The Characterization of the Alcoholic Fermentation Process in Wine Production Based on Acoustic Emission Analysis

Abstract

The present experimental study assessed the viability of utilizing an acoustic emission signal as a monitoring instrument to predict the chemical characteristics of wine throughout the alcoholic fermentation process. The purpose of this study is to acquire the acoustic emission signals generated by CO₂ bubbles to calculate the must density and monitor the kinetics of the alcoholic fermentation process. The kinetics of the process were evaluated in real time using a hydrophone immersed in the liquid within the fermentation tank. The measurements were conducted in multiple fermentation tanks at a winery engaged in the production of wines bearing the Rioja Denomination of Origin (D.O.) designation. Acoustic signals were acquired throughout the entirety of the fermentation process, via a sampling period of five minutes, and stored for subsequent processing. To validate the results, the measurements obtained manually in the laboratory by the winemaker were collected during this stage. Signal processing was conducted to extract descriptors from the acoustic signal and evaluate their correlation with the experimental data acquired during the process. The results of the analyses confirm that there is a high linear correlation between the density data obtained from the acoustic analysis and the density data obtained at the laboratory level, with determination coefficients exceeding 95%. The acoustic emission signal is a valuable decision-making tool for technicians and winemakers due to its sensitivity when describing variations in kinetics and density during the alcoholic fermentation process.

1. Introduction

The process of alcoholic fermentation is important in numerous procedures undertaken within the food industry. Monitoring this process enables the implementation of control measures, thereby facilitating the production of superior-quality products. The intricate nature of fermentation processes is attributable to the convergence of biochemical, chemical, and physicochemical processes, with complex interactions occurring among these elements. During the production of wine, carbon dioxide is generated and rises to the surface in the form of bubbles. The bubbles exhibit a range of sizes, and their dynamics are contingent upon the specific stage of the fermentation process [1].

The intricate nature of the alcoholic fermentation process produces challenges regarding the measurement of process variables. This is largely due to the high time constants, time-varying nature, and non-linearity of the system. In many wineries, the wine fermentation process is typically controlled manually through periodic sampling and off-line laboratory analysis [2,3]. One of the most crucial variables monitored during the alcoholic fermentation process in wine production is the density of the fermenting must and its alcohol content. Its variation over time serves as an indicator of the process’s behavior from start to finish, and understanding these changes enables winemakers to make informed decisions related to process control [3].

The conventional approach to monitoring the progression of wine fermentation entails the manual gauging of wine density (which is correlated with the concentration of sugars and ethanol) as the must undergoes transformation into wine. These measurements are taken from samples collected and analyzed a minimum of one time per day during the alcoholic fermentation process. These measurements cause the process to be time-consuming and laborious for the winemakers [3,4,5]. In large wineries, the large number of tanks involved makes the process highly complex and tedious. Consequently, there is a constant need to develop new measurement methods in order to reliably quantify the density as an important chemical property [2].

The implementation of in-line sensors would effectively circumvent the issue of time-consuming manual operations, thereby facilitating the real-time assessment of the fermenting medium during a process that is typically conducted manually by the winemaker on a daily basis.

To address this issue, research is being conducted to monitor and model the fermentation dynamics [2]. The methods employed are diverse and encompass a range of techniques, including spectroscopy [4,6,7,8,9,10], the use of sensors based on ultrasonic transmission [11,12], differential pressure measurements [3,5,13], the measurement of the CO₂ generated [14,15,16], the use of chemical sensors as an electronic nose [17,18,19,20], measurement by weight loss [21], the use of optical devices [22,23,24,25,26,27,28], the modeling of the process [29,30,31,32,33,34,35,36], and others. In this regard, Lachenmeier et al. [22] propose an infrared (IR) sensor for in situ alcohol content monitoring during the fermentation process. The results obtained during the validation of the proposed method demonstrated a high degree of accuracy, with a mean repeatability of 0.05% vol and a relative standard deviation of less than 0.2%. Another optical fermentation monitoring method is presented by Jimenez et al. [23]. The authors propose the use of optical methods, based on three wavelengths, to determine the chromatic characteristics of the red wine obtained from the refractive index and optical densities of the sample. However, the method is limited in that samples must be isolated from the sensor to avoid the formation of potassium bitartrate crystals.

