Intelligent tea rolling integrated system and method based on near infrared spectrum

Abstract

The invention discloses an intelligent tea rolling integrated system and method based on a near infrared spectrum, and relates to the technical field of tea preparation. After a tea rolling machine is started, a PLC collects initial pressure data and initial rotating speed data and sends the initial pressure data and the initial rotating speed data to a finished product prediction module; after a fixed time interval, a near infrared spectrum sampling module collects near infrared spectrum data of the tea leaves, and the near infrared spectrum data is sent to a finished product prediction module and a similarity judgment module through a PLC; the finished product prediction module predicts the quality condition, if the quality requirement is not met, the similarity judgment module compares the similarity between the near infrared spectrum data and the historical spectrum data, the pressure and rotating speed data corresponding to the finished tea meeting the quality are selected, and the initial pressure data and the initial rotating speed data of the tea rolling machine are adjusted. According to the invention, the information during tea rolling is collected through the infrared spectrum, so that the rotating speed and pressure of the tea rolling machine are controlled and adjusted, and the quality of the obtained finished tea is improved.

Description

Technical Field

[0001] The invention relates to the technical field of tea preparation, and more particularly to an intelligent tea rolling integrated system and method based on near infrared spectroscopy. Background Art

[0002] At present, with the development of tea processing scale and the increasing demand for high-quality tea, the quality of tea from raw materials to processing to dried tea has attracted more and more attention. The quality of tea can be judged by observing its color, appearance, smelling, etc.

[0003] During the processing of tea, it will go through multiple processing steps. As the processing steps proceed, the condition of the tea changes. When judging the quality of tea, it is generally aimed at the finished tea after processing. The main purpose is to understand the condition of the finished tea after processing. It is difficult to know the impact of the processing process on the quality of the finished tea. The existing tea processing technology mainly relies on experience to set up, which is not easy to adjust and has poor controllability. At this stage, infrared spectroscopy has become the first choice for detection due to its pollution-free, non-destructive sampling, real-time, and easy use. At the same time, the rapid development and widespread penetration of information science and technology has become an important feature of the times in today's society. Various identification theories and technologies have emerged and have been widely used.

[0004] Therefore, how to improve the quality of finished tea products based on infrared spectroscopy is an urgent problem that technicians in this field need to solve. Summary of the invention

[0005] In view of this, the present invention provides an intelligent tea rolling integrated system based on near-infrared spectroscopy, which collects information on tea rolling through infrared spectroscopy, thereby controlling and adjusting the rotation speed and pressure of a tea rolling machine, thereby improving the quality of the resulting finished tea.

[0006] In order to achieve the above object, the present invention adopts the following technical solution:

[0007] A tea intelligent rolling integrated system based on near-infrared spectroscopy is arranged on a tea rolling machine, comprising: a near-infrared spectroscopy sampling module, a PLC controller, a database, a finished product prediction module, and a similarity judgment module. After the tea rolling machine is started, the PLC controller collects initial pressure data and initial speed data and sends them to the finished product prediction module, and sends a sampling signal to the near-infrared spectroscopy sampling module after a fixed time interval; the near-infrared spectroscopy sampling module collects near-infrared spectroscopy data of rolled tea leaves, and sends the near-infrared spectroscopy data to the finished product prediction module and the similarity judgment module through the PLC controller; the finished product prediction module presets a finished product prediction model, and according to the finished product prediction model, the finished product prediction module generates a finished product prediction signal. The prediction model combines the near-infrared spectrum data, the initial pressure data and the initial rotation speed data to predict the finished tea product. If the prediction does not meet the quality requirements, an adjustment signal is sent to the similarity judgment module through the PLC controller; the database pre-stores historical sample data, including historical tea near-infrared spectrum data, and corresponding pressure data and rotation speed data; the similarity judgment module compares the near-infrared spectrum data with the historical tea near-infrared spectrum data for similarity, selects the pressure data and rotation speed data corresponding to the historical tea near-infrared spectrum data with a similarity greater than a preset threshold and with the highest similarity, and sends them to the tea rolling machine through the PLC controller to adjust the initial pressure data and initial rotation speed data.

