bitterness due to foaming caused by high fermentation pressures. Subsequently, the problem was characterized by the monitoring of the foams formed in the fermen- 3.3.3. Improving Phase (Trial Phase I) tations. For the same approximate concentration In order to assess whether the increased concen- of anti-foams, the fermentations were divided into tration of anti-foaming agent allows the better use those that presented foams and those that did not. of hops, without interfering with the quality of the The average and standard deviation of bitterness product, a pilot test was carried out. and maximum pressure reached in the fermenta- Two fermentations were performed with the same tions for the two sets were determined. Some raw materials, recipe, yeast origin and approxi- of the possible causes of the problem in ques- mately the same amount of yeast. The tested con- tion have been proposed. However, there was not centration of the additive was the same in both - 5 enough time to confirm all of them. mL/hL. Firstly, to evaluate the efficiency of the wort produc- 4. Results & Discussion tion the following characteristics were evaluated: 4.1. Priorization of case studies original extract, bitterness, limit attenuation and pH First of all, in order to identify the quality parame- from the samples collected from cold wort. The ters that mostly influenced the indicator, it was nec- cold wort sample is defined as the wort that has al- essary to analyze the indicator results for weeks 3 ready been cooled by passing the heat exchanger. to 14 of 2019. Sampling is defined as the in line collection via The quality department of the company analyses a Scandibrew tap during the transfer between the the samples and then inserts the results into the cooler and the fermentation tank. As one tank re- SAP software. The RFT is determined weekly and ceives four batches of cold wort, the results of the the target is 88%. For that period, 25% had a result four parameters were calculated from the average below the target and the average was 91%. Sepa- of the samples taken. rately, for the fabrication the average was 92% and Foams produced during fermentation were eval- for filtration 91%. uated. The quality of the product (concentrated The selection of the case studies was done accord- beer) was also tested by the usual sampling car- ing to the Pareto diagram, for fabrication and filtra- ried out at the company – extract, alcohol, attenuation RFT, for that period of time. In this way, were tion, bitterness, diacetyl concentration, sulfur diox- considered for the selection the prevalence of each ide concentration, pH, cell viability. The concen- quality characteristic in both indicators. The bitter- trated beer sample is defined as the beer after mat- ness in maturing beer; the dissolved oxygen and uration at 7◦C. The sampling was made by means the original extract in filtered beer, were considered of a Scandibrew tap in the tank. The methods for investigation. of analysis were those usually used by the quality department and the analyses was performed by 4.2. Case study I– Original extract in filtered beer them. 4.2.1. Problem Identification The identification of the prevalence of the products 3.3.4. Causes Identification in the indicator taking into account the quality characteristic in question was carried out by a Pareto di- Once the fermentations were finished the results agram. The most problematic products were prod- were analysed and a possible cause for the differ- ucts X and Y. Also, the main problems were defined ences was proposed. as filtrations in filtration line 2 and at the end of filtration cycles. 3.3.5. Trial Phase II To validate that the higher concentration of the 4.2.2. Current Situation anti–foaming agent had no negative impact on the The description of the filtration line and the filtration yeast, a new test was performed. For this purpose, process was necessary to understand the current the yeast that had fermented at high pressures in situation. the previous assay was used. Two fermentations The key variable to control the blending process were performed with the same concentration of the during filtration is the original extract but other char- anti-foaming agent (4,3 mL/hL). acteristics are also evaluated such as the alcohol, The usal quality parameters after fermentation in the carbon dioxide and temperature. Different in- the company were evaluated, the same as in 3.3.3. struments are used and they are located in the fil-

4 tration line. 4.3. Case Study II- dissolved oxygen in the filtered To quantify the deviations between the online to of- beer fline measurements, data was collected for prod- 4.3.1. Problem identification ucts X and Y. The results point to a meaning- Using the data from the company it was possible ful deviation for product X compared to Y and to to identify and prioritize the main problem of the the accuracy of the evaluation unit outputs (0,10 present case study, the increment of oxygen by the %). Also, samples were collected for the purpose final beer present in the fermentation tank. of comparing the results of the measurements of product X in the two lines. Therefore as the re- 4.3.2. Current Situation sult for the filtration line 1 was 1.36 ±0.01%, it was The evolution of oxygen has been monitored dur- possible to a confirm that the inaccurate measure- ing the draining of the beer (from the tank to the ments were not the cause for the problem. filter) and during the final push. Considering the results obtained for the three followed drainages, it is possible to confirm the increase of dissolved Table 4: Set of samples collected. oxygen during the last hectoliters of beer and high- Product Deviation (%) est levels during the final push. The group of 1000E X 2.02 ±0.01 and 3000E hL showed higher increments of oxygen Y 0.77 ±0.01 compared to the 3000 hL.

