Black tea continuous fermentation drying equipment and control method

Abstract

The invention provides black tea continuous fermentation drying equipment and a control method, and relates to the technical field of tea processing, and the method comprises the steps: collecting related data based on a data collection module, preprocessing the collected related data based on a preprocessing module, and constructing an intelligent parameter prediction model; and performing optimal process parameter prediction on the black tea continuous fermentation drying equipment based on the parameter intelligent prediction model, and controlling the black tea continuous fermentation drying equipment based on the predicted optimal process parameters. According to the method, the optimal process is customized for different tea leaves through the intelligent prediction model, and high-precision closed-loop control execution is combined, so that precise and intelligent production of black tea fermentation and drying is realized, and the uniformity and stability of tea leaf quality are remarkably improved; through continuous automatic operation, the efficiency is greatly improved, and the energy consumption and the labor cost are reduced; and modern upgrading of black tea processing from experience-driven processing to data-driven processing is promoted.

Description

Technical Field

[0001] This invention relates to the field of tea processing technology, and in particular to a continuous fermentation and drying equipment and control method for black tea. Background Technology

[0002] Black tea is the most consumed tea category among the six major tea categories in terms of both volume and area. Its English name, Black tea, originates from the red color of both the tea liquor and the infused leaves after brewing. As a fully fermented tea, black tea is made from suitable new buds and leaves of tea trees, refined through a series of processes including withering, rolling, fermentation, and drying. During processing, black tea undergoes a chemical reaction centered on the enzymatic oxidation of tea polyphenols, leading to significant changes in the composition of the fresh leaves: the polyphenol content decreases by more than 90%, while new components such as theaflavins and thearubigins, as well as aroma substances, are generated. These changes give black tea its characteristic red liquor, red leaves, and sweet, mellow flavor.

[0003] In the processing of black tea, fermentation and drying are key steps that directly affect the quality of the tea. Traditional processing methods have many drawbacks, such as separating fermentation and drying equipment, requiring frequent manual transfer of tea leaves, which is not only inefficient but also prone to tea contamination; during fermentation, the tea leaves pile up and are difficult to turn evenly, affecting the fermentation effect; and the inaccurate control of parameters such as temperature and humidity makes the quality of black tea unstable.

[0004] Existing black tea processing equipment, such as some fermentation machines, can only achieve simple tea leaf movement or turning, which cannot meet the needs of continuous production, and has shortcomings in parameter control and equipment integration. Therefore, it is of great significance to develop a device that can achieve continuous fermentation and drying, with all components working together and parameters precisely controllable.

[0005] In recent years, various mechanized processes in black tea production have been widely applied. However, as a core element determining the color, aroma, and flavor of black tea, the fermentation process still largely follows traditional methods, employing direct fermentation. Currently, black tea fermentation in tea factories mostly takes place indoors, creating a suitable fermentation environment by controlling temperature, humidity, and oxygen levels. Some small and micro-enterprises and cooperatives still use more traditional fermentation methods, such as fermentation indoors with damp cloths, which have relatively limited precision in temperature and humidity control. During tea fermentation, operators need to periodically turn the tea leaves to ensure all parts are in full contact with air. However, manual turning often results in insufficient uniformity, which not only reduces the overall uniformity of fermentation and affects the stability of the finished product quality but also prolongs the fermentation cycle, leading to low production efficiency. Furthermore, the process parameters of continuous fermentation and drying equipment for black tea are often determined manually, relying excessively on experience and posing significant risks.

[0006] Therefore, it is essential to design a continuous fermentation and drying equipment and control method for black tea. Summary of the Invention

[0007] In order to overcome the shortcomings of the prior art, the purpose of this invention is to provide a continuous fermentation and drying equipment and control method for black tea.

[0008] To achieve the above objectives, the present invention provides the following solution: This invention provides a continuous fermentation and drying apparatus for black tea, comprising: The data acquisition module is used to collect process parameters and tea leaf properties. The data preprocessing module, which communicates with the data acquisition module, is used to clean and fill in the acquired data; The parameter intelligent prediction module communicates with the data preprocessing module and is used to predict the optimal process parameters based on the intrinsic property data of the cleaned tea leaves. The equipment control module communicates with the parameter intelligent prediction module and is used to control the fermentation and drying equipment in real time based on the predicted optimal process parameters. The drum fermentation and drying unit, the temperature, speed, and time control unit, and the humidity control unit are all electrically connected to the equipment control module and are used to execute the instructions issued by the equipment control module.

[0009] Preferably, the data acquisition module includes: Multiple temperature sensors are installed at the front, middle, and rear positions inside the drum; A humidity sensor is located inside the drum. The speed sensor is installed on the rotating shaft of the drum; A power monitoring unit is used to collect the real-time power of the heating and ventilation devices. A human-computer interaction interface is used to input or obtain information about tea types and harvesting seasons; A moisture meter is used to collect the initial moisture content of tea leaves. An industrial camera used to capture image color information of tea leaves.

[0010] Preferably, the drum fermentation and drying unit includes: The food-grade stainless steel drum has spiral stirring blades welded to its inner wall and is set at an angle. A drive unit, which is connected to the food-grade stainless steel roller, is used to drive the roller to rotate; The heating device includes a heating jacket disposed on the outer wall of the drum and a heating tube disposed inside the drum.