[0008] Preferably, the near-infrared spectrum sampling module collects near-infrared spectrum data of the rolled tea leaves, including tea moisture data, tea polyphenols data, and theanine data.

[0009] Preferably, the finished product prediction model in the finished product prediction module specifically includes: a first neural network module, a second neural network module, and a similarity measurement module. The first neural network module has the same structure as the second neural network module, both of which are GRU layers, multi-head attention mechanism layers, and fully connected layers. The GRU layers, multi-head attention mechanism layers, and fully connected layers are connected in sequence; the similarity measurement module is connected to the first neural network module and the second neural network module, respectively.

[0010] Preferably, the finished product prediction model training in the finished product prediction module specifically includes:

[0011] A data acquisition module, wherein the data acquisition module acquires historical near-infrared spectral data of tea leaves in the database under conditions where tea leaves are rolled in compliance with or not in compliance with quality requirements, and corresponding pressure data and rotation speed data;

[0012] A sample pair construction module, which constructs training sample pairs based on the historical tea near-infrared spectrum data acquired by the data acquisition module, and the corresponding pressure data and rotation speed data;

[0013] The model training module extracts the tea change characteristics of each type of historical sample data through the GRU layer, and learns the cumulative change characteristics of multiple tea change characteristics coupled with each other over time; captures the weighted weights of each tea change characteristic through the multi-head attention mechanism layer; and based on the training sample pairs, implements the similarity update parameters through the similarity measurement module to converge the finished product prediction model.

[0014] Preferably, the similarity determination module specifically includes:

[0015] A moisture discrimination module, wherein the moisture discrimination module performs similarity calculation on the tea moisture data in the near-infrared spectrum data at the last sampling moment and the tea moisture data in the historical tea near-infrared spectrum data, and selects moisture similarity values;

[0016] A tea polyphenols identification module, wherein the tea polyphenols identification module calculates the similarity between the tea polyphenols data in the near infrared spectrum data at the last sampling moment and the tea polyphenols data in the historical tea near infrared spectrum data, and screens the tea polyphenols similarity value;

[0017] Theanine discrimination module, wherein the theanine discrimination module performs similarity calculation on the theanine data in the near-infrared spectrum data at the last sampling moment and the theanine data in the historical tea near-infrared spectrum data, and screens the theanine similarity value;

[0018] A comprehensive identification module, wherein the comprehensive identification module combines the moisture similarity value, the tea polyphenol similarity value, and theanine similarity value to obtain a feature vector of the tea to be identified, combines the historical tea near-infrared spectrum data in the same manner, and recommends the pressure data and rotation speed data corresponding to the tea near-infrared spectrum data in the historical successful cases of the database by comparing the historical feature vectors with the feature vectors of the tea to be identified.

[0019] Preferably, it also includes a secondary adjustment module, which synchronously receives the sampling signal sent by the PLC controller, and after a preset time interval, sends a repeated sampling signal to the PLC control module, so that the PLC control module adjusts the secondary pressure data and the speed data.

[0020] Preferably, the sample pair construction module includes sample pairs of the same type and sample pairs of different types divided according to the status of the finished tea leaves.

[0021] A tea intelligent rolling integrated method based on near infrared spectroscopy, comprising: after the tea rolling machine is started, the PLC controller collects initial pressure data and initial speed data and sends them to the finished product prediction module, and sends a sampling signal to the near infrared spectrum sampling module after a fixed time interval; the near infrared spectrum sampling module collects near infrared spectrum data of the rolled tea leaves, and sends the near infrared spectrum data to the finished product prediction module and the similarity discrimination module through the PLC controller; the finished product prediction module presets a finished product prediction model, and according to the finished product prediction model, combines the near infrared spectrum data and the initial pressure data to obtain the finished product prediction model. The finished tea leaves are predicted based on the initial rotation speed data, and if the prediction does not meet the quality requirements, an adjustment signal is sent to the similarity judgment module through the PLC controller; the database pre-stores historical sample data, including historical tea near-infrared spectrum data, and corresponding pressure data and rotation speed data; the similarity judgment module compares the near-infrared spectrum data with the historical tea near-infrared spectrum data for similarity, selects the pressure data and rotation speed data corresponding to the historical tea near-infrared spectrum data with a similarity greater than a preset threshold and with the highest similarity, and sends them to the tea rolling machine through the PLC controller to adjust the initial pressure data and initial rotation speed data.

