Online Identification of Beer Fermentation Phases

Abstract

Over the last two decades, the craft beer industry has significantly developed with the emergence of thousands of microbreweries all over the world. These are mostly small companies that cannot afford the cost of the process monitoring systems that are usually embedded in the machinery used by industrial breweries, but they need to monitor and control the production process in order to guarantee a constant quality of beer. The development of low-cost systems for monitoring the production process would help microbreweries obtain the desired product quality consistency. In this paper, the authors propose a low-cost system for the real-time identification of the different phases of the alcoholic fermentation of beer. The first results prove the ability of the proposed system to monitor the fermentation and to detect anomalies in the process promptly.

1. Introduction

In recent years, craft beer has been gaining popularity thanks to the greater variety of styles and ingredients compared to the more standardized industrial products. The number of microbreweries producing craft beer is increasing in several countries [1,2]. These companies are often small, with a limited financial capacity, making it difficult to invest in the quality control systems that are usually installed in industrial breweries. The measurement of the main quality parameters is based on international standards provided by several institutions, such as the American Society of Brewing Chemists (ASBC), the Central European Commission for Brewing Analysis (MEBAK), and the European Brewery Convention (EBC). These standards require expensive instrumentation and skilled personnel. For this reason, craft breweries are looking for cheaper and easy-to-use solutions for monitoring and controlling the production processes (milling of malts, wort production, fermentation, packaging) and testing the quality of the raw materials (water, malt, hops, yeast) and the beers. In this context, several studies have been conducted to develop methods of detecting the quality of the production processes and products suitable for craft beer producers [3]. Machine learning techniques with electronic noses and robotics have been used to propose a low-cost system that is able to assess some of the quality parameters in beers, like the type of fermentation and sensory evaluation [4]. In a similar way, near-infrared spectroscopy techniques and hardware, using an ad hoc approach for the beer application, have been integrated into a low-cost electronic nose based on gas sensors and machine learning to detect off-flavors in beer [5]. Sensory analysis has also been evaluated using gas chromatography/mass spectrometry, which is able to detect several volatile compounds present in beer [6]. In addition to sensory analysis, a system by which to predict the four physicochemical parameters (real extract, alcohol, polyphenol content, and bitterness) of beer using an electronic tongue based on eighteen potentiometric chemical sensors has also been proposed [7]. Image processing techniques have been used to evaluate the visual properties of beer, such as color and foam stability [8]. As for the production process, a low-cost automated system based on an Arduino board has been proposed to control the temperature steps during the mashing stage of the grains in which the saccharification of the starches by enzymes takes place [9]. The authors were involved in a research project for the development of low-cost systems for quality control in craft beer production. In this context, they proposed a device for the color measurement of beer based on the spectrophotometric method [10]. A reader for pH strips based on image analysis was also proposed [11]. Finally, a preliminary work on fermentation monitoring, based on the Internet of Things (IoT), was proposed [12].

Fermentation is a key step during beer production; therefore, in several studies, different techniques have been taken into account to verify the correct evolution of the process. The use of near-infrared spectroscopy (NIRS) has been proposed to monitor ethanol, specific gravity, and free amino nitrogen (FAN) [13]. A hybrid electronic tongue based on potentiometric and voltammetric tongues has been employed to distinguish samples taken during beer fermentation [14]. A transmission-based ultrasonic sensor, in conjunction with machine learning techniques, has been used to predict the volume of alcohol during beer fermentation [15,16]. The measurement of the soluble solid content (SSC) and the pH during beer fermentation has been evaluated using an optical non-destructive device based on vis-NIR spectroscopy [17]. The development of a glucose and ethanol amperometric biosensor has also been proposed for beer fermentation monitoring [18]. In this paper, the authors propose a low-cost, online system for identifying the different phases of alcoholic fermentation during beer production. The system is suitable for the real-time monitoring of the fermentation process, particularly with respect to identifying the fermentation phase and signaling anomalies during the process. It does not require any extraction, and the measurements are made without any contact with the beer, eliminating the need for sanitization since it makes the measurements using sensors placed outside the fermenter. Furthermore, it utilizes load cells placed below the fermentation tank, allowing it to be used with many types of fermenters.

