Mass flow detection model training and liquid-solid two-phase flow mass flow detection method

Abstract

The invention discloses a mass flow detection model training and liquid-solid two-phase flow mass flow detection method, which is applied to the field of mass flow detection, and comprises the following steps: dividing a liquid-solid two-phase flow measurement signal set with a mass flow label into a plurality of signal subsets; the liquid-solid two-phase flow measurement signals in each signal subset have the same mass flow label; constructing a circular system limited crossing visual graph of each liquid-solid two-phase flow measurement signal to obtain a visual graph set corresponding to the signal subset; constructing a mass flow detection model; and training a quality flow detection model based on the visual image set to obtain a target quality flow detection model. Based on a deep learning method, a mass flow detection model is trained through liquid-solid two-phase flow measurement signals under each mass flow, the mass flow of the liquid-solid two-phase flow is measured through the mass flow detection model, and due to the fact that the liquid-solid two-phase flow measurement signals are sensitive to weak changes of the mass flow, compared with the prior art, the mass flow measurement accuracy is greatly improved. According to the method, the accuracy of the measurement result is improved.

Description

Technical Field

[0001] The present invention relates to the field of mass flow detection, and in particular to a mass flow detection model training method, a liquid-solid two-phase flow mass flow detection method, a device, an electronic device and a computer-readable storage medium. Background Art

[0002] Liquid-solid two-phase flow is one of the most representative flow forms in multiphase flow, and is commonly found in industrial production in many industries such as chemical industry, metallurgy, and power generation. The transportation of materials such as ore mud is a typical application of liquid-solid two-phase flow, especially for mass flow measurement in cement slurry. If the mass flow rate of liquid-solid two-phase flow suddenly decreases during the production process, the particle flow will be unstable, especially for modified pneumatic transportation; if the mass flow rate suddenly increases, more energy will be consumed, which will cause serious pipeline wear and particle breakage. In addition, since the shape and distribution of the phase interface are randomly variable in time and space, the flow pattern of liquid-solid two-phase flow is complex and diverse, and the complexity and variability of the flow pattern brings great difficulties to the accurate measurement of the mass flow rate of liquid-solid two-phase flow.

[0003] The existing technology generally uses instruments to measure the mass flow rate of liquid-solid two-phase flow. It is difficult to capture subtle changes in the mass flow rate of liquid-solid two-phase flow through instrument measurement, and the error is large and the measurement accuracy is low. Summary of the invention

[0004] The purpose of the present invention is to provide a mass flow detection model training method, a liquid-solid two-phase flow mass flow detection method, a device, an electronic device and a computer-readable storage medium, which are applied to the field of mass flow detection. The method is based on a deep learning method, and the mass flow detection model is trained by the liquid-solid two-phase flow measurement signal under each mass flow rate. The mass flow rate of the liquid-solid two-phase flow is measured by the mass flow detection model. Since the liquid-solid two-phase flow measurement signal is more sensitive to the slight change of the mass flow rate, compared with the prior art, the method of the present invention improves the accuracy of the measurement result.

[0005] In order to solve the above technical problems, the present invention provides a mass flow detection model training method, comprising:

[0006] Dividing a liquid-solid two-phase flow measurement signal set with a mass flow label into a plurality of signal subsets; wherein the liquid-solid two-phase flow measurement signals in each of the signal subsets have the same mass flow label;

[0007] Constructing a circle system finite-travel visual graph of each of the liquid-solid two-phase flow measurement signals to obtain a visual graph set corresponding to the signal subset;

[0008] Construct a mass flow detection model;

[0009] Training the mass flow detection model based on the visual graph to obtain a target mass flow detection model;

[0010] The mass flow detection model includes a first feature extraction layer, a second feature extraction layer, a Flatten layer, a Concatenate layer and a fully connected layer connected in sequence;

[0011] The first feature extraction layer is used to extract the features of the circle system finitely crossing the visible graph in the visible graph set to obtain a visible graph feature set;

[0012] The second feature extraction layer is used to extract the visible image features in the visible image feature set by grouping to obtain a plurality of group feature sets;

[0013] The Flatten layer is used to convert the grouping features in the grouping feature set into one-dimensional grouping features;

[0014] The Concatenate layer is used to concatenate all the one-dimensional grouping features to obtain fused features;

[0015] The fully connected layer is used to receive the fusion features and output the mass flow prediction value; the activation function of the fully connected layer is the sigmoid function.

