Intelligent monitoring method and device for reaction kettle operation process based on visual analysis

Abstract

The application discloses a reaction kettle operation process intelligent monitoring method and device based on visual analysis. The method comprises the following steps: acquiring a video sequence of the inside of a reaction kettle in a current period collected by a 5G explosion-proof camera arranged in the reaction kettle according to a preset period, so as to obtain a to-be-analyzed kettle video sequence; inputting the to-be-analyzed kettle video sequence into a chemical product visual analysis model for fuzzy video processing and extracting product visual information of the current period, so as to obtain the product visual information; determining whether the time distribution of a process operation node in the current period is consistent with the time distribution in preset process operation node information, so as to obtain an analysis result; when the analysis result is inconsistent, creating corresponding operation process prompt information and sending the operation process prompt information to a client. The method can effectively improve the precision control of the production process and the stability of the product quality, and is particularly suitable for the production requirements of high-end chemical products.

Description

TECHNICAL FIELD

[0001] The present application relates to computers, and more particularly to a reaction kettle operation process intelligent monitoring method and device based on visual analysis. BACKGROUND

[0002] In the field of chemical production, the reaction kettle, as the core device of chemical reactions and production processes, has an extremely complex and crucial internal operation process. These operations include the addition of various raw materials and the precise control of reaction conditions such as temperature, pressure, and stirring speed. Accurate control and timely adjustment of these factors directly affect the quality of the final product.

[0003] In current technical practice, operators usually rely on observing the transparent viewing window installed on the reaction kettle to monitor the specific process. This method highly depends on the professional experience and personal attention level of the operator, and there is a risk of inaccurate monitoring due to operator fatigue or negligence. In addition, this visual-based manual monitoring method is difficult to accurately capture and record key details in the operation process, limiting the accuracy and timeliness of monitoring data, thereby affecting the consistency and stability of product quality.

[0004] Therefore, it is necessary to design a new method to effectively improve the precise control of the production process and the stability of product quality, especially suitable for the production requirements of high-end chemical products; to solve the technical problems in the prior art that the method of monitoring the operation process of the reaction kettle by the transparent viewing window through manual operation not only highly depends on the experience and attention of the operator, but also is difficult to accurately capture and record key process details in real time, thereby affecting the consistency and stability of product quality. SUMMARY

[0005] The purpose of the present application is to overcome the defects of the prior art and provide a reaction kettle operation process intelligent monitoring method and device based on visual analysis.

[0006] To achieve the above-mentioned purpose, the present application adopts the following technical solution: a reaction kettle operation process intelligent monitoring method based on visual analysis, comprising:

[0007] acquiring a video sequence of the inside of the reaction kettle in the current period collected by a 5G explosion-proof camera deployed in the reaction kettle according to a preset period, to obtain a to-be-analyzed kettle video sequence;

[0008] inputting the to-be-analyzed kettle video sequence into a chemical product visual analysis model for fuzzy video processing and extracting product visual information of the current period, to obtain product visual information; wherein the chemical product visual analysis model is obtained by model training using a neural network technology with operation process video segments of each process operation node in the historical video sequence as a sample set;

[0009] analyzing the process operation nodes in the current period based on the preset process operation node information and the product visual information to determine whether the time distribution of the process operation nodes in the current period is consistent with the time distribution in the preset process operation node information, to obtain an analysis result;

[0010] when the analysis result is that the time distribution of the process operation nodes in the current period is inconsistent with the time distribution in the preset process operation node information, creating corresponding operation process prompt information and sending it to the client.

[0011] A further technical solution thereof is that the training process of the chemical product visual analysis model comprises:

[0012] acquiring a historical video sequence reflecting the whole-period operation process of the reaction kettle;

[0013] determining the time period of each historical process operation node and intercepting the corresponding operation process video segment from the historical video sequence;

[0014] constructing a sample set for model training based on the operation process video segment;

[0015] creating an initial model using neural network technology;

[0016] inputting the sample set into the initial model to train the initial model, adjusting the initial model until the loss value is minimized, to obtain a chemical product visual analysis model.

[0017] A further technical solution thereof is that the preset period is a specific time period formed by dividing the production process by time, or a time period set by customization.

[0018] A further technical solution thereof is that the process operation node includes a time point with specific product visual information or an operation switching time point; one preset period contains one or more process operation nodes.

[0019] A further technical solution thereof is that the analysis of the process operation nodes in the current period based on the preset process operation node information and the product visual information to determine whether the time distribution of the process operation nodes in the current period is consistent with the time distribution in the preset process operation node information, to obtain an analysis result, comprises:

[0020] analyzing the process operation nodes in the current period based on the preset process operation node information and the product visual information to obtain process operation node information; wherein the process operation node information comprises process operation nodes, time positions, and product visual description parameters corresponding to the process operation nodes;

[0021] According to the process operation node information, whether the time distribution of the process operation node in the current period is consistent with the time distribution in the preset process operation node information is determined to obtain an analysis result.

[0022] Further, the product visual description parameters include color parameters and state parameters, wherein the color parameters include RGB values and color change rates, and the state parameters include state labels and state change rates.

[0023] Further, according to the process operation node information, whether the time distribution of the process operation node in the current period is consistent with the time distribution in the preset process operation node information is determined to obtain an analysis result, including:

[0024] According to the process operation node information, the product visual description parameters corresponding to the process operation node are compared with the parameter range corresponding to the preset process operation node information, and a matching result of each index is calculated.

[0025] According to the matching result, whether the time distribution of the process operation node in the current period is consistent with the time distribution in the preset process operation node information is comprehensively evaluated to obtain an analysis result.

[0026] Further, according to the process operation node information, the product visual description parameters corresponding to the process operation node are compared with the parameter range corresponding to the preset process operation node information, and a matching result of each index is calculated, including:

[0027] According to the process operation node information, whether the color RGB value of the current material is in the color parameter range corresponding to the preset process operation node information is compared, and a color matching result is generated.

[0028] According to the process operation node information, whether the color change rate of the current material conforms to the color change rate range corresponding to the preset process operation node information is judged, and a color change rate matching result is generated.

[0029] According to the process operation node information, whether the current material temperature is in the temperature range corresponding to the preset process operation node information is determined to obtain a temperature matching result.

[0030] According to the process operation node information, a matching degree between the current material state label and the state label corresponding to the preset process operation node information is calculated to obtain a state label matching result.

[0031] According to the process operation node information, it is confirmed whether the state change rate of the current material is within the state change rate range corresponding to the preset process operation node information, and a state change rate matching result is obtained.

[0032] The matching results of the indicators include a color matching result, a color change rate matching result, a temperature matching result, a state label matching result, and a state change rate matching result.

[0033] Further technical solutions of the application are as follows: after analyzing the process operation node in the current period based on the preset process operation node information and the product visual information to determine whether the time distribution of the process operation node in the current period is consistent with the time distribution in the preset process operation node information to obtain an analysis result, the method further includes:

[0034] When the analysis result is that the time distribution of the process operation node in the current period is consistent with the time distribution in the preset process operation node information, a video sequence of the inside of the reaction kettle in the current period collected by the 5G explosion-proof camera arranged in the reaction kettle is obtained according to a preset period to obtain a to-be-analyzed kettle video sequence.

[0035] The application also provides a reaction kettle operation process intelligent monitoring device based on visual analysis, which includes:

[0036] An acquisition unit is configured to obtain a video sequence of the inside of the reaction kettle in the current period collected by the 5G explosion-proof camera arranged in the reaction kettle according to a preset period to obtain a to-be-analyzed kettle video sequence.

[0037] An extraction unit is configured to input the to-be-analyzed kettle video sequence into a chemical product visual analysis model to perform fuzzy video processing and extract product visual information of the current period to obtain the product visual information, wherein the chemical product visual analysis model is obtained by using a neural network technology to train a model with operation process video segments of each process operation node in a historical video sequence as a sample set.

[0038] An analysis unit is configured to analyze a process operation node in the current period based on preset process operation node information and the product visual information to determine whether the time distribution of the process operation node in the current period is consistent with the time distribution in the preset process operation node information to obtain an analysis result.

[0039] A creation unit is configured to create corresponding operation process prompt information and send the information to a client when the analysis result is that the time distribution of the process operation node in the current period is inconsistent with the time distribution in the preset process operation node information.