[0011] The present invention also provides a control method for a continuous fermentation and drying device for black tea, applied to the aforementioned continuous fermentation and drying device for black tea, comprising: Step 1: Collect relevant data using the data acquisition module; Step 2: Preprocess the collected relevant data using the preprocessing module; Step 3: Construct a parameter intelligent prediction model and predict the optimal process parameters for the continuous fermentation and drying equipment for black tea based on the parameter intelligent prediction model; Step 4: Control the continuous fermentation and drying equipment for black tea based on the predicted optimal process parameters.

[0012] Preferably, in step 1, relevant data is collected based on the data acquisition module, specifically as follows: The relevant data includes process parameters and tea-based property parameters; The process parameters include temperature, humidity, drum speed, heating device power, ventilation device status, and running time. The intrinsic properties of the tea include tea variety, harvesting season, initial moisture content, and image and color information.

[0013] Preferably, in step 2, the collected relevant data is preprocessed based on the preprocessing module, specifically as follows: Based on temperature, humidity, drum speed, and heating device power data, primary filtering and noise reduction, outlier detection and removal, and missing data filling are performed sequentially. Rule-based verification is performed based on tea type and harvesting season.

[0014] Preferably, in step 3, a parameter intelligent prediction model is constructed, and the optimal process parameters for the continuous fermentation and drying equipment of black tea are predicted based on the parameter intelligent prediction model, specifically as follows: A parameter intelligent prediction model is constructed based on a hybrid CNN-LSTM attention mechanism model optimized by the dung beetle optimization algorithm; A training dataset was generated based on historical production batch data. The parameter intelligent prediction model is trained based on the training dataset to obtain the trained parameter intelligent prediction model. The relevant data to be detected is input into the trained intelligent prediction model of parameters to obtain the optimal process parameters.

[0015] According to specific embodiments provided by the present invention, the present invention discloses the following technical effects: This invention provides a continuous fermentation and drying equipment and control method for black tea. The equipment includes a data acquisition module for collecting process parameters and tea leaf properties; a data preprocessing module, communicatively connected to the data acquisition module, for cleaning and filling the acquired data; a parameter intelligent prediction module, also communicatively connected to the data preprocessing module, for predicting optimal process parameters based on the cleaned tea leaf properties; and an equipment control module, communicatively connected to the parameter intelligent prediction module, for real-time control of the fermentation and drying equipment based on the predicted optimal process parameters. A drum fermentation and drying unit, a temperature, speed, and time control unit, and a humidity control unit are all electrically connected to the equipment control module to execute commands issued by the equipment control module. The method includes: collecting relevant data using the data acquisition module; preprocessing the collected data using the preprocessing module; constructing a parameter intelligent prediction model; predicting the optimal process parameters for the continuous fermentation and drying equipment for black tea based on the parameter intelligent prediction model; and controlling the continuous fermentation and drying equipment for black tea based on the predicted optimal process parameters. Compared with traditional technologies, this invention has the following significant advantages: 1. This invention achieves intelligent and precise control, significantly improving tea quality. By introducing a hybrid CNN-LSTM attention mechanism model optimized by the dung beetle algorithm, it can accurately predict and output the optimal set of process parameters based on the inherent attributes of tea, such as type, season, and moisture content. This model overcomes the arbitrariness of traditional methods relying on human experience and ensures the quality of input data through adaptive DBSCAN-LOF and random forest algorithms. During the execution phase, the PLC uses a fuzzy PID algorithm to perform closed-loop control of temperature, humidity, and rotation speed, ensuring that the entire fermentation and drying process always runs on the optimal trajectory. This guarantees uniform and sufficient fermentation of the tea, stable theaflavins and thearubigins content in the finished tea, pure aroma, and good quality consistency. 2. Significantly improves production efficiency and reduces overall costs. The equipment adopts an integrated drum design, realizing continuous and automated production of fermentation and drying processes. This avoids the manpower and time wasted on frequent tea transfers in traditional processes, resulting in a significant improvement in production efficiency. At the same time, intelligent control avoids batch defects caused by improper process parameters, reducing the waste of raw materials and energy. Precise temperature and humidity control also effectively shortens the total process time, further reducing the energy consumption and labor costs per unit product. 3. Ensuring production hygiene and safety, and promoting industrial upgrading: The main body of the equipment is made of food-grade stainless steel, with a reasonable structural design that is easy to clean and maintain. It meets modern food safety production standards. The entire system transforms the "experience" of master craftsmen into a replicable and scalable "data model," reducing reliance on skilled workers and providing reliable technical equipment support for promoting the standardization, large-scale operation, and intelligent transformation and upgrading of black tea processing. Attached Figure Description

[0016] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0017] Figure 1 This is a schematic diagram of the method flow of the present invention; Figure 2 This is a schematic diagram illustrating the principle of DBSCAN. Figure 3 A schematic diagram of the algorithm optimization process for dung beetles; Figure 4 This is a schematic diagram of the hybrid CNN-LSTM attention mechanism model structure. Detailed Implementation

[0018] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0019] The purpose of this invention is to provide a continuous fermentation and drying equipment and control method for black tea. By using an intelligent prediction model to tailor the optimal process for different teas and combining it with high-precision closed-loop control, the invention achieves precise and intelligent production of black tea fermentation and drying, significantly improving the uniformity and stability of tea quality. Through continuous automated operation, it greatly improves efficiency, reduces energy consumption and labor costs, and promotes the modernization upgrade of black tea processing from "experience-driven" to "data-driven".