The density of wine or must is determined by the mass per unit volume at 20 °C and is expressed in g/L. Its measurement is used to determine the evolution of alcoholic fermentation during the wine production process. Variations in density over time provide insights into the process from inception to completion. By understanding these variations, winemakers can make informed decisions regarding process control. Some researchers have also measured liquid densities during the alcoholic fermentation process of wine.

Density can be calculated from differential pressure measurements. In this case, it is necessary to place more than one sensor equidistantly, so that their location has a significant effect on the measurement results. Nelson et al. [3] combined in situ density measurements using differential pressure sensors with a kinetic model of the process to estimate the evolutionary behavior of the fermentation. In this case, the differential pressure density calculations showed good correlation with the manual measurements taken with a density meter. The results showed that the automated measurement method, combined with a kinetic model of wine, can be used in wineries to control and predict the evolution of the alcoholic fermentation process. A more comprehensive study was carried out by Nerantzis et al. [14] to evaluate the evolution of the fermentation dynamics. In this case, they used a mass flow meter in the differential pressure measurement to measure the CO₂ production rate. Working from these measurements and the temperature of must in fermentation, the authors calculated the density and the evolution of the process in real time. The data obtained can be used to model the duration of fermentation and monitor fermentation dynamics in order to decide whether or not to add nutrients.

Reports on the use of acoustic signals in wine alcoholic fermentation processes are mainly based on the use of ultrasound to characterize the medium [11,12]. The main limitations of these methods lie in their dependence on the geometrical characteristics of the fermenting tank and on the liquid and volume contained therein. These conditions mean that the systems must be calibrated beforehand to ensure their effectiveness.

Previous studies by this research team [1] demonstrated the validity of using acoustic emission as a predictive tool for the anticipatory control of temperature variation. After analyzing the scientific literature and a recent review of oenological monitoring equipment [1], we can conclude that no research has used acoustic emission to estimate the physico-chemical properties of must during the wine fermentation process or any of its production phases.

This study proposes the use of the acoustic emissions generated by CO₂ bubbles during the alcoholic fermentation of wine in order to correlate the acquired signal with real-time variations in the density of must seen during the alcoholic fermentation process. The aim of this study was to demonstrate the feasibility of using the sound generated by CO₂ bubbles as a tool for monitoring and controlling fermentation dynamics. The proposed method, which uses novel and highly sensitive techniques, will allow important technical decisions to be made in enough time to achieve the oenological objectives of wine production. A novel method is proposed in order to assess the process dynamics and the evolution of the density during the alcoholic fermentation process. This would be determined based on the sound emitted by the CO₂ bubbles generated in the liquid.

2. Materials and Methods

During the process of alcoholic fermentation, carbon dioxide is released into the atmosphere as a consequence of the transformation of sugars (principally fructose and glucose) into ethanol and carbon dioxide. Due to this, the CO₂ that has been formed within the liquid rises to the surface in the form of bubbles. The rate of CO₂ generation during this process is closely related to process dynamics. In order to obtain the acoustic signal produced by the CO₂ bubbles during the fermentation process, a hydrophone was introduced into each of the stainless-steel fermentation tanks. These had volumes of 20,000 L. Figure 1a depicts the hydrophone model utilized and its position within the tank (Figure 1b).

As illustrated in Figure 1b, the hydrophone is situated at the center of the tank at a height of 2 m from the bottom. This positioning of the hydrophone ensures that the signal captured predominantly pertains to the impact of CO₂ bubbles on its surface, with minimal interference from external noise. A Pt100 temperature sensor was attached to the hydrophone to monitor the temperature of the liquid in the vicinity.

Table 1. Hydrophone specifications.

	Specifications
Model	A5
Sensitivity	−173 dBV re: 1 uPa (−193 sensor + 20 dB integrated signal conditioning)
Linear bandwidth	20 Hz–10 kHz (+/−4 dB)
Directivity	Omnidirectional (<20 kHz)
Max depth	100 m
Supply current	2–20 mA IEPE constant current

outlines the specifications of the hydrophone employed.

