A Rapid Method for Testing Filtration Performance of Malt and the Optimization of the Method

Abstract

A rapid filtration test method that predicts malt filtration efficiency has been developed in this research. This rapid filtration test method has an advantage over existing beer filtration tests, as it can be easily operated and there is no need to brew. Six kinds of malts with distinct filtration performance were selected, and the filtration performance was determined by measuring the turbidity of their clarified wort through simple simulation of mashing. The results indicate the filtering performance ranking as given: Copeland > Pearl malt 1 > Pearl malt 2 > Synergy > Planet > Wheat malt. When a sample is purchased for six times of parallelization, RSD is less than 10%, and the experimental results are significantly correlated with viscosity and β-glucan content. Therefore, this method can effectively predict the filtration performance of malt. The response surface design 3 × 3 factor experiment was used to optimize the fast detection method, which increased the sample differentiation (F-value) and made it easier to judge the experimental results. Response surface experiment results show the optimal reaction conditions for enhancing differentiation, i.e., an enzyme ratio of 0.5 (α-amylase, β-glucanase, neutral protease = 1:2:1), a holding period of 38 min, and a reaction temperature of 44 °C. The F-value is 147.00.

1. Introduction

The filterability of beer has long been studied since the invention of beer filtration; studies have included the identification of beer components that reduce filterability and new methods for the prediction of filterability at a production stage [1]. Beer clarification is a physical process of removing suspended particles by means of filter mediums [2]. A rest period of several days is usually demanded to remove most of the yeast and cellular tissues [3]. Nevertheless, after this period, some colloid particles still remain and cause a significant turbidity. Insoluble absorbents are used to remove haze precursors [4]. The low filtration efficiency of beer may lead to poor abiotic stability of the body and turbidity of the finished beer after packaging, resulting in quality problems. It also reduces filtration flow, which leads to filter replacement and complete discharge of cake and filter aid, which cause production costs [5]. During the fermentation, glycogen spontaneously released was strongly dependent on the yeast strain, its physiological status, the gravity of the wort, and the fermentation temperature. Glycogen particles could be present in every beer but at different levels to impact on the filterability and clarity of beer [6]. Nevertheless, non-starch polysaccharides, for instance, the arabinoxylan and β-glucan in beer, can cause filtration problems, such as a reduced extraction rate of wort, increased viscosity of wort, reduced filtration efficiency, and turbidity during beer storage [7]. The long pectin is not fully decomposed in the brewing process, the viscosity of fermentation liquid will increase, and the filtration will be affected [8]. Protein and polyphenols play a determinant role in turbidity formation [9].

In the filtration test, the filterability of wort was predicted by the method of viscosity measurement. The results showed that the viscosity of wort had no relation to the filterability of beer, but the viscosity of beer was related to the kieselguhr filtration test [10]. The filter cake factor, with higher values denoting worse filterability, was used to evaluate the filtration properties of the diatomaceous earth (DE) precoat filter test by Raible. It could be used to modify the dosage of filter aid and was ideal for testing the interactions between DE and beer haze particles [11]. In addition to the filter cake factor, the calculation of the specific filtrate volume (Fspez) has been suggested as a measure of filtration output per hour and the filter area (m²) at a constant filter pressure [12]. This test used membrane filtration with pores of 0.2 m in size, without the use of filter aids, at a sample temperature of 0 °C and a pressure of 2 bar. In order to calculate the maximum filtered volume at infinity (Vmax or Gmax), the filtered volume was measured twice. The quantified filtering performance ranged from “very bad” to “very good” [13]. Although the literature contains a variety of descriptions, a thorough comparison of the evaluation areas with DE filtration was reported. However, this filter test is extremely sensitive to the components of beer, particularly yeast, haze particles, and β-glucan gel. The homogeneity of the small samples and the membrane material both had an impact on reproducibility. As a result, there was a lack of comparability between laboratories and significant predictability variations between 40 and 90 percent, depending on the brewery [14]. A small-scale wort rapid filtration test (SWIFT) that predicts beer filtration efficiency has been developed by Stewart. This syringe test was performed with a 13 mm diameter 0.45 μm polyamide membrane, which used a vacuum generated by withdrawing a 10 mL syringe plunger. Test time was typically between 30 s and 3 min depending on the sample material. The results correlated negatively with wort viscosity (p < 0.05) and permitted a better prediction for filter performance than total β-glucan measurements. The advantage of this test is that it can be tested directly with malt or degassed beer. However, a large-scale evaluation of the laboratory data did not take place [15].

