Boiling detection method, device and system in ligusticum wallichii extraction process and storage medium

Abstract

The invention provides a boiling detection method, device and system in a ligusticum wallichii extraction process and a storage medium, and the method comprises the steps: synchronizing temperature distribution images, bubble stereo images and chemical component change characteristics collected at different time points by using a unified timestamp, and obtaining multi-modal data; and training and optimizing the physical information neural network model by using the corresponding relationship between the multi-modal data and the boiling state, and predicting the boiling state of the liquid in the extraction equipment by using the physical information neural network model. According to the boiling detection method, device and system in the ligusticum wallichii extraction process and the storage medium, intelligent monitoring of the boiling state in the traditional Chinese medicine extraction process can be achieved, the boiling state of internal liquid in the extraction process can be monitored and analyzed in real time, the defects of a traditional detection method are effectively overcome, the product quality is improved, and the detection efficiency is improved. Powerful support is provided for intelligent manufacturing of traditional Chinese medicine extraction.

Description

Technical Field

[0001] The present invention belongs to the technical field of intelligent manufacturing, and particularly relates to a boiling detection method, device, system and storage medium in the process of Ligusticum chuanxiong extraction. Background Art

[0002] Traditional Chinese medicine components are complex and there are significant differences in components between different batches. Therefore, there are often various negative impacts in the extraction of traditional Chinese medicine. Among them, the boiling state is an important stage in the process of traditional Chinese medicine extraction. Controlling the boiling state well can ensure that the active ingredients are fully extracted, thereby improving the extraction efficiency. If the boiling state is unstable, it may lead to uneven components or impurities in the extract, thus affecting the quality of the final product; too high temperature or violent boiling may cause damage to the equipment or occurrence of safety accidents. Therefore, precise control of the boiling state is an important measure to ensure the safety of the production process.

[0003] Traditional methods for extracting traditional Chinese medicine are difficult to intuitively reflect the true boiling state of the extraction, which may lead to temperature fluctuations during the extraction process and affect the extraction effect. In order to monitor the boiling state during the extraction of traditional Chinese medicine in real time, existing methods generally use machine vision technology to monitor the extraction process. Through high-performance imaging devices and image processing algorithms, the boiling phenomenon can be analyzed and identified. However, since measuring the boiling state inside the extraction equipment is a complex process, involving various influencing factors such as temperature distribution, solution composition, and thermal state, relying solely on images for monitoring and analysis cannot obtain an accurate boiling state, and thus cannot achieve precise control of traditional Chinese medicine extraction, affecting the quality of the final product. Summary of the Invention

[0004] In view of this, the present invention aims to provide a boiling detection method, device, system and storage medium in the process of Ligusticum chuanxiong extraction to solve the problem that in the existing process of traditional Chinese medicine extraction, only using the image analysis method cannot obtain an accurate boiling state, which affects the quality of traditional Chinese medicine extraction products.

[0005] To achieve the above object, the technical solution of the present invention is realized as follows:

[0006] In the first aspect

[0007] An embodiment of the present invention provides a boiling detection method in the process of Ligusticum chuanxiong extraction, including:

[0008] Performing thermal imaging acquisition on the boiling liquid in the extraction equipment to obtain the temperature distribution image of the liquid during the boiling process;

[0009] Image acquisition is performed on the surface of the boiling liquid in the extraction device from different angles to obtain the generation and movement characteristics of bubbles on the liquid surface during the boiling process. Based on the generation and movement characteristics of bubbles on the liquid surface during the boiling process, a three-dimensional image of the bubbles on the liquid surface during the boiling process is generated using holographic imaging technology;

[0010] Based on the registration of the position information in the temperature distribution image and the three-dimensional bubble image, a spatial correspondence relationship between temperature and bubbles is established, and different boiling regions are divided according to a preset temperature range;

[0011] Chemical composition data acquisition is performed on different boiling regions of the liquid in the extraction device to obtain the characteristics of chemical composition changes of the liquid during the boiling process in different boiling regions;

[0012] Synchronize the temperature distribution images, three-dimensional bubble images, and chemical composition change characteristics collected at different time points using a unified timestamp, and establish the relationship between the chemical composition change characteristics and the temperature distribution characteristics of the corresponding boiling regions in the temperature distribution images, as well as the relationship between the chemical composition change characteristics and the generation and movement characteristics of the bubbles in the corresponding boiling regions in the three-dimensional bubble images for each time point to obtain multi-modal data;

[0013] After performing dimensionality reduction processing on the multi-modal data using the principal component analysis algorithm, use the clustering algorithm to identify different boiling states and establish the corresponding relationship between the multi-modal data and the boiling states;

[0014] Based on the Navier-Stokes equation, fluid mechanics, and the corresponding relationship between the multi-modal data and the boiling states, construct a physical information neural network model for the liquid in the extraction device, and use the corresponding relationship between the multi-modal data and the boiling states to train and optimize the physical information neural network model to obtain a trained physical information neural network model;

[0015] Real-time obtain any one or more of the temperature distribution image, three-dimensional bubble image, and chemical composition change characteristics of the liquid in the extraction device, and input them into the trained physical information neural network model to predict the boiling state of the liquid in the extraction device.

[0016] Furthermore, the image acquisition of the surface of the boiling liquid in the extraction device from different angles to obtain the generation and movement characteristics of bubbles on the liquid surface during the boiling process, and based on the generation and movement characteristics of bubbles on the liquid surface during the boiling process, using holographic imaging technology to generate a three-dimensional image of the bubbles on the liquid surface during the boiling process includes:

[0017] Image acquisition is performed on the liquid surface in the extraction device from different angles, and the collected images are preprocessed to obtain preprocessed images;

[0018] Perform edge detection on the preprocessed image based on the Canny edge detection algorithm to obtain the bubble edges;

[0019] Segment the preprocessed image based on the Otsu algorithm and the bubble edges to obtain a number of bubbles, and track the movement trajectories of the bubbles based on the inter-frame difference method to obtain the generation and movement characteristics of the bubbles on the liquid surface during the boiling process;

[0020] Generate a three-dimensional image of the bubbles on the liquid surface during the boiling process using holographic imaging technology based on the generation and movement characteristics of the bubbles on the liquid surface during the boiling process.

[0021] Furthermore, register based on the position information in the temperature distribution image and the three-dimensional bubble image, establish the spatial correspondence between temperature and bubbles, and divide different boiling regions according to a preset temperature range, including:

[0022] Register based on the position information in the temperature distribution image and the three-dimensional bubble image, and establish the spatial correspondence between temperature and bubbles;

[0023] Based on the spatial correspondence, overlay the three-dimensional bubble image and the temperature distribution map, and divide the boiling liquid into a high-temperature boiling region, a medium-temperature boiling region, and a low-temperature boiling region according to a preset temperature range.

[0024] Furthermore, after performing dimensionality reduction processing on the multi-modal data using the principal component analysis algorithm, use the clustering algorithm to identify different boiling states and establish the correspondence between the multi-modal data and the boiling states, including:

[0025] Select the first k principal components according to the eigenvalue magnitudes;

[0026] Project the multi-modal data into the selected principal component space to generate the dimensionality-reduced data;

[0027] Perform clustering on the dimensionality-reduced data to generate the classification results of different boiling states

[0028] Use the eigenvectors in the principal component analysis to analyze the contribution degree of each modal data to the identification of the boiling state.