[0022] It can be seen from the above technical scheme that, compared with the prior art, the present invention discloses an intelligent tea rolling integrated system based on near-infrared spectroscopy, which evaluates and sorts tea through near-infrared spectroscopy technology and finished product prediction models combined with historical data, and monitors and predicts tea in real time, thereby realizing an intelligent production process, ensuring that the quality of the produced tea is stable and meets the standards, and can automatically adjust the pressure data and speed data of the rolling machine through data comparison by a similarity judgment module and a database, thereby improving production efficiency and product consistency. The system can pre-store historical sample data and compare real-time data to provide support and basis for intelligent decision-making. The collected data includes multi-dimensional information such as tea moisture, tea polyphenols, and theanine, which can comprehensively judge the quality of tea and improve the comprehensiveness and accuracy of the production process. It also includes a secondary adjustment module, which can perform secondary data adjustment after collecting data to ensure the accuracy and stability of production parameters. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on the provided drawings without paying creative work.

[0024] Figure 1 A schematic diagram of the structure provided by the present invention;

[0025] Figure 2 This is a schematic diagram of the finished product prediction model provided by the present invention. DETAILED DESCRIPTION

[0026] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0027] The embodiment of the present invention discloses an intelligent tea rolling integrated system based on near infrared spectroscopy, which is arranged on a tea rolling machine. Figure 1 As shown, it includes: a near-infrared spectrum sampling module, a PLC controller, a database, a finished product prediction module, and a similarity judgment module. After the tea rolling machine is started, the PLC controller collects initial pressure data and initial speed data and sends them to the finished product prediction module, and sends a sampling signal to the near-infrared spectrum sampling module after a fixed interval; the near-infrared spectrum sampling module collects near-infrared spectrum data of the rolled tea leaves, and sends the near-infrared spectrum data to the finished product prediction module and the similarity judgment module through the PLC controller; the finished product prediction module presets a finished product prediction model, and predicts the finished tea situation according to the finished product prediction model in combination with the near-infrared spectrum data, the initial pressure data and the initial speed data. If the prediction does not meet the quality requirements, the PLC controller sends an adjustment signal to the similarity judgment module; the database pre-stores historical sample data, including historical tea near-infrared spectrum data, and the corresponding pressure data and speed data; the similarity judgment module compares the near-infrared spectrum data with the historical tea near-infrared spectrum data for similarity, selects the pressure data and speed data corresponding to the historical tea near-infrared spectrum data with a similarity greater than a preset threshold and the highest similarity, and sends them to the tea rolling machine through the PLC controller to adjust the initial pressure data and initial speed data.

[0028] In a specific embodiment, the near-infrared spectrum sampling module collects near-infrared spectrum data of the rolled tea leaves, including tea moisture data, tea polyphenols data, and theanine data.

[0029] In a specific embodiment, the finished product prediction model in the finished product prediction module specifically includes: a first neural network module, a second neural network module, and a similarity measurement module. The first neural network module and the second neural network module have the same structure, both of which are GRU layers, multi-head attention mechanism layers, and fully connected layers. The GRU layers, multi-head attention mechanism layers, and fully connected layers are connected in sequence; the similarity measurement module is connected to the first neural network module and the second neural network module, respectively.

[0030] Each neural network module includes at least a hierarchically connected GRU layer and a multi-head attention mechanism layer. Learning the cumulative change characteristics of multiple tea change characteristics coupled with each other in time and the weighted weights of each meteorological change characteristic includes: extracting the cumulative change characteristics through the GRU layer during the learning process; capturing the weighted weights of each tea change characteristic through the multi-head attention mechanism layer. The weights of the first neural network module and the second neural network module are shared.

[0031] In a specific embodiment, Figure 2 As shown in the figure, the input of each neural network module is one model training sample among the model training samples. After the model training sample is input into the network, it passes through the GRU layer in sequence to extract the time accumulation features of multiple categories of time series data. After the multi-head attention mechanism layer, the pressure data and speed data corresponding to the historical tea near-infrared spectrum data are adaptively weighted fused. Then, the data is mapped to the feature space through the fully connected layer (Fully connected, FC), and the similarity measurement module calculates the Euclidean distance of the sample pair through the Euclidean distance calculation layer to determine the similarity of the model training samples.