2. Materials and Methods

2.1. Proposed Measurement Method

The proposed monitoring system employs a measurement method based on the fermentation reaction. This reaction generates a significant amount of carbon dioxide (CO₂), which escapes through an airlock in the fermentation tank, resulting in a measurable decrease in the tank’s weight. To accurately monitor this process, it is critical to maintain the fermentation vessel’s integrity by ensuring it remains airtight, with no possibility of air leaks or unintended air entry. Measuring the fermenter weight in real time allows us to track the evolution of the fermentation process. The scale used for this purpose must have a resolution fine enough to discern the gradual weight reduction, especially during the final stages of the fermentation process, where the weight decrease is minimal. Furthermore, the scale must have a maximum capacity that exceeds the total weight of the fermenter and the wort. Standard commercial scales often fail to meet these opposite requirements, as they typically offer a resolution too coarse relative to the needed full-scale capacity. To meet both requirements, a custom weight module has been engineered to provide a fine resolution and full-scale capacity of up to 200 kg.

Here, we propose a method to measure the quantity of carbon dioxide produced during alcoholic fermentation. Before this process starts, the disaccharides in the malt—maltose and maltotriose—are hydrolyzed in glucose by the a-glucosidase enzymes in the yeast. Additionally, any portion of the maltose or maltotriose that does not convert into glucose should be considered part of the biomass [19]. Then, during beer fermentation, the glucose obtained in the wort is converted into ethanol and carbon dioxide [20,21] according to the following reaction:

$${\mathrm{C}}_6{\mathrm{H}}_{12}{\mathrm{O}}_6\stackrel{\mathrm{y}\mathrm{e}\mathrm{a}\mathrm{s}\mathrm{t}}{\to }2{\mathrm{C}}_2{\mathrm{H}}_5\mathrm{O}\mathrm{H}+2\mathrm{C}{\mathrm{O}}_2$$

(1)

According to Equation (1), the alcoholic fermentation converts, thanks to the action of yeast, 1 kg of glucose (C₆H₁₂O₆) into 0.51 kg of ethanol (C₂H₅OH) and 0.49 kg of carbon dioxide (CO₂). The original extract of the wort is related to the alcohol and the carbon dioxide according to the Balling equation [20,21,22]:

$$\underset{\mathrm{e}\mathrm{x}\mathrm{t}\mathrm{r}\mathrm{a}\mathrm{c}\mathrm{t}}{\underbrace{2.0665 \mathrm{g}}}\to \underset{\mathrm{a}\mathrm{l}\mathrm{c}\mathrm{o}\mathrm{h}\mathrm{o}\mathrm{l}}{\underbrace{1.0000 \mathrm{g}}}+\underset{\mathrm{c}\mathrm{a}\mathrm{r}\mathrm{b}\mathrm{o}\mathrm{n}\mathrm{d}\mathrm{i}\mathrm{o}\mathrm{x}\mathrm{i}\mathrm{d}\mathrm{e}}{\underbrace{0.9565 \mathrm{g}}}+\underset{\mathrm{b}\mathrm{i}\mathrm{o}\mathrm{m}\mathrm{a}\mathrm{s}\mathrm{s}}{\underbrace{0.1100 \mathrm{g}}}$$

(2)

This relation suggests that by weighing the tank in which the fermentation reaction occurs, it is possible to follow the progress of the process. As mentioned before, the fermentation reaction starts thanks to the presence of yeast that is inoculated into the beer wort. Fermentation affects the flavor and taste of the beer. It is necessary to inoculate the correct number of viable yeast cells to obtain a product with the desired characteristics. Moreover, each type of yeast has an optimal temperature range, so it is necessary to control the temperature of the fermentation tank. The fermentation process can be divided into three phases: lag, exponential, and stationary. During the first phase, which can last up to 24 h, the yeast cells consume the oxygen contained in the wort, then the number of cells grows exponentially, and there is a rapid increase in the alcohol concentration. This phase can last several days, after which the fermentation ends, obtaining a maximum ethanol concentration and yeast biomass [23,24,25]. The weighing system aims to detect the three phases of the fermentation and signal possible anomalies at the end of the process.