[0029] Furthermore, construct a neural network model for extracting the physical information of the liquid in the extraction device based on the Navier-Stokes equation, fluid mechanics, and the correspondence between the multi-modal data and the boiling states, and use the correspondence between the multi-modal data and the boiling states to train and optimize the physical information neural network model to obtain the trained physical information neural network model, including:

[0030] Construct a neural network model for extracting the physical information of the liquid in the extraction device based on the Navier-Stokes equation, fluid mechanics, and the correspondence between the multimodal data and the boiling state. The formula is as follows:

[0031]

[0032] Wherein, is the rate of change of the fluid velocity u with respect to time t; is the convection term, which is used to represent the acceleration generated by the fluid due to its own motion; is the pressure gradient term, which is used to represent the acceleration generated due to the non-uniform distribution of the pressure p, and ρ is the density of the fluid; is the viscous term, which is used to represent the acceleration generated due to the fluid viscosity v, is the Laplace operator, which is used to represent the second derivative of the velocity; g is the acceleration due to gravity, which is used to represent the acceleration generated due to gravity.

[0033] The motion of the bubble is described by the Rayleigh-Plesset equation. The formula is as follows:

[0034]

[0035] Wherein, represents the acceleration, represents the velocity, represents the rate of change of pressure, ρ represents the density of the liquid, P 0 represents the pressure inside the bubble, σ represents the surface tension of the liquid, and η represents the viscosity of the liquid;

[0036] Use the correspondence between the multimodal data and the boiling state to train and optimize the physical information neural network model to obtain the trained physical information neural network model.

[0037] Furthermore, the use of the correspondence between the multimodal data and the boiling state to train and optimize the physical information neural network model to obtain the trained physical information neural network model includes:

[0038] Based on the physical information neural network model, combine the physical constraints of the multimodal data and the data-driven loss function. The formula is as follows:

[0039] Loss = α 1 L d (u, y) + α 2 L p (u)

[0040] Wherein, α 1 is the weight of the data loss; α 2is the weight of physical loss; L d (u, y) is the data loss function, which measures the difference between the model prediction u and the observed data y; L p (u) is the physical loss function, which measures whether the model prediction u satisfies the physical laws;

[0041] The bubble image features, chemical composition features, and temperature features in the multimodal data are weighted and fused through a multi-head attention mechanism to obtain fused features;

[0042] The physical information neural network model is trained and optimized using the fused features so that the physical information neural network model can focus on the most important part of information for the boiling state.

[0043] Further, the step of weighting and fusing the bubble image features, chemical composition features, and temperature features in the multimodal data through a multi-head attention mechanism to obtain fused features includes:

[0044] Query, Key, and Value are respectively mapped into multiple different subspaces, and the attention weights are independently calculated and weighted fusion is performed within each subspace; among them, the generation of Query is to splice the bubble image features, chemical composition features, and temperature features into a long vector; corresponding Key vectors are generated for the bubble image features, chemical composition features, and temperature features respectively and spliced into a matrix; corresponding Value vectors are generated for each of the bubble image features, chemical composition features, and temperature features respectively and spliced into a matrix;

[0045] The calculation formula for the attention weight Ai is: where Dk is the dimension of the Key vector;

[0046] The weighted fusion calculation process is as follows:

[0047] The Value is weighted using the attention weight Ai to obtain the weighted fusion feature for each head i, and the formula is: Fi = Ai × Vi;

[0048] The weighted fusion features Fi of all heads i are spliced together and fused through an additional linear layer Wo to obtain the final fused feature, and the formula is as follows:

[0049]

[0050] where x is the input feature, zq represents Query, k is the index of Key, q is the index of Query, M represents the number of heads of multi-head attention, m represents the m-th attention head, Ai represents the attention weight of the m-th head, W G M X OActually, it is Value, W G M It is the result of applying attention to Value and then performing a linear transformation to obtain the output results of different heads.

[0051] Second aspect

[0052] An embodiment of the present invention provides a detection device, including:

[0053] A thermal imaging module for performing thermal imaging acquisition on the boiling liquid in the extraction device to obtain the temperature distribution image of the liquid during the boiling process;

[0054] An image acquisition module for performing image acquisition on the surface of the boiling liquid in the extraction device from different angles to obtain the generation and movement characteristics of bubbles on the liquid surface during the boiling process, and generating a three-dimensional bubble image on the liquid surface during the boiling process based on the generation and movement characteristics of bubbles on the liquid surface during the boiling process using holographic imaging technology;

[0055] A processing module for registering based on the position information in the temperature distribution image and the three-dimensional bubble image, establishing the spatial correspondence between temperature and bubbles, and dividing different boiling regions according to a preset temperature range;

[0056] A chemical composition acquisition module for collecting chemical composition data of different boiling regions of the liquid in the extraction device to obtain the chemical composition change characteristics of the liquid during the boiling process in different boiling regions;

[0057] A synchronization module for synchronizing the temperature distribution image, the three-dimensional bubble image, and the chemical composition change characteristics collected at different time points with a unified timestamp, and establishing the relationship between the chemical composition change characteristics and the temperature distribution characteristics of the corresponding boiling region in the temperature distribution image, as well as the relationship between the chemical composition change characteristics and the generation and movement characteristics of bubbles in the corresponding boiling region in the three-dimensional bubble image at each time point to obtain multi-modal data;

[0058] An analysis module for performing dimensionality reduction processing on the multi-modal data using the principal component analysis algorithm, then using the clustering algorithm to identify different boiling states, and establishing the correspondence between the multi-modal data and the boiling states;

[0059] A building module for constructing a physical information neural network model of the liquid in the extraction device based on the Navier-Stokes equation, fluid mechanics, and the correspondence between the multi-modal data and the boiling states, and training and optimizing the physical information neural network model using the correspondence between the multi-modal data and the boiling states to obtain the trained physical information neural network model;

[0060] A prediction module, configured to obtain in real time any one or more of the temperature distribution image of the liquid in the extraction device, the three-dimensional image of the bubbles, and the characteristics of the chemical composition change, and input them into the trained physical information neural network model to predict the boiling state of the liquid in the extraction device.

[0061] In a third aspect, an embodiment of the present invention further provides a detection system, including:

[0062] One or more processors;

[0063] A storage device, configured to store one or more programs;

[0064] A camera, configured to collect images;

[0065] A chemical sensor, configured to collect chemical components;

[0066] A display, configured to display detection results;

[0067] When the one or more programs are executed by the one or more processors, the one or more processors implement the boiling detection method in the Chuanxiong extraction process provided in the above embodiment.

[0068] In a fourth aspect, an embodiment of the present invention further provides a storage medium containing computer-executable instructions, and the computer-executable instructions are used to execute the boiling detection method in the Chuanxiong extraction process provided in the above embodiment when executed by a computer processor.