[0032] GRU has an update gate and a reset gate. The gating unit can control the flow of information, determine which information should pass and which information should be discarded, and regulate the interaction between input data and previous hidden states. By using the gating unit, GRU can effectively transfer and update information between different time steps, and can better capture the cumulative effect of multiple types of time series data (tea moisture data, tea polyphenols data, theanine data) in time, that is, capture the cumulative change characteristics in time under the mutual coupling of multiple tea change characteristics.

[0033] In a specific embodiment, the tea quality prediction result output by the finished product prediction model is a time accumulation result, and the training method further includes:

[0034] Performing window smoothing on the tea quality prediction result according to a sliding time window of a first preset length to obtain tea quality prediction results within multiple window intervals;

[0035] The tea quality prediction results are corrected according to the maximum voting principle based on the sample sequence of tea quality prediction results within each window interval.

[0036] In order to ensure the interpretability of predicting the state of the tea rolling machine at the future moment, the above scheme finally smoothes the prediction result using a sliding window of size M to obtain the tea quality prediction results within multiple window intervals; the tea quality prediction results are corrected based on the sample sequence of the tea quality prediction results within each window interval.

[0037] Among them, the similarity between the same type of sample pairs and different types of sample pairs is determined by calculating the Euclidean distance or cosine distance between the sample pairs. The Euclidean distance can be used as an example to illustrate: during the training process, the parameters are updated based on the similarity between the same type of sample pairs and different types of sample pairs, so that the Euclidean distance of the same type of sample pairs is as small as possible, that is, the similarity is as large as possible, and the Euclidean distance of the different types of sample pairs is as large as possible, that is, the similarity is as small as possible, so that the finished prediction model converges.

[0038] The loss function uses Contrastive loss, which can effectively handle the paired data relationship of the twin network; the specific expression of Contrastive loss is as follows:

[0039]

[0040]

[0041] Among them, D w is the Euclidean distance between two samples in the Euclidean distance calculation layer; when the model training sample pair is of the same type, Y = 1, otherwise Y = 0; m is the set threshold used to divide whether the similarity measurement belongs to the range of consistent labels; N is the number of samples.

[0042] It should be understood that the finished product prediction model trained by the model training method of extracting the tea change characteristics of each type of time series data, learning the cumulative change characteristics over time under the mutual coupling of multiple tea change characteristics and the weighted weights of each tea change characteristic, when making finished product predictions, compared with the method of using the changes of a single indicator data, or using the change training model of multiple types of time series data to predict the finished product situation separately, can consider the time accumulation effect and time series dependency between multiple types of time series data through the cumulative change characteristics over time under the mutual coupling of multiple tea change characteristics and the weighted weights of each tea change characteristic, and has higher accuracy.

[0043] In a specific embodiment, the finished product prediction model training in the finished product prediction module specifically includes:

[0044] A data acquisition module, which acquires historical near-infrared spectral data of tea leaves in a database under conditions where tea leaves are rolled in compliance with or not in compliance with quality requirements, and corresponding pressure data and rotation speed data;

[0045] A sample pair construction module constructs training sample pairs based on the historical tea near-infrared spectrum data obtained by the data acquisition module, and the corresponding pressure data and rotation speed data;

[0046] The model training module extracts the tea change characteristics of each type of historical sample data in the time series through the GRU layer, and learns the cumulative change characteristics over time under the mutual coupling of multiple tea change characteristics; captures the weighted weights of each tea change characteristic through the multi-head attention mechanism layer; and based on the training sample pairs, the similarity measurement module is used to implement the similarity update parameters to converge the finished product prediction model.

[0047] Because the tea rolling machine is rolling the tea leaves in a cumulative change process over time under certain conditions, and there are complex coupling relationships between various subtle conditions, the finished product cannot be judged simply by the rolling at a certain moment and the conditions of the tea leaves themselves.