4.3.3. Possible Causes 4.2.3. Possible Causes After analyzing the samples and by the daily obser- vation of the production, the causes for the problem Possible causes were identified and organized in a were identified by the 5 Whys’ analysis. cause and effect diagram. Most of the causes come from the use of com- The main causes pointed were the incorrect mea- pressed air, what causes the mix of beer with air surements at 0◦P; excess of concentrated beer (oxygen). The counter pressure of carbon dioxide added by the program during Y filtration; incorrect in the 3000 hL tanks may be insufficient for ade- maintenance and calibration of in-line instruments; quate dragging of air from the beer. The tanks that a set of product adjustment instructions is inac- receive beer (1000 and 3000 hL) may have air in- curate; temperature fluctuation at start of filtration E E side before the transfer and besides that, for these (after periods of pause, like recirculation). How- two, there is not a carbon dioxide counter pressure ever, these presented causes can be interpreted, procedure. For tanks distant from the filtration ma- as most and less relevant for the problem. trix, sometimes it’s necessary to add air pressure Moreover, it was concluded that the incorrect mea- to take-out the yeast from the tank, which may also surements at 0◦P and the excess of concentrated cause the incorporation of oxygen within the beer. beer added are the most likely causes for the de- There’s a greater change that the air will blend with fects. Besides that, the others should also con- the beer, when beer storage in the tanks is longer. tribute in part for the deviations between the final The final push allows the last beer present in the average measured online and offline. tank to be filtered, this beer is the one that was in contact with air, so it will contain oxygen. Aldo, is 4.2.4. Improvements likely to have incorporation of air within the beer Due to the causes it was possible to recommend in the line by the delay of the closing of the valve some improvement possibilities. present on the fermentation tank. As the main cause for the incorrect periods at 0◦P was identified by the rapid reaction of the pump af- 4.3.4. Solutions Prioritization ter a buffer tank, it is imperative to change the reg- Once the causes have been identified, they have ulation of that pump, which nowadays is controlled been prioritized with the aim of eliminating or re- through a level-indicator in the beer tank. Then it is ducing their effect. important to further investigate the excess of con- As already stated, the main cause is the utiliza- centrated beer added to the filter, regulated by the tion of compressed air during the process, but its program during the filtration of Y. replacement is out of the question due to the op- Also, once a year it is expected to have a preventive eration costs. It has been found that the increase maintenance of the instruments and calibration. in oxygen is mostly observed during the final push, These two operations prevent the chance of inac- but not performing it is not a viable hypothesis ei- curate measurements. Besides that, the method of ther. The choice for further investigation was to im- calibration, regarding product adjustments, was to plement some procedures to reduce or eliminate be reviewed. As well as the control of the devia- the contact between the product and air (present tions which has to be weekly. in the tank or compressed air).

5 4.3.5. Cause: Air present at 1000E hL 4.3.6. Cause: Absence of counter-pressure of After the final push, the thanks are complete dis- carbon dioxide in the 1000E hL charged and depressurized. If the CIP is valid, the Since, for these tanks, there is no counter pres- tanks are refilled when necessary or if it is not, the sure during the transfer to the filter, the tanks were CIP is executed and then the beer is transferred. In pressurized before the transfer. The results point both, there is the possibility of the presence of air to a meaningful reduction of oxygen during the dis- inside the tank. Which means that there is some charge (Figure 3) and specially in the final push air (and also carbon dioxide) after the depressur- (Figure 4). However, as expected, the results ob- ization. With that, a possibility was tested and a tained previously are more favorable. method has been proposed for dealing with this situation – pressurization, expulsion of air by the lower valve present in the tank, and further pressurization. It is thought that counter pressure allows the separation between air and beer as carbon dioxide is heavier. The results obtained for this test presented a significant reduction of the problem, both when the beer is discharged into the tank (Figure 1) and during the final push (Figure 2). Therefore, it was concluded that the expulsion of the air from the tank followed by pressurization solves the problem for this type of tanks.