[0040] Compared with the prior art, the present application has the beneficial effects that: the present application automatically acquires video sequences inside the reaction kettle according to a preset period, and processes these video data using a deep learning algorithm to extract key product visual information. Based on pre-defined process operation node information, the system can analyze the operation node time distribution in the current production cycle in real time, judge whether it meets the expected standard, and generate and send operation prompt information to relevant personnel immediately when deviation is detected, to ensure timely adjustment. Compared with the traditional method relying on manual monitoring, this automatic solution not only reduces the dependence on the experience and attention of the operator, but also can accurately capture and record key process details, realize precise control of the production process and significant improvement of product quality stability, and is particularly suitable for strict production requirements of high-end chemical products. Therefore, the scheme effectively solves the problem of difficulty in ensuring product quality consistency and stability in the prior art, represents a more efficient and reliable production process monitoring means, and realizes automatic and intelligent monitoring of the operation process in the reaction kettle.

[0041] The present application will be further described below in combination with the drawings and specific embodiments. BRIEF DESCRIPTION OF DRAWINGS

[0042] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings needed in the embodiment description will be briefly introduced. Obviously, the drawings in the following description are some embodiments of the present application, and other drawings can also be obtained by those skilled in the art without creative labor.

[0043] Figure 1 The flowchart of the reaction kettle operation process intelligent monitoring method based on visual analysis provided by the embodiment of the present application is shown in the figure.

[0044] Figure 2 The schematic diagram of the video sequence provided by the embodiment of the present application is shown in the figure.

[0045] Figure 3 The sub-flowchart of the reaction kettle operation process intelligent monitoring method based on visual analysis provided by the embodiment of the present application is shown in the figure. Figure 1

[0046] Figure 4 The sub-flowchart of the reaction kettle operation process intelligent monitoring method based on visual analysis provided by the embodiment of the present application is shown in the figure. Figure 2

[0047] Figure 5 The sub-flowchart of the reaction kettle operation process intelligent monitoring method based on visual analysis provided by the embodiment of the present application is shown in the figure. Figure 3

[0048] Figure 6 ​​​A schematic block diagram of the computer device provided by the embodiment of the present application is shown in FIG. 6.

[0049] Figure 7 A schematic block diagram of the computer device provided by the embodiment of the present application is shown in FIG. 6. DETAILED DESCRIPTION

[0050] The technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the accompanying drawings in the embodiments of the present application. Obviously, the described embodiments are only some of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative efforts should fall within the protection scope of the present application.

[0051] It should be understood that, when used in the specification and the appended claims, the terms "comprise" and "include" indicate the presence of the described features, integers, steps, operations, elements, and / or components, but do not exclude one or more other features, integers, steps, operations, elements, components, and / or groups thereof.

[0052] It should also be understood that the terms used in the specification of the present application are only for the purpose of describing the specific embodiments and are not intended to limit the present application. As used in the specification and the appended claims of the present application, the singular forms "a", "an" and "the" are intended to include the plural forms, unless the context clearly indicates otherwise.

[0053] It should be further understood that the term "and / or" used in the specification and the appended claims of the present application means any combination of one or more of the associated listed items and all possible combinations, and includes these combinations.

[0054] Please refer to Figure 1 , Figure 1The schematic flowchart of the intelligent monitoring method for the reaction kettle operation process based on visual analysis provided by the embodiment of the present application. The intelligent monitoring method for the reaction kettle operation process based on visual analysis is applied to a server which interacts with a camera. The server automatically acquires video sequences inside the reaction kettle and uses neural network technology to process fuzzy videos and extract key product visual information, thereby realizing accurate identification of each process operation node in the production process and real-time analysis of the time distribution. The method can effectively improve the accurate control of the production process and the stability of the product quality, and is particularly suitable for the production requirements of high-end chemical products. Compared with the traditional method relying on manual monitoring, the present application not only reduces the dependence on the experience of the operator, but also accurately captures and records the key process details in real time, thereby overcoming the problems of poor product quality consistency and insufficient stability caused by human factors in the prior art. In addition, when the actual operation does not meet the preset standard, the system will automatically generate prompt information to feed back to the operator, further ensuring the accuracy and safety of the production.

[0055] Figure 1 The flowchart of the intelligent monitoring method for the reaction kettle operation process based on visual analysis provided by the embodiment of the present application. As shown in Figure 1 , the method comprises the following steps S110-S140.

[0056] S110, acquire the video sequence inside the reaction kettle in the current period collected by the 5G explosion-proof camera deployed in the reaction kettle according to a preset period, to obtain a to-be-analyzed kettle video sequence.

[0057] In the present embodiment, the to-be-analyzed kettle video sequence refers to the video stream inside the reaction kettle collected by the 5G explosion-proof camera in the preset period, which is used to monitor and analyze the process operation node state change in the chemical product production process. As shown in the picture. Figure 2

[0058] Considering that the chemical production environment usually has the risk of being flammable and explosive, a 5G camera with explosion-proof function is used for video acquisition. The camera has the functions of custom cooling back-blowing, circulating cooling water, supporting variable focal mode, explosion-proof, etc. Such a camera not only ensures safe work in a high-risk environment, but also realizes real-time and efficient monitoring of the internal condition of the reaction kettle by using the high speed and low delay characteristics of the 5G network.

[0059] ​According to the requirements of the actual production process, the entire production cycle is divided into several specific time periods (i.e., preset periods), and different process operation nodes or product state changes are focused on in each period. The setting of these time periods can be based on time proportioning or adjusted flexibly according to specific process operation needs. For example, some key process nodes may need more frequent monitoring, while other stages can appropriately extend the monitoring period.

[0060] In each preset period, the 5G explosion-proof camera automatically captures the video stream inside the reaction kettle and transmits it to the server as the data source for subsequent analysis. These video sequences contain rich visual information, such as color changes and physical state transitions of materials, which are important basis for evaluating whether the process operation is normal.

[0061] The kettle video sequences obtained through the above steps will be further input into a pre-trained chemical product visual analysis model to identify and extract key visual information in the video, thereby achieving precise control and quality assurance of the production process. This method is particularly suitable for the production monitoring of high-end chemical products, as the production of such products has extremely strict requirements on environmental conditions and operation precision.

[0062] In this embodiment, the preset period is a specific time period formed by dividing the production process by time, or a custom-set time period. A series of specific time periods are obtained by systematically analyzing and dividing the entire production cycle. These time periods can be set according to the characteristics of the production process, the change law of the product state, and the monitoring needs. The establishment of the preset period aims to effectively capture and record the key visual information changes at each stage of the production process, facilitating subsequent analysis and quality control.

[0063] Specifically, the preset period can be determined in two ways: one is to proportionally divide the entire production cycle based on time, which is suitable for scenarios with relatively uniform state change patterns or low requirements for time distribution; the other is to customize according to specific process operation nodes, which is more flexible and can be customized according to different production process characteristics. For example, in a typical chemical production process, process operation node 1 to process operation node 3 may correspond to white, light yellow to other color changes, respectively, and the required attention time and monitoring frequency between each sub-node may be different. Therefore, the time period for process operation node 1 to process operation node 3 can be divided into period 1, while the operation of process operation node 4 is separately as a period 2, and so on.

[0064] By setting the preset period in this way based on the combination of time and process operation nodes, not only can the monitoring efficiency be improved and unnecessary computing resource consumption be reduced, but also the most accurate production status information can be obtained at critical moments, thereby providing strong protection for high-quality production of high-end chemical products. In addition, compared to the high resource consumption and frequent false positives caused by real-time monitoring and early warning, periodic monitoring and analysis is more economical and efficient, especially suitable for chemical reaction processes that require extremely high production control and frequent changes in product visual information.

[0065] In summary, by scientifically and reasonably setting the preset period, the product quality control level can be greatly improved while ensuring production efficiency, meeting the strict requirements of high-end chemical product production.

[0066] S120, input the to-be-analyzed video sequence in the reactor into a chemical product visual analysis model for fuzzy video processing and extract the product visual information of the current period to obtain product visual information; wherein the chemical product visual analysis model is obtained by using the operation process video segments of each process operation node in the historical video sequence as a sample set and using neural network technology for model training.

[0067] In this embodiment, product visual information refers to the key visual features and state change information extracted by the pre-trained chemical product visual analysis model after processing the video sequence in the reactor.

[0068] The product visual description parameters include color parameters and state parameters, wherein the color parameters include RGB values and color change rates, and the state parameters include state labels and state change rates.

[0069] In the chemical industry, monitoring the color parameters and state parameters of the materials in the reactor is crucial for understanding the progress of the chemical reaction. These parameters not only reflect changes in reaction conditions, but also indicate specific chemical processes or stages.