In all of the above methods for measuring filtration efficiency, attention was mainly on the filtration efficiency of wort and beer. However, the production of wort and beer went through multiple processes, so it may be the production process that affected their filtration efficiency, but malt, as the main raw material of beer, also had its own filtration performance. Therefore, it was also important to estimate the potential filtration performance of the malt. There was a method to test filtration performance of wort in a laboratory, but it needed the help of small equipment [15]. In this study, a new rapid test method was developed to detect the filtration performance of malt. The evaluation of the filtration performance of malt was obtained by detecting the wort turbidity through the imitation of the mashing procedure to retain the colloid particles that affected the filtration efficiency in the laboratory. This new method did not need any self-made equipment, and it was convenient to operate.

2. Materials and Methods

2.1. Materials

One wheat malt (Xinlei) and two commercial mixture malt samples were obtained from Pearl River Co., Ltd. (Pearl malt 1 and Pearl malt 2, Guangzhou, China), two from Malteurop (Copeland and Planet) and one from Yuehai Yongshuntai Group Co., Ltd. (Synergy, Guangzhou, China). Malt fermentation broth and malt basic indices’ basic values have been provided (

Table 1. Malt specifications for pilot brewing.

Malt Sample	Viscosity (mPa·s)	β-Glucan (mg/L)	Arabinoxylan (mg/L)	Protein (mg/L)	Polyphenol (mg/L)
Wheat malt	1.56	87.56	419.17	578.23	135
Pearl malt 1	1.46	72.94	983.25	755.46	169
Pearl malt 2	1.48	38.3	909.92	714.06	171
Copeland	1.40	25.81	684.05	594.32	163
Planet	1.43	29.56	863.57	743.64	202
Synergy	1.45	48.45	879.34	627.48	175

and

Table 2. Indicators relating to the filtration performance of malt.

Malt Sample	β-Glucan (%)	Arabinoxylan (%)	Protein (%)
Wheat malt	0.05	0.32	1.66
Pearl malt 1	0.18	0.59	1.46
Pearl malt 2	0.16	0.45	1.39
Copeland	0.12	0.43	1.36
Planet	0.13	0.44	1.43
Synergy	0.13	0.47	1.42

). α-amylase (800FAU-F/g), β-glucanase (250FXU-S/g; 700EGU/g), and neutral proteinase (0.8 AU/mL) were provided by Novozymes Co., Ltd. (Tianjin, China). In this study, all enzymes were added to be prepared against the malt’s weight.

2.2. Determination of Filtration Behavior

The malt was crushed with a fine grind setting (0.2 mm), and the water/malt ratio was 5:1. After mixing water, malt, and enzymes well, the heat was kept at 44 °C for 30 min, and then, the enzyme was inactivated for 30 min at 80 °C. Mash was micro-filtered through a 3 μm drainage filter membrane and cooled down to room temperature. Clarification of wort was determined by the light absorption value of the supernatant, as measured by an ultraviolet spectrophotometer at a wavelength of 680 nm as the turbidity, which allows one to assess the effectiveness of malt filtration. All measurements were repeated six times and stated in average values. For the statistical analysis of all data measured in this study, SPSS (Statistics Package for the Social Science, ver. 21) and Microsoft Excel 2013 were used. Analysis of variance (ANOVA) was used to analyze the significant differences (p < 0.05) between each sample. The flow chart of the new method is as follows (Figure 1):

2.3. Experimental Design and Statistical Analysis

In this study, the F-value of each group of experiments was used as the response value for the response surface experiment to enhance sample differentiation and judgement accuracy. A full 3 × 3 factorial design with three variables and three levels were employed to optimize the method for rapid detection of the filtration properties of malt, as depicted in

Table 3. Experimental range and levels of independent variables.