2.2. Proposed Experiments

The monitoring system has been evaluated through two separate classes of tests. The first focused on monitoring the weight during fermentation and was conducted on the fermentation tanks used in the production line of the Birring microbrewery in Penta (SA, Italy). The tanks used have a cylindrical shape with a capacity of 150 L; they are equipped with an airlock on the lid, allowing carbon dioxide to be released. Furthermore, thanks to the anaerobic nature of the process and the fact that the system is closed, it is possible to consider the water evaporation as negligible. A weighing module was installed under these cylindrical fermentation tanks to monitor the fermentation process. With this equipment, tests were carried out with six types of beer that differed in recipes and initial densities. Several batches of 90 L volume of the other beers were produced using a 150 L brewing system. In beer production, wort is fundamental because the sweet liquid yeast ferments into beer. The density of the wort at the start (original gravity or OG) and at the end (final gravity or FG) of the fermentation was measured with a densimeter (see

Table 1. Beer types with the original gravity (OG) and final gravity (FG).

Beer Type	Yeast Strain	OG (g/cm3)	FG (g/cm3)
British Golden Ale	SafAle™ S-04	1.045	1.010
Apple Wit Ale	Lalbrew Wit™	1.048	1.010
Lemon Ale	SafAle™ S-04	1.036	1.010
Belgian Pale Ale	Safbrew™ BE-256	1.054	1.014
English IPA	SafAle™ S-04	1.058	1.012
British Strong Ale	Safale™ S-04	1.073	1.018

). OG is related to the amount of sugars in the wort. At the same time, the difference between OG and FG indicates the amount of sugar converted into alcohol and carbon dioxide during the fermentation process. These values are influenced by several factors, including the composition of the wort, the yeast strain used, and the temperature at which fermentation occurs.

The second test class was conducted in the laboratory using the proposed system and a cylindrical fermentation tank with a 50 L capacity. Comparable to the big one, this tank was also equipped with an airlock on the lid to release carbon dioxide. The beer wort for these tests was prepared by mixing 4 kg of liquid malt extract with 20 L of water, resulting in an original density of 1.054 g/cm³. Two fermentations with identical wort were established for comparison: one using 22 g of Safale™ S-04 yeast (Fermentis By Lesaffre, 90 Rue de Lille, 59520 Marquette-lez-Lille, France) and the other using only 6 g of the same yeast. This setup was designed to evaluate the impact of under-pitching yeast. This situation arises when the quantity of yeast used is insufficient relative to the amount and density of the wort. The results from this comparison provided valuable insights into how such a scenario can influence the rate and efficiency of the fermentation process, thereby informing the optimal yeast quantity for efficient fermentation.

These two tests comprehensively evaluated the monitoring system and its effectiveness in different brewing scenarios, contributing to the optimization of the brewing process

2.3. Monitor Module

The monitoring system, capable of working simultaneously on different fermentation processes, comprises N-acquiring modules and a central computing server. The diagram is shown in Figure 1, and the parts of the module are described below:

•

ESP8266 microcontroller with an integrated Wi-Fi module; used as an acquiring and control module, capable of communicating with the server.

•

MCP9808 digital temperature sensor placed on the lateral surface of the fermenter. The sensor leads are designed to be long enough to reach the microcontroller without affecting the measure. The metrological specifications are listed here:

○

Accuracy:

• ±0.25 (typical) from −40 °C to +125 °C;
• ±0.5 °C (maximum) from −20 °C to 100 °C;
• ±1 °C (maximum) from −40 °C to +125 °C.