[0070] Color parameters: including RGB values (intensity of red, green, and blue primary colors) and their change rates. RGB values represent the intensity of red (Red), green (Green), and blue (Blue) respectively, and the intensity range of each color is usually 0 to 255. Color change rate refers to the speed of color change per unit time, which can be obtained by calculating the difference in RGB values between consecutive video frames.

[0071] State parameters: including state labels and state change rates. State labels are used to describe the physical state of the material, such as liquid, solid, emulsion state, etc.; state change rate refers to the speed of the material changing from one state to another, such as the crystallization process from liquid to solid.

[0072] The process operation nodes include time points with specific product visual information or operation switching time points; a preset cycle contains one or more process operation nodes.

[0073] Specifically, the time point for specific product visual information refers to the point in time within a production cycle when the material's color parameters (such as RGB values ​​and their rate of change) and state parameters (such as physical state labels and their rate of change) undergo significant changes. These changes are usually closely related to the progress of chemical reactions, reflecting changes in reaction conditions or phased progress.

[0074] Color change: For example, in a redox reaction, the color of a material may change from light to dark, and this color change can be detected by a significant change in the RGB values.

[0075] State transition: Changes in physical state, such as the crystallization process from liquid to solid, can also be used as important time points for marking.

[0076] For example, the current product visual information is shown in Table 1.

[0077] Table 1. Current Product Visual Information

[0078]

[0079] Operation changeover points refer to the moments when different operational steps in a production process transition between each other. These include, but are not limited to, changes in raw material addition, temperature adjustment, pressure regulation, and stirring speed. These operation changeovers not only mark different stages of the process but also serve as important indicators for assessing whether the production process is proceeding as planned.

[0080] Raw material addition: Record when new raw materials are added to the reactor, as this is crucial for subsequent analysis of the reaction results.

[0081] Temperature / pressure adjustment: Monitoring and recording changes in temperature or pressure can help understand how these variables affect the rate and outcome of chemical reactions.

[0082] Changes in stirring speed: Adjusting the stirring speed may affect the degree of mixing of materials, thereby affecting the quality of the final product.

[0083] Within a pre-defined production cycle, there can be one or more such process operation nodes. Each node represents a significant event or point of change in the production process. By identifying and recording these nodes, we can not only better understand the production process but also provide a foundation for subsequent data analysis.

[0084] The chemical product visual analysis model is a trained deep learning model specifically designed to analyze video data within a reaction vessel, automatically identifying and extracting key visual features from the video. Specifically, the model can process the following information:

[0085] RGB values: The RGB values obtained through the camera can directly reflect the color changes of the materials, which is very useful for monitoring the progress of chemical reactions.

[0086] Color change rate: By calculating the change of RGB values between consecutive video frames, the color change rate can be determined, which helps to understand the occurrence of redox reactions and other processes.

[0087] State label and state change rate: The physical state of the material is another important visual feature in the reaction process. Changes in state can indicate different stages or processes of the reaction, such as phase transitions or crystallization processes.

[0088] Although the camera can obtain more types of video data, such as color uniformity, bubble size and generation rate, the running state of the scraper and stirring device, temperature, pressure value, etc., this application specifically selects RGB values and color change rate, state label and state change rate as key monitoring parameters. This is because:

[0089] Intuitiveness: Color and state are one of the most easily observed and recorded visual features.

[0090] Relevance: These parameters are closely related to specific stages of chemical reactions and can sensitively reflect changes in reaction conditions.

[0091] Cost-effectiveness: Selecting the most representative and sensitive parameters can reduce data processing complexity while ensuring monitoring effectiveness.

[0092] Ease of integration and expansion: Simplifying the complexity of the model and reducing the demand for computing resources is conducive to the integration and subsequent expansion of the system.

[0093] The training process of the chemical product visual analysis model includes:

[0094] Obtain historical video sequences reflecting the full-cycle operation process of the reaction vessel;

[0095] Determine the time period of each historical process operation node and extract the corresponding operation process video segments from the historical video sequences;

[0096] Based on the operation process video segments, build a sample set for model training;

[0097] Create an initial model using neural network technology;

[0098] The sample set is inputted to the initial model for training of the initial model, adjusting the initial model until the loss value is minimized to obtain a chemical product visual analysis model.

[0099] In this embodiment, first, a series of historical video sequences need to be acquired, which record the operation process of the reaction kettle in each historical reaction cycle. The video content includes important visual information such as color change of materials and physical state transition. Before formal use, the quality of the video must be checked to ensure that there is no frame loss or damage phenomenon, so as to ensure the accuracy of subsequent analysis.

[0100] Based on the preset process operation node information, such as shown in Table 2, the specific time period of each process operation node is analyzed and determined. This step involves analyzing key events in historical videos, such as color changes and physical state transitions.

[0101] Table 2. Preset process operation node information

[0102]

[0103] For the determined time period, the corresponding operation process video clips are precisely cut from the complete historical video sequence, which will be used to construct the model training sample set.

[0104] Then, relevant features are extracted from each video clip, including but not limited to material color (RGB value and its change rate), physical state (liquid, solid, etc. Label and its change rate), etc.

[0105] Each extracted feature is labeled to indicate its corresponding process operation node. This is an important part of supervised learning, as it provides the "correct answer" needed for model learning.

[0106] According to the task requirements, select the appropriate neural network architecture, such as convolutional neural networks (CNNs) are suitable for processing image data, while recurrent neural networks (RNNs) or their variants LSTM, GRU are suitable for processing time series data.

[0107] Set the number of network layers, the number of neurons in each layer, and the weight initialization strategy, etc.

[0108] Through the backpropagation algorithm, continuously adjust the model parameters, so that the error (i.e. loss value) between the model prediction result and the actual label gradually decreases. This process may require multiple iterations until a predetermined stopping condition is reached, such as the loss value no longer significantly decreases or reaches the maximum number of iterations.

[0109] To evaluate the performance of the model, the dataset is usually divided into training, validation, and test sets. After training, the model hyperparameters are further tuned using the validation set, and finally, the test set is used to evaluate the generalization ability of the model.

[0110] The above steps constitute the complete training process of the visual analysis model for chemical products. This process not only covers the data preparation phase but also includes the key links of model creation, training, and evaluation, ensuring that the generated model can accurately identify the state changes of different process operation nodes in the reactor, thereby supporting the optimization of the automated production process.

[0111] In addition, during the training process, first, the historical video sequences collected for the historical operation process of the reactor are obtained, which reflect the visual information of the reactor during its historical reaction cycle. Then, the specific time period of each historical process operation node is analyzed from the historical reaction cycle, and the operation process video clips of each historical process operation node corresponding to the time period are extracted from the historical video sequences as its operation process video clips. Based on these operation process video clips, samples are constructed, an initial model is created by using a neural network, and the samples are input into the initial model to output the model loss value. When the loss value reaches the minimum, the final model is generated.

[0112] For how to analyze the historical time period of each historical process operation node, first, the historical video sequences are frame-extracted in chronological order to obtain multiple original video frames carrying timestamps. Then, the historical operation steps and visual description parameters of the historical chemical products of each historical process operation node are obtained according to the preset process operation node information. Subsequently, video frames matching the historical operation steps and historical chemical product visual description parameters of each historical process operation node are found in these original video frames to determine the starting frame and ending frame of each historical process operation node. Finally, the timestamps of these frames are used to determine the historical time period of each historical process operation node.

[0113] Further, to find the video frames related to each historical process operation node from the original video frames, an inter-frame difference algorithm is used to calculate the pixel change rate between adjacent video frames, and those video frames with pixel change rates exceeding a preset threshold are selected to form a sequence of video frames to be analyzed. For each video frame in this sequence, operation tool information and operation gesture information are defined to obtain first operation step information, and first chemical product visual description information is analyzed. By comparing these information with the related parameters of each historical process operation node, correlation analysis is performed to determine which video frames to be analyzed are associated with a specific historical process operation node, and finally, the video frames matching the historical operation steps and chemical product visual description parameters are identified.

[0114] Finally, to construct the model training samples from the operation process video segments of each historical process operation node, all video frames need to be extracted from each operation process video segment. For each frame, the historical color parameters of the material color (including historical RGB values, color change rate) and the historical state parameters (such as state label and state change rate) are calculated. Using these data as labels, each video frame is labeled to construct the model training samples. This process ensures that the training samples accurately reflect the color and state changes of the material in the historical operation process, providing a solid foundation for model training.