[0069] Compared with the prior art, the boiling detection method, device, system, and storage medium in the Chuanxiong extraction process of the present invention have the following advantages:

[0070] The boiling detection method, device, system, and storage medium in the Chuanxiong extraction process of the present invention improve the accuracy and robustness of boiling state detection by performing multi-modal feature fusion of image features and physical parameters and combining simulation calculations, thereby realizing intelligent monitoring of the boiling state in the traditional Chinese medicine extraction process, can monitor and analyze the boiling state of the internal liquid in real time during the extraction process, effectively overcome the deficiencies of traditional detection methods, are beneficial to improving product quality, and provide strong support for the intelligent manufacturing of traditional Chinese medicine extraction. BRIEF DESCRIPTION OF THE DRAWINGS

[0071] The drawings constituting a part of the present invention are used to provide a further understanding of the present invention. The schematic embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute an improper limitation of the present invention. In the drawings:

[0072] Figure 1 It is a schematic flowchart of the boiling detection method in the Chuanxiong extraction process according to Embodiment 1 of the present invention;

[0073] Figure 2 Schematic diagram of the multi-modal data processing flow in the boiling detection method during the extraction process of Ligusticum chuanxiong according to the first embodiment of the present invention;

[0074] Figure 3 Schematic diagram of the structure of a detection device according to the second embodiment of the present invention;

[0075] Figure 4 Schematic diagram of the structure of a system provided by the third embodiment of the present invention. Detailed implementation manners

[0076] The present invention will be further described in detail below with reference to the drawings and embodiments. It can be understood that the specific embodiments described herein are only used to explain the present invention, rather than limiting the present invention. Additionally, it should be noted that for the sake of description, only parts related to the present invention are shown in the drawings, rather than all structures.

[0077] Embodiment 1

[0078] Figure 1 Schematic diagram of the flow of the boiling detection method during the extraction process of Ligusticum chuanxiong according to the first embodiment of the present invention. Figure 2 Schematic diagram of the multi-modal data processing flow in the boiling detection method during the extraction process of Ligusticum chuanxiong according to the first embodiment of the present invention. This detection method can be used to identify the boiling state during the important extraction process, so as to realize the dynamic evaluation of the extraction process state and improve the quality of the final product. See Figure 1 and Figure 2 , this detection method specifically includes the following steps:

[0079] Step 101: Perform thermal imaging acquisition on the boiling liquid in the extraction device to obtain the temperature distribution image of the liquid during the boiling process.

[0080] By analyzing the relationship between the temperature distribution and the generation and movement of bubbles, the influence of heat transfer on bubble behavior can be inferred. Specifically, an infrared camera can be used to perform thermal imaging acquisition on the boiling liquid in the extraction device to obtain a temperature distribution image with time points. For example, an infrared thermal imager can be installed outside the extraction device to ensure that it can cover the key areas inside the device, and the infrared thermal imager is used to collect the temperature distribution image inside the device in real time, and the collected image is transmitted to a computer, and image processing software is used to analyze the temperature distribution.

[0081] In the actual application process, the extraction device can also be monitored through an infrared camera, and the temperature data can be displayed on the front-end interface. Users can view the temperature distribution of each extraction device. At the same time, the historical temperature data of each extraction device will be recorded and a time series curve will be generated to help users view the temperature change trend.

[0082] Exemplarily, an infrared thermal imager can be used to continuously collect the temperature distribution images inside the device. The operator can view the images in real time through the display screen of the instrument. During the real-time collection process, ensure that the data recording function is enabled. Most infrared thermal imagers have a built-in storage function and can automatically save the image data. Use software tools to generate a temperature distribution map, which shows the temperature distribution in different areas inside the device. The temperature range and color mapping can be set to more intuitively observe the temperature changes.

[0083] Step 102: Collect images of the boiling liquid surface inside the extraction device from different angles to obtain the generation and movement characteristics of the bubbles on the liquid surface during the boiling process, and based on the generation and movement characteristics of the bubbles on the liquid surface during the boiling process, use holographic imaging technology to generate a three-dimensional image of the bubbles on the liquid surface during the boiling process.

[0084] To monitor the boiling state in real time, machine vision technology can be used to monitor the extraction process. Through high-performance imaging devices and image processing algorithms, the boiling phenomenon can be analyzed and identified. This technology can capture and analyze the bubble growth rate and movement characteristics, thereby identifying different boiling states.

[0085] In the actual application process, a network high-speed camera can be installed; the high-performance imaging device is set at different angles of the extraction device to capture high-frame-rate images of the liquid surface during the boiling process, and record the generation and movement characteristics of the bubbles, such as the growth rate and the interaction between bubbles. At the same time, based on the high-speed camera system, holographic imaging technology is introduced. The holographic imaging system works synchronously with the data of the high-speed camera, and uses the two-dimensional image data captured by the high-speed camera to generate a three-dimensional image of the bubbles on the liquid surface.

[0086] Specifically, step 102 can be specifically implemented as follows:

[0087] Step 1021: Collect images of the liquid surface inside the extraction device from different angles and preprocess the collected images to obtain preprocessed images.

[0088] Specifically, the camera operates at a frame rate of 5000 frames per second (fps) to ensure that it can capture the generation and movement details of extremely small bubbles.

[0089] Step 1022: Perform edge detection on the preprocessed image based on the Canny edge detection algorithm to obtain the bubble edges. The process of performing edge detection on the preprocessed image based on the Canny edge detection algorithm to obtain the bubble edges includes: performing smoothing processing on the preprocessed image based on a Gaussian filter to obtain a smoothed image; calculating the gradient of the smoothed image to obtain the gradient magnitude and gradient direction corresponding to the pixels; screening the pixels based on the gradient magnitude and gradient direction corresponding to the pixels using the non-maximum suppression method to obtain edge points; calculating the bubble edges based on the double-threshold algorithm for a number of the edge points.

[0090] Step 1023: Segment the preprocessed image based on the Otsu algorithm and the bubble edges to obtain a number of bubbles, and track the motion trajectories of the bubbles based on the inter-frame difference method to obtain the generation and motion characteristics of the bubbles on the liquid surface during the boiling process.

[0091] First, the method for judging the flow pattern change in flow boiling. The process of segmenting the preprocessed image includes calculating the global threshold of the image based on the Otsu algorithm, and the formula is as follows:

[0092] In the formula, i is the gray level currently considered, w(i) is the cumulative probability up to the gray level i, u(i) is the cumulative mean up to the gray level i, and u ] is the global mean of the image.

[0093] Second, denote the nth and (n - 1)th frame images in the video sequence as fn and fn-1, and denote the gray values of the corresponding pixel points of the two frames as fn(x, y) and fn-1(x, y). Subtract the gray values of the corresponding pixel points of the two frames and take the absolute value to obtain the difference image Dn, and the formula is as follows: Dn(x, y) = |fn(x, y) - fn-1(x, y)|.

[0094] Third, set a threshold T and perform binary processing on each pixel point one by one to obtain a binary image Rn'. Among them, the point with the initial gray value is the foreground (moving target) point, and the point with the gray value of 0 is the background point; perform connectivity analysis on the image Rn' to finally obtain an image Rn containing the complete moving target, and the formula is:

[0095] Step 1024: Generate a three-dimensional image of the bubbles on the liquid surface during the boiling process based on the generation and motion characteristics of the bubbles on the liquid surface during the boiling process using holographic imaging technology.