[0048] Since the rolling and making of tea leaves needs to meet certain conditions and accumulate changes in time, and there are complex coupling relationships under various subtle conditions in the above-mentioned cumulative changes, in order to determine the cumulative changes in the above-mentioned time and the time accumulation effect with complex coupling relationships of the data (tea moisture data, tea polyphenols data, theanine data) representing each type of time series, in the present invention, when training the model, the gated recurrent unit (GRU) is used to extract the characteristics of the tea change characteristics of each type of time series. The retention of historical information is mainly achieved through the update gate and the reset gate.

[0049] Since there are complex coupling relationships in various subtle conditions during the tea rolling process, a multi-head attention mechanism is used to process the complex relationship of feature data in time series. The weighted weights of tea change features are adjusted through multiple groups of attention mechanisms to obtain the relationship between different feature sequences.

[0050] During use, the finished product prediction model obtains the near-infrared spectral data, initial pressure data and initial rotation speed data of the rolled tea leaves and the historical tea near-infrared spectral data stored in the database, as well as the corresponding pressure data and rotation speed data, to construct a prediction sample pair; based on the time accumulation characteristics and similarity of the prediction sample pairs, the quality prediction results of the finished tea leaves are obtained.

[0051] In a specific embodiment, the similarity determination module specifically includes:

[0052] The moisture discrimination module calculates the similarity between the tea moisture data in the near-infrared spectrum data at the last sampling moment and the tea moisture data in the historical tea near-infrared spectrum data, and selects the moisture similarity value;

[0053] The tea polyphenols identification module calculates the similarity between the tea polyphenols data in the near-infrared spectrum data at the last sampling moment and the tea polyphenols data in the historical tea near-infrared spectrum data, and selects the tea polyphenols similarity value;

[0054] Theanine discrimination module: The theanine discrimination module calculates the similarity between theanine data in the near-infrared spectrum data at the last sampling moment and theanine data in the historical tea near-infrared spectrum data, and selects theanine similarity value;

[0055] The comprehensive discrimination module combines the moisture similarity value, tea polyphenols similarity value and theanine similarity value to obtain the feature vector of the tea to be discriminated, combines the historical tea near-infrared spectrum data in the same way, and compares the historical feature vectors with the feature vectors of the tea to be discriminated, and recommends the pressure data and rotation speed data corresponding to the tea near-infrared spectrum data in the historical successful cases of the database.

[0056] When the extracted moisture similarity value, tea polyphenols similarity value and theanine similarity value are combined into a feature vector, the combination order is the same as the combination order of the feature vectors of the above historical scheme.

[0057] The similarity comparison in the comprehensive discrimination module can be calculated using the cosine similarity algorithm. The principle of the cosine similarity algorithm is to treat the two feature vectors as two line segments in space. Both of these line segments start from the origin and point to different directions, that is, an angle is formed between the two line segments. The two vectors start from the same point and point to different directions, and the angle in between is θ. The smaller this angle is, the closer the directions of the two feature vectors are, that is, the more similar the two vectors are.

[0058] In a specific embodiment, a secondary adjustment module is also included. The secondary adjustment module synchronously receives the sampling signal sent by the PLC controller, and after a preset time interval, sends a repeated sampling signal to the PLC control module, so that the PLC control module adjusts the secondary pressure data and the speed data.

[0059] Among them, the secondary adjustment module is to re-collect samples based on the speed and pressure data adjusted by the tea rolling machine for the first time, and combine them with the speed and pressure after the first adjustment to make a secondary adjustment to ensure the quality of the finished tea.

[0060] In a specific embodiment, the sample pair construction module includes sample pairs of the same type and sample pairs of different types divided according to the status of the finished tea leaves.

[0061] Among them, the same type of sample pairs can be a sample pair constructed by the near-infrared spectral data at a certain moment when the quality of the finished tea meets the quality requirements and the pressure and speed data of the corresponding tea rolling machine, and a sample pair constructed by the near-infrared spectral data when the quality of the finished tea meets the quality requirements and the pressure and speed data of the corresponding tea rolling machine. One of the different types of sample pairs is a sample constructed by the near-infrared spectral data when the quality of the finished tea meets the quality requirements and the pressure and speed data of the corresponding tea rolling machine, and the other is a sample constructed by the near-infrared spectral data when the quality of the finished tea does not meet the quality requirements and the pressure and speed data of the corresponding tea rolling machine.