[0115] In an embodiment, the chemical product visual analysis model also has the function of processing blurred video sequences. Specifically, the model includes an inter-frame pixel analysis layer, a historical reactor video sequence acquisition layer, a first pixel feature point extraction layer, a second pixel feature point extraction layer, a pixel feature offset vector analysis layer, an optical flow estimation layer, an image pixel parameter application layer, and a visual information extraction layer.

[0116] First, the to-be-analyzed reactor video sequence is input into the chemical product visual analysis model, and the image pixels between adjacent frames are compared using the inter-frame pixel analysis layer to calculate the image pixel contrast value. This step is used to detect blurring or significant changes in the video sequence. If the calculated image pixel contrast value is greater than a preset threshold, it indicates that the current video sequence may be blurred or have other interference factors, and the historical reactor video sequence acquisition layer needs to be called to find the nearest clear historical video sequence as a reference. Using this clear historical video sequence, the first pixel feature points of each first image are extracted from it by the first pixel feature point extraction layer, and the second pixel feature points of each second image are extracted from the to-be-analyzed video sequence by the second pixel feature point extraction layer.

[0117] Next, the pixel feature offset vector analysis layer is used to calculate the differences between the first pixel feature points and the second pixel feature points, generating pixel feature offset vectors. These offset vectors are then used in the optical flow estimation layer to perform optical flow estimation on the to-be-analyzed video sequence using the optical flow method to predict the image pixel parameters in the reactor. Then, the predicted image pixel parameters are applied to the to-be-analyzed video sequence by the image pixel parameter application layer to generate a clear video sequence. Finally, the visual information extraction layer extracts and outputs the product visual information of the current period from this clear video sequence, such as the RGB values of the material color, the color change rate, the material physical state label, and the state change rate.

[0118] If the image pixel contrast value is not greater than the preset threshold value, the product visual information of the current period is directly extracted and output from the to-be-analyzed video sequence in the reactor. It should be noted that for the case of blurring, in addition to judging whether it is caused by environmental factors, further analysis is needed to determine whether the blurring is caused by inconsistent nodes (i.e., differences in time periods or operation steps between different process operation nodes). For this purpose, the specific reason for the blurring can be distinguished by adding a timestamp comparison analysis or consistency verification of the process operation steps, so that targeted measures can be taken for correction or adjustment to ensure that the finally extracted product visual information is accurate. This method not only improves the reliability and accuracy of the monitoring system, but also has important application value in complex chemical production environments.

[0119] Through the clear processing, the details in the video can be more obvious. The clear video sequence can more accurately identify these visual features of the product, thereby improving the accuracy of the chemical product visual analysis model in extracting product visual information. For example, when some small-volume product flocculation is generated in the reaction kettle, the blurred video may make the monitoring system unable to accurately judge the size and distribution of the flocculation, and after the clear processing, the system can more reliably monitor these conditions, avoid taking wrong operation measures due to misjudgment, and enhance the reliability of the entire monitoring system.

[0120] The method provided in the present application can correct the blurred video sequence by using the clear historical video sequence under the condition of blurring through the combination of inter-frame comparison and historical video sequence, thereby effectively solving the common visual interference problem in the chemical scene. Compared with the traditional video data processing method, the data is more robust and interpretable, and is more suitable for result-oriented chemical production processes in combination with the production process. In particular, the intelligent triggering method is more suitable for long-term monitoring scenarios in chemical production, greatly reducing the computational overhead and improving the efficiency.

[0121] Further, in the chemical reaction process, the color and state of the material are the key visual features for monitoring. The method can more accurately capture the changes in these visual features through multi-layer analysis (such as pixel feature point extraction, optical flow estimation, etc.) for clear processing of the video sequence. This comprehensive processing method is more comprehensive and effective than a single image sharpening technique (such as a simple de-fogging or de-blurring algorithm), and can better cope with complex chemical reaction scenarios.

[0122] It should be noted that the blurred to-be-analyzed video sequence in the reactor can be sharpened using the above technical means, and other image sharpening strategies in existing technologies can also be used for sharpening, which is not limited in the present application.

[0123] Of course, the above-mentioned fuzzy situation also needs to be distinguished whether it is due to the improper control of the complex environment in the reaction kettle, such as temperature, pressure, material input, humidity and other factors. The camera back-blowing device cannot cope with the large-scale state change in the reaction kettle, resulting in the fuzzy phenomenon of the video sequence collected by the camera, rather than the fuzzy phenomenon caused by the misalignment of the nodes. Specifically, ensure that enough sensors are deployed inside the reaction kettle to monitor temperature, pressure, humidity and other key environmental parameters in real time, and record these data synchronously with the video frame timestamps.

[0124] A fast response module is added to the model, which can evaluate whether the current environmental state is likely to cause video quality degradation (such as lens fogging, air turbulence, etc.) at the moment of fuzzy phenomenon according to the real-time collected environmental data. This can be achieved through pre-set thresholds or machine learning models, which analyze which specific condition combinations are prone to cause fuzzy phenomena based on historical data.

[0125] A fuzzy type classifier is constructed to automatically analyze the fuzzy features in the video sequence using image processing techniques. For example, through texture analysis of the fuzzy area, it can be distinguished whether it is caused by optical factors (such as lens contamination) or motion blur (such as liquid flow during stirring). This classification helps to quickly locate the root cause of the problem. Once the fuzzy phenomenon caused by environmental factors is detected, the system should be able to provide immediate maintenance suggestions, such as adjusting the ventilation rate, optimizing the temperature control settings, etc., to reduce the occurrence of similar problems in the future.

[0126] For those fuzzy situations that cannot be immediately explained by environmental factors, they are marked for further analysis. When the operation period ends and all node information has been collected, a detailed node consistency check is performed. If misalignment of nodes is found, it will be considered as one of the potential causes of the fuzzy phenomenon.

[0127] This method can accurately identify and solve most video blur problems caused by environmental factors as much as possible without relying on immediate node alignment information, while retaining the ability to explore other possible causes (such as node misalignment). This can both improve the response speed of the system and ensure the comprehensiveness and accuracy of the final diagnosis.

[0128] S130, based on the preset process operation node information and the product visual information, analyzing the process operation nodes in the current period to determine whether the time distribution of the process operation nodes in the current period is consistent with the time distribution in the preset process operation node information, to obtain an analysis result.

[0129] In this embodiment, the analysis result refers to determining the time distribution of the process operation nodes in the current period according to the time matching degree, and judging whether it meets the preset standard.

[0130] In an embodiment, referring to Figure 3 , the step S130 can include steps S131-S132.

[0131] S131, based on the preset process operation node information and the product visual information, analyzing the process operation node in the current cycle to obtain process operation node information; wherein the process operation node information includes process operation node, time position and product visual description parameters corresponding to the process operation node;

[0132] In this embodiment, the preset process operation node information includes the specific operation steps (such as "add raw material A", "stir for 3 minutes") that should be performed at each process operation node, the corresponding chemical product visual description parameters (such as color range, color change rate, etc.), and the chemical reaction conditions (such as temperature range).

[0133] Using video monitoring or image processing technology, key visual description parameters are extracted from the product in the current cycle. This includes but is not limited to the color RGB value of the material, the state label (liquid, solid, etc.) and its change rate, etc.

[0134] The product visual description parameters extracted in the current cycle are matched with the preset chemical product visual description parameters to calculate the matching degree between them. This matching degree reflects the consistency degree of the actual production process and the pre-designed plan.

[0135] After the above steps, a detailed information set containing process operation node names, their time positions and corresponding product visual description parameters is obtained.

[0136] S132, according to the process operation node information, determining whether the time distribution of the process operation node in the current cycle is consistent with the time distribution in the preset process operation node information to obtain an analysis result.

[0137] By analyzing the real-time collected product visual information and comparing it with the preset process operation node information, a matching degree on the time division is obtained. If the actual visual description parameters in a certain time period are highly consistent with the preset visual description parameters, it is considered that the operation node in this time period is consistent with the preset; otherwise, there may be deviations.

[0138] Based on the results of the above matching degree, the actual time distribution of each process operation node in the current cycle is redefined. For example, an operation node in the original plan should be completed within 0-1 minutes, but if the analysis result shows that the node is actually completed within 0-2 minutes, it means that there is a delay or other problems in this node.

[0139] When the time distribution of the actual process operation node is found to be inconsistent with the preset, further analysis can be performed to find the specific problem. For example, in the example mentioned above, if the originally planned node 1 time is 0-1 minute, but the actual time is 0-2 minutes, it can be preliminarily determined that there is a problem with node 1.