[0096] Since three-dimensional images can provide more comprehensive bubble motion information and can also perform spatial and temporal correspondence analysis with chemical composition characteristics and temperature characteristics, analyzing the regional relationship between the distribution of bubbles in three-dimensional space and changes in chemical composition, it is necessary to establish a three-dimensional image of bubbles. For example, subsequently, chemical composition data can be recorded as concentration-time-position data ((c,t,x,y,z)), where (c) is the chemical composition concentration, (t) is the time, and ((x,y,z)) is the position. The temperature characteristics are the same and will not be elaborated here.

[0097] In the actual application process, it is necessary to synchronize the images captured by the high-speed camera with the three-dimensional data generated by the holographic imaging system to ensure temporal consistency. At the same time, those skilled in the art can also use the method of combining multi-view geometry and deep learning, make full use of the geometric relationships of multiple views, and combine the three-dimensional structures predicted by the deep learning model to achieve high-precision three-dimensional reconstruction, fuse the two-dimensional image data captured by the high-speed camera with the holographic data, and reconstruct the three-dimensional image of the bubbles on the liquid surface, that is, the three-dimensional image of the bubbles.

[0098] Specifically, a deep learning model can be used for feature matching to ensure accurate matching of feature points between multiple views. d(xi,xi') = min(j)~d(xi,xj') (defines the distance from point xi to its nearest neighbor point xi'). Combining the depth prediction results and geometric registration information, a three-dimensional point cloud is generated, and a three-dimensional model is generated through a surface reconstruction algorithm. The formula is: where (u,v) are pixel coordinates, (cx,cy) are the optical center coordinates, (fx,fy) are the focal lengths, and d is the depth value. Those skilled in the art can also classify the characteristics of chemical composition changes in different boiling regions through a support vector machine (SVM) and perform time series prediction on the pH value through a long short-term memory network (LSTM), which will not be elaborated here.

[0099] Step 103: Register based on the position information in the temperature distribution image and the three-dimensional image of the bubbles, establish the spatial correspondence relationship between temperature and bubbles, and divide different boiling regions according to the preset temperature range.

[0100] Specifically, a computer can be used to digitally reconstruct the collected hologram to obtain a three-dimensional stereoscopic image of the bubbles, register the temperature distribution map generated from the infrared thermal imaging data with the position information of the holographic image, and establish a spatial correspondence relationship. Superimpose the three-dimensional image of the bubbles on the temperature distribution map to visually display the thermal distribution during the boiling process. Set a preset temperature range according to the temperature distribution map, and divide the boiling region into different temperature ranges, such as a high-temperature boiling region, a medium-temperature boiling region, and a low-temperature boiling region. Among them, the preset temperature range can be adjusted according to different traditional Chinese medicine extractions and will not be elaborated here.

[0101] In the actual application process, registration can be performed first based on the position information in the temperature distribution image and the bubble three-dimensional image to establish the spatial correspondence between temperature and bubbles. Then, based on the spatial correspondence, the bubble three-dimensional image and the temperature distribution map are superimposed, and the boiling liquid is divided into a high-temperature boiling zone, a medium-temperature boiling zone, and a low-temperature boiling zone according to a preset temperature range. By dividing different boiling regions, chemical composition characteristics can be collected specifically, and a strong correlation between temperature characteristics - chemical composition characteristics - bubble generation and movement characteristics can be established, which is convenient for improving the training effect of the subsequent multi-modal data on the model.

[0102] Step 104: Collect chemical composition data for different boiling regions of the liquid in the extraction device to obtain the chemical composition change characteristics of the liquid during the boiling process in different boiling regions.

[0103] Exemplarily, a chemical sensor can be installed inside the extraction device, and the chemical sensor is installed near the boiling region to collect the chemical composition data in the liquid during the boiling process in real time. A real-time analysis software is used to perform real-time analysis on the collected chemical sensor data to identify the change characteristics of the chemical components in the solution.

[0104] Specifically, taking the Chuanxiong extraction process as an example, the chemical sensor can be installed near the boiling region to ensure that the change of the chemical composition of the liquid during the boiling process can be accurately captured. Generally, it is recommended to install it in a region where the liquid flow is stable to avoid bubble interference. Among them, the region where the liquid flow is stable usually refers to the region where the liquid flow rate is low, the flow direction is consistent, and there is no violent vortex or turbulence. Such a region can reduce the generation and disturbance of bubbles, thereby improving the accuracy of the chemical sensor measurement. Secondly, avoiding bubble interference is because a large number of bubbles are generated during the boiling process, and these bubbles will interfere with the measurement of the sensor, especially near the liquid surface. Therefore, the sensor should be installed as far as possible from these regions when installing. Specific installation position example: the stable flow region near the container wall. Near the container wall, the liquid flow is usually relatively stable, and the sensor can be installed at a certain distance from the container wall (such as 5 - 10 cm) to ensure that the liquid flow rate here is low.

[0105] During the extraction process of Ligusticum chuanxiong, the chemical sensors applied include pH sensors and optical sensors. The pH sensor is used to detect the hydrogen ion concentration in the solution and identify changes in acidity and alkalinity. The optical sensor is used to detect color or fluorescence changes in the solution and identify specific compounds (such as ferulic acid). Among them, each sensor and device needs to be equipped with a high-precision time synchronization module, synchronized using GPS time or NTP (Network Time Protocol), and the output data of the sensors should all contain timestamps to record the time points of data acquisition. The data of all sensors can be centrally recorded through a central data recording system, so that for each time point, the corresponding three-dimensional bubble images, temperature distribution images, and chemical composition characteristic data can be aligned to form a multi-modal data record, and the multi-modal data is used for pattern recognition, machine learning and other analyses to identify different boiling states.

[0106] Step 105: Synchronize the temperature distribution images, three-dimensional bubble images, and chemical composition change characteristics collected at different time points using a unified timestamp, and establish the relationship between the chemical composition change characteristics and the temperature distribution characteristics of the corresponding boiling region in the temperature distribution image, as well as the relationship between the chemical composition change characteristics and the generation and movement characteristics of the bubbles in the corresponding boiling region in the three-dimensional bubble image for each time point to obtain multi-modal data.

[0107] Since there are many types of collected data and the data volume is also large, it is necessary to synchronize the different data collected at different time points using a unified timestamp through simulation, so as to construct a physical information neural network model of the boiling process by using fluid mechanics models, heat transfer and mass transfer theories to achieve dynamic evaluation of the extraction process state.

[0108] Among them, the combination with chemical composition adjustment is manifested in synchronizing the image data and chemical composition data with timestamps, extracting the bubble movement speed in different frame images, and analyzing whether the time points of bubble generation and the time points of chemical composition changes (the Ligusticum chuanxiong extract is acidic, with changes in hydrogen ion concentration; the optical sensor detects the presence and concentration of target compounds by measuring the absorbance of the solution) are consistent or correlated.

[0109] In the actual application process, a pH sensor and an optical sensor can be used to collect the hydrogen ion concentration and the concentration change of specific compounds in the solution respectively. The data collection frequency should be consistent with the collection frequency of the bubble image and the temperature image to ensure the time synchronization of multi-modal data. Then, a high-speed camera or a multi-view camera is used to collect the stereoscopic image of the bubble, capturing the generation, movement, and bursting process of the bubble. The image data should contain a timestamp to ensure the time alignment with the chemical sensor data. Then, by matching the images between consecutive frames, the movement speed and direction of the bubble are calculated, and features such as the bubble generation frequency and bubble bursting time at different time points are statistically analyzed. Finally, the chemical sensor data, bubble feature data, and temperature image data are aligned according to the timestamp to ensure the consistency of multi-modal data at each time point, and the chemical sensor data and bubble feature data are integrated into a multi-modal data set to form a feature vector containing multiple features.