[0062] The invention discloses an intelligent tea rolling integrated method based on near infrared spectroscopy, comprising: after a tea rolling machine is started, a PLC controller collects initial pressure data and initial rotation speed data and sends them to a finished product prediction module, and sends a sampling signal to the near infrared spectroscopy sampling module after a fixed time interval; the near infrared spectroscopy sampling module collects near infrared spectroscopy data of rolled tea leaves, and sends the near infrared spectroscopy data to a finished product prediction module and a similarity discrimination module through a PLC controller; a finished product prediction model is preset in the finished product prediction module, and the finished tea condition is predicted according to the finished product prediction model in combination with near infrared spectroscopy data, initial pressure data and initial rotation speed data, and if the prediction does not meet the quality requirements, an adjustment signal is sent to the similarity discrimination module through the PLC controller; a database pre-stores historical sample data, including historical tea near infrared spectroscopy data, and corresponding pressure data and rotation speed data; the similarity discrimination module performs similarity comparison between the near infrared spectroscopy data and the historical tea near infrared spectroscopy data, selects the pressure data and rotation speed data corresponding to the historical tea near infrared spectroscopy data with a similarity greater than a preset threshold and having the highest similarity, and sends them to the tea rolling machine through the PLC controller to adjust the initial pressure data and initial rotation speed data.

[0063] In the present invention, the near-infrared spectrum data of the rolled tea leaves collected by the near-infrared spectrum sampling module used in the similarity measurement module is a certain period within the rolling time, and the data used in the similarity judgment module is the near-infrared spectrum data of the last moment of a certain period within the rolling time.

[0064] In this specification, each embodiment is described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The same or similar parts between the embodiments can be referred to each other. For the device disclosed in the embodiment, since it corresponds to the method disclosed in the embodiment, the description is relatively simple, and the relevant parts can be referred to the method part.

[0065] The above description of the disclosed embodiments enables one skilled in the art to implement or use the present invention. Various modifications to these embodiments will be apparent to one skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention will not be limited to the embodiments shown herein, but rather to the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. An intelligent tea rolling integrated system based on near infrared spectroscopy, arranged on a tea rolling machine, characterized in that: include: A near-infrared spectrum sampling module, a PLC controller, a database, a finished product prediction module, and a similarity judgment module. After the tea rolling machine is started, the PLC controller collects initial pressure data and initial speed data and sends them to the finished product prediction module, and sends a sampling signal to the near-infrared spectrum sampling module after a fixed time interval; the near-infrared spectrum sampling module collects near-infrared spectrum data of the rolled tea leaves, and sends the near-infrared spectrum data to the finished product prediction module and the similarity judgment module through the PLC controller; the finished product prediction module presets a finished product prediction model, and according to the finished product prediction model, a combination of the near-infrared spectrum data, the initial speed data, and the similarity judgment module calculates the finished product prediction model. The initial pressure data and the initial rotation speed data are used to predict the finished tea condition, and if the prediction does not meet the quality requirements, an adjustment signal is sent to the similarity judgment module through the PLC controller; the database pre-stores historical sample data, including historical tea near-infrared spectrum data, and corresponding pressure data and rotation speed data; the similarity judgment module compares the near-infrared spectrum data with the historical tea near-infrared spectrum data for similarity, selects the pressure data and rotation speed data corresponding to the historical tea near-infrared spectrum data with a similarity greater than a preset threshold and with the highest similarity, and sends them to the tea rolling machine through the PLC controller to adjust the initial pressure data and initial rotation speed data.

2. The intelligent tea rolling integrated system based on near infrared spectroscopy according to claim 1 is characterized in that: The near-infrared spectrum sampling module collects near-infrared spectrum data of the rolled tea leaves, including tea moisture data, tea polyphenols data, and theanine data.

3. The intelligent tea rolling integrated system based on near infrared spectroscopy according to claim 2 is characterized in that: The finished product prediction model in the finished product prediction module specifically includes: a first neural network module, a second neural network module, and a similarity measurement module. The first neural network module has the same structure as the second neural network module, both of which are GRU layers, multi-head attention mechanism layers, and fully connected layers. The GRU layers, multi-head attention mechanism layers, and fully connected layers are connected in sequence; the similarity measurement module is connected to the first neural network module and the second neural network module, respectively.