Variables		Ranges and Levels
		Low (−1)	Middle (0)	High (+1)
X1	Residence time (min)	22	30	38
X2	Reaction temperature (°C)	38	44	50
X3	Enzyme ratio	0.5	1	1.5

. Fifteen experiments were carried out, and the factors affecting the optimization process were investigated effectively, which ensured the system convergence to the best advantage. A full factor design has the advantage of providing complete information about the effects of the experimental factors under study and the interactions between the factors. The following three parameters were considered for this study: reaction temperature, residence time, and enzyme ratio. The general form of the second-degree polynomial equation is as follows:

$$Y=B_0+\sum _{i-1}^k{B}_i{X}_i+\sum _{i-1}^k{B}_{ii}{X}_i^2+\sum _{\begin{array}{c}i-1\\ i<j\end{array}}^k{B}_{ij}{X}_i{X}_j$$

where Y is the response function (F-value), B₀ is the center point of the system, ε is the random error, and B_i, B_ii, and B_ij represent the coefficients of the linear, quadratic, and interactive effects, respectively. X_i, X_i2, and X_iX_j represent the linear, quadratic, and interactive effects of the independent variables, respectively. Design Expert 11.0.1.0 software was used for the regression and graphical analysis of the experimental data. The operating conditions (reaction temperature, residence time, and enzyme ratio) are rearranged in Table 3. The statistical analysis of the model was performed to evaluate the analysis of variance (ANOVA). The empirical mathematical model was tested by ANOVA, which was used to test the significance of the second-order model. The determination coefficient R² was used to evaluate the accuracy and general ability of the above polynomial model.

2.4. Optimization of Reaction Conditions and Experimental Verification

To decrease the values of indexes that were unexpected and increase the values of indexes that were predicted. The ranges of three reaction variables were optimized to balance the content of all indexes. We looked at the discrepancies between the projected value and the verification value. To verify the effectiveness of the ideal conditions, experiments in triplicate were conducted.

3. Results and Discussion

3.1. Assessment of the Rapid Detection of Malt Filtration Performance Test

The factory production conditions were used to determine the filtration performance of six malts, and Copeland, which had the best filtration performance, was chosen as the standard malt for testing the filtration performance of malt. From best to worst, they are Copeland, Pearl malt 1, Pearl malt 2, Synergy, planet, and Wheat malt. By repeating the experiment six times for each malt, the standard error was found to be less than 10%, indicating that the method was stable, and the variation between repeated experiments was minimal.

To increase the reliability, a rapid fermentation experiment (200 mL wort was fermented at 28 °C for 48 h) was adopted to measure the properties of wort and fermented wort. Six different types of malt data were analyzed using an ANOVA (p < 0.05, data not displayed), which indicates that there were significant differences between the samples. Correlation analysis (Figure 2) showed that the results of a rapid test method were correlated with wort viscosity (r > 0.95, p < 0.05 ), wort filtration time (r > 0.97, p < 0.05 ), wort β-glucan (r > 0.98, p < 0.05), beer viscosity (r > 0.98, p < 0.05), beer filtration time (r > 0.86, p < 0.05), beer β-glucan (r > 0.92, p < 0.05), and beer turbidity (r > 0.86, p < 0.05). Therefore, this method has been proven to be successful for the rapid detection of malt filtration performance.

3.2. Influence of Various Parameters on Method

3.2.1. Influence of Reaction Temperature

In this study, α-amylase, β-glucanase, and neutral protease were selected as the main enzymes used in the process of saccharification, and they are also the three of the major mashing enzymes used in industrial production, with wide access and high stability [16]. Amylose reaction with amylase primarily produced maltose, glucose, and tiny molecules called dextrin. Amylopectin was acted upon to create boundary dextrin, maltose, glucose, and isomaltose. In comparison to β-amylase, the thermal stability was stronger [17], the action time was shorter, and the operation was easier to detect quickly.

The impact of the F-value on the reaction temperature is shown in Figure 3, where a direct relationship is observed. Each group’s F-value was significantly higher than 1, demonstrating that the variance between their mean values was statistically significant. Mostly, the optimal activity of α-amylase was measured at different temperatures ranging from 30 to 90 °C. The optimum temperature range lay between 40 to 60 °C. At 50 °C, it had 100% relative activity. As the temperature rose from 60 to 80 °C, it rapidly abated the relative activity [18]. The numerical optimization study suggested that the desired maximum activity and stability of dextranase can be obtained at a temperature of 40 °C. In addition, stability tests have shown that dextranase was stable, without compromising its activity, up to 55 °C and for pH ranging from 5–8 [19]. The optimum activity temperature for neutral proteases was around 55 °C, but at 20 °C the activity dropped by half [20]. Combining the reaction temperatures of the three enzymes, for ease of operation, the reaction temperature ranging from 32–62 °C was taken as the test temperature.

