Industrial distillation process boiling degree judgment method and device based on machine vision

Abstract

The invention provides a method and a device for judging the boiling degree of an industrial distillation process based on machine vision. The method comprises the following steps: acquiring a target image sequence of a boiling process in a boiling reaction kettle within a target duration; extracting feature data of each image in the target image sequence, and obtaining boiling process feature data of the target image sequence; and based on the boiling process feature data, classifying the boiling degree of the boiling process within the target duration to obtain a boiling degree classification result. The technical problems that the boiling state cannot be reflected in real time and the distillation process cannot be optimally controlled according to the boiling state in the prior art are solved.

Description

Technical Field

[0001] The present invention relates to the field of artificial intelligence technology, and in particular to a method and device for judging boiling degree in an industrial distillation process based on machine vision. Background Art

[0002] In the industrial distillation process, boiling heat transfer is a key step in achieving efficient heat transfer. Accurately determining the boiling level is crucial for ensuring distillation efficiency, product quality, and safe equipment operation. Existing technologies primarily rely on indirect measurements using sensors such as temperature and pressure. These methods suffer from poor timeliness and inaccuracy, making it difficult to accurately and timely reflect the boiling state. Summary of the Invention

[0003] The present invention provides a method and device for judging the boiling degree of an industrial distillation process based on machine vision, which is used to solve the technical problems in the prior art that the boiling state cannot be reflected in real time and the distillation process cannot be optimized and controlled according to the boiling state.

[0004] The present invention provides a method for determining boiling degree in an industrial distillation process based on machine vision, the method comprising:

[0005] Acquire a target image sequence of the boiling process within a target time period in a boiling reactor;

[0006] Extracting feature data of each image in the target image sequence, and obtaining boiling process feature data of the target image sequence;

[0007] Based on the boiling process characteristic data, the boiling degree of the boiling process within the target duration is classified to obtain a boiling degree classification result.

[0008] In some embodiments, acquiring a target image sequence of a boiling process in a boiling reactor specifically includes:

[0009] Use a high-speed camera to capture the boiling surface in real time during industrial distillation, obtaining high-frame-rate image sequences of the boiling process.

[0010] The collected images are normalized to obtain the target image sequence.

[0011] In some embodiments, extracting feature data of each image in the target image sequence specifically includes:

[0012] Inputting each frame image in the target image sequence into a pre-trained feature extraction model in sequence to obtain feature data corresponding to each frame image output by the feature extraction model;

[0013] Arranging the characteristic data obtained from all frame images in time series to obtain boiling process characteristic data of the target image sequence;

[0014] The feature extraction model is obtained by training a convolutional neural network using the image sequence and visual features of the sample.

[0015] In some embodiments, based on the boiling process characteristic data, the boiling degree of the boiling process within the target duration is classified to obtain a boiling degree classification result, specifically including:

[0016] Inputting the boiling process characteristic data into a pre-trained degree classification model to obtain a boiling degree classification result output by the degree classification model;

[0017] The degree classification model is trained using the boiling process feature data of the sample and the corresponding boiling degree category label.

[0018] In some embodiments, the boiling degree classification results specifically include four levels: natural convection, nucleate boiling, transition boiling, and film boiling.

[0019] In some embodiments, obtaining a boiling degree classification result further includes:

[0020] According to the boiling degree classification result and based on a preset control strategy, the opening of the heat medium regulating valve of the boiling reactor is controlled.

[0021] In some embodiments, the preset control strategy includes:

[0022] The boiling degree classification result is used as a basis for setting the heat medium flow rate or pressure, and the opening of the regulating valve on the heat medium pipeline is controlled according to the setting value of the heat medium flow rate or pressure.

[0023] In some embodiments, the preset control strategy includes:

[0024] According to the boiling degree classification results, a fuzzy controller is established to directly control the opening of the heat medium regulating valve.

[0025] In some embodiments, an image acquisition device is installed on the boiling reactor, and the image acquisition device is installed on the top or side of the boiling reactor.