[0140] Through such an analysis process, not only can the current state of the reaction kettle be accurately identified, but also the operation steps can be dynamically adjusted to reduce errors caused by manual intervention, thereby improving the automation degree and accuracy of production. In addition, this method helps to find problems in time and take corrective measures to ensure the improvement of product quality and production efficiency.

[0141] In summary, the S130 step realizes accurate monitoring and management of each stage in the production process by combining the preset process operation node information and the real-time acquired product visual information, which is of great significance for modern chemical production.

[0142] In an embodiment, referring to Figure 4 The above step S132 can include steps S1321-S1322.

[0143] S1321, according to the process operation node information, comparing the product visual description parameters corresponding to the process operation node with the parameter range corresponding to the preset process operation node information, and calculating the matching result of each index.

[0144] In this embodiment, the matching result refers to the evaluation conclusion generated by comparing each index of the current material with the preset standard range, indicating whether each index meets the requirements.

[0145] In an embodiment, referring to Figure 5 The above step S1321 can include steps S13211-S13215.

[0146] S13211, according to the process operation node information, comparing whether the color RGB value of the current material is within the color parameter range corresponding to the preset process operation node information, and generating a color matching result.

[0147] In this embodiment, the color matching result refers to the result obtained by comparing the color difference between the standard color sample and the material to be tested to confirm whether the color difference is within an acceptable range.

[0148] Specifically, the real-time color RGB value of the product in the current period is acquired; the RGB value is compared with the color parameter range specified in the preset process operation node information; if it is within the range, a positive color matching result is generated; otherwise, it is marked as not matching.

[0149] S13212, determine whether the color change rate of the current material meets the color change rate range corresponding to the preset process operation node information according to the process operation node information, and generate a color change rate matching result.

[0150] In this embodiment, the color change rate matching result refers to the result of evaluating whether the speed of the color change of the material over time meets the preset standard.

[0151] Specifically, the change rate of the current material color is calculated; the change rate is compared with the preset standard range; and the corresponding color change rate matching result is generated according to the comparison result.

[0152] S13213, determine whether the temperature of the current material meets the temperature range corresponding to the preset process operation node information according to the process operation node information, and obtain a temperature matching result.

[0153] In this embodiment, the temperature matching result refers to the result of confirming whether the temperature performance of the material under a specific condition falls within the specified safe range.

[0154] Specifically, real-time temperature data of the reaction kettle in the current period is obtained; it is checked whether the temperature falls within the preset temperature range; and the temperature matching result is generated according to the checking result.

[0155] S13214, calculate the matching degree between the current material state label and the state label corresponding to the preset process operation node information according to the process operation node information, and obtain a state label matching result.

[0156] In this embodiment, the state label matching result refers to the result of verifying whether the current physical or chemical state label of the material is consistent with the expected state.

[0157] Specifically, the state label (such as liquid, solid, etc.) of the current material is determined; the state label is compared with the state label in the preset process operation node information; and the matching degree between the state labels is calculated and generated as the matching result.

[0158] S13215, determine whether the state change rate of the current material meets the state change rate range corresponding to the preset process operation node information according to the process operation node information, and obtain a state change rate matching result.

[0159] In this embodiment, the state change rate matching result refers to the result of measuring whether the speed of the material from one state to another state meets the established requirements.

[0160] Specifically, the change rate of the current material state is analyzed; the change rate is compared with the preset allowable range; and the state change rate matching result is obtained according to the analysis result.

[0161] The matching results include color matching results, color change rate matching results, temperature matching results, state label matching results, and state change rate matching results.

[0162] After all the above steps are completed, a series of matching results will be obtained, including but not limited to color matching results, color change rate matching results, temperature matching results, state label matching results, and state change rate matching results.

[0163] S1322, according to the matching results, comprehensively evaluate whether the time distribution of the process operation node in the current period is consistent with the time distribution in the preset process operation node information to obtain an analysis result.

[0164] Using the matching results obtained in step S1321, the time distribution of the process operation node in the current period is comprehensively evaluated to determine whether it is consistent with the preset information.

[0165] First, all matching results from S1321 are summarized. This usually involves converting each individual matching result into a quantifiable numerical value or scoring system to facilitate subsequent comprehensive evaluation.

[0166] If one of them does not meet the requirements, it is determined that the time distribution of the process operation node in the current period is inconsistent with the preset. Otherwise, it is consistent.

[0167] In this way, the S132 step can not only accurately identify potential problems in the current production process, but also dynamically adjust the operation steps, thereby improving the automation and accuracy of production, ensuring product quality and production efficiency.

[0168] In this embodiment, the method of the present embodiment matches the preset product visual description parameters and the product visual description parameters in the current period to determine the matching degree of the visual description parameters in time division, and determines the time distribution of the current process operation node according to the above time matching degree. Specifically, if a period is divided into different time intervals, for example, 0-1 minutes is process operation node 1, 1-2 minutes is process operation node 2, and 2-5 minutes is process operation node 3, each process operation node has corresponding operation steps, control conditions and video information, then by analyzing the actual time distribution of these process operation nodes can be redefined, for example, 0-2 minutes is process operation node 1, 2-2.5 minutes is process operation node 2, and 2.5-4 minutes is process operation node 3, thereby identifying possible problem areas, selecting the first process operation node that cannot be aligned as the process operation node where the problem is, as error information.

[0169] S140, when the analysis result is that the time distribution of the process operation nodes in the current cycle is inconsistent with the time distribution in the preset process operation node information, creating corresponding operation process prompt information and sending it to the client.

[0170] When the analysis result is that the time distribution of the process operation nodes in the current cycle is consistent with the time distribution in the preset process operation node information, the step S110 is executed.

[0171] The system first compares the time distribution of the process operation nodes identified in the current cycle with the time distribution specified in the preset process operation node information. This comparison not only involves the starting and ending time points of each node, but also includes the time interval between the conversion of each node.

[0172] Difference detection: If it is found that the actual operation time node (such as adding raw material A, stirring for 3 minutes, etc.) deviates from the preset time arrangement, whether it is ahead of time or delayed, it is considered as inconsistent in time distribution.

[0173] Once it is confirmed that there is inconsistency, the system will automatically generate one or more operation process prompt information. These prompt information includes but is not limited to:

[0174] Specific which process operation node has time deviation;

[0175] The specific value of the deviation, such as how many time earlier or later than expected;

[0176] Possible cause analysis and suggested adjustment measures, such as whether to speed up or slow down a certain step to restore normal flow.

[0177] According to the user's preference settings or specific requirements, the prompt information can be further customized to ensure that it has guiding significance to specific operators.

[0178] The generated prompt information will be immediately sent to the relevant client through the pre-set method. This usually means that the information will be pushed to the display screen of the on-site monitoring system, or directly sent to the handheld device (such as tablet computer, smart phone, etc.) of the operator responsible for the production line.

[0179] In order to increase reliability, the system may use multiple communication channels to send notifications at the same time, such as SMS, email or special APP push messages, to ensure that relevant personnel can receive important operation instructions in time.

[0180] When the analysis result is that the time distribution of the process operation nodes in the current cycle is consistent with the time distribution in the preset process operation node information, the step S110 is executed.

[0181] If, through analysis, it is found that the time distribution of the process operation nodes in the current cycle fully meets the preset information, that is, there is no deviation or only a slight fluctuation within an acceptable range, the system will continue to operate according to the predetermined plan for the next cycle, repeating step S110. This means that the entire production process is running efficiently according to the plan, without the need for additional manual intervention or adjustment.

[0182] In this way, not only can the production process be effectively monitored to ensure that each link is strictly in accordance with the preset standards, but also the system can respond quickly when an abnormality occurs, reducing quality problems or production delays caused by operational errors.

[0183] The above-mentioned intelligent monitoring method for the reaction kettle operation process based on visual analysis automatically acquires video sequences inside the reaction kettle according to a preset cycle, and processes these video data using a deep learning algorithm to extract key product visual information. Based on the pre-defined process operation node information, the system can analyze the operation node time distribution in the current production cycle in real time, determine whether it meets the expected standard, and immediately generate and send operation prompt information to relevant personnel when a deviation is detected, ensuring timely adjustment. Compared with traditional methods that rely on manual monitoring, this automated solution not only reduces the dependence on the experience and attention of the operator, but also accurately captures and records key process details, achieving precise control of the production process and significant improvement in product quality stability, especially suitable for strict production requirements of high-end chemical products. Therefore, this solution effectively solves the problem of difficulty in ensuring product quality consistency and stability in the prior art, represents a more efficient and reliable production process monitoring method, and realizes automatic and intelligent monitoring of the operation process in the reaction kettle.