The enzyme activity increased as the temperature rose. The F-value increased from 31.34 to 183.62 with the greatest difference between the groups; as the reaction temperature rose from 32 °C to 38 °C, α-amylase, β-glucanase, and neutral protease began to work. When the temperature reached 50 °C, the decomposition peaked, and the enzymes’ ideal reaction temperatures was achieved. As the variance between the sample decreased, the F-value also decreased to 153.7. The temperature increase led to a decreasing differentiation between the samples until it reached the neutral protease’s optimum reaction temperature; at this point, protease acted on both the malt substance and the other two enzymes, resulting in enzymatic hydrolysis and reaching its optimal reaction temperature.

3.2.2. Influence of Residence Time

In the contemporary brewing of beer, mashing takes about 60 to 90 min [21]. To account for the impact of variable time on enzyme activity, a reaction time range of 22 min to 70 min was chosen. The reaction time of the enzyme used in this study was limited to 90 min to simplify the detection process. The impact of the residence time on the F-value is presented in Figure 4, which shows fluctuation with increasing time. A double peak in the effect of holding time on the F-value indicated that time did not have a stable impact on the relationship between enzyme activity and F-value. The differentiation between samples increased once the holding time exceeded 46 min, but the stability was poor, and the trend changed in comparison to that of the experimental group, whose holding time was less than 46 min. It was hypothesized that the prolonged use of the enzyme group may lead to excessive decomposition of the substance in malt.

3.2.3. Influence of Enzyme Ratio

The quantity of enzymes was added according to the quantity of malt, and the total amount was controlled. The proportion of three enzymes was changed. The basic amount of each enzyme was added first, and the extra amount of each enzyme was added and reduced successively, so that each enzyme could play its full role. The impact of enzyme ratio on the F-value is shown in Figure 5, This exhibited an increasing trend followed by a decline. As the amount of neutral protease rose, it affected the effectiveness of both α-amylase and β-glucanase, resulting in reduced differentiation between the samples with a corresponding F-value of 75.52. The balance between the three enzyme ratios, as well as the balance between the decomposition of enzymes, allowed for the greatest differentiation between malts, with an F-value of 176.92. The decomposition of amylopectin was significantly accelerated, the differentiation between malts shrank, and the F-value kept falling as the proportion of α-amylase rose.

3.3. RSM Results

3.3.1. Model Fitting

ANOVA was used to statistically evaluate the adequacy of a quadratic model. According to the experimental results, a quadratic regression model was established using Expert Design software to represent the influence of various variables on F-value prediction, as shown in Equation (1):

Y = −1158.71434 + 46.57930 X₁ + 26.24509 X₂ + 59.18908 X₃ − 0.222969 X₁X₂ − 2.61813 X₁X₃ − 0.339167 X₂X₃ − 0.535629 X₁² − 0.233340 X₂² + 5.35900 X₃²

(1)

where Y is the estimate response of F-value. X₁, X₂, and X₃ are the residence time, reaction temperature, and enzyme ratio, respectively.

Table 4. ANOVA of quadratic model for F-value.

Sources	Sum of Squares	Degree of Freedom	Mean Square	F-Value	p-Value	Remarks
Model	9977.19	9	1108.58	16.19	0.0007	***
A	2074.32	1	2074.32	30.29	0.0009	***
B	499.60	1	499.60	7.29	0.0306	*
C	1110.15	1	1110.15	16.21	0.0050	**
AB	458.17	1	458.17	6.69	0.0361	*
AC	438.69	1	438.69	6.41	0.0392	*
BC	4.14	1	4.14	0.0605	0.8128
A2	4947.94	1	4947.94	72.24	0.0001	***
B2	297.11	1	297.11	4.34	0.0758
C2	7.56	1	7.56	0.1103	0.7495
Residual	479.42	7	68.49
Lack of Fit	49.03	3	16.34	0.1519	0.9232
Pure Error	430.39	4	107.60
Cor Total	10,456.60	16

R² = 0.9542; adjusted R² = 0.8952; predicted R² = 0.8607; Adeq precision = 11.5306. * (p < 0.05), ** (p < 0.01), *** (p < 0.001).

shows the results of the quadratic surface response model analysis given by ANOVA and response surface experiment result is shown in

Table 5. Response surface experiment results.