○

User-Selectable Measurement Resolution:

• +0.5 °C, +0.25 °C, +0.125 °C, +0.0625 °C.

○

Communication protocol:

• I2C.

•

HX711 24-bit analog to digital converter (ADC) for the weigh scales

•

Load cell (Qt. 4):

○

Full scale: 50 kg;

○

Sensibility: 1.0 ± 0.15 mV/V;

○

Linearity: 0.2% F. S.;

○

Hysteresis: 0.2% F. S.;

○

Creep: 0.1% F. S. (3 min).

As shown in Figure 1, each monitoring module is connected to a server that collects the information through wireless communication. Each module comprises a microcontroller, a temperature sensor, four load cells that convert the force applied to them into electrical signals, the necessary circuitry to condition the acquired signals, and the mechanics required to be placed under the fermenter.

The load cells are arranged in a Wheatstone bridge configuration and measure the weight of the object placed on top of the steel frames. The electrical signals transduced by load cells are amplified and converted into the digital domain by a specific device (HX711) connected to a microcontroller unit (MCU) by a serial bus. Figure 2b shows how the load cells are connected to the HX711 and the MCU.

Each load cell has a maximum capacity of 50 kg and an output sensitivity equal to 1.0 ± 0.1 mV/V. Four cells were used to create the custom scale; they were placed at the vertices of the metal frame and connected in a Wheatstone bridge configuration (Figure 2a). This configuration enabled a maximum measurement capacity of 200 kg with a resolution of 1 g.

The HX711 module amplifies and digitalizes the V₀ voltage and gives it to the microcontroller via serial port. The conversion factor from digital code to weight is obtained through a specific adjustment process using the known weights placed on the scale. This factor is then used to convert subsequent codes to accurate weight measurements. This procedure is intended to be conducted every time the measurement setup changes. The module has been designed for a 150 L fermenter, but the same architecture can be used with different fermenter capacities by simply changing the load cells.

All data collected from the fermentations are stored in databases containing the batch name and fermentation data. The entire system is designed to be Wi-Fi enabled in order to be easily used from anywhere in the brewery and for remote data monitoring; the server also takes care of the fermentation control algorithms.

3. Results

3.1. Calibration Procedures and Parasitic Effect Analysis

Before conducting the main tests, a calibration procedure was carried out using a reference scale (see

Table 2. Reference scale.

Working Temperature	[−10,40] °C, max humidity (85%)
Full scale	150 kg
Linearity	<0.01% of full scale
Resolution	15 gr
Cell Supply	5 Vcc 150 mA
Max cell number	4 (350 Ω), 8 (700 Ω)
Sampling frequency	20 Hz
Serial Port	RS232
Baud Rate	1200, 2400, 4800, 9600

for more details). The setup involved placing the fermenter on our scale, which was then positioned on the reference scale. As the fermenter was filled with different volumes of water (steps of 1 L until 115 L), data from both scales were recorded simultaneously.

The measurements were repeated 20 times per step. The obtained calibration curve is reported in Figure 3. The least square regression curve showed a ${\mathrm{R}}^2$ = 0.9999; using the so obtained parameters, all the measured values were corrected as:

$${\mathrm{V}\mathrm{a}\mathrm{l}}_{\mathrm{C}\mathrm{o}\mathrm{r}\mathrm{r}\mathrm{e}\mathrm{c}\mathrm{t}\mathrm{e}\mathrm{d}}={1.0000\times \mathrm{V}\mathrm{a}\mathrm{l}}_{\mathrm{m}\mathrm{e}\mathrm{a}\mathrm{s}}-0.0735$$

(3)

where ${\mathrm{V}\mathrm{a}\mathrm{l}}_{\mathrm{C}\mathrm{o}\mathrm{r}\mathrm{r}\mathrm{e}\mathrm{c}\mathrm{t}\mathrm{e}\mathrm{d}}$ is the final weight value corrected using the regression coefficient and ${\mathrm{V}\mathrm{a}\mathrm{l}}_{\mathrm{m}\mathrm{e}\mathrm{a}\mathrm{s}}$ is the measured value.