[0184] Figure 6 is a schematic block diagram of an intelligent monitoring device 300 for the reaction kettle operation process based on visual analysis provided by an embodiment of the present application. As shown in Figure 6 Corresponding to the above-mentioned intelligent monitoring method for the reaction kettle operation process based on visual analysis, the present application also provides an intelligent monitoring device 300 for the reaction kettle operation process based on visual analysis. The intelligent monitoring device 300 for the reaction kettle operation process based on visual analysis includes units for performing the above-mentioned intelligent monitoring method for the reaction kettle operation process based on visual analysis, and the device can be configured in a server. Specifically, please refer to Figure 6 The intelligent monitoring device 300 for the reaction kettle operation process based on visual analysis includes an acquisition unit 301, an extraction unit 302, an analysis unit 303, and a creation unit 304.

[0185] The acquisition unit 301 is configured to acquire, according to a preset period, a video sequence of an inside of a reaction kettle in a current period collected by a 5G explosion-proof camera deployed in the reaction kettle, to obtain a to-be-analyzed kettle video sequence; the extraction unit 302 is configured to input the to-be-analyzed kettle video sequence into a chemical product visual analysis model for fuzzy video processing and extraction of product visual information of the current period, to obtain the product visual information; the chemical product visual analysis model is obtained by using a neural network technology to train a model by taking operation process video clips of each process operation node in a historical video sequence as a sample set; the analysis unit 303 is configured to analyze a process operation node in the current period based on preset process operation node information and the product visual information, to determine whether a time distribution of the process operation node in the current period is consistent with a time distribution in the preset process operation node information, to obtain an analysis result; when the analysis result is that the time distribution of the process operation node in the current period is consistent with the time distribution in the preset process operation node information, the acquisition unit 301 is configured to acquire, according to a preset period, a video sequence of an inside of a reaction kettle in a current period collected by a 5G explosion-proof camera deployed in the reaction kettle, to obtain a to-be-analyzed kettle video sequence.

[0186] The creation unit 304 is configured to, when the analysis result is that the time distribution of the process operation node in the current period is inconsistent with the time distribution in the preset process operation node information, create corresponding operation process prompt information and send the operation process prompt information to a client.

[0187] In an embodiment, the analysis unit 303 includes:

[0188] The node analysis subunit is configured to analyze a process operation node in a current period based on preset process operation node information and product visual information, to obtain process operation node information; the process operation node information includes a process operation node, a time position, and a product visual description parameter corresponding to the process operation node; and the time distribution analysis subunit is configured to determine whether a time distribution of the process operation node in the current period is consistent with a time distribution in the preset process operation node information, to obtain an analysis result.

[0189] In an embodiment, the time distribution analysis subunit includes:

[0190] The parameter matching module is configured to compare product visual description parameters corresponding to the process operation nodes with preset parameter ranges corresponding to the preset process operation node information according to the process operation node information, and calculate matching results of the indexes.

[0191] In an embodiment, the parameter matching module comprises:

[0192] The color matching sub-module is configured to compare whether the color RGB value of the current material is within the preset color parameter range corresponding to the preset process operation node information according to the process operation node information, and generate a color matching result. The color change rate matching sub-module is configured to determine whether the color change rate of the current material conforms to the preset color change rate range corresponding to the preset process operation node information according to the process operation node information, and generate a color change rate matching result. The temperature matching sub-module is configured to determine whether the current temperature is within the preset temperature range corresponding to the preset process operation node information according to the process operation node information, and generate a temperature matching result. The state label matching sub-module is configured to calculate the matching degree between the current material state label and the preset state label corresponding to the preset process operation node information according to the process operation node information, and generate a state label matching result. The state change rate matching sub-module is configured to determine whether the state change rate of the current material is within the preset state change rate range corresponding to the preset process operation node information according to the process operation node information, and generate a state change rate matching result.

[0193] The matching results of the indexes comprise the color matching result, the color change rate matching result, the temperature matching result, the state label matching result, and the state change rate matching result.

[0194] It should be noted that the specific implementation process of the reaction kettle operation process intelligent monitoring device 300 and each unit based on visual analysis described above can be clearly understood by those skilled in the art, and can refer to the corresponding description in the foregoing method embodiments. For the convenience and brevity of description, it will not be repeated here.

[0195] The reaction kettle operation process intelligent monitoring device 300 based on visual analysis described above can be implemented in the form of a computer program, which can run on a computer device as shown in the computer device. Figure 7

[0196] Please refer to Figure 7 , Figure 7 ​is a schematic block diagram of a computer device provided by an embodiment of the present application. The computer device 500 can be a server, wherein the server can be a stand-alone server or a server cluster composed of multiple servers.

[0197] Referring to Figure 7 The computer device 500 includes a processor 502, a memory, and a network interface 505 connected through a system bus 501, wherein the memory can include a non-volatile storage medium 503 and an internal memory 504.

[0198] The non-volatile storage medium 503 can store an operating system 5031 and a computer program 5032. The computer program 5032 includes program instructions which, when executed, can cause the processor 502 to perform a visual analysis-based intelligent monitoring method for a reaction kettle operation process.

[0199] The processor 502 is configured to provide computing and control capabilities to support the operation of the entire computer device 500.

[0200] The internal memory 504 provides an environment for the running of the computer program 5032 in the non-volatile storage medium 503, which, when executed by the processor 502, can cause the processor 502 to perform a visual analysis-based intelligent monitoring method for a reaction kettle operation process.

[0201] The network interface 505 is configured to perform network communication with other devices. Those skilled in the art can understand that Figure 7 The structure shown in the figure is only a block diagram of part of the structure related to the scheme of the present application, and does not constitute a limitation on the computer device 500 to which the scheme of the present application is applied. The specific computer device 500 can include more or fewer components than those shown in the figure, or combine certain components, or have a different arrangement of components.

[0202] The processor 502 is configured to run the computer program 5032 stored in the memory to implement the following steps:

[0203] According to a preset period, a video sequence of a reaction kettle inside in a current period collected by a 5G explosion-proof camera arranged in the reaction kettle is acquired to obtain a to-be-analyzed kettle video sequence; the to-be-analyzed kettle video sequence is input into a chemical product visual analysis model for fuzzy video processing and extraction of product visual information of the current period to obtain product visual information; wherein the chemical product visual analysis model is obtained by model training using a neural network technology with operation process video clips of each process operation node in a historical video sequence as a sample set; based on preset process operation node information and the product visual information, process operation nodes in the current period are analyzed to determine whether the time distribution of the process operation nodes in the current period is consistent with the time distribution in the preset process operation node information to obtain an analysis result; when the analysis result is that the time distribution of the process operation nodes in the current period is inconsistent with the time distribution in the preset process operation node information, corresponding operation process prompt information is created and sent to a client.

[0204] The preset period is a specific time period formed by dividing a production process by time, or a time period set by customization.

[0205] The process operation node includes a time point with specific product visual information or an operation switching time point; one preset period contains one or more process operation nodes.

[0206] The product visual description parameter includes a color parameter and a state parameter, wherein the color parameter includes an RGB value and a color change rate, and the state parameter includes a state label and a state change rate.

[0207] In an embodiment, the processor 502 specifically implements the following steps when implementing the training step of the chemical product visual analysis model:

[0208] A historical video sequence reflecting the whole cycle operation process of the reaction kettle is acquired; the time period of each historical process operation node is determined, and the corresponding operation process video clip is intercepted from the historical video sequence; a sample set for model training is constructed based on the operation process video clip; an initial model is created using a neural network technology; the sample set is input into the initial model to train the initial model, and the initial model is adjusted until the loss value is minimized to obtain a chemical product visual analysis model.

[0209] In an embodiment, the processor 502, when implementing the step of analyzing the process operation node in the current period based on the preset process operation node information and the product visual information to determine whether the time distribution of the process operation node in the current period is consistent with the time distribution in the preset process operation node information to obtain an analysis result, specifically implements the following steps:

[0210] analyzing the process operation node in the current period based on the preset process operation node information and the product visual information to obtain process operation node information; wherein the process operation node information comprises a process operation node, a time position, and a product visual description parameter corresponding to the process operation node; and determining whether the time distribution of the process operation node in the current period is consistent with the time distribution in the preset process operation node information according to the process operation node information to obtain an analysis result.