[0020] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0021] This invention provides a continuous fermentation and drying apparatus for black tea, comprising: The data acquisition module is used to collect process parameters and tea leaf properties. The data preprocessing module, which communicates with the data acquisition module, is used to clean and fill in the acquired data; The parameter intelligent prediction module communicates with the data preprocessing module and is used to predict the optimal process parameters based on the intrinsic property data of the cleaned tea leaves. The equipment control module communicates with the parameter intelligent prediction module and is used to control the fermentation and drying equipment in real time based on the predicted optimal process parameters. The drum fermentation and drying unit, the temperature, speed, and time control unit, and the humidity control unit are all electrically connected to the equipment control module and are used to execute the instructions issued by the equipment control module.

[0022] The data acquisition module includes: Multiple temperature sensors, using Pt100 platinum resistance temperature sensors, are used. There are 3-5 sensors installed at the front, middle, and rear positions inside the drum, as well as near the heating jacket and heating tube. The sensor accuracy is ±0.5℃, and the response time is ≤10s. They are connected to the controller through shielded wires. The humidity sensor is a capacitive humidity sensor, installed in the middle of the drum. It has a measurement range of 20%-90%RH, an accuracy of ±3%RH, an operating temperature range of 0-60℃, and outputs a 4-20mA standard signal. The speed sensor, a Hall effect type, is installed on the side of the roller's rotating shaft and works in conjunction with a magnet on the shaft. The measurement range is 0-50 r / min, and the accuracy is [insert accuracy here]. The output signal is a pulse signal; A power monitoring unit is used to collect the real-time power of the heating and ventilation devices. A human-computer interaction interface is used to input or obtain information about tea types and harvesting seasons; A moisture meter is used to collect the initial moisture content of tea leaves. An industrial camera used to capture image color information of tea leaves.

[0023] The drum fermentation and drying unit includes: Food-grade stainless steel drum, with spiral stirring blades welded to its inner wall and set at an angle, is made of food-grade 304 stainless steel. It is 6-10m long and 0.8-1.5m in diameter, and is set at an angle of [insert angle here]. The inner wall of the drum is uniformly welded with spiral stirring blades, the blade height is 8-15cm and the pitch is 0.5-1m. The stirring blades are fully welded to the inner wall of the drum, and the welding strength is not less than 50MPa. A feed hopper is welded to the feed end of the drum. The feed hopper is funnel-shaped, with an upper diameter of 0.5-0.8m and a lower diameter that matches the feed end of the drum. The discharge end is provided with a discharge port, and an adjustable gate is installed at the discharge port. The gate is made of stainless steel plate with a thickness of 3-5mm and is controlled to open and close by an electric push rod. The drive unit, which is connected to the food-grade stainless steel roller and is used to drive the roller to rotate, includes a drive motor and a reducer. The drive motor is a three-phase asynchronous motor with a power of 5.5-11kW and a rated speed of 1450r / min. The reducer is a hardened cylindrical gear reducer with a transmission ratio of 10:1-30:1. The motor output shaft is connected to the reducer input shaft through a coupling. The reducer output shaft is connected to the roller's rotating shaft through a chain drive. The chain is a single-row roller chain, model 16A-1. The heating device consists of a heating jacket on the outer wall of the drum and heating tubes inside the drum. The heating jacket is made of aluminum silicate insulation cotton wrapped around the electric heating wire. The thickness of the heating jacket is 50-80mm. The electric heating wire is made of nickel-chromium alloy, with a diameter of 2-3mm and a power density of [missing information]. The heating tubes are seamless stainless steel tubes with a diameter of 15-20mm, evenly distributed inside the drum, numbering 4-8 tubes, arranged along the drum axis. The heating tubes are also equipped with nickel-chromium alloy electric heating wires. The real-time working power of the heating jacket and internal heating tubes can be directly read through the analog output module of the PLC to reflect the heating status of the system.

[0024] The temperature, speed, and time control unit includes: Controller: The controller is an S7-200 SMART PLC controller, equipped with a 16-point input / 16-point output expansion module. It receives signals from temperature and humidity sensors through the analog input module, controls the output power of the heating device, humidifier and dehumidifier through the analog output module, and controls the start and stop of each motor through the digital output module. Timer: Integrated into the PLC controller, with a timing range of 0-999 minutes and a timing accuracy of ±1 second. The time parameters for the fermentation and drying stages can be set via a touch screen.

[0025] The humidity control unit includes: Humidifier: An ultrasonic humidifier with a humidification capacity of 1-3 kg / h is used. It is installed on the outside of the feed end of the drum and is connected to the inside of the drum through a PVC pipe with a diameter of 20-30 mm. A solenoid valve is installed on the pipe and the start and stop are controlled by the controller.