We calculated the regression and adjusted the measured values using the obtained coefficients to assess the accuracy. Subsequently, we examined the residuals, removing all the outliers and obtaining a maximum residual of 0.13 kg, corresponding to 0.07% of the full scale.

The accuracy of the linear approximation was validated by measuring the non-linearity error. This error was less than 0.02 kg, indicating a high level of precision in the approximation. Such a low non-linearity error confirms that the linear model is highly reliable for our measurements, ensuring that the deviations from the actual values are minimal and within acceptable limits. After the calibration procedure, the dependence of the proposed scale on the environmental and parasite effect of the load cells was investigated. To this purpose, a fermenter was filled with 20 L of water, and the system was left in situ for three days. An initial and final weighing was conducted using a reference scale to validate the results.

In addition to these steps, an in-depth analysis was conducted to observe how our scale reacts to temperature changes. Maintaining the same setup described in Section 3.1’s paragraph, we thoroughly investigated the potential drift caused by the parasitic effects, with particular attention given to the temperature influences. A detailed characterization was performed despite fermentation occurring in environments where the temperature is controlled to ensure stability for the scale in contact with the fermenter. The scale underwent temperature variations, ranging from 10 °C to 20 °C, allowing the acquisition of corrective parameters to mitigate any heat-related interferences. This temperature range corresponds to the effective working conditions of a brewing room equipped with temperature control systems. Temperature-related phenomena are negligible within this range, even considering the previously calculated accuracy.

Furthermore, the creep rate was considered as the change in load cell output under a constant load with all environmental conditions remaining unchanged. Our findings indicated no observable weight drift. During the test, the room temperature and the weights were recorded. Figure 4 shows the daily trend, while the synthetic values are listed here:

• The mean weight recorded was 20.128 kg;
• The minimum was 20.035 kg;
• The maximum was 20.176 kg;
• The standard deviation was 0.028 kg.

The value is within the previously calculated limits of uncertainty. Therefore, the parasitic effects of the load cells may be neglected.

3.2. Experimental Setup

3.2.1. First Test—Weight Monitoring

As said before, the fermenters used for the first test have a capacity of 150 L. However, each batch of wort has a volume of only 90 L, which leaves some headspace in the fermenters. This is necessary to allow the carbon dioxide produced by the yeast to escape and to prevent the foam (called “krausen”)—that forms on top of the wort—from overflowing.

The fermenters have been equipped with the monitoring system and placed in the fermentation room at 16 °C. The density of each batch has been measured at the start (original gravity OG) and the end (final gravity FG) using a densimeter with a 0.001 g/cm³ resolution. The final product is characterized by the same final gravity with a different original gravity. A higher original gravity means a higher sugar content in the wort that can be converted into alcohol and carbon dioxide during fermentation. The difference between the FG and OG can be related to the alcohol content in the final product.

Our module has been positioned directly beneath the fermenter, ensuring the precise detection of weight variations throughout the fermentation process. Regarding temperature monitoring, the sensor has been placed in contact with the external wall of the fermenter. This positioning was chosen for its non-invasive nature and compliance with hygiene regulations.

