Abstract
In order to explore the segmented relationship between the segmented rating results of Baijiu and the segmented characteristics of alcohol content and cumulative flow in the distillation process and to verify whether different liquor categories can achieve rating through the segmented characteristics of alcohol content and cumulative flow, in view of this, combined with the Industrial Internet of Things (IIoT) technology and online detection and analysis of alcohol and cumulative traffic, its accuracy can reach ±0.5%, ±1%, with the integration of Baijiu categories, sources, liquor characteristic conditions, and other multisource data, to achieve Baijiu segmented rating data reconstruction, the use of standard error and standard deviation as evaluation indicators, quantification of distilled liquor alcohol, and cumulative flow segmented characteristics to form a liquor rating strategy, so as to use the Arduino platform control motor to achieve automatic grading of Baijiu. Experiments show that the relative error between automatic rating and manual rating is less than 10%, which shows that automatic rating can be better applied to the actual brewing process. It provides a solution for the digitization and standardization of Baijiu grading.
1. Introduction
The “Baijiu rating” in traditional Baijiu brewing refers to the prediction and judgment of Baijiu by the grading experts based on the previous data such as liquor categories and fermentation parameters, combined with the size of the hops at the time of flow, the dissipation rate of the hops, the measurement of the flow time, and the taste of the Baijiu, so as to rate the liquor to distinguish different liquor segments [1, 2]. This method relies too much on the experience of the Baijiu grading expert and does not have a unified quantitative index, resulting in unstable product quality.
In terms of hops recognition, Tian Zichen et al. used convolutional neural network to extract the multiscale feature algorithm of hops contour, and its classification accuracy was as high as 97.59% [3]. In terms of substance content detection, Fan Mingming et al. used near-infrared spectroscopy to detect the content of hexyl caproate and other substances online and established an alcohol prediction model, which can meet the grading standards of Baijiu, thereby simulating the characteristics of artificial grading judgment [4]. At the Baijiu brewing site, fog and vibration are inevitable, which easily affects the accuracy of hop identification and material content measurement, and the cost is so high that it is difficult to apply. The application of IIoT technology to Baijiu brewing will help enterprises digitally transform and improve enterprise production efficiency and will also create new opportunities to meet the needs of differentiated enterprises [5, 6]. Embedding IIoT technology in the manufacturing industry has significantly increased the level of digitalization of enterprises and shortened production cycles [7].
In order to integrate the artificial grading characteristics and quantify the grading criteria of Baijiu, the Baijiu grading system based on IIoT technology is a real-time analysis of the temperature, alcohol content, and cumulative flow data of distilled liquor [8]; combined with the segmented results of Baijiu grading experts looking at hops, tasting tastes, and counting time, the alcohol content and cumulative flow are associated, so as to establish the corresponding relationship characteristics between artificial grading and alcohol content and cumulative flow.
In this experiment, the study object was based on the strong aromatic Chinese Baijiu type and the mild aromatic Chinese Baijiu, and the sources of the mash were ZL1-ZL8 (the proportion of raw materials was α, the fermentation cycle was N days) and CJ1-CJ8 (the proportion of raw materials was β, the fermentation cycle was P day), the raw materials added to the bottom pot were QS, HS, FJY, and WJ, and according to the different raw materials and volume added to the bottom pot, the grading data of Baijiu enterprises was reconstructed [9], and it was divided into different process types.
2. Methodology
Integrating multisource data [10] such as Baijiu categories and fermentation parameters and analyzing the relationship between the segmented characteristics of alcohol content and cumulative flow and the manual rating results under different process types, Baijiu grading experts still use hops as segmented rating criteria and alcohol content and cumulative flow during distillation in the Baijiu production process as automatic rating characteristic indicators.
When the steam is distilled out of the pot and the steam is condensed into a liquid through the steam pipe to the condenser, the Baijiu grading system samples by bypass [11], and the L-Dens2300 density sensor measures the temperature and density of the Baijiu online, and the conversion and compensation of the density, temperature, and alcohol are completed by the host computer program algorithm [12]. The cumulative flow rate is measured directly by a turbine flow sensor installed in the distillate mains, and the flow value is filtered by the program [13]. The Baijiu grading system uses IIoT technology to automatically read the type of Baijiu brewing process and analyzes the alcohol content size, the change trend before and after the alcohol content, and the accumulated flow of the Baijiu segment when grading the Baijiu under different process types by the grading experts in real time. In order to ensure the accuracy of statistical results, five groups of manual rating and automatic rating data under the same process type are counted, and the average value is taken after removing the extremum value to remove accidental errors [14], and the segmented characteristics of liquor are formed by analyzing the data.