[0026] The present invention also provides a device for determining boiling degree in an industrial distillation process based on machine vision, the device comprising:

[0027] An image acquisition unit is used to obtain a target image sequence of the boiling process within a target time period in the boiling reactor;

[0028] a feature extraction unit, configured to extract feature data of each image in the target image sequence and obtain boiling process feature data of the target image sequence;

[0029] The classification generating unit is configured to classify the boiling degree of the boiling process within the target duration based on the boiling process characteristic data to obtain a boiling degree classification result.

[0030] The machine vision-based method for determining the boiling degree of an industrial distillation process provided by the present invention obtains a target image sequence of the boiling process within a target time period in a boiling reactor, extracts feature data for each image in the target image sequence, and obtains boiling process feature data for the target image sequence; based on the boiling process feature data, the boiling degree of the boiling process within the target time period is classified to obtain a boiling degree classification result. Thus, the boiling degree determination method provided by the present invention can accurately monitor the boiling state of an industrial distillation process in real time, provide a scientific basis for optimizing the distillation process, improve distillation efficiency and product quality, and ensure equipment safety. This solves the technical problem in the prior art of being unable to reflect the boiling state in real time and unable to optimize and control the distillation process based on the boiling state. BRIEF DESCRIPTION OF THE DRAWINGS

[0031] In order to more clearly illustrate the technical solutions in the present invention or the prior art, a brief introduction is given below to the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0032] Figure 1 This is one of the flow charts of the method for determining boiling degree in an industrial distillation process based on machine vision provided by the present invention;

[0033] Figure 2 A schematic structural diagram of a reactor used in the method provided by the present invention;

[0034] Figure 3 This is the second flow chart of the method for determining boiling degree in an industrial distillation process based on machine vision provided by the present invention;

[0035] Figure 4 This is the third flow chart of the method for determining boiling degree in an industrial distillation process based on machine vision provided by the present invention;

[0036] Figure 5 This is the fourth flow chart of the method for determining boiling degree in an industrial distillation process based on machine vision provided by the present invention;

[0037] Figure 6This is a structural block diagram of a device for determining boiling degree in an industrial distillation process based on machine vision. DETAILED DESCRIPTION

[0038] To make the objectives, technical solutions, and advantages of the present invention more clear, the technical solutions of the present invention will be clearly and completely described below in conjunction with the accompanying drawings. Obviously, the embodiments described are only some of the embodiments of the present invention, not all of them. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts shall fall within the scope of protection of the present invention.

[0039] In order to solve the problems existing in the prior art, the present invention provides a method for judging and controlling the boiling degree of an industrial distillation process based on machine vision. The method realizes real-time recognition and classification of the boiling state through high-speed camera and deep learning technology, and controls the amount of heat medium through complex control methods based on the results of recognition and classification. The method has the advantages of high accuracy and good real-time performance, and is suitable for monitoring and control optimization of industrial distillation processes.

[0040] In a specific embodiment, the method for determining boiling degree in an industrial distillation process based on machine vision provided by the present invention comprises the following steps:

[0041] S110: Acquire a target image sequence of the boiling process within a target time period in the boiling reactor;

[0042] S120: extracting feature data of each image in the target image sequence, and obtaining boiling process feature data of the target image sequence;

[0043] S130: Classifying the boiling degrees of the boiling process within the target duration based on the boiling process characteristic data to obtain a boiling degree classification result.

[0044] After the boiling degree classification result is obtained, the classification result can be output and a control strategy can be determined based on the classification result.

[0045] In step S110, in order to ensure the image effect for feature extraction, a target image sequence of the boiling process in the boiling reactor is obtained, which specifically includes:

[0046] Use a high-speed camera to shoot the boiling surface in the industrial distillation process in real time and obtain a high frame rate image sequence of the boiling process; in specific usage scenarios, such as Figure 2As shown, an image acquisition device is installed on the boiling reactor, and the image acquisition device is installed on the top or side of the boiling reactor; wherein, the image acquisition device can be specifically a camera, and a reasonable installation method of the camera needs to be considered in actual application. For example, installation method 1 is top-mounted, which is suitable for reactors with inconvenient side openings; installation method 2 is side-mounted, which is preferred.