[0026] This invention provides an embodiment in which the high-efficiency continuous fermentation and drying equipment for black tea has a drum length of 8m, a diameter of 1.2m, and an inclination angle of [missing information]. The spiral stirring blades are 12cm high with a pitch of 0.8m. The drive motor has a power of 7.5kW and a reducer ratio of 20:1. The heating jacket is 60mm thick, and the electric heating wire has a power density of [missing information]. The heating element consists of 6 heating tubes with a diameter of 18mm, a tilting motor with a power of 1.1kW, a rotating shaft with a diameter of 40mm, and 12 tilting claws with a length of 15cm each. The device contains four temperature sensors, a humidity sensor with a measurement range of 20%-90%RH, a humidifier with a humidification capacity of 2 kg / h, a dehumidifier with a dehumidification capacity of 1.5 kg / h, and a ventilation system with a fan power of 1.1 kW and an airflow of [missing information]. ; The conveyor belt of the weighing machine is 500mm wide and has a belt speed of 0.5m / s; the weighing hopper has a capacity of 80L; the S-type tension and compression sensor has a range of 0-100kg; the electric push rod has a rated thrust of 800N and a stroke of 120mm.

[0027] When the equipment is working, the fermentation temperature is set via the PLC controller. Fermentation time is 3-4 hours, drying temperature is The drying time is 1-2 hours, and the humidity is controlled at 60%-70%RH. After being weighed by the lifting and weighing machine, the fresh tea leaves enter the drum. Driven by the drive device, the drum rotates, and the tea leaves are turned over by the spiral blades and turning claws and move towards the discharge end. At the same time, the heating device, humidity control unit and ventilation device work according to the set parameters to complete the fermentation and drying process, and finally discharged from the discharge port.

[0028] like Figure 1 As shown, the present invention also provides a control method for a continuous fermentation and drying device for black tea, comprising: Step 1: Collect relevant data using the data acquisition module; Step 2: Preprocess the collected relevant data using the preprocessing module; Step 3: Construct a parameter intelligent prediction model and predict the optimal process parameters for the continuous fermentation and drying equipment for black tea based on the parameter intelligent prediction model; Step 4: Control the continuous fermentation and drying equipment for black tea based on the predicted optimal process parameters.

[0029] In step 1, relevant data is collected based on the data acquisition module, specifically as follows: The relevant data includes process parameters and tea-based property parameters; The process parameters include temperature, humidity, drum speed, heating device power, ventilation device status, and running time; wherein, the ventilation device status is assessed by monitoring the start / stop status and operating frequency of the fan connected to the drum to evaluate the air circulation and oxygen supply inside the drum, and the running time is obtained by a timer. The tea's intrinsic attributes include tea variety, harvesting season, initial moisture content, and image and color information. The tea variety is entered by the operator via a human-machine interface (HMI) before production begins, or automatically obtained by binding it to the incoming batch using QR code / RFID technology, for example: Zhuye variety, Yunkang No. 10, etc.; the harvesting season is entered manually or automatically associated by the system, for example: pre-Qingming spring tea, summer tea, autumn tea; initial moisture content: an online near-infrared moisture meter is equipped at the feeding end to perform non-contact rapid detection of the rolled leaves before they enter the drum, obtaining their initial moisture content; image and color information: an industrial color camera is installed in the observation window or inside the drum to capture images of the tea leaves at regular intervals. Through image processing algorithms, the average color value of the tea surface (e.g., ...) is extracted. color space The values ​​(representing red-green hue) and texture features serve as visual quantitative indicators of the degree of fermentation.