[0016] Optionally, the method further includes:

[0017] Segmenting the liquid-solid two-phase flow measurement signal by a sliding window of a preset time to obtain a plurality of segmented signals;

[0018] Constructing the circle system of each segmented signal to finitely cross the visible graph to obtain the visible graph set;

[0019] The mass flow detection model is trained based on the visual graph to obtain the target mass flow detection model.

[0020] Optionally, the first feature extraction layer includes a convolutional layer and a pooling layer connected in sequence.

[0021] Optionally, the second feature extraction layer includes multiple feature extraction branches, each of the feature extraction branches includes multiple feature extraction sub-branches connected in sequence, and the feature extraction sub-branches include convolutional layers and pooling layers connected in sequence.

[0022] Optionally, the size of the convolution kernel in the first feature extraction layer is 16, the size of the convolution kernel of the last feature extraction sub-branch in each feature extraction branch of the second feature extraction layer is 64, and the size of the convolution kernel of the remaining feature extraction sub-branches is 32.

[0023] Optionally, the training the mass flow detection model based on the visible graph to obtain a target mass flow detection model includes:

[0024] The mass flow detection model is trained based on the visual set to obtain a target mass flow detection model. During the training process, the accuracy of the target mass flow detection model is verified by a ten-fold cross validation method.

[0025] In order to solve the above technical problems, the present invention provides a liquid-solid two-phase flow mass flow detection method, comprising:

[0026] Acquire a liquid-solid two-phase flow measurement signal to be detected;

[0027] Acquire a mass flow prediction value based on the liquid-solid two-phase flow measurement signal to be detected and a target mass flow detection model;

[0028] Wherein, the target mass flow detection model is a model trained according to any one of the mass flow detection model training methods.

[0029] In order to solve the above technical problems, the present invention provides a detection model training device, comprising:

[0030] The first module is used to divide the liquid-solid two-phase flow measurement signal set with a mass flow label into a plurality of signal subsets; wherein the liquid-solid two-phase flow measurement signal in each of the signal subsets has the same mass flow label;

[0031] The second module is used to construct a circle system finite traversal visual graph of each of the liquid-solid two-phase flow measurement signals to obtain a visual graph set corresponding to the signal subset;

[0032] The third module is used to build a mass flow detection model;

[0033] A fourth module is used to train the mass flow detection model based on the visual graph to obtain a target mass flow detection model;

[0034] The mass flow detection model includes a first feature extraction layer, a second feature extraction layer, a Flatten layer, a Concatenate layer and a fully connected layer connected in sequence;

[0035] The first feature extraction layer is used to extract the features of the circle system finitely crossing the visible graph in the visible graph set to obtain a visible graph feature set;

[0036] The second feature extraction layer is used to extract the visible image features in the visible image feature set by grouping to obtain a plurality of group feature sets;

[0037] The Flatten layer is used to convert the grouping features in the grouping feature set into one-dimensional grouping features;

[0038] The Concatenate layer is used to concatenate all the one-dimensional grouping features to obtain fused features;

[0039] The fully connected layer is used to receive the fusion features and output the mass flow prediction value; the activation function of the fully connected layer is the sigmoid function.

[0040] In order to solve the above technical problems, the present invention provides an electronic device, comprising:

[0041] Memory, for storing computer programs;

[0042] A processor is used to implement any one of the mass flow detection model training methods and / or the liquid-solid two-phase flow mass flow detection methods when executing the computer program.

[0043] In order to solve the above technical problems, the present invention provides a computer-readable storage medium, in which computer-executable instructions are stored. When the computer-executable instructions are executed by a processor, any one of the mass flow detection model training methods and / or the liquid-solid two-phase flow mass flow detection method is implemented.

[0044] It can be seen that this method divides the liquid-solid two-phase flow measurement signal set with mass flow labels into multiple signal subsets; the liquid-solid two-phase flow measurement signals in each signal subset have the same mass flow label; constructs a circular system finite traversal visual graph of each liquid-solid two-phase flow measurement signal to obtain a visual graph set corresponding to the signal subset; constructs a mass flow detection model; trains the mass flow detection model based on the visual graph set to obtain a target mass flow detection model; based on a deep learning method, trains the mass flow detection model through the liquid-solid two-phase flow measurement signals at each mass flow rate, and measures the mass flow rate of the liquid-solid two-phase flow through the mass flow detection model. Since the liquid-solid two-phase flow measurement signal is more sensitive to slight changes in the mass flow rate, the accuracy of the measurement result is improved compared with the existing technology. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on the provided drawings without paying creative work.