[0047] Normalize the captured images to obtain the target image sequence. Image preprocessing is completed by normalizing the captured images. The preprocessing process includes adjusting the image size, contrast enhancement, and noise removal to improve image quality and ensure the accuracy of subsequent image analysis.

[0048] In step S120, in order to improve the efficiency and accuracy of feature extraction, a pre-trained convolutional neural network can be used to implement feature extraction. Specifically, extracting feature data of each image in the target image sequence includes the following steps:

[0049] Inputting each frame image in the target image sequence into a pre-trained feature extraction model in sequence to obtain feature data corresponding to each frame image output by the feature extraction model;

[0050] Arranging the characteristic data obtained from all frame images in time series to obtain boiling process characteristic data of the target image sequence;

[0051] The feature extraction model is obtained by training a convolutional neural network using the image sequence and visual features of the sample.

[0052] That is to say, a convolutional neural network (CNN) is used to extract features from the preprocessed image. The extracted features mainly include boiling process features such as bubbles and vapor films. After training, the CNN model can identify key visual features of the boiling process.

[0053] Specifically, during the training process of the feature extraction model, a convolutional neural network (CNN) is used as the feature extraction model, and the ResNet-50 architecture is specifically selected. ResNet-50 contains multiple convolutional layers and residual blocks, which can automatically learn complex features in images. It performs well in image feature extraction tasks and has good generalization ability.

[0054] During the training process, the CNN model is trained using the training set data. The cross-entropy loss function is used to measure the difference between the model prediction value and the true label. The optimization algorithm adopts the Adam optimizer, with an initial learning rate of 0.001 and a learning rate decay strategy of 0.1 times every 10 epochs. During the training process, verification is performed on the validation set every 5 epochs, and the loss value and accuracy of the validation set are recorded to monitor the training status of the model.

[0055] During the parameter tuning process, we adjusted the learning rate through multiple experiments and found that an initial learning rate of 0.001 resulted in faster model convergence and higher accuracy. We also adjusted the batch size to 32, which was verified to strike a good balance between training speed and model performance. We also used L2 regularization with a parameter of 0.0001 to prevent overfitting. We also set the training cycle to 50 epochs, which showed that the model's accuracy on the validation set stabilized at this time.

[0056] In step S130, based on the boiling process characteristic data, the boiling degree of the boiling process within the target duration is classified to obtain a boiling degree classification result, which specifically includes:

[0057] The boiling process characteristic data is input into a pre-trained degree classification model to obtain a boiling degree classification result output by the degree classification model; Figure 3 As shown, the degree classification model is trained using the boiling process feature data of the sample and the corresponding boiling degree category label.

[0058] Specifically, during the training process of the degree classification model, the degree classification model adopts a long short-term memory network (LSTM) combined with a fully connected layer architecture. LSTM can process time series data and is suitable for modeling characteristic data of the boiling process; the number of hidden units in the LSTM layer is set to 128, and the fully connected layer outputs 4 categories (natural convection, nucleate boiling, transition boiling, and film boiling).

[0059] During training, the degree classification model was trained using the training set data. The cross-entropy loss function was used to measure the difference between the model's predicted values and the true labels. The Adam optimizer was used for optimization, with an initial learning rate of 0.001 and a learning rate decay strategy of 0.1% every 10 epochs. During training, validation data was tested on the validation set every five epochs, and the validation set loss and accuracy were recorded to monitor the model's training status.

[0060] During the parameter tuning process, the learning rate was adjusted through multiple experiments. It was found that when the initial learning rate was 0.001, the model converged faster and had higher accuracy. The batch size was set to 32. Experimental verification showed that this size achieved a good balance between training speed and model performance. L2 regularization was used, and the regularization parameter was set to 0.0001 to prevent overfitting of the model. The training cycle (Epoch) was set to 50 epochs. After verification, the accuracy of the model on the validation set tended to be stable at this time.