[0030] In step 2, the collected relevant data is preprocessed using the preprocessing module, specifically as follows: Based on numerical process parameters (temperature, humidity, drum speed, heating device power data): This type of data is typical time series data, and its anomalies mainly manifest as instantaneous spikes, missing data, constant data, or anomalous distribution. A stratified cleaning strategy is adopted: 1. Primary filtering and noise reduction: For high-frequency sampled data (such as temperature), first use moving average filtering or Kalman filtering algorithms to smooth out random high-frequency noise and retain the true trend of data change; 2. Outlier detection and removal: Deep cleaning is performed using an adaptive DBSCAN-LOF combined algorithm; First, adaptive DBSCAN is used for clustering. This algorithm can adaptively determine parameters and divide data points into core points (normal operating area), boundary points and noise points (obvious isolated points). It can effectively identify and remove discrete anomalies (global anomalies) that are obviously deviated from the main data flow. Then, the LOF algorithm is used to identify local anomalies. Even within the same "normal" cluster identified by DBSCAN, there may be points with a density that is too low relative to their nearest neighbors, i.e., local anomalies. The LOF algorithm can accurately calculate the local anomaly factor of each core data point and effectively detect clustered anomalies (local anomalies) that are not easily detected due to slight sensor drift or minor process disturbances. These two steps can completely eliminate all kinds of outliers in numerical parameters; This section provides a detailed introduction to the adaptive DBSCAN-LOF combinatorial algorithm: First, let's introduce the DBSCAN algorithm: DBSCAN is a density-based clustering algorithm that is widely favored in many practical applications due to its simplicity, efficiency, and robustness to outliers. This method requires two input parameters to distinguish between high-density and low-density regions: (Neighborhood radius) and (Minimum number of points), see the principle of DBSCAN. Figure 2 As shown, let the sample set be... The density description of the DBSCAN algorithm is as follows: (1) Sample points Neighborhood: the neighborhood of data point p Neighborhood Included Data points within the distance range, i.e. ,in This represents the distance between points p and q; (2) Core point: If data point p is in The distance has greater than or equal to If the data point p is a core point, then the data point p is the core point, and the red circle represents the core point. distance; (3) Noise points: If data point p does not contain In the neighborhood Points, not including core points. If the neighborhood is considered, then the data point p is a noise point, and the red dashed circle represents the noise point. distance; (4) Boundary points: If data point p is in The neighborhood contains less than The point, and it is at any core point. If the data point p is within the neighborhood, then it is a boundary point. The boundary point r... Distance is represented by a dashed blue circle; (5) Density direct reach: If r is in q Within the neighborhood, and if q satisfies the core point condition, then data point r is directly reachable from point q by density. There is at least a minimum quantity within the range Points surround q, and r is located at one of these points. Similarly, point s is also directly accessible from point q. (6) Density attainable: For p and q, if there is still one core point sequence The points in the two sequences satisfy ,and Depend on If density is directly accessible, then p is said to be accessible by density q, and points s and q are also accessible by density, as are points q and s. (7) Density connectivity: If there exists a point s in the dataset, where p and q are relative to each other. and If the density of s is attainable, then the data point p is relative to and Density is connected to point q, and p and q are density-connected because p and q are density-reachable from point s; Next, the adaptive DBSCAN algorithm will be introduced: The adaptive DBSCAN algorithm is based on the DBSCAN algorithm. It optimizes key parameters using a parameter optimization strategy and generates candidate algorithms by leveraging the data distribution characteristics of the dataset itself. and The parameter list is generated, and the number of clusters is determined based on the selected parameters. When the number of clusters stabilizes, the minimum density threshold is used. and The parameters are used as the optimal parameters; Next, the principle of the LOF anomaly detection algorithm will be introduced: The LOF algorithm is an algorithm that determines the degree of data anomaly based on data density. First, based on a defined distance metric, the distance to the k-th point in the input data is determined. Then, the ratio of the density of neighboring points to the density of this point is calculated. This result is called the Local Outlier Factor (LOF) and is used to evaluate whether a sample point is anomaly. The specific definition is as follows: (1) The k-th distance (k-distance) of point p is the distance between points p and q in dataset D, denoted as k-distance(p), where k is a given positive integer. And it satisfies the following conditions: There are at least k points ,make Established; There are at most k-1 points ,make Established. (2) k-distance neighborhood of point p: the k-th distance neighborhood, denoted as neighborhood. It consists of points whose distance from point p is not greater than k, and is defined as follows: ; (3) The reachable distance of point p relative to point q is denoted as reach-distancek(p,q), which is the maximum value of k-distance(p) and the distance between point p and point q. The k-th reachable distance from point p to point q is defined as: ; (4) Locally accessible density The local reachability density of point p is equal to the reciprocal of the reachability distance of the average k nearest neighbor, defined as follows: ; In the formula, It is the set of k-neighborhood data of point p; (5) Local outlier The local outlier of point p is expressed as: ; This reflects the degree of anomaly of the current data point. Essentially, it compares the density of this sample point with the average density of surrounding points. If the comparison result is equal, the data point is considered normal; if it is greater than the average density of surrounding points, it may be an outlier. The larger the value, the higher the degree of anomaly of this data point.

[0031] 3. Imputing missing data: Outliers that have been removed and the original missing data need to be properly imputed to maintain the integrity of the dataset. Random forest regression algorithm is used for imputation. Principle: Random forest is an ensemble learning algorithm that constructs multiple decision trees and combines their predictions. During imputation, the feature column containing missing values ​​is used as the prediction target (label), while other complete and related features (such as temperature and humidity at the same time, sensor readings at other locations, etc.) are used as input features to train the model, thereby predicting the missing values. It can capture complex nonlinear relationships between variables, the imputed values ​​more closely match the true distribution of the data, and it is not sensitive to the missing parts of the features, exhibiting strong robustness.

[0032] Cleaning of tea attribute data (type, season): This type of data is categorized or labeled data, and its anomalies mainly manifest as incorrect values ​​or inconsistencies with the process logic (for example, a certain variety of tea is processed using a typical process for another variety). The main cleaning methods are: Rule verification: Establish a knowledge base that corresponds tea varieties, seasons and reasonable process parameter ranges. When the entered attribute data does not match the real-time collected process data in a serious way, the system will issue an alarm to prompt the operator to check the batch information.

[0033] Image data cleaning: Images captured by industrial cameras may suffer from quality degradation due to changes in lighting, lens contamination, or steam interference. Abnormal image removal is performed by automatically filtering out excessively blurry or overly dark / bright invalid images by calculating image sharpness (such as Laplacian variance) and mean brightness. For the retained valid images, algorithms such as histogram equalization or homomorphic filtering are used to enhance contrast and mitigate the impact of uneven lighting, providing a high-quality image foundation for subsequent feature extraction.

[0034] In step 3, a parameter intelligent prediction model is constructed, and the optimal process parameters for the continuous fermentation and drying equipment of black tea are predicted based on the parameter intelligent prediction model. Specifically: A parameter intelligent prediction model is constructed based on a hybrid CNN-LSTM attention mechanism model optimized by the dung beetle optimization algorithm; A training dataset was generated based on historical production batch data. The parameter intelligent prediction model is trained based on the training dataset to obtain the trained parameter intelligent prediction model. The relevant data to be detected is input into the trained intelligent prediction model of parameters to obtain the optimal process parameters.