[0046] Figure 1 A flow chart of a mass flow detection model training method provided by an embodiment of the present invention;

[0047] Figure 2A flow chart of a liquid-solid two-phase flow mass flow detection method provided by an embodiment of the present invention;

[0048] Figure 3 A structural schematic diagram of a mass flow detection model provided by an embodiment of the present invention;

[0049] Figure 4 This is a structural block diagram of a mass flow detection model training device provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0050] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0051] The following combination Figure 1 , Figure 1 A flow chart of a model training method provided by an embodiment of the present invention, the method may include:

[0052] S101: Divide a liquid-solid two-phase flow measurement signal set with a mass flow label into a plurality of signal subsets; wherein the liquid-solid two-phase flow measurement signals in each signal subset have the same mass flow label.

[0053] This embodiment can obtain a liquid-solid two-phase flow measurement signal set with a mass flow label, and the liquid-solid two-phase flow measurement signal set includes liquid-solid two-phase flow measurement signals at each mass flow rate. This embodiment does not limit the method of obtaining the liquid-solid two-phase flow measurement signal, which can be a voltage signal read from a sensor. The liquid-solid two-phase flow measurement signal set in this embodiment can be used as a data set for training a mass flow detection model.

[0054] Grouped convolution can effectively reduce the amount of calculation and model parameters, and save computing resources. In this embodiment, the idea of ​​grouped convolution can be introduced. In order to perform grouping in the model later, this embodiment can first divide the liquid-solid two-phase flow measurement signal set with mass flow labels into multiple signal subsets; wherein the liquid-solid two-phase flow measurement signals in each signal subset have the same mass flow label. In order to ensure the accuracy of the model, the number of signals in each signal subset in this embodiment can be the same.

[0055] S102: constructing a circle system finite-travel visual graph of each liquid-solid two-phase flow measurement signal to obtain a visual graph set corresponding to the signal subset.

[0056] As an important method of complex network analysis, visibility graph theory is widely used in univariate time series analysis. Liquid-solid two-phase flow measurement signals are typical time series signals. This embodiment can convert liquid-solid two-phase flow measurement signals into complex networks through visibility graphs to obtain characteristic degree sequences. The circular limited penetrable visibility graph (CLPVG) can effectively capture the important features of time series and has better noise resistance than ordinary visibility graphs. This embodiment constructs a circular limited penetrable visibility graph of each liquid-solid two-phase flow measurement signal, so that the model can learn the features of time series more efficiently, thereby further improving the accuracy of the trained mass flow detection model.

[0057] This embodiment can construct a circular finite-crossing visible graph of each liquid-solid two-phase flow measurement signal. The liquid-solid two-phase flow measurement signal in each signal subset corresponds to a circular finite-crossing visible graph. The circular finite-crossing visible graphs corresponding to the liquid-solid two-phase flow measurement signals in each signal subset are organized into a visible graph set, and the visible graph set corresponding to each signal subset can be obtained.

[0058] S103: Construct a mass flow detection model.

[0059] This embodiment can construct a mass flow detection model and measure the mass flow rate of liquid-solid two-phase flow through a deep learning method. The mass flow detection model in this embodiment includes a first feature extraction layer, a second feature extraction layer, a Flatten layer, a Concatenate layer and a fully connected layer connected in sequence.

[0060] In this embodiment, the first feature extraction layer can be used to extract features of a circle system finitely passing through a visible graph in a visible graph set to obtain a visible graph feature set, and the visible graph feature set includes visible graph features corresponding to the circle system finitely passing through the visible graph in the visible graph set. This embodiment does not limit the specific method of extracting features by the first feature extraction layer. Generally, the first feature extraction layer may include a convolution layer and a pooling layer connected in sequence. In this embodiment, the size of the convolution kernel of the convolution layer in the first feature extraction layer is not limited, and generally can be 16.

[0061] In this embodiment, the second feature extraction layer can be used to group the visible graph features in the visible graph feature set to extract features and obtain multiple grouped feature sets. The second feature extraction layer can group the visible graph features in the visible graph feature set to obtain multiple visible graph feature subsets. The second feature extraction layer can include multiple feature extraction branches, each feature extraction branch is used to extract features of the corresponding visible graph feature subset, thereby realizing grouped feature extraction and reducing the amount of calculation and parameters of the model. In order to ensure the accuracy of the model, in this embodiment, in actual applications, the visible graph features in the visible graph feature set should be evenly divided, and the number of visible graph feature subsets after equal division should correspond to the number of feature extraction branches, that is, each feature extraction branch is used to extract features of a visible graph feature subset, and then iterate multiple grouped feature sets.