[0061] In this way, by collecting image data under different boiling states, the deep learning model is trained and verified to ensure that the model has good generalization ability and accuracy.

[0062] During the training process of the feature extraction and degree classification models, the sample data was derived from boiling reactors used in actual industrial distillation processes. By operating the boiling reactors under different operating conditions, image data was collected at different boiling levels (natural convection, nucleate boiling, transition boiling, and film boiling). At least 1,000 images were collected at each boiling level to ensure sample diversity. The acquisition device was a high-speed camera with a frame rate of 1,000 fps and a resolution of 1,920 × 1,080 pixels.

[0063] In the training process of the above feature extraction model and degree classification model, the sample data preprocessing process includes the following steps:

[0064] 1. Normalize the collected images, including adjusting the image size to 256×256 pixels, contrast enhancement, and noise removal;

[0065] 2. Label the image, including the boiling degree category (natural convection, nucleate boiling, transition boiling, and film boiling);

[0066] 3. Divide the labeled image data into a training set (70%), a validation set (15%), and a test set (15%).

[0067] In the training process of the above feature extraction model and degree classification model, model verification includes the following steps:

[0068] Use the test set to verify the trained feature extraction model and degree classification model; verification indicators include accuracy, recall, precision, and F1 score;

[0069] The confusion matrix is used to analyze the classification effect of the model on different categories to ensure that the model has good recognition ability for each boiling degree category.

[0070] The verification results are:

[0071] The feature extraction model achieved an accuracy of over 95% on the test set, with recall, precision, and F1 scores all above 90%;

[0072] The accuracy of the degree classification model on the test set reached over 90%, and the recall rate, precision rate and F1 score were all above 85%;

[0073] The confusion matrix shows that the model has the best classification effect on nucleate boiling and film boiling, and the classification effect on transition boiling is slightly worse, but still within an acceptable range.

[0074] The boiling degree classification results specifically include four levels: natural convection, nucleate boiling, transition boiling and film boiling. Among them, natural convection includes 1 degree; nucleate boiling includes 10 degrees, which are further subdivided into the early, middle and late stages of nucleate boiling; transition boiling includes 2 degrees, identifying the beginning and end of transition boiling; film boiling includes 1 degree, indicating the critical state of the boiling process.

[0075] Furthermore, the boiling degree classification results are obtained, and then the following are also included:

[0076] According to the boiling degree classification result and based on a preset control strategy, the opening of the heat medium regulating valve of the boiling reactor is controlled.

[0077] Wherein, the preset control strategy includes cascade control and fuzzy control; Figure 4 As shown, in the cascade control, the boiling degree classification result is used as the basis for the set value of the heat medium flow or pressure, and the opening of the regulating valve on the heat medium pipeline is controlled according to the set value of the heat medium flow or pressure.

[0078] For ease of understanding, the following describes the cascade control strategy. This advanced control method utilizes two or more controllers to achieve precise control of complex processes. In this specific embodiment, the cascade control strategy is used to adjust the opening of the heat medium control valve based on the boiling degree classification results to optimize the distillation process. Specifically, the boiling degree classification results serve as the basis for setting the heat medium flow rate or pressure, and the boiling state is controlled by adjusting the opening of the control valve on the heat medium pipeline.

[0079] The corresponding relationship between the boiling degree classification results and the regulating valve opening is as follows: According to the boiling degree classification results, the boiling process is divided into four main levels: natural convection, nucleate boiling, transition boiling and film boiling. Each level is further subdivided into multiple degrees, as shown in Table 1:

[0080] Natural convection: 1 degree;

[0081] Nucleate boiling: 10 levels (early, middle, late);

[0082] Transition boiling: 2 degrees (beginning, end);

[0083] Film boiling: 1 degree.