[0110] Step 106: After performing dimensionality reduction processing on the multi-modal data using the principal component analysis algorithm, use the clustering algorithm to identify different boiling states and establish the corresponding relationship between the multi-modal data and the boiling states.

[0111] Since the multi-modal data is still relatively large, the principal component analysis algorithm (i.e., PCA) can be used to perform dimensionality reduction on the multi-modal data to extract the main features. PCA can map high-dimensional data to a low-dimensional space and retain the main change features of the data. In the feature space after PCA dimensionality reduction, a clustering algorithm (such as K-means) is used to identify different boiling states. According to the clustering results or classification results, different boiling states are classified, such as initial boiling, stable boiling, transitional boiling, etc. Through the above steps, the chemical sensor data can be combined with the bubble generation and movement characteristics to identify different boiling states.

[0112] Specifically, step 106 can specifically include the following steps:

[0113] Step 1061: Select the first k principal components according to the eigenvalue size.

[0114] Step 1062: Project the multi-modal data into the selected principal component space to generate the dimensionality-reduced data.

[0115] Step 1063: Cluster the dimensionality-reduced data to generate the classification results of different boiling states.

[0116] Exemplarily, taking the Chuanxiong extraction process as an example, first ensure the synchronization of the acquisition times of different modality data for joint analysis. Then, according to the eigenvalue magnitudes, select the top k principal components (k is usually the dimension after data dimensionality reduction). After that, project the original multi-modal data into the selected principal component space to generate the dimensionality-reduced data. The dimensionality-reduced data retains the main variation characteristics of the multi-modal data and can be used for subsequent clustering analysis. Finally, cluster the dimensionality-reduced data to generate classification results for different boiling states.

[0117] In the actual application process, K-means clustering can be performed on the dimensionality-reduced data, which is divided into three boiling states, including the low boiling point state, the high boiling point state, and the transition state. Among them, in the low boiling point state, the bubble generation frequency is low, the characteristic movement of the bubbles is slow, the pH value is stable, and the change in the content of the preset chemical components is not obvious; in the high boiling point state, the bubble generation frequency is high, the characteristic movement of the bubbles is fast, and both the pH value and the content of the preset chemical components show linear changes; in the transition state, the bubble generation frequency fluctuates between high and low, and both the pH value and the content of the preset chemical components show non-linear changes.

[0118] Step 1064: Use the eigenvectors in principal component analysis to analyze the contribution degree of each modality data to the identification of the boiling state.

[0119] Exemplarily, taking the Chuanxiong extraction process as an example, the eigenvectors in principal component analysis (PCA) can be used to analyze the contribution degree of each modality data to the identification of the boiling state. K-means clustering is performed on the dimensionality-reduced data, which is divided into three boiling states: "low boiling point state", "high boiling point state", and "transition state". Among them, in the "low boiling point state": the bubble generation frequency is low, the characteristic movement is slow, the pH value is stable, and the ferulic acid content slightly increases. In the "high boiling point state": the bubble generation frequency is high, the characteristic movement is fast, the pH value rapidly decreases, and the ferulic acid content significantly increases. In the "transition state": the bubble generation frequency fluctuates greatly, and both the pH value and the ferulic acid content show non-linear changes.

[0120] Step 107: Construct a physical information neural network model for the liquid in the extraction device based on the Navier-Stokes equation, fluid mechanics, and the corresponding relationship between the multi-modal data and the boiling state, and use the corresponding relationship between the multi-modal data and the boiling state to train and optimize the physical information neural network model to obtain the trained physical information neural network model.

[0121] Due to the simulation calculation technology based on physical modeling, the thermal state of the solvent can be evaluated in real time during the extraction process. Therefore, a fluid mechanics model can be used to perform simulation analysis on parameters such as the temperature, pressure, and steam volume inside the liquid in the extraction device in combination with the theory of heat transfer and mass transfer. This simulation calculation can reduce the uncertainty during the experiment and provide data support for the intelligent control of the device.

[0122] To construct a physical model of the boiling process, we can base it on the following three basic theories. The Navier-Stokes equation: describes the flow behavior of the liquid, including the bubble movement speed and evaporation rate. The heat conduction equation: describes the heat transfer in the liquid, considering the influence of liquid flow on heat transfer. Describes the mass transfer of compounds in the liquid, considering the influence of bubble generation on mass transfer. Then, input multi-modal data (chemical sensor data, bubble image data, temperature image data) into the physical model. Simulate the liquid flow, heat transfer, and mass transfer behaviors during the boiling process through the physical model, predict the changes in the liquid chemical composition under different boiling states, analyze the simulated data, extract key features (such as liquid flow rate, temperature distribution, compound concentration, etc.), and elaborate on their functions and application processes.

[0123] In the actual application process, a neural network model for extracting the physical information of the liquid in the extraction device can be constructed based on the Navier-Stokes equation, fluid mechanics, and the corresponding relationship between the multi-modal data and the boiling state. The formula is as follows:

[0124]

[0125] Where, is the rate of change of the fluid velocity u with respect to time t; is the convection term, used to represent the acceleration generated by the fluid due to its own motion; is the pressure gradient term, used to represent the acceleration generated due to the non-uniform distribution of the pressure p, and ρ is the density of the fluid; is the viscous term, used to represent the acceleration generated due to the viscosity v of the fluid, is the Laplace operator, used to represent the second derivative of the velocity; g is the acceleration due to gravity, used to represent the acceleration generated due to gravity.

[0126] The movement of the bubble is described by the Rayleigh-Plesset equation. The formula is as follows:

[0127]

[0128] Where, represents the acceleration, represents the velocity, represents the rate of change of pressure,... represents the density of the liquid, P 0 represents the pressure inside the bubble, σ represents the surface tension of the liquid, η represents the viscosity of the liquid. In addition, those skilled in the art can also construct a neural network model for extracting the physical information of the liquid in the extraction device in combination with specific existing extraction devices, which will not be elaborated here.

[0129] In addition, after establishing the physics-informed neural network model, the established physics-informed neural network model can be trained and optimized by using the corresponding relationship between the multi-modal data and the boiling state to obtain a trained physics-informed neural network model, which specifically includes the following steps:

[0130] First, based on the physics-informed neural network model, combining the physical constraints of multi-modal data and the data-driven loss function, the formula is as follows:

[0131] Loss = α 1 L d (u, y) + α 2 L p (u)

[0132] where α 1 is the weight of the data loss; α 2 is the weight of the physical loss; L d (u, y) is the data loss function, which measures the difference between the model prediction u and the observed data y; L p (u) is the physical loss function, which measures whether the model prediction u satisfies the physical law.

[0133] Then, the bubble image features, chemical composition features, and temperature features in the multi-modal data are weighted and fused through the multi-head attention mechanism to obtain the fused features.