The experiments were conducted in a winery situated in Aldeanueva de Ebro, La Rioja, Spain, within the Rioja Denomination of Origin (D.O.). Throughout the course of the tests, the temperature of the fermentation process was monitored and regulated through the utilization of external cooling jackets on the tanks.

2.1. Experimental Setup

Figure 2 shows a blueprint of the experimental setup used for this research.

As illustrated in Figure 2, the hydrophone signal was linked to a Siemens SIPLUS CMS1200 SM1281 condition monitoring module, which was in turn connected to a SIMATIC S7-1200 programmable logic controller (PLC). The SM1281 module was equipped with four channels for the measurement of vibration signals, with the option of connecting IEPE (Integrated Electronics Piezoelectric) sensors. Additionally, a Pt100 module for the measurement of temperature within the tanks was connected to the PLC. The data from both the hydrophone signals and the temperature sensors were stored in a database via the PLC. The data were exported to a computer for subsequent analysis and processing.

The acoustic and temperature signals were acquired in real time with a sampling period of five minutes. The monitoring of the entire process spanned approximately ten days, with six days devoted to the alcoholic fermentation of the wine. The measurements were initiated at the start of the experiment, when the tank was empty (0–45 h), and were then taken at key intervals throughout the alcoholic fermentation process (45–200 h) and at its conclusion (200–250 h). The experiments were conducted in three fermentation tanks for subsequent processing. To validate the tests, the must density in the three tanks was manually measured by the winemaker throughout the duration of the trials. The manual sampling of densities was performed once a day for a period of 24 h.

2.2. Analysis of Acoustic Signals

The considerable divergence of the acoustic data, seen due to the intricate nature of the alcoholic fermentation process, requires the use of digital filtering techniques to remove outliers and thus rectify the aberrations in the signal. In this instance, a moving average filter was employed through the implementation of a robust quadratic regression on each window. The filter is particularly useful when processing real-time signals.

Once the acoustic signals have been filtered, it is possible to obtain the behavior of the dynamics of the fermentation process from the variation in the speed with which the CO₂ bubbles impact on the hydrophone. This will be more intense at the beginning of the fermentation process and will subsequently decrease as the sugars are converted into ethanol as a consequence of the reduction in CO₂ generation and the decrease in the diameter of the bubbles.

Another investigation based on the CO₂ weight loss analyses demonstrated that the expulsion of CO₂ into the environment during fermentation exhibits a linear relationship with the density of the fermenting liquid (14). In light of the aforementioned considerations, it is possible to derive the fermentation rate from the first derivative of the density compared to time. This may be understood as the dynamic behavior of CO₂ production. Similarly, Nelson et al. [3] employ the first derivative of the density curve to derive fermentation rate curves.

In light of the findings presented in Buonocore et al. [21] and Nelson et al. [3], it is possible to derive the density curve from the CO₂ production rate, which is closely linked to the acoustic signal captured by the hydrophone during the alcoholic fermentation process. Buonocore et al. [21] evaluate the changes in fermentation during beer fabrication and its termination based on the measurement of the weight loss of the deposit due to CO₂ emissions. In this case, the fermentation process rate can be measured as the rate of change in weight over time. The authors show how, during the process, the rate of change changes according to the stage of fermentation. On the other hand, Nelson et al. [3] show that the rates of heat and CO₂ production during fermentation are related to the fermentation rate or the first derivative of the density curve. The authors show that the first derivative of the density curve has a similar shape to a Gaussian distribution curve and that the maximum usually occurs between the first third and the middle of fermentation process. Working from the data obtained, they produce a kinetic model of the wine that can predict the future fermentation course and the emission of heat and CO₂ reliably.