[0084] For each boiling degree, the specific control valve opening adjustment strategy is set as follows:

[0085] Natural convection means the liquid has not yet reached boiling, relying primarily on natural convection for heat transfer. The control strategy in this case is that when natural convection is detected, it indicates that the system has not yet reached the ideal boiling state and that the heat medium flow rate needs to be increased to accelerate the temperature rise. In this case, the flow rate / pressure is set to a value, such as A. Based on the set flow rate / pressure value A, the PID control valve opening is adjusted. The specific increase is determined by the difference between the actual flow rate / pressure value and the target value.

[0086] Nucleate boiling occurs when bubbles begin to form in the liquid, forming and detaching on the heating surface. This is a highly efficient heat transfer stage. The control strategy for this stage is as follows: Initial stage: Bubbles begin to form, and heat transfer efficiency gradually increases, but it has not yet reached optimal levels. At this stage, the heat transfer medium flow rate needs to be moderately increased to maintain a stable boiling state. During this stage, the flow / pressure ratio is set to a single value, such as B, and the valve opening is adjusted using PID control based on the set flow / pressure ratio. Mid-stage: Bubble formation and detachment are relatively stable, and heat transfer efficiency is high. During this stage, the heat transfer medium flow rate is maintained stable, and the valve opening is fine-tuned to optimize heat transfer. During this stage, the flow / pressure ratio is set to a single value, such as C, and the valve opening is adjusted using PID control based on the set flow / pressure ratio. Late-stage: Bubble count decreases, and heat transfer efficiency begins to decline. At this stage, the heat transfer medium flow rate needs to be appropriately reduced to prevent overheating. During this stage, the flow / pressure ratio is set to a single value, such as D, and the valve opening is adjusted using PID control based on the set flow / pressure ratio.

[0087] Transition boiling refers to the unstable boiling state of the liquid, with bubbles and vapor films appearing alternately and large fluctuations in heat transfer efficiency. At this time, the flow rate / pressure is set to a value, such as E, and the valve opening is adjusted through PID control according to the set flow rate / pressure value E.

[0088] Film boiling occurs when a vapor film forms on the surface of a liquid, resulting in extremely low heat transfer efficiency and potentially causing equipment dryness and damage. When film boiling is detected, it indicates a critical system state and requires immediate reduction of the heat transfer flow to prevent equipment damage. In this case, the flow / pressure is set to a value, such as F. The PID controller adjusts the valve opening based on the set flow / pressure value, E, and issues an alarm prompting the operator to take further action.

[0089] Table 1 Correspondence between boiling state and degree

[0090]

[0091]

[0092] In fuzzy control, according to the boiling degree classification results, a fuzzy controller is established, and the opening of the heat medium regulating valve is directly controlled by the fuzzy controller. Figure 5 As shown in the figure, establishing a fuzzy controller includes the steps of fuzzy quantization processing, establishing fuzzy control rules, establishing a fuzzy control table, and finally defuzzification.

[0093] For ease of understanding, the fuzzy control strategy is described below.

[0094] Fuzzy quantization processing is the process of converting the precise value of the input into a fuzzy set. In this specific implementation, the input variable is the boiling degree classification result. The input variable and the output variable are fuzzy quantized separately, and the output variable is the control valve opening adjustment instruction.

[0095] The input variables are boiling degree classification results, which are divided into four main levels and multiple sub-levels:

[0096] Natural convection (N)

[0097] Nucleate boiling (NB: early stage, NM: middle stage, NL: late stage)

[0098] Transition boiling (TB: beginning, TE: end)

[0099] Film boiling (F)

[0100] In order to fuzzy quantify these classification results, the fuzzy set and its membership function are defined:

[0101] Natural convection (N): membership function is 1

[0102] Nucleate boiling initial stage (NB): membership function is 0.8

[0103] Nucleate boiling mid-stage (NM): membership function is 0.6

[0104] Late nucleate boiling (NL): membership function is 0.4

[0105] Transition boiling start (TB): membership function is 0.2

[0106] End of transition boiling (TE): membership function is 0.1

[0107] Film boiling (F): membership function is 0

[0108] The output variable is the control valve opening adjustment instruction, which is divided into the following fuzzy sets:

[0109] Significant increase (BI)

[0110] Moderate increase (MI)

[0111] Slight increase (SI)

[0112] Remain unchanged

[0113] Slight decrease (SD)

[0114] Moderate decrease (MD)

[0115] Significant reduction (BD)

[0116] Membership function definition:

[0117] Substantial increase (BI): membership function is 0.9

[0118] Moderate increase (MI): membership function is 0.7

[0119] Slight increase (SI): membership function is 0.5

[0120] Remain unchanged (H): the membership function is 0.3

[0121] Slight decrease (SD): membership function is 0.2

[0122] Moderate reduction (MD): membership function is 0.1

[0123] Greatly reduced (BD): The membership function is 0.

[0124] Fuzzy control rules define the fuzzy set of output variables based on the fuzzy set of input variables. Table 2 is a table of specific fuzzy control rules:

[0125] Table 2 Fuzzy control rules table

[0126]

[0127] A fuzzy control table is a two-dimensional table used to store fuzzy control rules. The rows of the table correspond to the fuzzy sets of the input variables, and the columns correspond to the fuzzy sets of the output variables. Table 3 is an example of a fuzzy control table:

[0128] Table 3 Fuzzy control table

[0129]

[0130]

[0131] The fuzzy reasoning process includes the following steps:

[0132] Fuzzification: Map the input boiling degree classification results to fuzzy sets.

[0133] Rule matching: Calculate the matching degree of each rule according to the rules in the fuzzy control table.

[0134] Composite reasoning: synthesize the outputs of all matching rules to obtain fuzzy output.

[0135] In one embodiment, assuming that at a certain moment, the boiling degree classification result is nucleate boiling mid-stage (NM), the processing process of the fuzzy controller is as follows:

[0136] Fuzzification:

[0137] The input variable is the nucleate boiling mid-stage (NM), and the membership function is 0.6.

[0138] Rule matching:

[0139] According to the fuzzy control table, the rule corresponding to the middle stage of nucleate boiling (NM) is slight increase (SI), and the membership degree is 0.6.

[0140] Synthetic reasoning:

[0141] The result of synthetic reasoning is a slight increase (SI) with a membership of 0.6.

[0142] Defuzzification is to convert the fuzzy output into a specific control valve opening adjustment instruction. The center of gravity method (COG) is used for defuzzification to calculate the center of gravity position of the fuzzy output and obtain the specific control valve opening adjustment value.

[0143] In one embodiment, assuming the fuzzy output is a small increase (SI) with a membership of 0.6, the defuzzification process is as follows:

[0144] Define the membership function of the fuzzy set:

[0145] Substantial increase (BI): membership function is 0.9

[0146] Moderate increase (MI): membership function is 0.7

[0147] Slight increase (SI): membership function is 0.5

[0148] Remain unchanged (H): the membership function is 0.3

[0149] Slight decrease (SD): membership function is 0.2

[0150] Moderate reduction (MD): membership function is 0.1

[0151] Greatly reduced (BD): the membership function is 0

[0152] Compute the centroid of the fuzzy output:

[0153] Assume that the fuzzy output is small increase (SI) and the membership degree is 0.6.

[0154] The formula for calculating the center of gravity is:

[0155] Center of gravity = ∑ membership ∑ (membership × adjustment value)

[0156] Assume that the adjustment values are:

[0157] BI: 20%

[0158] MI: 15%

[0159] SI: 10%

[0160] H: 0%

[0161] SD: -5%

[0162] MD: -10%

[0163] BD: -15%

[0164] The calculation results are:

[0165] Center of gravity = 0.60.6 × 10% = 10%

[0166] Control valve opening adjustment:

[0167] According to the calculation results, the opening of the regulating valve increases by 10%.