Run	X1 Residence Time (min)	X2 Reaction Temperature (°C)	X3 Enzyme Ratio	F-Value
1	−1	0	1	83.25
2	0	−1	1	129.46
3	0	0	0	125.65
4	0	1	1	110.36
5	0	0	0	127.45
6	0	1	−1	131.25
7	1	−1	0	127.35
8	−1	1	0	81.49
9	1	1	0	91.4
10	−1	−1	0	74.63
11	0	0	0	147.34
12	0	−1	−1	146.28
13	−1	0	−1	90.57
14	0	0	0	134.58
15	0	0	0	146.97
16	1	0	−1	144.61
17	1	0	1	95.4

. These studies were carried out: a p-value of less than 0.05 showed that the linear influence of the parameters X₁ (residence time), X₂ (reaction temperature), and X₃ (enzyme ratio) was statistically significant for F-value. Since the lack of fit is not significant, the model is appropriate for the response. The factors of X₁, X₂, and X₃ and the interaction terms of X₁X₂ and X₁X₃ were significant in the quadratic polynomial regression model for F-value, indicating that the response was interactive. Plotting the predicted values as a function of the observed values allowed us to express the correlation factor R² and assess the reliability of the derived polynomial. The accuracy of the model to forecast the actual experimental data was assessed using the R² coefficient of the determination. The R² will often be higher when more variables and their interactions are taken into account. In this study, the model for the response had an R² of more than 0.9, showing the model’s acceptable accuracy of predicting the data and their potential to direct test experiments. The plot in Figure 6 compares the expected response and the actual response value obtained from tests using the quadratic model.

3.3.2. Response Surface/Contour Plots

In order to identify the most influencing parameters considered in the ongoing work, an analysis of variance was performed on the results. ANOVA was used to evaluate the validity and adequacy of the model effects of the operating parameters and their interactions. The effects of three factors on response were found by RSM using three-dimensional plots and contour plots.

The effects of variables on the F-value could be observed from the response surface plots. As illustrated in Figure 7, F-value as the response variable investigated the impact of changing two out of three factors on it. All the response surfaces exhibited concave shapes with both center and edges falling within the studied domain. This behavior indicates that there is an optimum response for the F-value. Figure 7 indicates that the impact of reaction temperature and enzyme ratio on the F-value is greater than that of residence time, emphasizing the significant influence of reaction temperature and enzyme ratio on F-value.

Thus, when the residence time was between 22 min and 34 min and the enzyme ratio was between 0.5–0.9, the F-value was significantly increased (Figure 7b). At the reaction temperature of 38–50 °C, α-amylase did not reach the optimal reaction temperature, and the residence time was short, so the decomposition effect of amylase on starch particles was unsatisfactory [22]. At temperatures exceeding 50 °C during mashing, β-glucanase activity was quickly diminished, while the solubilization of β-glucans from intact cell walls persisted [23,24,25]. This could be explained by the presence of carboxypeptidase and other highly thermoresistant enzymes in barley malt [26]. As the thermal insulation temperature in this study was within the effective range, the effects of such enzymes were not considered. There was a two-step degradation of β-glucan, first by a lichenase to β-glucooligosaccharides and finally by a β-glucosidase to fermentable glucose units [27]. Because of the short residence time, β-glucan may only have been decomposed into β-oligosaccharides, but it had no significant effect on the turbidity of the clarified solution. The decomposition of β-glucan in malt was limited, and the remaining β-glucan will cause turbidity in less solution, which can also be compared and predicted from the side of different filtration performances between malts. Apart from beta glucanase, Arabian xylanase was also present. This enzyme hydrolyzes arabinoxylan and reduces the viscosity of wort [28]. The reduction in viscosity had a certain effect on the fermented liquid filtration but did not affect the turbidity of the clarifying liquid. The activity of the neutral protease peaked at 30–55 °C within 10–30 min of reaction time [29]. This experiment achieved the optimal activity temperature and reaction time for neutral protease, resulting in its highest level of activity. Therefore, the efficiency of proteolytic hydrolysis was higher than that of amylase and glucanase. In case the proteins are not decomposed, they can give rise to turbid particles resulting in turbid wort [30]. Thus, the filtration efficiency of malt can be evaluated and predicted based on this observation. The optimal proportion of the enzymes with the highest F-value was between 0.5 and 0.9, which shows the proportion of β-glucanase was gradually increased and the proportion of amylase and protease decreased. However, since the temperature and time of protease and glucanase action were within the optimal range, it was speculated that because of the differences among different malts, only limited starch, β-glucan, and protein could be hydrolyzed, while the remaining substances would cause turbidity, so as to quickly detect and predict the filtration performance of malt.