2.1. High-Precision Online Monitoring Solutions
Baijiu grading equipment is connected in the middle of the condenser and the wine plate classifier system containing sensor online detection, data analysis, and processing, and program algorithm control; Baijiu is from the condenser through the servo drive to the wine plate in each grading pipeline. Accurate measurement of alcohol content online should avoid the formation of bubbles in the U-shaped tube in the density sensor to affect the density, thereby affecting the accuracy of alcohol content [15]. In actual production, there is a certain pressure in the condenser, and bubbles will inevitably be generated to affect the measurement results. In order to avoid bubbles entering the U-shaped pipe, the density sensor adopts the bypass small pipe connection mode; in order to ensure the real-time measurement, and to avoid the alcohol solution in the cache barrel to indulge in the front and rear sections of the liquor, the booster pump is added in the bypass system to establish a linear relationship [16] between the flow rate of the liquor and the operating speed of the vacuum pump, to avoid the pump running speed being too fast, bubbles are generated in the alcohol density sensor, or the pump running speed is too slow to affect the real-time detection of alcohol. Since the pressure generated by the condenser will cause the turbine flowmeter impeller to idle to cause inaccurate flow counting, an exhaust device is added to the cache barrel to make the pressure of the turbine flowmeter consistent before and after without liquid, so that the online monitoring accuracy of alcohol content and cumulative flow can reach ±0.5% and ±1%, respectively. The overall design is shown in Figure 1.
2.2. Grading Devices Based on Industrial Internet of Things (IIoT) Technology
The architecture of the Baijiu automatic rating system based on IIoT technology includes the perception layer, communication layer, and application layer [17]. The perception layer mainly includes L-Dens 2300 density sensor, flow sensor, near-infrared online sensor, etc. The communication layer uses Zigbee technology for wireless transmission, and at the same time, it is equipped with MODBUS protocol to connect to other devices [18]. The application layer is to use the industrial computer for human-computer interaction to understand the status of the Baijiu rating system in real time.
The Baijiu grading system is combined with IIoT technology, which is composed of four parts: host computer, input module, output module, and data exchange module. IIoT [19] is the use of miniature sensors and wireless networks to monitor equipment status data, in-depth analysis of the data, thereby forming a decision control actuator. The host computer is an industrial all-in-one machine with a touch display, on-board Intel SkylakeCorei5 processor, for the data analysis and processing, the actuator control. The host computer analyzes and processes the data, and interacts with the input module and PLC through the Internet of Things to read the type data of the brewing process. At the same time, the Arduino platform [20] is used to control the servo motor, so as to drive the liquor tube to achieve automatic grading of Baijiu. The input module includes liquid turbine flow sensor, L-Dens2300 online density sensor, etc. The L-Dens2300 in-line density sensor uses the reliable oscillating U-tube principle for density measurement [21] and an integrated temperature sensor for reading the temperature of the measured solution. Density calculation can be done using the following formula to calculate density from oscillation period and temperature [22]:where Dt is the density measured at temperature, the unit is [kg/m3], and it has been corrected for the temperature effect of the sensor, P is the oscillation period, the unit is [μs], T is the temperature, the unit is [°C], and A, B, C, and D are the calibration normal number.
According to GB5009.225-2016 alcohol aqueous solution density and alcohol content (ethanol content) control table (20°C) and alcohol meter temperature and 20°C alcohol content (ethanol content) conversion table, MATLAB is used to establish a density and alcohol content relationship fitting model:where ρ represents the standard density of aqueous alcohol solution at 20°C, A, B, C, and m are the coefficients, i = n = 5 is the constant, p indicates the mass concentration of alcohol, and t represents the temperature of the aqueous alcohol solution. The system uses a program to convert the measured density data into alcohol in real time for transmission to the host computer interface. The turbine flowmeter transmits the flow data to the upper unit state interface through RS485 to display and statistically record the cumulative flow of different liquor segments [23]. The output module contains a vacuum pump and a servo motor, and when the flowmeter recognizes that there is a flow of liquor, it controls the opening of the vacuum pump and adjusts the flow rate accordingly. By analyzing the manual rating data, the host computer forms a liquor segmentation strategy and sends it to the servo drive in the form of RS485 to perform the grading operation. A hardware block diagram of the hierarchical device is shown in Figure 2.
2.3. Multisource Data Preprocessing for Data Reconstruction
In the actual Baijiu brewing production process, on the basis of retaining the structure, process flow, and operating procedures of its original brewing equipment, only with a small amount of transformation of the pipeline of the original manual grading equipment, the installation of automatic grading equipment, external signal acquisition circuit, control system, and actuator and other components, you can achieve online collection and control, and its actual installation effect is shown in Figure 3.
In order to clearly understand the influence of different processes on the grading results, the experiment prescreened the factor indicators affecting the grading results of liquor and reconstructed them into AB with multisource data such as liquor category, mash source, HS amount, FJY amount, and WJ amount. P16 process types are shown in Table 1. Using IIoT technology, Baijiu grading experts evaluate the operation configuration screen, collect manual rating data in real time, and synchronize the brewing characteristics data of the flow sensor, density sensor, and on-site PLC. The results of the grading of Baijiu according to different process types are shown in Figures 4–15. It can be concluded that the segmented characteristics of the A-H process type of strong aromatic Chinese Baijiu are as follows: the first and second stages of Baijiu can be graded according to the flow rate, the third stage of Baijiu can be graded according to the slope of the change in alcohol, and the fourth and fifth sections of Baijiu can be graded according to the size of the alcohol. The segmented characteristics of the I–P process type of mild aromatic Chinese Baijiu are as follows: the first stage of Baijiu can be graded according to flow, the second stage of Baijiu can be graded according to the slope of alcohol change, and the third and fourth stages of Baijiu can be graded according to the size of alcohol.