[0035] A parameter intelligent prediction model is constructed based on a hybrid CNN-LSTM attention mechanism model optimized by the dung beetle optimization algorithm, specifically as follows: First, let's introduce the Dung Beetle Optimization Algorithm (DBO). This algorithm is a global exploration and local exploitation optimization algorithm designed based on the dung beetle's ball-rolling, dancing, foraging, stealing, and reproductive behaviors. Dung beetles feed on animal feces, rolling it into balls. These dung balls not only serve as eggs and a means of raising offspring but also as food for the dung beetles. Its workflow diagram is shown below. Figure 3 As shown.

[0036] 1. Dung beetle rolling a ball During the rolling process, if there are no obstacles, the dung beetle will use the sun for navigation, and its position on the rolling ball will be updated, which can be represented as: ; Where t represents the current iteration number, This represents the position information of the i-th dung beetle at the t-th iteration. The constant value representing the deflection coefficient, b is a constant between (0, 1), and α is a natural coefficient of -1 or 1. Indicates the worst position globally. Used to simulate changes in light intensity; 2. Dung beetle reproduction Dung beetles roll their dung balls to a safe place, providing a secure environment for their offspring. The formula for the dung beetle's egg-laying area is: ; Indicates the optimal position for the current population. and These represent the lower and upper limits of the spawning area, respectively. , Let Lb represent the maximum number of iterations, and Ub represent the lower and upper bounds of the optimization problem, respectively. 3. Dung beetles foraging The optimal foraging area boundary for the dung beetle is defined as follows: ; in, Indicates the globally optimal position. and These represent the lower and upper limits of the optimal foraging area, respectively; 4. Dung beetle stealing Within the population, dung beetles also steal food from other dung beetles. During the iteration process, the formula for the thief's location information is: ; in, Let g represent the location information of the i-th th th th th iteration, g is a random vector of size 1×D that follows a normal distribution, and S represents a constant; 5. Population segmentation In a population, dung beetles are divided into different roles in a ratio of 6:6:7:11. Taking 30 individuals as an example, this invention assigns 6 dung beetles to ball-rolling behavior, 6 dung beetles to reproductive behavior, 7 dung beetles to forage for food, and 11 dung beetles to steal.

[0037] The following section introduces the hybrid CNN-LSTM attention mechanism model, whose structural diagram is shown below. Figure 4 As shown, the hybrid CNN-LSTM attention mechanism model consists of CNN, LSTM, Dropout, attention mechanism, and fully connected layers; 1. CNN network: CNN network is a deep learning model that is particularly suitable for processing grid-like data. Its structure mainly includes convolutional layers, pooling layers, and fully connected layers. Convolutional layers are mainly used to extract local features of the network, pooling layers are used to compress feature information and simplify the network's calculations, and finally, fully connected layers flatten all the processed effective features and output the final calculation result. ; In the formula, "dot" represents the dot product of two vectors; W conv Represents the weights of the convolutional kernel in a convolutional neural network; 2. LSTM Network: LSTM is a special type of RNN (Recurrent Neural Network). A single LSTM unit includes a forget gate, an input gate, and an output gate. The forget gate determines whether the input information can be forgotten; the input gate determines the storage of information and the updating of the state; the output gate, together with the updated state of the unit, determines the final output state and output value. Assuming an LSTM contains H hidden layers, where the memory unit, input gate, forget gate, output gate, and candidate state are represented by... The LSTM update process is as follows: ; In the formula, LSTM is an iterative step of an LSTM network; LSTM_Cell is the computation process of a cell; 3. Attention Mechanism: In the attention mechanism, A is defined as the attention weight; Hatt is the weight output layer. The attention calculation process is as follows: ; In the formula, softmax represents the SoftMax function; This represents the sum of weights; 4. Fully Connected Layer: The final output is achieved through a fully connected layer, defined as follows: This represents the weights and biases of the fully connected layer, and the final output is: .