Figure 3: Comparison of the average dissolved oxygen during beer discharge between the current situation (data collected at 4.3.2) and the test (counter-pressure of 0.5 bar).

Figure 1: Comparison of the average dissolved oxygen during beer discharge between the current situation (data collected at 4.3.2) and the test (elimination of air present in the tank).

Figure 4: Comparison of the average dissolved oxygen during the ﬁnal push between the current situation (data collected at 4.3.2) and the test (counter-pressure of 0.5 bar).

4.3.7. Cause: Insufﬁcient counter-pressure of CO2 in 3000E hL The hypothesis that increased carbon dioxide counter pressure could improve the results in the 3000 hL tanks was tested. To verify this, two different tests were made – Test A: from an apparent extract of 4.5◦P increase the pressure set point to 0.5 bar; Test B: from an apparent extract of 4.5◦P increase the pressure set point to 0.6 bar. Figure 2: Comparison of the average dissolved oxygen during This value of apparent extract was chosen since it the ﬁnal push between the current situation (data collected at represents an advanced stage of fermentation and 4.3.2) and the test (elimination of air present in the tank). does not interfere with it.

6 By the same analysis made previously, for this test the first method to be carried out, more manual was concluded that the results do not reduce the work is required. It should also be noted that dur- increment of oxygen (Figures 5 and 6). ing the tests were identified some tanks that lose pressure after pressurization so the method is not feasible until the situation is rectified. It has also been determined the amount of carbon dioxide necessary for the expulsion of oxygen from a bright beer tank (BBT), when it has oxygen above the desired limit. The result was 0.07 ton of CO2 to reduce the oxygen from 120 to 85 ppb from a BBT with 1450 hL of beer. It is important to note that this is an approximated value. The amount of gas needed depends on the oxygen concentration and the volume present in the BBT. It was concluded that in monetary terms the proposed methods, while helping to reduce and eliminate the problem, were not viable due to their cost. Figure 5: Comparison of the average dissolved oxygen during Giving the number of existing tanks and the per- beer discharge between the current situation (data collected at centage of off-specification results. Nevertheless, 4.3.2) and test A (counter-pressure of 0.5 bar) and B (counter- pressure of 0.6 bar). the loss of quality associated with the oxygen re- moval must be considered.

4.4. Case Study III- Bitterness in maturing beer 4.4.1. Problem Identification The results of RFT were used to check which products had relevance in the RFT indicator, taking into account the bitterness. With this, the selected product was X. The bitterness in maturing beer depends on the performance of isomerisation and the loss of bitterness in fermentation by foaming and absorption of bitter substances by the yeast. Since the yield of isomerisation is already controlled by the pH, temperature and time of boiling and, besides that, the absorption of bitter substances is a very difficult process to control, the problem to Figure 6: Comparison of the average dissolved oxygen during be considered was defined as loss of bitterness by the final push between the current situation (data collected at foaming during fermentation. 4.3.2) and test A and B. 4.4.2. Current Situation 4.3.8. Validation Firstly, the bitterness results in maturing beer were After testing the hypotheses, the next phase was compared during periods among fermentations to evaluate the applicability of the proposed meth- where anti-foam agents were and were not used. ods. Their effectiveness in eliminating or reducing Secondly, the bitterness in the cold wort was de- the problem and the associated costs were consid- termined for both periods. For periods when anti- ered. As the method of presented in section 4.3.7 foam was added the average values obtained were was not efficient, it was discarded for future analy- 37 ± 2 EBU (cold wort) and 26 ± 2 EBU (matur- sis. ing beer), without anti-foam the results were 37 ± 3 Firstly, from the ideal gas equation, the amount of EBU (cold wort) and 24 ± 2 EBU (maturing beer). gas (CO2) was calculated for each procedure. For The loss of bitterness was 11 ± 4 and 13 ± 5 EBU, the first one (section 4.3.5) the quantity of carbon respectively. In summary, despite the uncertainty dioxide obtained was 0.54 ton and for the other associated with the results in cold wort, the differ- method (section 4.3.6) 0.27 ton. Regarding the ef- ence between the two periods regarding with bit- fectiveness of the method by the analysis made in terness loss is 2 EBU. the previous section, the expulsion of air followed by pressurization, had the best results for reduc- 4.4.3. Optimization the use of hops– Phase I ing the increase of oxygen levels in the beer. For Currently, the concentration of anti-foam agent both methods, manpower is necessary since there used is the 4.3 mL/hL. The concentration rec- is not an automatic program for that. However, for ommended by the supplier of 2-10 mL/hL. The