[0134] Specifically, Query, Key, and Value can be mapped to multiple different subspaces respectively, and the attention weights are calculated independently and weighted fusion is performed within each subspace; among them, the generation of Query is to splice the bubble image features, chemical composition features, and temperature features into a long vector; corresponding Key vectors are generated for the bubble image features, chemical composition features, and temperature features respectively and spliced into a matrix; corresponding Value vectors are generated for each of the bubble image features, chemical composition features, and temperature features respectively and spliced into a matrix; the calculation formula of the attention weight Ai is: where Dk is the dimension of the Key vector.

[0135] Among them, the weighted fusion calculation process is as follows: the Value is weighted by the attention weight Ai to obtain the weighted fusion feature of each head i, and the formula is: Fi = Ai × Vi; the weighted fusion features Fi of all heads i are spliced together and fused through an additional linear layer Wo to obtain the final fused feature, and the formula is as follows: where x is the input feature, zq represents Query, k is the index of Key, q is the index of Query, M represents the number of heads of the multi-head attention, m represents the mth attention head, Ai represents the attention weight of the mth head, and W GM X O Actually, it is Value, W G M It is the result after the attention is applied to Value, which is linearly transformed to obtain the output results of different heads.

[0136] Finally, use the fused features to train and optimize the physics-informed neural network model, so that the physics-informed neural network model can focus on the most important part of the information for the boiling state.

[0137] As can be seen from the above, in order to further improve the expressive power and robustness of the physics-informed neural network model, this embodiment introduces the multi-head attention mechanism. The core idea of this mechanism is to map Query, Key, and Value into multiple different subspaces respectively, and independently calculate the attention weights and perform weighted fusion within each subspace. In this way, the model can capture the complex relationships between data from multiple perspectives. In the multi-head attention mechanism, each head generates a set of independent attention weights and fused features. Finally, these features are concatenated together and fused through an additional linear layer to produce the final output. Through the above steps, the features of the three modalities of image, chemical composition, and temperature are successfully fused together, and the attention mechanism is used to dynamically adjust their importance. This fusion method not only improves the performance of the physics-informed neural network model, but also enhances the understanding ability of the physics-informed neural network model for multi-modal data.

[0138] Step 108: Real-time obtain any one or more of the temperature distribution image of the liquid in the extraction device, the three-dimensional bubble image, and the chemical composition change characteristics, and input them into the trained physics-informed neural network model to predict the boiling state of the liquid in the extraction device.

[0139] By training and optimizing the physics-informed neural network model in the above manner, the physics-informed neural network model can identify any one or more of the temperature distribution image of the liquid in the extraction device, the three-dimensional bubble image, and the chemical composition change characteristics, and give a prediction of the boiling state of the liquid in the extraction device, with the advantage of high prediction accuracy. Operators can accurately control the extraction process according to the prediction results, thereby improving the quality of the final product.

[0140] This embodiment studies the application of multi-modal technology in boiling state detection, establishes a physics-informed neural network model based on boiling detection, and provides systematic support for the control of the boiling state in the extraction process.

[0141] Embodiment 2

[0142] Figure 3Schematic structural diagram of a detection device according to Embodiment 2 of the present invention, see Figure 3 , this detection device includes:

[0143] A thermal imaging module 201, configured to perform thermal imaging acquisition on the boiling liquid in the extraction device to obtain a temperature distribution image of the liquid during the boiling process.

[0144] An image acquisition module 202, configured to perform image acquisition on the surface of the boiling liquid in the extraction device from different angles to obtain the generation and movement characteristics of bubbles on the liquid surface during the boiling process, and generate a three-dimensional bubble image on the liquid surface during the boiling process by using holographic imaging technology based on the generation and movement characteristics of bubbles on the liquid surface during the boiling process;

[0145] A processing module 203, configured to perform registration based on the position information in the temperature distribution image and the three-dimensional bubble image, establish a spatial correspondence relationship between temperature and bubbles, and divide different boiling regions according to a preset temperature range.

[0146] A chemical composition acquisition module 204, configured to perform chemical composition data acquisition on different boiling regions of the liquid in the extraction device to obtain the chemical composition change characteristics of the liquid during the boiling process in different boiling regions.

[0147] A synchronization module 205, configured to synchronize the temperature distribution image, the three-dimensional bubble image, and the chemical composition change characteristics collected at different time points with a unified timestamp, and establish a relationship between the chemical composition change characteristics and the temperature distribution characteristics of the corresponding boiling region in the temperature distribution image, and between the chemical composition change characteristics and the generation and movement characteristics of bubbles in the corresponding boiling region in the three-dimensional bubble image for each time point, to obtain multimodal data.

[0148] An analysis module 206, configured to perform dimensionality reduction processing on the multimodal data by using a principal component analysis algorithm, then use a clustering algorithm to identify different boiling states, and establish a correspondence relationship between the multimodal data and the boiling states.

[0149] A construction module 207, configured to construct a physical information neural network model of the liquid in the extraction device based on the Navier-Stokes equation, fluid mechanics, and the correspondence relationship between the multimodal data and the boiling states, and use the correspondence relationship between the multimodal data and the boiling states to train and optimize the physical information neural network model to obtain a trained physical information neural network model.

[0150] A prediction module 208 is configured to obtain in real time any one or more of a temperature distribution image of the liquid in the extraction device, a three-dimensional bubble image, and a chemical composition change feature, and input the obtained feature(s) into the trained physical information neural network model to predict the boiling state of the liquid in the extraction device.

[0151] The detection device described in this embodiment is an optimization based on the above embodiment, and can execute the detection method provided in any embodiment of the present invention, and has the corresponding functional modules and beneficial effects for executing the method.

[0152] Embodiment III

[0153] Figure 4 It is a schematic structural diagram of a system provided in Embodiment III of the present invention; Figure 4 It shows a block diagram of an exemplary terminal system suitable for implementing the embodiments of the present invention. Figure 4 The terminal system shown is merely an example and should not impose any limitation on the functions and usage scope of the embodiments of the present invention.

[0154] As Figure 4 shown, the terminal 12 is presented in the form of a general-purpose computing device. The components of the terminal 12 may include, but are not limited to: one or more processors or processing units 16, a system memory 28, and a bus 18 connecting different system components (including the system memory 28 and the processing unit 16).

[0155] The bus 18 represents one or more of several types of bus structures, including a memory bus or a memory controller, a peripheral bus, a graphics acceleration port, a processor, or a local bus using any of the multiple bus structures. For example, these architectures include, but are not limited to, Industry Standard Architecture (ISA) bus, Micro Channel Architecture (MAC) bus, Enhanced ISA bus, Video Electronics Standards Association (VESA) local bus, and Peripheral Component Interconnect (PCI) bus.

[0156] The terminal 12 typically includes a variety of computer system-readable media. These media can be any available media accessible by the terminal 12, including volatile and non-volatile media, removable and non-removable media.