4. The intelligent tea rolling integrated system based on near infrared spectroscopy according to claim 3 is characterized in that: The finished product prediction model training in the finished product prediction module specifically includes: A data acquisition module, wherein the data acquisition module acquires historical near-infrared spectral data of tea leaves in the database under conditions where tea leaves are rolled in compliance with or not in compliance with quality requirements, and corresponding pressure data and rotation speed data; A sample pair construction module, which constructs training sample pairs based on the historical tea near-infrared spectrum data acquired by the data acquisition module, and the corresponding pressure data and rotation speed data; The model training module extracts the tea change characteristics of each type of historical sample data through the GRU layer, and learns the cumulative change characteristics of multiple tea change characteristics coupled with each other over time; captures the weighted weights of each tea change characteristic through the multi-head attention mechanism layer; and based on the training sample pairs, implements the similarity update parameters through the similarity measurement module to converge the finished product prediction model.

5. The intelligent tea rolling integrated system based on near infrared spectroscopy according to claim 2 is characterized in that: The similarity determination module specifically includes: A moisture discrimination module, wherein the moisture discrimination module performs similarity calculation on the tea moisture data in the near-infrared spectrum data at the last sampling moment and the tea moisture data in the historical tea near-infrared spectrum data, and selects moisture similarity values; A tea polyphenols identification module, wherein the tea polyphenols identification module calculates the similarity between the tea polyphenols data in the near-infrared spectrum data at the last sampling moment and the tea polyphenols data in the historical tea near-infrared spectrum data, and screens the tea polyphenols similarity value; Theanine discrimination module, wherein the theanine discrimination module performs similarity calculation between the theanine data in the near-infrared spectrum data at the last sampling moment and the theanine data in the historical tea near-infrared spectrum data, and screens the theanine similarity value; A comprehensive identification module, wherein the comprehensive identification module combines the moisture similarity value, the tea polyphenol similarity value, and theanine similarity value to obtain a feature vector of the tea to be identified, combines the historical tea near-infrared spectrum data in the same manner, and recommends the pressure data and rotation speed data corresponding to the tea near-infrared spectrum data in the historical successful cases of the database by comparing the historical feature vectors with the feature vectors of the tea to be identified.

6. The intelligent tea rolling integrated system based on near infrared spectroscopy according to claim 1 is characterized in that: It also includes a secondary adjustment module, which synchronously receives the sampling signal sent by the PLC controller, and after a preset time interval, sends a repeated sampling signal to the PLC control module, so that the PLC control module adjusts the secondary pressure data and the speed data.

7. The intelligent tea rolling integrated system based on near infrared spectroscopy according to claim 4 is characterized in that: The sample pair construction module includes sample pairs of the same type and sample pairs of different types divided according to the status of finished tea leaves.

8. An intelligent tea rolling integrated method based on near infrared spectroscopy, applied to an intelligent tea rolling integrated system based on near infrared spectroscopy according to any one of claims 1 to 7, characterized in that: include: After the tea rolling machine is started, the PLC controller collects initial pressure data and initial speed data and sends them to the finished product prediction module, and sends a sampling signal to the near-infrared spectrum sampling module after a fixed time interval; the near-infrared spectrum sampling module collects near-infrared spectrum data of the rolled tea leaves, and sends the near-infrared spectrum data to the finished product prediction module and the similarity judgment module through the PLC controller; the finished product prediction module presets a finished product prediction model, and predicts the finished tea condition according to the finished product prediction model in combination with the near-infrared spectrum data, the initial pressure data and the initial speed data. If the prediction does not meet the quality requirements, an adjustment signal is sent to the similarity judgment module through the PLC controller; the database pre-stores historical sample data, including historical tea near-infrared spectrum data, and corresponding pressure data and speed data; the similarity judgment module compares the near-infrared spectrum data with the historical tea near-infrared spectrum data for similarity, selects the pressure data and speed data corresponding to the historical tea near-infrared spectrum data with a similarity greater than a preset threshold and the highest similarity, and sends them to the tea rolling machine through the PLC controller to adjust the initial pressure data and initial speed data.

Images