An example of acquisition is shown in Figure 5:

During the fermentation process, a loss of wort weight is expected due to the conversion of sugars into alcohol and carbon dioxide, which is released through the fermenter airlock. Simultaneously, there is an increase in temperature because fermentation is an exothermic reaction; this one in the figure represents a preliminary acquisition in which the temperature is also measured since the quality of beer depends on the temperature reached during fermentation; the latter is controlled to keep it constant throughout the process. Again, in Figure 5, the three main stages of fermentation can be appreciated:

• Lag phase (1): During this phase, yeast acclimates to the environment, absorbs nutrients, and prepares for cellular division with minimal to no fermentation. Depending on the recipe (type and quantity of yeast used), this phase can range from 5–6 h up to 24 h.
• Exponential phase (2) corresponds to rapid yeast growth, doubling cells at a defined rate. At this stage, we have the maximum gradient in weight loss.
• The stationary phase (3) is a halt in yeast reproduction due to a lack of nutrients, which defines the end of fermentation. A negligible weight loss characterizes this phase.

Consequently, by analyzing the weight loss rate, it is possible to distinguish the three phases.

The fermentations were then observed to collect data. Analyses were performed on different recipes to highlight the differences. The final weight loss at the end of the fermentation and the obtained ABV (alcohol by volume) are given in the next table (

Table 3. Final weight loss and alcohol by volume (ABV).

	Figure	Weight Loss [kg]	OG [g/cm3]	FG [g/cm3]	ABV [%]	CO2 Produced [kg]
Lemon Ale	Figure 6a	2.64	1.036	1.010	3.413	2.347
Belgian Pale Ale	Figure 6b	3.95	1.054	1.014	5.250	3.611
British Pale Ale	Figure 6c	3.31	1.045	1.010	4.594	3.160
Apple Wit Ale	Figure 6d	3.54	1.048	1.010	4.988	3.431
Lemon Ale	Figure 6e	2.37	1.036	1.010	3.413	2.347
British Pale Ale	Figure 6f	3.43	1.045	1.010	4.594	3.160

).

Referring to Table 1, it is possible to obtain the alcohol by volume using the formula:

$$\mathrm{A}\mathrm{B}\mathrm{V}=\left(\mathrm{O}\mathrm{G}-\mathrm{F}\mathrm{G}\right)\times 131.25$$

(4)

where (OG − FG) is the difference in density, and 131.25 is a constant and comes from the ratio between the number of grams of ethanol generated per gram (1.05) and the approximate density of ethanol (0.8) [27].

The fermentation progress over six days for six different beer batches is shown in Figure 6. The three phases of the fermentation are visible. In the first one or two days, the weight does not change during the lag phase, and then it decreases exponentially. Lastly, it settles to a value corresponding to the end of fermentation (stationary phase). In more detail, Figure 6a,e refer to different batches of the same type of beer (lemon ale), which has the lowest original gravity (OG) among the beers taken into account (1.036 g/cm³). This means that it contains a lower wort extract than the other types of beer and, as expected, exhibited the lowest quantity of released carbon dioxide. A greater weight decrease can be seen in Figure 6c,f, corresponding to the fermentation of two batches of British golden ale (OG 1.045 g/cm³) and more in Figure 6b, corresponding to the Belgian pale ale with an OG of 1.054 g/cm³. Another important aspect of the fermentation evolution is the duration of the three phases, particularly the lag and the exponential phases. These strongly depend on the characteristics of the yeast strain, the pitching rate (the quantity of yeast inoculated in the wort), temperature, and the original gravity of the wort. In this regard, the fermentation evolution shown in Figure 6d exhibited the shortest lag time. This was due to the characteristics of the Lallemand LalBrew Wit™ yeast (Lallemand, 1620 Rue Préfontaine, Montréal, QC, Canada) used in the apple with ale recipe.

3.2.2. Fermentation Rate

A new parameter, the fermentation rate (FR), has been introduced to highlight the evolution of the three phases of the fermentation process. This rate is defined as the variation of the weight per second. Figure 7 and Figure 8, respectively, show the fermentation evolution and the related fermentation rates of four batches. These batches used two types of yeast: Fermentis SafAle™ S-04 (Fermentis By Lesaffre, 90 Rue de Lille, 59520 Marquette-lez-Lille, France) for batches one, two, and four, and Fermentis Safbrew™ BE-256 (Fermentis By Lesaffre, 90 Rue de Lille, 59520 Marquette-lez-Lille, France) for batch three. In this context, the fermenter capacity is 150 L, the beer quantity is 95 L, and the room temperature is 19 °C. The results of this experiment, including the original gravity for each batch, are listed in

Table 4. Summary of the comparisons.