[0157] The system memory 28 may include computer system-readable media in the form of volatile memory, such as random access memory (RAM) 30 and / or cache memory 32. The terminal 12 may further include other removable / non-removable, volatile / non-volatile computer system storage media. By way of example only, a storage system 34 may be used to read and write non-removable, non-volatile magnetic media ( Figure 4 not shown, commonly referred to as a "hard disk drive"). Although Figure 4Not shown in the figure, a disk drive for reading and writing to a removable non-volatile disk (such as a "floppy disk") and an optical disk drive for reading and writing to a removable non-volatile optical disk (such as a CD-ROM, DVD-ROM or other optical medium) can be provided. In these cases, each drive can be connected to the bus 18 through one or more data medium interfaces. The memory 28 may include at least one program product having a set (such as at least one) of program modules configured to perform the functions of the embodiments of the present invention.

[0158] A program / utilities 40 having a set (at least one) of program modules 42 can be stored, for example, in the memory 28. Such program modules 42 include, but are not limited to, an operating system, one or more application programs, other program modules, and program data. Each or some combination of these examples may include the implementation of a network environment. The program modules 42 generally perform the functions and / or methods in the embodiments described in the present invention.

[0159] The terminal 12 can also communicate with one or more external devices 14 (such as a keyboard, a pointing device, a display 24, etc.), and can also communicate with one or more devices that enable a user to interact with the terminal 12, and / or communicate with any device that enables the terminal 12 to communicate with one or more other computing devices (such as a network card, a modem, etc.). Such communication can be carried out through the input / output (I / O) interface 22. Also, the terminal 12 can communicate with one or more networks (such as a local area network (LAN), a wide area network (WAN), and / or a public network, such as the Internet) through the network adapter 20. As shown in the figure, the network adapter 20 communicates with other modules of the terminal 12 through the bus 18. It should be understood that although not shown in the figure, other hardware and / or software modules can be used in combination with the terminal 12, including but not limited to: microcode, device drivers, redundant processing units, external disk drive arrays, RAID systems, tape drives, and data backup storage systems, etc.

[0160] The processing unit 16 executes various functional applications and data processing by running programs stored in the system memory 28, such as implementing the boiling detection method in the Chuanxiong extraction process provided by the embodiments of the present invention.

[0161] Embodiment Four

[0162] Embodiment Four of the present invention also provides a storage medium containing computer-executable instructions that, when executed by a computer processor, are used to execute any one of the boiling detection methods in the Chuanxiong extraction process provided by the above embodiments.

[0163] The computer storage medium of the embodiments of the present invention may adopt any combination of one or more computer-readable media. The computer-readable media may be computer-readable signal media or computer-readable storage media. The computer-readable storage media may be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination of the above. More specific examples (non-exhaustive list) of the computer-readable storage media include: an electrical connection having one or more wires, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the above. In this document, the computer-readable storage media may be any tangible medium that contains or stores a program, and the program may be used by or in combination with an instruction execution system, apparatus, or device.

[0164] The computer-readable signal media may include data signals propagated in a baseband or as part of a carrier wave, which carry computer-readable program codes. Such propagated data signals may take various forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination of the above. The computer-readable signal media may also be any computer-readable media other than the computer-readable storage media, and the computer-readable media may send, propagate, or transmit a program for use by or in combination with an instruction execution system, apparatus, or device.

[0165] The program codes contained on the computer-readable media may be transmitted by any appropriate medium, including but not limited to wireless, wire, optical cable, RF, etc., or any suitable combination of the above.

[0166] The computer program codes for performing the operations of the present invention may be written in one or more programming languages or combinations thereof. The programming languages include object-oriented programming languages such as Java, Smalltalk, C++, and also include conventional procedural programming languages such as the "C" language or similar programming languages. The program codes may be executed entirely on the user's computer, partially on the user's computer, executed as an independent software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In the case of a remote computer, the remote computer may be connected to the user's computer through any type of network, including a local area network (LAN) or a wide area network (WAN), or may be connected to an external computer (for example, by using an Internet service provider to connect through the Internet).

[0167] Note that the above is only a preferred embodiment of the present invention and the technical principles applied. Those skilled in the art will understand that the present invention is not limited to the specific embodiments described herein, and various obvious changes, re-adjustments and substitutions can be made by those skilled in the art without departing from the protection scope of the present invention. Therefore, although the present invention has been described in more detail through the above embodiments, the present invention is not limited to the above embodiments only. Without departing from the concept of the present invention, more other equivalent embodiments can be included, and the scope of the present invention is determined by the scope of the appended claims.

Claims

1. A boiling detection method in the process of extracting Ligusticum chuanxiong, characterized in that: include: Perform thermal imaging of the boiling liquid in the extraction device to obtain a temperature distribution image of the liquid during the boiling process; Capturing images of the boiling liquid surface in the extraction device from different angles to obtain the generation and movement characteristics of bubbles on the liquid surface during the boiling process, and generating a three-dimensional image of bubbles on the liquid surface during the boiling process using holographic imaging technology based on the generation and movement characteristics of bubbles on the liquid surface during the boiling process; Based on the position information of the temperature distribution image and the bubble stereo image, the spatial correspondence between the temperature and the bubble is established, and different boiling areas are divided according to the preset temperature range; Collect chemical composition data of different boiling areas of the liquid in the extraction device to obtain the chemical composition change characteristics of the liquid during the boiling process in different boiling areas; The temperature distribution images, bubble stereo images, and chemical composition change characteristics collected at different time points are synchronized using a unified timestamp, and a relationship between the chemical composition change characteristics and the temperature distribution characteristics of the corresponding boiling area in the temperature distribution image, and between the chemical composition change characteristics and the generation and movement characteristics of the bubbles in the corresponding boiling area in the bubble stereo image is established for each time point to obtain multimodal data; After performing dimensionality reduction processing on the multimodal data using a principal component analysis algorithm, a clustering algorithm is used to identify different boiling states, and a corresponding relationship between the multimodal data and the boiling state is established; Based on the Navier-Stokes equations, fluid mechanics, and the correspondence between the multimodal data and the boiling state, a physical information neural network model of the liquid in the extraction device is constructed, and the correspondence between the multimodal data and the boiling state is used to train and optimize the physical information neural network model to obtain a trained physical information neural network model; The temperature distribution image, the bubble stereo image, and any one or more of the chemical composition change characteristics of the liquid in the extraction device are acquired in real time and input into the trained physical information neural network model to predict the boiling state of the liquid in the extraction device.

2. The method according to claim 1, characterized in that The image acquisition is performed on the surface of the boiling liquid in the extraction device from different angles to obtain the generation and movement characteristics of bubbles on the surface of the liquid during the boiling process, and based on the generation and movement characteristics of bubbles on the surface of the liquid during the boiling process, a three-dimensional image of bubbles on the surface of the liquid during the boiling process is generated using holographic imaging technology, including: Capturing images of the liquid surface in the extraction device from different angles, and preprocessing the captured images to obtain preprocessed images; Performing edge detection on the preprocessed image based on the Canny edge detection algorithm to obtain the bubble edge; The pre-processed image is segmented based on the Otsu algorithm and the bubble edge to obtain a number of bubbles, and the movement trajectory of the bubbles is tracked based on the inter-frame difference method to obtain the generation and movement characteristics of the bubbles on the liquid surface during the boiling process; Based on the generation and movement characteristics of bubbles on the liquid surface during the boiling process, a three-dimensional image of bubbles on the liquid surface during the boiling process is generated using holographic imaging technology.