[0038] The following is a detailed introduction to the overall model: A smart prediction model is constructed that integrates deep learning and optimization algorithms to predict a set of optimal fermentation and drying process parameters for each batch of tea before it is put into production, based solely on its inherent properties. The core of this model is a hybrid CNN-LSTM attention mechanism model optimized by the dung beetle optimization algorithm. Model Input Layer: The model input is a static feature vector, consisting entirely of the inherent properties of tea leaves: Tea variety coding: for example, Zhuye variety is coded as 1, Yunkang 10 is coded as 2, etc. (using One-hot coding); Harvesting season coding: for example, pre-Qingming spring tea is coded as 1, summer tea is coded as 2, autumn tea is coded as 3 (using One-hot coding); Initial moisture content: a continuous numerical feature representing the moisture content of the leaves before they enter the rolling drum (e.g., 60%). Feature enhancement and dimensionality expansion: Since the input features have low dimensionality and are static data, a fully connected layer is first used to perform nonlinear transformation and dimensionality expansion on them to generate a feature vector richer in semantic information, providing richer input for subsequent time series models; Convolution and pooling modules (used to learn abstract feature patterns): The enhanced feature vectors are reshaped into time-series-like data and fed into a one-dimensional convolutional layer. The role of the convolutional layer is no longer to process real-time sensor sequences, but to learn deep, abstract intrinsic relationships between different attribute combinations from static inputs. For example, it may learn to identify the specific processing requirement patterns implied by the combination of "spring tea + oak leaf variety + high moisture content". Pooling layers further compress and highlight these abstract features; Long Short-Term Memory Network Module (used to simulate the sequential dependence of process decisions): The abstract feature sequence extracted by CNN is input into LSTM; the role of LSTM here is to simulate a virtual, experience-based "decision-making process". It treats different attribute features as a logical sequence and learns how they jointly affect the sequential decision-making on parameters such as fermentation time and drying temperature, capturing the inherent temporal logical dependencies between process parameters (such as deciding the fermentation temperature first, and then deciding the drying temperature accordingly). Attention Mechanism Module: The attention mechanism is applied to all hidden states of the LSTM. It can automatically evaluate the importance of different parts of the input features to the final decision on each process parameter. For example, the model may learn to pay more attention to the harvest season when predicting the fermentation temperature, and to pay more attention to the initial moisture content when predicting the drying time. This greatly improves the interpretability and prediction accuracy of the model. Fully connected output layer: The context vector, which is weighted and summarized by the attention mechanism, is used to regress and output the optimal set of process parameters: [target temperature for fermentation stage, target temperature for drying stage, target ambient humidity, target drum speed, recommended fermentation time, and recommended drying time]; Training dataset construction: Model training relies on a high-quality historical database; Data source: Accumulated from a large amount of historical production batch data from this equipment and similar processes; Sample composition: Each sample represents a successful black tea production batch, including: Input data (features): tea variety, harvesting season, initial moisture content; Label data (target): Optimal process parameter set: When the final tea produced in this batch is rated as excellent by quality assessment (such as theaflavins / thearubigins content, sensory evaluation score), the corresponding actual process parameter set (fermentation temperature, drying temperature, humidity, rotation speed, duration of each stage) serves as the "standard answer" for model learning.

[0039] In step 4, the continuous fermentation and drying equipment for black tea is controlled based on the predicted optimal process parameters, specifically as follows: This step is the final execution stage of the intelligent control method. Its core task is to transform the "optimal process parameter set" output by the predictive model into precise actions of each execution unit of the equipment, and to perform real-time closed-loop control and dynamic fine-tuning during production to ensure that the actual fermentation and drying process always runs on the optimal trajectory. This step is completed by the PLC controller as the "brain," coordinating all sensors and actuators. 1. Parameter setting and task assignment Parameter reception: The PLC controller receives the optimal set of process parameters customized for this batch of tea from the intelligent prediction model, including: target temperature for the fermentation stage, target temperature for the drying stage, target ambient humidity, target drum speed, recommended fermentation time, and recommended drying time; The PLC's internal program automatically divides and manages the "fermentation" and "drying" processes based on the target temperatures for the fermentation and drying stages. Information such as the current stage and remaining time is displayed in real-time on the human-machine interface.

[0040] 2. Multi-loop adaptive closed-loop control The PLC performs independent closed-loop control on each key variable based on the set target value; Temperature closed-loop control loop: Controlled object: Total output power of the heating jacket and internal heating tubes; Sensor: Pt100 platinum resistance temperature sensor (multi-point front, middle, and rear); Control Algorithm: Fuzzy PID control is adopted. Traditional PID algorithms adjust based on the deviation between the temperature setpoint and the sensor feedback value. Fuzzy logic is introduced to further refine the PID parameters (K0) online based on the magnitude and trend of the deviation. p K i K d For example, when the temperature is much lower than the set value, a larger K value is used. p To achieve rapid temperature rise; when approaching the set value, Kp is reduced to avoid overshoot, which makes temperature control faster, more stable, and with less overshoot; Execution unit: The PLC adjusts the power supply of the heating jacket and the electric heating wire inside the heating tube through the analog output module.

[0041] Humidity closed-loop control loop: Controlled object: Relative humidity inside the drum; Sensor: Capacitive humidity sensor; Control algorithm: Fuzzy PID control is also used, which dynamically controls the intensity of humidification and dehumidification based on the deviation between the humidity setpoint and the measured value. Execution unit: Humidification: Controlling the start / stop and power of the ultrasonic humidifier and its pipeline solenoid valve; Dehumidification: Controlling the speed and start / stop of ventilation devices (fans) to regulate humidity by introducing or exhausting air.

[0042] Speed ​​closed-loop control loop: Controlled object: The rotational speed of the drum; Sensor: Hall effect speed sensor; Control algorithm: Conventional PID control is used to ensure that the drum speed is stable at the target drum speed recommended by the model; Execution unit: Adjusts the operating frequency of the drive motor through a frequency converter.