	Batch1: Belgian Pale Ale	Batch2: British Golden Ale	Batch3: IPA	Batch4: British Strong Ale
Yeast	SafAle™ S-04	SafAle™ S-04	Safbrew™ BE-256	SafAle™ S-04
Original gravity	1.054 g/cm3	1.045 g/cm3	1.058 g/cm3	1.073 g/cm3

.

The fermentation data, collected over six days, reveal the three phases of the fermentation process. The weight initially decreases and stabilizes after a few days. In the initial 5–24 h, the weight shows minimal variations. Subsequently, the active fermentation phase starts, leading to a rapid weight decrease. Finally, as the fermentation process ends, the weight loss slows down. Each of the four batches has a different final weight, with batch 3 (IPA), which used a different yeast in the recipe, showing a slower weight loss.

Regarding temperature, it rises as the weight decreases and reaches the maximum value when the weight decrease is at its maximum rate; by the end of the fermentation process, it tends to approach room temperature.

The change in the fermentation over time and its completion can be measured by a fermentation rate (FR) parameter (5):

$$\mathrm{F}\mathrm{R}={\displaystyle \frac{∆\mathrm{W}}{∆\mathrm{t}}}$$

(5)

where ∆W is the change in weight in the time interval ∆t, which is 10 s. The fermentation rate is almost zero in the lag phase and at the end of fermentation, and it is negative when the fermentation is active, with the maximum value indicating a faster weight decrease during the exponential phase. Figure 8 shows the fermentation rate for the four batches.

To improve the clarity of the data, the time window in which the fermentation is active and has its maximum variation has been considered, i.e., the first two days have been considered for all batches. The fermentation process varies for the four batches under test. Batch 3, which used a different yeast and a higher temperature, had the highest fermentation rate, while batch 2 had the lowest fermentation rate, even compared to the batch with the same recipe.

However, this relation is not universally valid as the quantity of carbon dioxide released during fermentation depends on the volume of the fermentation tank, the amount of wort, and the temperature since part of the CO₂ remains dissolved in the wort.

3.2.3. Second Test—Relation between Density and Weights

Other tests were conducted to quantify the correlation between wort density and weight reduction during fermentation. It is well-established that the original gravity allows for predicting the potential final alcohol content. Identifying a relationship between wort density and weight reduction during fermentation makes monitoring the gravity possible.

Another batch was brewed using a known recipe, adhering strictly to best brewing practices. The initial data for this batch were as follows:

• Maximum capacity: 50 L;
• Quantity of malt: 4 kg;
• Quantity of water: 20 L;
• OG: 1.054 g/cm³;
• FG: 1.014 g/cm³;
• Yeast: SafAle™ S-04;
• Batch yeast quantity: 22 g.

The density was measured by taking samples from the fermenter during the process using a standard brewer’s triple-scale hydrometer with the following characteristics:

• ABV range: 0–20%;
• Specific gravity range: 0.99–1.17 [g/cm³];
• BRIX: 0–35 [°Br].

The weight and density measurements were as follows (

Table 5. Sample from fermenter.

Weight [kg]	Density [g/m3]
25.08	54
24.93	50
24.89	47
24.80	43
24.28	19
24.18	16
24.08	14

):

As shown in Figure 9, the results revealed a strong correlation between these two variables, obtaining the following equation:

$$\mathrm{y}=42.654\mathrm{x}-1014.8$$

(6)

The coefficient of determination (R²) was found to be 0.9948, indicating a nearly perfect linear relationship. This high R² value suggests that the initial gravity of the wort is a reliable predictor of the weight loss that occurs during fermentation.