3. The method according to claim 1, characterized in that The registering based on the temperature distribution image and the position information in the bubble stereo image, establishing a spatial correspondence between temperature and bubbles, and dividing different boiling areas according to preset temperature intervals, includes: Based on the temperature distribution image and the position information in the bubble stereo image, a spatial correspondence between temperature and bubbles is established; Based on the spatial correspondence, the bubble stereoscopic image is superimposed on the temperature distribution map, and the boiling liquid is divided into a high-temperature boiling zone, a medium-temperature boiling zone and a low-temperature boiling zone according to a preset temperature range.

4. The method according to claim 1, characterized in that: After the multimodal data is subjected to dimensionality reduction processing using the principal component analysis algorithm, different boiling states are identified using a clustering algorithm, and a corresponding relationship between the multimodal data and the boiling state is established, including: According to the size of the eigenvalue, select the first k principal components; Project the multimodal data into the selected principal component space to generate dimensionally reduced data; Clustering the dimensionally reduced data to generate classification results of different boiling states; The contribution of each modal data to the boiling state identification is analyzed using the eigenvectors in principal component analysis.

5. The method according to claim 1, characterized in that The physical information neural network model of the liquid in the extraction device is constructed based on the Navier-Stokes equations, fluid mechanics, and the corresponding relationship between the multimodal data and the boiling state, and the physical information neural network model is trained and optimized using the corresponding relationship between the multimodal data and the boiling state to obtain the trained physical information neural network model, including: Based on the Navier-Stokes equations, fluid mechanics, and the corresponding relationship between the multimodal data and the boiling state, a physical information neural network model of the liquid in the extraction device is constructed, and the formula is as follows: in, is the rate of change of the velocity u of the fluid with time t; is the convection term, which is used to represent the acceleration of the fluid due to its own motion; is the pressure gradient term, which is used to represent the acceleration due to the uneven distribution of pressure p, and ρ is the density of the fluid; is the viscous term, which is used to represent the acceleration due to the viscosity v of the fluid, is the Laplace operator, which is used to represent the second derivative of velocity; g is the gravitational acceleration, which is used to represent the acceleration due to gravity. The motion of the bubbles is described by the Rayleigh-Plesset equation, which is as follows: in, represents acceleration, Indicates speed, represents the pressure change rate, ... represents the density of the liquid, P0 represents the pressure inside the bubble, σ represents the surface tension of the liquid, and η represents the viscosity of the liquid; The physical information neural network model is trained and optimized using the correspondence between the multimodal data and the boiling state to obtain a trained physical information neural network model.

6. The method according to claim 5, characterized in that The method of training and optimizing the physical information neural network model by using the correspondence between the multimodal data and the boiling state to obtain the trained physical information neural network model includes: Based on the physical information neural network model, combined with multimodal data physical constraints and data-driven loss function, the formula is as follows: Loss=α1L d (u,y)+α2L p (u) Among them, α1 is the weight of data loss; α2 is the weight of physical loss; L d (u,y) is the data loss function, which measures the difference between the model prediction u and the observed data y; L p (u) is the physical loss function, which measures whether the model's prediction u satisfies the laws of physics; The bubble image features, chemical composition features and temperature features in the multimodal data are weightedly fused through a multi-head attention mechanism to obtain fused features; The physical information neural network model is trained and optimized by utilizing the fusion features, so that the physical information neural network model can focus on a part of information that is most important to the boiling state.

7. The method according to claim 6, characterized in that The multi-head attention mechanism is used to weightedly fuse the bubble image features, chemical composition features and temperature features in the multimodal data to obtain fused features, including: Map the query, key and value to multiple different subspaces, and calculate the attention weight and perform weighted fusion independently in each subspace; the query is generated by concatenating the bubble image features, chemical composition features and temperature features into a long vector; generate corresponding key vectors for the bubble image features, chemical composition features and temperature features, and concatenate them into a matrix; generate corresponding value vectors for each bubble image feature, chemical composition feature and temperature feature, and concatenate them into a matrix; The calculation formula of attention weight Ai is: Where Dk is the dimension of the Key vector; The weighted fusion calculation process is as follows: Use the attention weight Ai to weight Value and obtain the weighted fusion feature of each head i. The formula is: Fi = Ai × Vi; The weighted fusion features Fi of all heads i are concatenated together and fused through an additional linear layer Wo to obtain the final fusion feature. The formula is as follows: Among them, x is the input feature, zq represents Query, k is the index of Key, q is the index of Query, M represents the number of heads of multi-head attention, m represents the number of attention heads, Ai represents the weight of the mth attention head, and W G MX O Actually it is Value, W G M is the result after attention is applied to Value, which is linearly transformed to obtain the output results of different heads.

8. A detection device, characterized in that: include: A thermal imaging module is used to collect thermal images of the boiling liquid in the extraction device to obtain a temperature distribution image of the liquid during the boiling process; An image acquisition module is used to acquire images of the surface of the boiling liquid in the extraction device from different angles to obtain the generation and movement characteristics of bubbles on the surface of the liquid during the boiling process, and based on the generation and movement characteristics of bubbles on the surface of the liquid during the boiling process, a three-dimensional image of bubbles on the surface of the liquid during the boiling process is generated by using holographic imaging technology; A processing module, used for registering the temperature distribution image with the position information in the bubble stereo image, establishing a spatial correspondence between temperature and bubbles, and dividing different boiling areas according to preset temperature intervals; A chemical composition acquisition module is used to collect chemical composition data of different boiling areas of the liquid in the extraction device to obtain the chemical composition change characteristics of the liquid during the boiling process in different boiling areas; A synchronization module is used to synchronize the temperature distribution images, bubble stereo images, and chemical composition change characteristics collected at different time points using a unified timestamp, and establish a relationship between the chemical composition change characteristics and the temperature distribution characteristics of the corresponding boiling area in the temperature distribution image, and between the chemical composition change characteristics and the generation and movement characteristics of bubbles in the corresponding boiling area in the bubble stereo image at each time point, to obtain multimodal data; An analysis module, configured to use a principal component analysis algorithm to perform dimensionality reduction processing on the multimodal data, and then use a clustering algorithm to identify different boiling states, and establish a corresponding relationship between the multimodal data and the boiling state; Establishing a module for constructing a physical information neural network model of the liquid in the extraction device based on the Navier-Stokes equations, fluid mechanics, and the corresponding relationship between the multimodal data and the boiling state, and training and optimizing the physical information neural network model using the corresponding relationship between the multimodal data and the boiling state to obtain a trained physical information neural network model; The prediction module is used to obtain in real time any one or more of the temperature distribution image, bubble stereo image, and chemical composition change characteristics of the liquid in the extraction device, and input them into the physical information neural network model after training to predict the boiling state of the liquid in the extraction device.

9. A detection system, characterized in that: include: one or more processors; A storage device for storing one or more programs; A camera, used to collect images; Chemical sensors, used to collect chemical components; A display, used for displaying the test results; When the one or more programs are executed by the one or more processors, the one or more processors implement the boiling detection method in the Ligusticum chuanxiong extraction process as described in any one of claims 1-7.

10. A storage medium comprising computer executable instructions, wherein the computer executable instructions are used to execute the boiling detection method in the extraction process of Chuanxiong as claimed in any one of claims 1 to 7 when executed by a computer processor.

Images