[0043] 3. Real-time monitoring and dynamic fine-tuning Process data synchronization: During the control process, the PLC continuously collects all sensor data (process parameters) and synchronizes them to the edge database. This data is not only used for control, but also serves as a production record for this batch, which can be used for subsequent model optimization. Online monitoring and feedback of key quality indicators: Images of tea leaves are captured periodically using an industrial camera installed inside the drum. Image processing algorithms analyze the color features of tea leaves in real time (e.g.) In color space The value (reflecting the degree of redness) is a key indicator for judging the degree of fermentation; When the system detects that the color of the tea leaves has not reached the preset "fully fermented" threshold within the expected time, it will trigger a fine-tuning signal. This signal is sent to the prediction model. The model, combined with the current actual process data, can make a small online adjustment to the remaining fermentation time or fermentation temperature and send the new set value to the PLC. This achieves adaptive optimization based on real-time quality feedback. 4. Safety interlock and alarm mechanism Parameter limit alarm: The system sets safe upper and lower limits for all critical parameters (such as temperature, humidity, and current). Once the sensor reading exceeds the limit, the system immediately issues an audible and visual alarm and displays specific fault information on the HMI.

[0044] Equipment interlock protection: When a serious fault is detected (such as motor overload or heater overheating), the PLC will immediately execute the safety interlock program, such as cutting off the heating power supply and stopping the drum rotation to prevent equipment damage and tea carbonization.

[0045] Batch completion and discharge: When the total running time reaches the recommended fermentation time + recommended drying time, or when the moisture content of the tea leaves at the drying stage is determined by the moisture meter to have reached the standard (e.g., <6%), the system determines that processing is complete. The PLC automatically opens the electric gate at the discharge port to complete the discharge, and resets the equipment to prepare for the next batch.

[0046] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on the differences from other embodiments. The same or similar parts between the various embodiments can be referred to each other.

[0047] This document uses specific examples to illustrate the principles and implementation methods of the present invention. The descriptions of the above embodiments are only for the purpose of helping to understand the method and core ideas of the present invention. Furthermore, those skilled in the art will recognize that, based on the ideas of the present invention, there will be changes in the specific implementation methods and application scope. Therefore, the content of this specification should not be construed as a limitation of the present invention.

Claims

1. A continuous fermentation drying apparatus for black tea, characterized by, The application relates to a continuous fermentation and drying device for black tea, which comprises the following: a data acquisition module for acquiring process parameters and tea body attribute parameters; a data preprocessing module in communication connection with the data acquisition module and used for cleaning and filling the acquired data; an intelligent parameter prediction module in communication connection with the data preprocessing module and used for predicting optimal process parameters based on the cleaned tea body attribute data; a device control module in communication connection with the intelligent parameter prediction module and used for controlling the fermentation and drying device in real time based on the predicted optimal process parameters; a drum fermentation and drying unit, a temperature, rotating speed and time control unit and a humidity control unit, all of which are electrically connected with the device control module and used for executing the instructions from the device control module.

2. The apparatus of claim 1, wherein, The data acquisition module comprises: a plurality of temperature sensors arranged at front, middle and rear positions in the drum; a humidity sensor arranged in the drum; a rotating speed sensor arranged on a rotating shaft of the drum; a power monitoring unit for acquiring real-time power of a heating device and a ventilation device; a man-machine interactive interface for inputting or acquiring tea type and picking season information; a moisture meter for acquiring initial moisture content of tea; an industrial camera for acquiring image and color information of tea.

3. The apparatus of claim 2, wherein, The drum fermentation and drying unit comprises: a food-grade stainless steel drum, the inner wall of which is welded with helical stirring blades and is arranged in an inclined manner; a driving device drivingly connected with the food-grade stainless steel drum and used for driving the drum to rotate; a heating device comprising a heating jacket arranged on the outer wall of the drum and a heating pipe arranged in the drum.

4. A control method of a black tea continuous fermentation drying apparatus, applied to the black tea continuous fermentation drying apparatus according to any one of claims 1 to 3, characterized in that, The application further relates to a method for controlling the continuous fermentation and drying device for black tea, which comprises the following steps: Step 1: collecting relevant data based on the data acquisition module; Step 2: preprocessing the collected relevant data based on the preprocessing module; Step 3: constructing an intelligent parameter prediction model and predicting optimal process parameters of the continuous fermentation and drying device for black tea based on the intelligent parameter prediction model; Step 4: controlling the continuous fermentation and drying device for black tea based on the predicted optimal process parameters.

5. The method of claim 4, wherein, In step 1, the relevant data collected based on the data acquisition module specifically comprises: process parameters and tea body attribute parameters; the process parameters comprise temperature, humidity, drum rotating speed, heating device power, ventilation device state and running time; the tea body attribute parameters comprise tea type, picking season, initial moisture content, image and color information.

6. The method of claim 5, wherein, In step 2, the preprocessing of the collected relevant data based on the preprocessing module specifically comprises: primary filtering and noise reduction, abnormal value detection and elimination and missing data filling based on temperature, humidity, drum rotating speed and heating device power data; rule verification processing based on tea type and picking season.

7. The method of claim 6, wherein, In step 3, the construction of the intelligent parameter prediction model and the prediction of optimal process parameters of the continuous fermentation and drying device for black tea based on the intelligent parameter prediction model specifically comprise: constructing the intelligent parameter prediction model based on a hybrid CNN-LSTM attention mechanism model optimized by a dung beetle optimization algorithm; generating a training data set based on historical production batch data; training the intelligent parameter prediction model based on the training data set to obtain a trained intelligent parameter prediction model; Input the relevant data to be detected into the trained parameter intelligent prediction model to obtain the optimal process parameters.

Images