As a counterproof, another batch was brewed using the same recipe but not the recommended yeast quantity. Despite the original gravity (OG) being the same as the previous batch, the final gravity (FG) was not reached, indicating incomplete fermentation. This is evident in Figure 10, which shows a significantly slower fermentation process compared to batch 1. With the proposed system, tracking and predicting fermentation dynamics is possible. This is a simple example of how the real-time measurements that the system provides can be used to understand the quality of ongoing fermentation.

4. Discussion

The authors have developed a system that uses four load cells mounted in a Wheatstone bridge configuration to monitor weight changes during fermentation. The system has a resolution of 1 g and a full-scale capacity of 200 kg, making it capable of accurately capturing small weight changes during fermentation. By observing the weight changes, it was possible to identify the three phases of fermentation: the lag phase, the exponential phase, and the stationary phase. This system provides a non-invasive method for real-time fermentation-process monitoring and represents a significant innovation in the brewing field.

Before conducting the main tests, the measurement system was calibrated using a reference scale. During the calibration, no systematic phenomena or other influencing parameters were observed, ensuring the system’s accuracy for the subsequent tests.

The system was then used to monitor the fermentation of various types of beer, each with different recipes and initial densities. Several batches of 90 L volumes of the different beers were produced using a 150 L brewing system. Figure 6 in the paper provides a visual representation of this process. It shows the weight variations over time for six different beer batches, clearly illustrating the three phases of fermentation. With the lowest original gravity (OG), the Lemon Ale (Figure 6a) produces the least CO₂ produced during fermentation. On the other hand, the Belgian Pale Ale, having the highest OG, leads to the greatest amount of CO₂ production. The fermentation rate, defined as the change in weight per second, was introduced to better understand the evolution of the three fermentation phases: it is almost zero in the lag phase and at the end of fermentation, and it is negative when the fermentation is active. Figure 7 and Figure 8 clearly illustrate that a higher original gravity (OG) corresponds to a greater fermentation rate during the exponential phase and, consequently, a larger weight loss.

The study conducted tests to establish a correlation between wort density and weight reduction during fermentation, revealing a strong linear correlation between these variables (Figure 9). Furthermore, the study demonstrated the system’s potential to identify errors in beer production, such as using less yeast, which resulted in a slower rate of weight reduction.

This research provides a novel approach to fermentation monitoring in the context of previous studies. Traditional methods often involve periodic sampling, which can be disruptive and may introduce contaminants into the fermentation [28,29,30]. This approach could potentially enhance monitoring efficiency and accuracy while also reducing the need for manual operation.

Looking forward, one possible direction is the development of applications for microbreweries. Given the system’s simplicity and effectiveness, it could be a valuable tool for small-scale breweries that lack the resources for more complex monitoring systems. As a future development, integrating this system with advanced data analytics and machine learning algorithms can be explored to predict fermentation outcomes more accurately. Additional data will be collected from new recipes to refine the model further and establish alarm thresholds, enabling the prompt signaling of problems to the brewmaster. By creating a labeled dataset from these collections, a machine learning model can be trained to assist in the analysis, aiming to identify any defects faster and more precisely.

5. Conclusions

The research conducted in this study has led to the successful development and evaluation of a non-invasive, low-cost online system for monitoring the different phases of beer fermentation. This study introduces a weight-based monitoring system that represents a noteworthy advancement in beer fermentation, facilitating and enabling the real-time early detection of potential issues. The system’s simplicity, accuracy, and scalability make it an asset, especially for microbreweries. The system’s scalability allows for its adaptation to diverse brewing scenarios, enhancing its versatility for breweries of varying sizes. Additionally, expanding the system’s application to other fermentation processes, such as wine or spirits production, could provide valuable insights and enhance production efficiency across the beverage industry.