3. Results and Discussion
Baijiu grading experts take hops, etc. as the Baijiu segmentation evaluation standard, Baijiu automatic grading to set the alcohol content, and cumulative flow as the basis for Baijiu segmentation; each process takes five groups of Baijiu grading data to the extreme value and then takes the average value, to the standard error, standard deviation as the alcohol degree, cumulative flow evaluation index [24, 25], to the absolute error, relative error as the automatic grading, and manual grading data error judgment [26].
3.1. Grading Strategy of Strong Aromatic Chinese Baijiu
As shown in Table 2, the segmented characteristics of the A-P process type of strong aromatic Chinese Baijiu are as follows: The cumulative flow of the first stage of liquor is stable at 3.5–4.4 L, which can be seen from the standard deviation and standard error, the cumulative flow data of the first period of liquor grading at the time of liquor has less fluctuation, the difference between the sample mean and the overall mean is small, and the relative error between the automatic grading of liquor and the manual grading is small; then the average value of the accumulated flow can represent the automatic segmentation index of the first period of Baijiu. It can be seen from the standard deviation and standard error of the cumulative flow rate of Baijiu in the second stage that the difference between the cumulative flow range of liquor in the second stage and the sample mean and the overall mean is large, and it is impossible to represent the sample mean through the overall mean, which does not have the characteristics of unified segmentation, but the automatic grading data of Baijiu and the manual grading data have relatively small errors, indicating that the second stage of Baijiu can be automatically graded according to the set flow value according to different process types; when the third stage of Baijiu is transferred to the fourth section of Baijiu, the alcohol content drops sharply and produces an inflection point; at this time, the relative error with the alcohol slope setting value is small, so it can be seen that the alcohol slope change can be used as the automatic segmentation index of the third stage of Baijiu; when the fourth stage of Baijiu is transferred to the fifth section of Baijiu, the error of alcohol standard error and standard deviation, automatic segmentation, and artificial segmentation relative error also indicates that the average alcohol content can represent the segmented index, so that the alcohol content can be used as the automatic grading index of the fourth stage of Baijiu.
3.2. Grading Strategy of Mild Aromatic Chinese Baijiu
As shown in Table 3, the unified segmental characteristics of the first stage of the I–P process type of mild aromatic Chinese Baijiu are consistent with the segmented indicators of strong aromatic Chinese Baijiu. The inflection point of the alcohol content is generated when the second stage of Baijiu is converted to the third stage of Baijiu, and the deviation of alcohol degree is relatively small in the average value of automatic segmentation and artificial segmentation data, indicating that the alcohol slope change can be unified as the second segmentation index of Baijiu. Similarly, according to the alcohol content, it can be used as the indicator of the third stage of Baijiu automatically turning into the fourth stage of Baijiu.
3.3. Discussion
This paper mainly introduces the automatic rating method of strong aromatic and mild aromatic Chinese Baijiu, uses the powerful information fusion ability of the Industrial Internet of Things, establishes a rating model based on alcohol content and cumulative flow, and realizes the automatic grading of Baijiu through the Control Motor of Arduino Platform, which improves the stability of Baijiu rating and also provides an idea for the automatic grading method of Maotai-flavor liquors and Fengxiang-type liquor. However, due to the failure to detect the physical and chemical properties of Baijiu, the automatic grading accuracy relies heavily on the manual experience grading results, and the next step will be to detect the content of acid, ester, and alcohol components of Baijiu and intelligently rate Baijiu in combination with neural networks.
4. Conclusion
This paper combines multisource data to reconstruct the segmented characteristics of Baijiu and obtains the factor data affecting Baijiu rating: Baijiu category, fermentation parameters, HS amount, FJY amount, WJ amount, and Baijiu manual rating data. Using the powerful information fusion capability of the Industrial Internet of Things, an automatic rating model based on alcohol content and cumulative traffic is established, and the data of automatic rating of Baijiu and manual rating results are compared, and the relative error is less than 10%, and the relative error between the automatic rating average and the manual rating average is less than 6%, indicating that the Baijiu automatic rating equipment based on IIoT technology can be better used in actual brewing.
Data Availability
The datasets used and/or analyzed during the current study are available from the corresponding author on reasonable request.
Conflicts of Interest
The authors declare that they have no conflicts of interest.
Acknowledgments
This paper was supported by Development and Industrialization of Intelligent Brewing Equipment for Baijiu (2020ZHCG0040), Research and Application Demonstration of Key Technologies of Baijiu Intelligent Brewing Equipment (2018GZDZX0045), and Research and Development and Industrialization Application of Intelligent Graded Baijiu Grading Equipment Based on Multi-Source Data Deep Learning (21ZYZFZDYF0021).