7 increase of anti-foam allowed better use of hops, of alcohol (low attenuation). However, it was later reducing the losses during fermentation, without concluded that this was not due to the formation of interfering on the product and, therefore, this strat- foams. egy for avoiding bitterness loss can be considered Also, two hours after the filling of the tank, the cell for standard process implementation. count made showed that CCV 89 presented a num- For the test, the same receipt of beer, the same ber of cells higher than desired. raw materials, the same yeast at approximately the Since two hours after filling the tank, the yeast is same amount were considered. The concentration still adapting to the environment and there is no of anti-foam in both was 5 mL/hL. cell growth, it has been concluded that the result Table 5 shows the results obtained for cold wort may be due to excessive amount of yeast added. in the tanks tested. The results obtained for the Although the yeast of CCV 88 and 89 is the same cold wort were satisfactory with the exception of and has the same cellular viability, in the filling of bitterness (deviation from the lower specification tank 89 more yeast has been added (98 kg), al- limit of 3% (CCV 88) and 6% (CCV 89)). This though the same amount of yeast had been de- was already to be expected due to the reduced vised for the trial. accuracy of the result due to the fact that the Finally, the cell viability after the two fermentations iso–α–acids are not homogeneously dispersed in was evaluated and the results were favorable for the wort. These values are usually only indicative. both.

4.4.4. Possible Causes- Differences Table 5: Results of cold wort variables for the two tanks, 88 and During fermentation, large quantities of carbon 89. dioxide are produced and released for later recov- CCV 88 CCV 89 ery. The pressure is released by the control of a Original Extract (◦P) 17.84 17.96 valve. The desired fermentation pressure is 0.4 Bitterness (EBU) 35 34 bar. Limit Attenuation (%) 70 71 It was identiﬁed that the tank that presented foams pH 5.02 5.04 (CCV 89) fermented at high pressures in the ﬁrst two days (Figure 7). In addition, it was speculated that the formation of Table 6 shows the results achieved for the beer foams was due to high pressures during fermenta- at the end of maturation in studied tanks. tion. Table 6: Results of beer at the end of its maturation period, CCV 88 and 89. CC 88 CC 89 Original Extract(◦P) 17.85 17.82 Real Extract (◦P) 6.4 6.02 Apparent Extract (◦P) 3.68 3.23 Alcohol (% v/v) 7.83 8.06 Bitterness (EBU) 28 27 Attenuation (%) 66 68 Cell count (× 106/mL) 25 39 Figure 7: Pressure (bar) as a function of the fermentation time (h) to the tank that presented foam formation (CCV 89). Diacetyl (mg/L) 0.07 0.07

SO2 (mg/L) 7 6 4.4.5. Phase II In order to check the behavior of the yeast after fer- pH 4.26 4.24 mentation at high pressures and with a higher concentration of anti-foam, the following fermentation In the tests, it was found that with the same con- of the yeast were monitored. This time the concen- centration of anti-foam, one of the tanks presented tration of anti-foam used was the normal one (4.3 foams (CCV 89). The bitterness values were quite mL/hL). Again, one of the fermentations presented similar- 28 (tank without foams) EBU, 27 (tank with foams and the other did not. The cause was ex- foams) EBU. Both results are within the quality actly the same, fermentations at higher pressure. speciﬁcation. Other quality parameters were also The bitterness values obtained were 24 EBU (for evaluated. In the tank without foams, the fermen- the fermentation at higher pressure) and 26 EBU tation did not go as expected given the production for the other. Cellular viability of the yeast were