Figure 8a shows the effects of reaction temperature, residence time, and enzyme ratio on the differentiation between samples. Enzyme ratio and reaction temperature had significant influence on the degree of differentiation. When the reaction temperature was 38–50 °C and the enzyme ratio was 0.5–1.3, the differentiation was greater than 130 °C. For greater differentiation, an enzyme ratio of 0.5 and an insulation temperature of 38–47 °C are recommended. In addition, as shown in the Figure 7, although the differentiation between samples was large within the experimental range, higher α-amylase ratio and reaction temperature would result in sufficient enzymatic hydrolysis reaction, and fewer colloidal particles would enter the wort, which will reduce the differentiation between samples. As shown in Figure 8b, c, holding time and holding temperature affect sample differentiation in a small range. If the holding time is 27–38 min and reaction time is 38–47 °C, then the differentiation will be greater than 130 °C. On the other hand, when the residence time is 22–27 min, then temperature change will have a very minute effect on the degree of differentiation. Therefore, only when the residence time is kept above 27 min can a higher degree of differentiation be obtained. According to these results, both high enzyme ratio and low residence time lead to decreased differentiation between samples. The reasons for the decrease in differentiation could be the uneven content of various substances between the given samples or the effects of holding time and holding temperature on enzyme activity that cause excessive decomposition, and the uniformity of turbidity between samples.

3.3.3. Optimization of Test Conditions and Experimental Verification

The optimization tool of the Expert Design software was used to determine the maximum F-value of samples, so as to improve the differentiation between samples and make it easier to judge the filtering performance. The optimal values of the experimental variables are shown in

Table 6. Optimum operating conditions; predicted and experimental values of F-value.

Optimum Operating Conditions			Predicted	Experimental	Error (%)
Residence Time (min)	Reaction Temperature (°C)	Enzyme Ratio	F-Value	F-Value	F-Value
38	44	0.5	141.81	147.00	3.7

. The optimal reaction conditions yield a maximum F-value, a residence time of 38 min, a reaction temperature of 44 °C, and an enzyme ratio of 0.5 (α-amylase, β-glucanase, neutral protease = 1:2:1). In order to verify the optimization of the prediction, three verification experiments were carried out under the optimal conditions of the prediction. The results depict that the experimental values are in good agreement with the prediction model within a 4% error. Therefore, RSM is a reliable and feasible tool to optimize the method for rapid detection of malt filtration performance.

4. Conclusions

We imitated a simple mashing process to measure the turbidity of wort to test the potential filtration performance of malt. The operation is operable and the residence time is short, which truly reflect the difference in filtration performance between malts, is consistent with the actual production situation in the factory, and was significantly correlated with glucan content in malt (p < 0.05). The findings demonstrated the method’s efficacy and dependability, with less than 10% RSD across six replicated trials. The experiment was planned so as to use the 3 × 3 factor design chosen by the response surface method (RSM), and the ideal experimental conditions were identified for the quick assessment of malt filtering performance. Analysis of the quadratic model variance revealed that reaction temperature is not as important for sample differentiation as reaction duration and enzyme ratio. Reaction time is the most crucial characteristic, according to the RSM approach. Using experimental data and variance analysis, the model equation is created. Optimal conditions for differentiation can be enhanced with the enzyme ratio of 0.5 and time period and temperature of 38 min and 44 °C, respectively. The F-value under these circumstances is 147.00.

A fermentation experiment is the best way to verify the new test method in the laboratory. Further studies will construct a model by combining commercial production procedures and filtration dates to improve reliability.