To determine the density using the acoustic signal, the signal of the fermentation dynamics of the process captured by the hydrophone is integrated, as shown in Equation (1).

d(t) = integral (a(t), 0, 250) + doffset,

(1)

The variables d(t) and a(t) have the following meanings: d(t) is the calculated density, a(t) is the acceleration signal captured by the hydrophone, 0.250 is the limit of integration from t = 0 h to t = 250 h, and offset is the initial density of the must at the start of alcoholic fermentation. Finally, to validate the results, the density values d(t) obtained from the acoustic signal are compared with the density values obtained manually by the winemaker. The X-Y plots and the coefficient of determination between the analyzed data are obtained. MATLAB^® software 2023a was used for the statistical analysis, the processing of the acoustic signals, and the obtention of the graphs.

3. Results and Discussion

The purpose of this initial stage of measurements is to evaluate the behavior of the acoustic signal during the fermentation process in order to correlate it with the chemical properties of the wine and the kinetics of the alcoholic fermentation process. Figure 3 illustrates the acceleration values acquired by the hydrophones positioned within the three fermentation tanks being analyzed.

As illustrated in Figure 3, the impact of a substantial number of CO₂ bubbles against the surface of the hydrophone results in a signal with markedly scattered values, thereby underscoring the intricate dynamics of the alcoholic fermentation process during winemaking. In order to encompass the entirety of the process, the signals were obtained prior to the transfer of the must into the tanks and the commencement of the alcoholic fermentation phase. To ensure consistency, the supply of the must to the three tanks being analyzed was performed on the same day. During the initial 50 h of measurement, no amplitude values were observed for the acoustic signal. In this instance, the low signal levels may be attributed to the background noise generated within the empty tank. This results from either echoes or activity within the cellar itself. The amplitudes observed between 40 and 50 h are indicative of the moment of tank filling. Once the tanks are full, the alcoholic fermentation process begins, whereby the sugars are converted into alcohol. During this phase, the chemical reaction depicted in Equation (2) is initiated [1].

C₆H₁₂O₆ (aq) → 2 CH₃CH₂OH (l) + 2 CO₂ ↑(g) + 98, 324 kJ,

(2)

As a consequence of the aforementioned reaction, a change in the dynamics of the process is observed after a specific period of time. This is associated with the commencement of the alcoholic fermentation process. This change is reflected in the behavior of the temperature values within the tank. Figure 4 presents a graph that shows the temperature behavior in order to show its temporal relationship with the kinetics of the process, which occurs as a consequence of the exothermic reaction that occurs in the tank.

The temperature shown from t = 0 to point (1) is in accordance with measurements obtained from the tank devoid of contents. The point marked (1) indicates the moment at which the tank is filled, thereby initiating the process of alcoholic fermentation. At this point in time, the temperature immediately reaches the temperature of the must that has been poured in (25.5 °C). Subsequently, the temperature of the must reaches equilibrium with the external temperature, at which point (2) is attained. From point (2), an increase in temperature is observed, which is indicative of the onset of the alcoholic fermentation process. This behavior is analogous with that observed in the acoustic signal captured by the hydrophone, which also exhibits a sudden increase in its values. The upward trajectory of the temperature is observed at point (3), which coincides with the attainment of the maximum hysteresis value for temperature control and the initiation of cooling for the external jackets of the tank, thereby ensuring that the temperature remains within the desired range. Upon reaching point (4), which marks the conclusion of the fermentation process, the temperature exhibits no discernible increase. This is attributable to the decline in fermentation kinetics.

Figure 5 illustrates the filtered signals, as outlined in Section 2. As can be observed, the measurement values demonstrate a highly variable dynamic, with behavior differing between tanks. This complexity makes their characterization a challenging and distinct process, varying from one tank to another. As illustrated in Figure 5a–c, the initial and final moments of the process exhibit minimal dispersion in terms of acoustic values. At these times, CO₂ production is negligible due to the absence of a chemical reaction and the subsequent completion of the conversion of sugars into alcohol.