[0211] In an embodiment, the processor 502, when implementing the step of determining whether the time distribution of the process operation node in the current period is consistent with the time distribution in the preset process operation node information according to the process operation node information to obtain an analysis result, specifically implements the following steps:

[0212] comparing the product visual description parameter corresponding to the process operation node with a parameter range corresponding to the preset process operation node information according to the process operation node information to calculate a matching result of each index; and comprehensively evaluating whether the time distribution of the process operation node in the current period is consistent with the time distribution in the preset process operation node information according to the matching result to obtain an analysis result.

[0213] In an embodiment, the processor 502, when implementing the step of comparing the product visual description parameter corresponding to the process operation node with a parameter range corresponding to the preset process operation node information according to the process operation node information to calculate a matching result of each index, specifically implements the following steps:

[0214] According to the process operation node information, whether the color RGB value of the current material is in the color parameter range corresponding to the preset process operation node information is compared, and a color matching result is generated; according to the process operation node information, whether the color change rate of the current material color conforms to the color change rate range corresponding to the preset process operation node information is judged, and a color change rate matching result is generated; according to the process operation node information, whether the current material temperature is in the temperature range corresponding to the preset process operation node information is determined, so as to obtain a temperature matching result; according to the process operation node information, the matching degree between the current material state label and the state label corresponding to the preset process operation node information is calculated, so as to obtain a state label matching result; according to the process operation node information, whether the state change rate of the current material is in the state change rate range corresponding to the preset process operation node information is confirmed, so as to obtain a state change rate matching result; wherein, the matching results of the indicators include the color matching result, the color change rate matching result, the temperature matching result, the state label matching result and the state change rate matching result.

[0215] In an embodiment, the processor 502, after analyzing the process operation node in the current period based on the preset process operation node information and the product visual information to determine whether the time distribution of the process operation node in the current period is consistent with the time distribution in the preset process operation node information to obtain an analysis result, further implements the following steps:

[0216] When the analysis result is that the time distribution of the process operation node in the current period is consistent with the time distribution in the preset process operation node information, the video sequence of the inside of the reaction kettle in the current period collected by the 5G explosion-proof camera deployed in the reaction kettle is acquired according to the preset period to obtain a to-be-analyzed kettle video sequence.

[0217] It should be understood that, in the embodiments of the present application, the processor 502 can be a central processing unit (CPU), and the processor 502 can also be other general-purpose processors, digital signal processors (DSP), application specific integrated circuits (ASIC), ready-to-program gate arrays (FPGA) or other programmable logic devices, discrete gates or transistor logic devices, discrete hardware components, etc. Among them, the general-purpose processor can be a microprocessor or the processor can also be any conventional processor, etc.

[0218] Those skilled in the art can understand that all or part of the processes in the method of implementing the above embodiments can be completed by instructing the relevant hardware through a computer program. The computer program includes program instructions, and the computer program can be stored in a storage medium, which is a computer readable storage medium. The program instructions are executed by at least one processor in the computer system to implement the process steps of the above method embodiments.

[0219] Therefore, the present application also provides a storage medium. The storage medium can be a computer readable storage medium. The storage medium stores a computer program, wherein the computer program is executed by a processor to make the processor execute the following steps:

[0220] According to a preset period, a video sequence of a reaction kettle inside in a current period collected by a 5G explosion-proof camera arranged in the reaction kettle is acquired to obtain a to-be-analyzed kettle video sequence; the to-be-analyzed kettle video sequence is input into a chemical product visual analysis model for fuzzy video processing and extraction of product visual information of the current period to obtain product visual information; wherein the chemical product visual analysis model is obtained by model training through operation process video clips of each process operation node in a historical video sequence as a sample set using a neural network technology; based on preset process operation node information and the product visual information, a process operation node in the current period is analyzed to determine whether a time distribution of the process operation node in the current period is consistent with a time distribution in the preset process operation node information to obtain an analysis result; when the analysis result is that the time distribution of the process operation node in the current period is inconsistent with the time distribution in the preset process operation node information, corresponding operation process prompt information is created and sent to a client.

[0221] The preset period is a specific time period formed by dividing the production process by time, or a time period set by the user.

[0222] The process operation node includes a time point with specific product visual information or an operation switching time point; one preset period contains one or more process operation nodes.

[0223] The product visual description parameter includes a color parameter and a state parameter, wherein the color parameter includes an RGB value and a color change rate, and the state parameter includes a state label and a state change rate.

[0224] In an embodiment, when the processor executes the computer program to implement the training step of the chemical product visual analysis model, the following steps are specifically implemented:

[0225] acquire a historical video sequence reflecting a whole-cycle operation process of a reaction kettle; determine a time period of each historical process operation node and intercept a corresponding operation process video clip from the historical video sequence; construct a sample set for model training based on the operation process video clip; create an initial model using a neural network technology; input the sample set into the initial model to train the initial model, adjust the initial model until a loss value is minimized, and obtain a chemical product visual analysis model.

[0226] In an embodiment, when the processor implements the step of analyzing the process operation node in the current cycle based on the preset process operation node information and the product visual information to obtain an analysis result by executing the computer program, the following steps are specifically implemented:

[0227] analyzing the process operation node in the current cycle based on the preset process operation node information and the product visual information to obtain process operation node information, wherein the process operation node information comprises a process operation node, a time position, and a product visual description parameter corresponding to the process operation node; and determining whether the time distribution of the process operation node in the current cycle is consistent with the time distribution in the preset process operation node information according to the process operation node information to obtain an analysis result.

[0228] In an embodiment, when the processor implements the step of determining whether the time distribution of the process operation node in the current cycle is consistent with the time distribution in the preset process operation node information according to the process operation node information to obtain an analysis result by executing the computer program, the following steps are specifically implemented:

[0229] comparing the product visual description parameter corresponding to the process operation node with a parameter range corresponding to the preset process operation node information according to the process operation node information, calculating a matching result of each index, and comprehensively evaluating whether the time distribution of the process operation node in the current cycle is consistent with the time distribution in the preset process operation node information according to the matching result to obtain an analysis result.

[0230] In an embodiment, when the processor implements the step of comparing the product visual description parameter corresponding to the process operation node with a parameter range corresponding to the preset process operation node information according to the process operation node information, calculating a matching result of each index by executing the computer program, the following steps are specifically implemented:

[0231] According to the process operation node information, whether the color RGB value of the current material is in the color parameter range corresponding to the preset process operation node information is compared, and a color matching result is generated; according to the process operation node information, whether the color change rate of the current material color conforms to the color change rate range corresponding to the preset process operation node information is judged, and a color change rate matching result is generated; according to the process operation node information, whether the current material temperature is in the temperature range corresponding to the preset process operation node information is determined, so as to obtain a temperature matching result; according to the process operation node information, the matching degree between the current material state label and the state label corresponding to the preset process operation node information is calculated, so as to obtain a state label matching result; according to the process operation node information, whether the state change rate of the current material is in the state change rate range corresponding to the preset process operation node information is confirmed, so as to obtain a state change rate matching result;

[0232] The matching results of the indicators include the color matching result, the color change rate matching result, the temperature matching result, the state label matching result, and the state change rate matching result.

[0233] In an embodiment, after the processor executes the computer program to analyze the process operation node in the current period based on the preset process operation node information and the product visual information to determine whether the time distribution of the process operation node in the current period is consistent with the time distribution in the preset process operation node information, the following steps are further implemented:

[0234] When the analysis result is that the time distribution of the process operation node in the current period is consistent with the time distribution in the preset process operation node information, the video sequence of the inside of the reaction kettle in the current period collected by the 5G explosion-proof camera deployed in the reaction kettle is acquired according to the preset period, so as to obtain the to-be-analyzed kettle inside video sequence.

[0235] The storage medium can be a U disk, a mobile hard disk, a read-only memory (ROM), a magnetic disk, or an optical disk, and various computer readable storage media that can store program codes.

[0236] Those skilled in the art can understand that the units and algorithm steps of each example described in combination with the embodiments disclosed herein can be realized in electronic hardware, computer software or a combination of both. In order to clearly illustrate the interchangeability of hardware and software, the components and steps of each example have been described in general terms in the above description. Whether the functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. A person skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of the present application.

[0237] In several embodiments provided by the present application, it should be understood that the disclosed apparatus and method can be implemented in other ways. For example, the apparatus embodiments described above are merely schematic. For example, the division of each unit is only a logical function division, and actual implementation can have another division manner. For example, a plurality of units or components can be combined or integrated into another system, or some features can be omitted or not executed.