[0168] In the above-mentioned specific embodiment, the method for determining the boiling degree of an industrial distillation process based on machine vision provided by the present invention obtains a target image sequence of the boiling process within a target time in a boiling reactor, extracts feature data of each image in the target image sequence, and obtains boiling process feature data of the target image sequence; based on the boiling process feature data, the boiling degree of the boiling process within the target time is classified to obtain a boiling degree classification result. In this way, the boiling degree determination method provided by the present invention can monitor the boiling state in the industrial distillation process in real time and accurately, providing a scientific basis for optimizing the distillation process, improving distillation efficiency and product quality, and ensuring equipment safety. This solves the technical problem in the prior art that the boiling state cannot be reflected in real time and the distillation process cannot be optimized and controlled according to the boiling state.

[0169] In addition to the above method, the present invention also provides a device for judging the boiling degree of industrial distillation process based on machine vision, such as Figure 6 As shown, the device includes:

[0170] The image acquisition unit 610 is used to obtain a target image sequence of the boiling process within a target time period in the boiling reactor;

[0171] a feature extraction unit 620, configured to extract feature data of each image in the target image sequence and obtain boiling process feature data of the target image sequence;

[0172] The classification generating unit 630 is configured to classify the boiling degree of the boiling process within the target duration based on the boiling process characteristic data to obtain a boiling degree classification result.

[0173] In some embodiments, acquiring a target image sequence of a boiling process in a boiling reactor specifically includes:

[0174] Use a high-speed camera to capture the boiling surface in real time during industrial distillation, obtaining high-frame-rate image sequences of the boiling process.

[0175] The collected images are normalized to obtain the target image sequence.

[0176] In some embodiments, extracting feature data of each image in the target image sequence specifically includes:

[0177] Inputting each frame image in the target image sequence into a pre-trained feature extraction model in sequence to obtain feature data corresponding to each frame image output by the feature extraction model;

[0178] Arranging the characteristic data obtained from all frame images in time series to obtain boiling process characteristic data of the target image sequence;

[0179] The feature extraction model is obtained by training a convolutional neural network using the image sequence and visual features of the sample.

[0180] In some embodiments, based on the boiling process characteristic data, the boiling degree of the boiling process within the target duration is classified to obtain a boiling degree classification result, specifically including:

[0181] Inputting the boiling process characteristic data into a pre-trained degree classification model to obtain a boiling degree classification result output by the degree classification model;

[0182] The degree classification model is trained using the boiling process feature data of the sample and the corresponding boiling degree category label.

[0183] In some embodiments, the boiling degree classification results specifically include four levels: natural convection, nucleate boiling, transition boiling, and film boiling.

[0184] In some embodiments, obtaining a boiling degree classification result further includes:

[0185] According to the boiling degree classification result and based on a preset control strategy, the opening of the heat medium regulating valve of the boiling reactor is controlled.

[0186] In some embodiments, the preset control strategy includes:

[0187] The boiling degree classification result is used as a basis for setting the heat medium flow rate or pressure, and the opening of the regulating valve on the heat medium pipeline is controlled according to the setting value of the heat medium flow rate or pressure.

[0188] In some embodiments, the preset control strategy includes:

[0189] According to the boiling degree classification results, a fuzzy controller is established to directly control the opening of the heat medium regulating valve.

[0190] In some embodiments, an image acquisition device is installed on the boiling reactor, and the image acquisition device is installed on the top or side of the boiling reactor.

[0191] In the above-mentioned specific embodiment, the device for judging the boiling degree of an industrial distillation process based on machine vision provided by the present invention obtains a target image sequence of the boiling process within a target time in a boiling reactor, extracts feature data of each image in the target image sequence, and obtains boiling process feature data of the target image sequence; based on the boiling process feature data, the boiling degree of the boiling process within the target time is classified to obtain a boiling degree classification result. In this way, the device for judging the boiling degree provided by the present invention can monitor the boiling state in the industrial distillation process in real time and accurately, providing a scientific basis for optimizing the distillation process, improving distillation efficiency and product quality, and ensuring equipment safety. This solves the technical problem in the prior art that the boiling state cannot be reflected in real time and the distillation process cannot be optimized and controlled according to the boiling state.