As illustrated in Figure 5, an intriguing acoustic phenomenon manifests during the interval between the commencement and conclusion of alcoholic fermentation. This phenomenon is closely associated with the dynamics and generation of CO₂ during the process, which are a consequence of the conversion of sugars into ethanol. In the initial stages of the process, the hydrophone demonstrates minimal acceleration values, as ethanol and CO₂ are not yet formed. As the conversion of sugars into ethanol begins, a marked surge in CO₂ production ensues until a maximum value is reached, whereby the rate of alcohol and CO₂ production stabilizes. The amplitude of these values is associated with the greater generation of CO₂ and the formation of CO₂ bubbles that impact against the hydrophone, thereby emitting a greater number of acceleration signals due to the greater energy of the process and the diameter of the bubbles.

As the fermentation process continues, its dynamics undergo a transformation. Following the peak in CO₂ generation, the intensity of these levels declines as a reduction in sugars and an increase in alcohol concentration occur. This causes less CO₂ to be produced until the process practically stops completely. This results in a reduction in both the levels of the acoustic signal generated and its dispersion.

As illustrated in Figure 5, throughout the fermentation process, the amplitude values of the acoustic signal are observed to be highly dispersed over short time intervals. This dispersion of values may be attributed to the fact that the diameter of the bubbles impacting the hydrophone can vary significantly. The formation of CO₂ throughout the fermenting liquid results in bubbles of varying diameters reaching the hydrophone and forming near it. The formation of bubbles with smaller diameters leads to the generation of low signal levels, while bubbles that arise at greater depths in the tank impact and coalesce with others on their way up; this forms bubbles with greater diameters that also impact against the hydrophone, producing signals with greater amplitudes.

Figure 6 illustrates the process dynamics curves derived from the acoustic acceleration signal captured by the hydrophone. The process dynamics represent a sample of the chemical reaction occurring within the fermentation tank during the conversion of sugars into ethanol. Following the filtration of the acoustic signals, the curves representing the dynamics of the fermentation process are obtained.

As illustrated in Figure 6, the acceleration signals acquired with the hydrophones are capable of representing the dynamics of the fermentation process in their entirety. This dynamic is associated with changes in the density of the medium. The densities at each instant of time depicted in Figure 6 are due to the integration of the acceleration signal acquired by the hydrophone Equation (1) over time. This integration corresponds to the speed of the reaction, which is associated with the conversion of sugars into ethanol.

To validate the results obtained, the calculated density values were compared with the real values recorded in the winery by the winemaker. Figure 7 illustrates the correlation between the real values and those obtained from the acoustic signal captured by the hydrophone.

As illustrated in Figure 7, the coefficient of determination (R²) demonstrates an almost perfect linear relationship, and the coefficients of determination range from 95 to 99%. The higher R² value indicates a better fit between the data measured by the winemaker and experimental data, showing that the use of the acoustic emission signal to estimate the density of the liquid during the alcoholic fermentation process is a reliable indicator of the density during fermentation and can be used for future predictions.

Table 2. Estimated and experimental densities.

Tank 24		Tank 25		Tank 26
Experimental (g/L)	Predicted (g/L)	Experimental (g/L)	Predicted (g/L)	Experimental (g/L)	Predicted (g/L)
1095	1084	1098	1092	1092	1084
1067	1067	1070	1070	1075	1067
1040	1047	1036	1039	1041	1049
1018	1026	1015	1021	1016	1033
1005	1010	1000	1006	1002	1014
997	999	997	995	997	1002
995	991	992	990	995	994

shows the actual density values measured in the winery and those obtained from the acoustic signal for the same time point during the alcoholic fermentation process.

4. Conclusions

In summary, in this study, a novel method based on the analysis of the acoustic signals generated within the liquid was developed to calculate the evolution of density in real time during the alcoholic fermentation process. This demonstrates the efficacy of the acoustic emission technique in estimating the chemical properties of must during the wine fermentation process. The technique used to integrate the acoustic signal representing the dynamics of fermentation can be linearly correlated with density as a chemical property of the must undergoing fermentation, exhibiting determination coefficient values exceeding 95%. The simplicity and sensitivity of the acoustic emission method make it a valid method for implementation in wineries when aiming to acquire the chemical properties of the must undergoing fermentation in real time. The implementation of the results in wineries will serve to significantly reduce operating times, alleviating manual work and providing winemakers with a reliable and quantitative tool for interventions such as the addition of nutrients or temperature adjustments.