[0062] This embodiment does not limit the structure of each feature extraction branch. Generally, each feature extraction branch may include multiple feature extraction sub-branches connected in sequence, and each feature extraction sub-branch includes a convolution layer and a pooling layer connected in sequence. This embodiment does not limit the number of feature extraction branches, the number of feature extraction sub-branches, or the size of the convolution kernel of each convolution layer in the feature extraction branch. Generally, the second feature extraction layer may include 3 feature extraction branches, each feature extraction branch includes 4 feature extraction sub-branches, and the convolution kernel size of the convolution layer of the last feature extraction sub-branch may be 64, and the convolution kernel size of the convolution layer of the remaining feature extraction sub-branches may be 32.

[0063] The Flatten layer in this embodiment is used to convert the grouping features in the grouping feature set into one-dimensional grouping features. The Flatten layer is a flattening layer, which is used to expand multidimensional data into a one-dimensional vector to facilitate subsequent neural network calculations. In this embodiment, the Flatten layer can be used to convert the grouping features in the grouping feature set into one-dimensional vectors to obtain one-dimensional grouping features.

[0064] The Concatenate layer in this embodiment is used to concatenate all one-dimensional grouping features to obtain fused features. The Concatenate layer is a very important operation in network structure design and is often used to concatenate features to perform feature fusion.

[0065] The fully connected layer in this embodiment is used to receive the fused features and output the mass flow prediction value, and the activation function of the fully connected layer is the sigmoid function.

[0066] S104: Training a mass flow detection model based on the visible graph to obtain a target mass flow detection model.

[0067] In this embodiment, a mass flow detection model can be trained based on a visible data set to obtain a target mass flow detection model, and a loss value is calculated by inputting each visible data set into the mass flow detection model and outputting the value and the label corresponding to the visible data set, and the model parameters are adjusted by the loss value, and the training is iterated until the model parameters converge to obtain a trained mass flow detection model, and the trained mass flow detection model is determined as the target mass flow model. In the training process, the ten-fold cross validation method can be used to verify the accuracy of the model.

[0068] In the process of deep learning, when the amount of data used to train the model is too small, the model will be overfitted, resulting in large errors in the final mass flow detection results. To solve the above problems, this embodiment can segment the liquid-solid two-phase flow measurement signal into a sliding window of a preset time to obtain multiple segmented signals. This embodiment does not limit the size of the preset time and can be set according to the actual application scenario. After segmentation, a circle system finitely crosses the visible graph of each segmented signal to obtain a visible graph set, and further train the mass flow detection model based on the visible graph set of the segmented signal to obtain the target mass flow detection model.

[0069] This embodiment is based on a deep learning method. The mass flow detection model is trained by the liquid-solid two-phase flow measurement signal at each mass flow rate. The mass flow rate of the liquid-solid two-phase flow is measured by the mass flow detection model. Since the liquid-solid two-phase flow measurement signal is more sensitive to slight changes in the mass flow rate, the accuracy of the measurement result is improved compared with the prior art.

[0070] The following combination Figure 2 , Figure 2 A flow chart of a liquid-solid two-phase flow mass flow detection method provided in an embodiment of the present invention, the method may include:

[0071] S201: Acquire a liquid-solid two-phase flow measurement signal to be detected.

[0072] S202: Obtaining a mass flow prediction value based on the liquid-solid two-phase flow measurement signal to be detected and a target mass flow detection model.

[0073] Wherein, the target mass flow detection model is a model trained according to any mass flow detection model training method.

[0074] This embodiment can obtain the measurement signal of the liquid-solid two-phase flow to be detected. When the target mass flow detection model is obtained by training when the amount of data in the data set is sufficient, this embodiment can construct a circular system finite-crossing visible graph of the liquid-solid two-phase flow signal to be measured, and replicate the circular system finite-crossing visible graph multiple times to obtain a visible graph set to be detected, and input the visible graph set to be detected into the target mass flow detection model to obtain the mass flow prediction value. When the target mass flow detection model is obtained by training when the amount of data in the data set is insufficient, the liquid-solid two-phase flow measurement signal to be detected is segmented by a sliding window of a preset time to obtain multiple segmented signals, and a circular system finite-crossing visible graph of each segmented signal is constructed to obtain a visible graph set to be detected, and the visible graph set to be detected is input into the target mass flow detection model to obtain the mass flow prediction value. This embodiment detects the mass flow of liquid-solid two-phase flow through a deep learning method, which improves the accuracy of the detection results, and does not require cleaning of the instrument.