[0192] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units, i.e., they may be located in one location or distributed across multiple network units. Some or all of the modules may be selected based on actual needs to achieve the objectives of the present embodiment. Persons of ordinary skill in the art will be able to understand and implement the present invention without inventive effort.

[0193] Through the description of the above embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus a necessary general hardware platform, or of course, by hardware. Based on this understanding, the essence of the above technical solution or the part that contributes to the existing technology can be embodied in the form of a software product. The computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, a magnetic disk, an optical disk, etc., and includes a number of instructions for enabling a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in each embodiment or certain parts of the embodiments.

[0194] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the various embodiments of the present invention.

Claims

1. A method for judging boiling degree in industrial distillation process based on machine vision, characterized in that: The method comprises: Acquire a target image sequence of the boiling process within a target time period in a boiling reactor; Extracting feature data of each image in the target image sequence and obtaining boiling process feature data of the target image sequence; Based on the boiling process characteristic data, the boiling degree of the boiling process within the target duration is classified to obtain a boiling degree classification result.

2. The method for judging boiling degree in industrial distillation process based on machine vision according to claim 1, characterized in that: Acquire the target image sequence of the boiling process in the boiling reactor, specifically including: Use a high-speed camera to capture the boiling surface in real time during industrial distillation, obtaining high-frame-rate image sequences of the boiling process. The collected images are normalized to obtain the target image sequence.

3. The method for judging boiling degree in industrial distillation process based on machine vision according to claim 1, characterized in that: Extracting feature data of each image in the target image sequence specifically includes: Inputting each frame image in the target image sequence into a pre-trained feature extraction model in sequence to obtain feature data corresponding to each frame image output by the feature extraction model; Arranging the characteristic data obtained from all frame images in time series to obtain boiling process characteristic data of the target image sequence; The feature extraction model is obtained by training a convolutional neural network using the image sequence and visual features of the sample.

4. The method for determining boiling degree in an industrial distillation process based on machine vision according to claim 1, wherein: Based on the boiling process characteristic data, the boiling degree of the boiling process within the target duration is classified to obtain a boiling degree classification result, specifically including: Inputting the boiling process characteristic data into a pre-trained degree classification model to obtain a boiling degree classification result output by the degree classification model; The degree classification model is trained using the boiling process feature data of the sample and the corresponding boiling degree category label.

5. The method for judging boiling degree in industrial distillation process based on machine vision according to claim 4, characterized in that: The boiling degree classification results specifically include four levels: natural convection, nucleate boiling, transition boiling and film boiling.

6. The method for judging boiling degree in industrial distillation process based on machine vision according to claim 1, characterized in that: Get the boiling degree classification results, followed by: According to the boiling degree classification result and based on a preset control strategy, the opening of the heat medium regulating valve of the boiling reactor is controlled.

7. The method for judging boiling degree in industrial distillation process based on machine vision according to claim 6, characterized in that: The preset control strategy includes: The boiling degree classification result is used as a basis for setting the heat medium flow rate or pressure, and the opening of the regulating valve on the heat medium pipeline is controlled according to the setting value of the heat medium flow rate or pressure.

8. The method for judging boiling degree in industrial distillation process based on machine vision according to claim 6, characterized in that: The preset control strategy includes: According to the boiling degree classification results, a fuzzy controller is established to directly control the opening of the heat medium regulating valve.

9. The method for determining boiling degree in an industrial distillation process based on machine vision according to any one of claims 1 to 8, characterized in that: An image acquisition device is installed on the boiling reactor, and the image acquisition device is installed on the top or side of the boiling reactor.

10. A device for judging boiling degree in industrial distillation process based on machine vision, characterized in that: The device comprises: An image acquisition unit is used to obtain a target image sequence of the boiling process within a target time period in the boiling reactor; a feature extraction unit, configured to extract feature data of each image in the target image sequence and obtain boiling process feature data of the target image sequence; The classification generating unit is configured to classify the boiling degree of the boiling process within the target duration based on the boiling process characteristic data to obtain a boiling degree classification result.

Images