[0238] The steps in the method embodiments of the present application can be adjusted, combined and deleted in sequence according to actual needs. The units in the apparatus embodiments of the present application can be combined, divided and deleted according to actual needs. In addition, each functional unit in each embodiment of the present application can be integrated in one processing unit, or each unit can exist physically, or two or more units can be integrated in one unit.

[0239] The integrated unit, if realized in the form of a software functional unit and sold or used as an independent product, can be stored in a storage medium. Based on such understanding, the technical solutions of the present application essentially or the part that contributes to the prior art, or the whole or part of the technical solutions can be embodied in the form of a software product. The computer software product is stored in a storage medium, and includes a plurality of instructions for causing a computer device (which can be a personal computer, a terminal or a network device, etc.) to execute all or part of the steps of the methods described in each embodiment of the present application.

[0240] The above description is only a specific implementation of the present application, but the protection scope of the present application is not limited thereto. Any person skilled in the art can easily think of various equivalent modifications or replacements within the technical range disclosed by the present application, and these modifications or replacements should be covered in the protection scope of the present application. Therefore, the protection scope of the present application should be subject to the protection scope of the claims.

Claims

1. A method for intelligent monitoring of reactor operation based on visual analysis, characterized in that, include: According to the preset cycle, the video sequence inside the reactor is acquired by the 5G explosion-proof camera deployed in the reactor within the current cycle, so as to obtain the video sequence inside the reactor to be analyzed. The preset cycle can be determined in two ways: one is to divide the entire production cycle proportionally based on time, and the other is to customize the setting according to the specific process operation node. The video sequence inside the vessel to be analyzed is input into the chemical product visual analysis model for blurred video processing and extraction of the product visual information for the current period to obtain the product visual information; the product visual information is the key visual information. The chemical product visual analysis model is obtained by training the model using neural network technology with video clips of the operation process of each process operation node in the historical video sequence as a sample set. The training process of the chemical product visual analysis model includes: acquiring historical video sequences reflecting the entire cycle operation process of the reactor; determining the time period of each historical process operation node and extracting the corresponding operation process video clips from the historical video sequence; constructing a sample set for model training based on the operation process video clips; creating an initial model using neural network technology; inputting the sample set into the initial model to train the initial model; and adjusting the initial model until the loss value is minimized to obtain the chemical product visual analysis model. Based on the preset process operation node information and the product visual information, analyze the process operation nodes in the current cycle to determine whether the time distribution of the process operation nodes in the current cycle is consistent with the time distribution in the preset process operation node information, so as to obtain the analysis results. Determine whether the time distribution of process operation nodes within the current cycle is consistent with the time distribution within the preset process operation node information to obtain analysis results, including: Based on the process operation node information, the product visual description parameters corresponding to the process operation node are compared with the parameter range corresponding to the preset process operation node information, and the matching results of each indicator are calculated. Based on the matching results, a comprehensive evaluation is conducted to determine whether the time distribution of process operation nodes in the current cycle is consistent with the time distribution in the preset process operation node information, so as to obtain the analysis results. If the analysis results show that the time distribution of process operation nodes in the current cycle is inconsistent with the time distribution in the preset process operation node information, then a corresponding operation process prompt message is created. The prompt message includes, but is not limited to: which specific process operation node has a time deviation; the specific value of the deviation; the cause analysis and suggested adjustment measures; and is sent to the client.

2. The intelligent monitoring method for reactor operation based on visual analysis according to claim 1, characterized in that, The process operation nodes include time points with specific product visual information or operation switching time points; a preset cycle contains one or more process operation nodes.

3. The intelligent monitoring method for reactor operation based on visual analysis according to claim 1, characterized in that, The process operation nodes within the current cycle are analyzed based on preset process operation node information and product visual information to determine whether the time distribution of the process operation nodes within the current cycle is consistent with the time distribution within the preset process operation node information, in order to obtain analysis results, including: Based on the preset process operation node information and the product visual information, the process operation nodes in the current cycle are analyzed to obtain the process operation node information; wherein, the process operation node information includes the process operation node, the time position, and the product visual description parameters corresponding to the process operation node. Based on the process operation node information, determine whether the time distribution of the process operation nodes in the current cycle is consistent with the time distribution in the preset process operation node information, so as to obtain the analysis results.

4. The intelligent monitoring method for reactor operation based on visual analysis according to claim 3, characterized in that, The product visual description parameters include color parameters and status parameters. The color parameters include RGB values ​​and color change rate, and the status parameters include status labels and status change rate.

5. The intelligent monitoring method for reactor operation based on visual analysis according to claim 1, characterized in that, The step of comparing the product visual description parameters corresponding to the process operation node with the preset parameter range corresponding to the process operation node information, and calculating the matching results of various indicators, includes: Based on the process operation node information, compare whether the RGB value of the current material's color is within the preset color parameter range corresponding to the process operation node information, and generate a color matching result; Based on the process operation node information, determine whether the current material color change rate conforms to the preset color change rate range corresponding to the process operation node information, and generate a color change rate matching result. Verify whether the current temperature is within the preset temperature range corresponding to the process operation node information based on the process operation node information, so as to obtain the temperature matching result; The matching degree between the current material status label and the status label corresponding to the preset process operation node information is calculated based on the process operation node information to obtain the status label matching result. Based on the process operation node information, confirm whether the current state change rate of the material is within the preset state change rate range corresponding to the process operation node information, and obtain the state change rate matching result. The matching results for each indicator include color matching results, color change rate matching results, temperature matching results, status label matching results, and status change rate matching results.

6. The intelligent monitoring method for reactor operation based on visual analysis according to claim 1, characterized in that, The step of analyzing the process operation nodes within the current cycle based on preset process operation node information and product visual information to determine whether the time distribution of the process operation nodes within the current cycle is consistent with the time distribution within the preset process operation node information, and obtaining the analysis results, further includes: If the analysis result shows that the time distribution of the process operation nodes in the current cycle is consistent with the time distribution in the preset process operation node information, then the process of acquiring the video sequence inside the reactor in the current cycle collected by the 5G explosion-proof camera deployed in the reactor according to the preset cycle is executed to obtain the video sequence inside the reactor to be analyzed.

7. An intelligent monitoring device for the operation process of a reaction vessel based on visual analysis, characterized in that, include: The acquisition unit is used to acquire video sequences inside the reactor within the current period, collected by the 5G explosion-proof camera deployed in the reactor, according to a preset period, so as to obtain the video sequence inside the reactor to be analyzed. The preset period can be determined in two ways: one is to divide the entire production cycle proportionally based on time, and the other is to customize the setting according to the specific process operation node. The extraction unit is used to input the video sequence inside the vessel to be analyzed into the chemical product visual analysis model for blurred video processing and to extract the product visual information of the current period to obtain the product visual information; the product visual information is the key visual information. The chemical product visual analysis model is obtained by training the model using neural network technology with video clips of the operation process of each process operation node in the historical video sequence as a sample set. The training process of the chemical product visual analysis model includes: acquiring historical video sequences reflecting the entire cycle operation process of the reactor; determining the time period of each historical process operation node and extracting the corresponding operation process video clips from the historical video sequence; constructing a sample set for model training based on the operation process video clips; creating an initial model using neural network technology; inputting the sample set into the initial model to train the initial model; and adjusting the initial model until the loss value is minimized to obtain the chemical product visual analysis model. The analysis unit is used to analyze the process operation nodes in the current cycle based on the preset process operation node information and the product visual information, so as to determine whether the time distribution of the process operation nodes in the current cycle is consistent with the time distribution in the preset process operation node information, so as to obtain the analysis results. Determine whether the time distribution of process operation nodes within the current cycle is consistent with the time distribution within the preset process operation node information to obtain analysis results, including: Based on the process operation node information, the product visual description parameters corresponding to the process operation node are compared with the parameter range corresponding to the preset process operation node information, and the matching results of each indicator are calculated. Based on the matching results, a comprehensive evaluation is conducted to determine whether the time distribution of process operation nodes in the current cycle is consistent with the time distribution in the preset process operation node information, so as to obtain the analysis results. A creation unit is used to create corresponding operation process prompt information when the analysis result shows that the time distribution of process operation nodes in the current cycle is inconsistent with the time distribution in the preset process operation node information. The prompt information includes, but is not limited to: which specific process operation node has a time deviation; the specific value of the deviation; the cause analysis and suggested adjustment measures; and sends it to the client.

Images