8 analyzed after fermentations and the yeast being 4.4.7. Causes Identification fermented consecutively at high pressure was dis- regarded for futher fermentations. Some causes have been pointed out but it was Finally, besides the contribution of higher pres- not possible to confirm all of them during the in- sures for the loss of bitterness, it is possible to ternship period. The causes identified are faulty speculate that higher pressure can affect cell prop- pressure transmitters; mechanical problems in the agation [8] [9], and prevent its re-use. tanks; pipe encrustations; suction pressure of the compressors were insufficient to recover all carbon 4.4.6. Problem Analysis dioxide; high foam tank pressure. The first three were tested and the conclusion was The loss of bitterness because of the fermenta- that they do not interfere with the observed high tions of at higher pressures was studied deeper. pressures. Therefore, the cause for the problem This was achieved by supervising the fermenta- may be the insufficient suction pressure of the tions. These were divided into 2 groups- foams (for compressors to collect all the carbon dioxide pro- that ones that produced foams) and without foams. duced in the process. For both groups the mean of bitterness was obtained. Plus, the average of maximum pressures. It is possible to observe a significant difference in 5. Conclusions the average bitterness for the cases that had veri- The work was able to point out some of the causes fied foams (Figure 8) and also a higher pressure in for the defects that negatively influence the pro- the tank (Figure 9). cess indicator Right First Time, more particularly, in relation to original extract, presence of oxygen in filtered beer and low bitterness in maturing beer. Also, considering the causes it was possible to point out some methods of improvement. In addition, with the intent of improving the project’s performance, DMAIC methodology, 5 Why’s analysis and the cause and effect diagram (continuous improvement tools) were used. These showed ad- vantages since they guided the project team and the causes were more easily identified. Future studies should consider the change of the pump controlling in the filtration line to solve the incorrect measurements by the inline instruments (0◦P). Another recommendation would be the revi- sion of the calibration methods regarding the speci- Figure 8: Average and standard deviation of bitterness results fication of the product’s group. Furthermore, taking obtained for matured beer relative to the set of fermentations into consideration the oxygen in the filtered beer, with and without foam production. the methods propose weren’t tested in the tanks of 3000E hL. As by the first analysis made for the current situation, it constitutes a type of tank with less favorable results, thus it is important to test the success of the method for these. Would be of interest for the case study III the additional study of the higher pressures during the fermentations, since it causes the presence of foams, that correspond to loss of bitterness and loss of volume. Besides that, the higher pressures could have a negative influence in the yeast and in the flavor and taste of the beer. As for future work, it will be interest- ing to study the effect of consecutive high pressure fermentations using the same yeast (re-pitching). Concerning the optimization of the use of hops, the increase of the use of anti-foam agents to improve the yield of hop utilization, with the main goal of de- Figure 9: Average and standard deviation of the maximum creasing the addition of hop in the process, tests pressure recorded during fermentation to the set with and with- should be resumed when the pressures along the out foam production. fermentations are standardized.

9 Acknowledgment I would like to thank my supervisor at Super Bock Group, Engineer Dulce Barata Feio and my supervisor at Instituto Superior Tecnico,´ Professor Mar´ılia Mateus, for the attention and availability. To the teams of collaborators, Engineer Seraﬁm Sales and Engineer Emanuel Silva, thank you for the help and demonstrated availability. I would also like to thank my colleagues, my fam- ily and my boyfriend for the support and for all the motivation. References [1] Castro, R.: Lean Six Sigma Para Qualquer Negocio´ . IST Press, 3a edic¸ao,˜ 2016.

[2] Linko, M., A. Haikara, A. Ritala e M. Penttila. Recent advances in the malting and brewing industry. Journal of Biotechnology, 65: 85–98, 1998. [3] Wolfgang,K.: Technology brewing and malting. VLB Berlin, 3a ediçao,˜ 2014. [4] Antonio,´ N., A. Teixeira e A. Rosa: Gestao˜ da Qualidade, De Deming ao modelo de ex- celenciaˆ da EFQM. Ediçoes˜ S´ılabo, 3a ediçao,˜ 2019.

[5] Singh, J. e H. Singh: Continuous improvement philosophy-literature review and direc- tions. Benchmarking: An International Jour- nal, 22 (1): 75–119, 2015. [6] Nave, D.: How to compare six sigma, lean and the theory of constraints. American Society of Quality, 73–78, 2002.

[7] Mowafy, A.: KPIs for Garment. Dispon´ıvel em linkedin.com/pulse/kpis-garment-ayman- mowafy Acedido em 20.08.2019.

[8] Renger, R., S. Van Hateren e K. Luyben: The Formation of Esters and Higher Alcohols Dur- ing Brewery Fermentation; The Effect of Car- bon Dioxide Pressure. Journal of the Institute of Brewing, 98:509–513, 1992.

[9] Bleoanca, I. e G. Bahrim: Overwiew on Brew- ing Yeast Stress Factors. Romanian Biotech- nological Letters, 18(5):8559–8572, 2013.