[0075] The following is a specific embodiment of a liquid-solid two-phase flow mass flow measurement method provided by an embodiment of the present invention. The structural schematic diagram of the mass flow detection model constructed by the specific implementation can be as follows: Figure 3 As shown, the first feature extraction layer of the mass flow detection model may include a convolution layer with a convolution kernel size of 16 and a pooling layer connected in sequence, the second feature extraction layer may include three feature extraction branches, each feature extraction branch includes four feature extraction sub-branches connected in sequence, each feature extraction sub-branch includes a convolution layer and a pooling layer connected in sequence, the convolution kernel size of the convolution layer of the last feature extraction sub-branch may be 64, the convolution kernel size of the convolution layer of the remaining feature extraction sub-branches may be 32, the output of each branch is connected to the Flatten layer, the output of the Flatten is connected to the Concatenate layer, the output of the Concatenate layer enters the fully connected layer, and the specific embodiment may be:

[0076] The liquid-solid two-phase flow measurement signal set with mass flow labels is divided into a plurality of signal subsets; wherein the liquid-solid two-phase flow measurement signals in each signal subset have the same mass flow label;

[0077] Construct a circle system finite-travel visual graph of each liquid-solid two-phase flow measurement signal to obtain a visual graph set corresponding to the signal subset;

[0078] Inputting the visible graph set into the first feature extraction branch of the mass flow detection model to obtain a visible graph feature set;

[0079] The visible graph feature set is divided into three groups, and each group of visible graph feature sets is subjected to the feature extraction branch of the second feature extraction layer to obtain a grouped feature set;

[0080] The grouping features in the grouping feature set are input into the Flatten layer for one-dimensional processing to obtain one-dimensional grouping features;

[0081] All one-dimensional grouping features are concatenated in the Concatenate layer to obtain fused features;

[0082] Input the fused features into the fully connected layer whose activation function is the sigmoid function to obtain the output value;

[0083] The loss value is calculated based on the output value and label value of each visible graph set, the model parameters are adjusted according to the loss value, and the training is iterated until the model parameters converge to obtain the trained target mass flow detection model;

[0084] Constructing a circle system finite-through visible graph of the liquid-solid two-phase flow measurement signal to be detected, and duplicating the circle system finite-through visible graph multiple times to construct a set of visible graphs to be detected;

[0085] The visual set to be detected is input into the target mass flow detection model to obtain the mass flow prediction value.

[0086] The following combination Figure 4 , Figure 4 A structural block diagram of a mass flow detection model training device provided by an embodiment of the present invention, the device may include:

[0087] The first module 100 is used to divide the liquid-solid two-phase flow measurement signal set with a mass flow label into a plurality of signal subsets; wherein the liquid-solid two-phase flow measurement signal in each of the signal subsets has the same mass flow label;

[0088] The second module 200 is used to construct a circle system finite-travel visual graph of each of the liquid-solid two-phase flow measurement signals to obtain a visual graph set corresponding to the signal subset;

[0089] The third module 300 is used to build a mass flow detection model;

[0090] The fourth module 400 is used to train the mass flow detection model based on the visual graph to obtain a target mass flow detection model;

[0091] The mass flow detection model includes a first feature extraction layer, a second feature extraction layer, a Flatten layer, a Concatenate layer and a fully connected layer connected in sequence;

[0092] The first feature extraction layer is used to extract the features of the circle system finitely crossing the visible graph in the visible graph set to obtain a visible graph feature set;

[0093] The second feature extraction layer is used to extract the visible image features in the visible image feature set by grouping to obtain a plurality of group feature sets;

[0094] The Flatten layer is used to convert the grouping features in the grouping feature set into one-dimensional grouping features;

[0095] The Concatenate layer is used to concatenate all the one-dimensional grouping features to obtain fused features;

[0096] The fully connected layer is used to receive the fusion features and output the mass flow prediction value; the activation function of the fully connected layer is the sigmoid function.

[0097] Based on the above embodiments, the present invention is based on a deep learning method, trains a mass flow detection model through liquid-solid two-phase flow measurement signals at various mass flow rates, and measures the mass flow rate of the liquid-solid two-phase flow through the mass flow detection model, which improves the accuracy of the measurement results compared to the prior art.

[0098] Based on the above embodiment, the device may further include:

[0099] A fifth module is used to segment the liquid-solid two-phase flow measurement signal into a sliding window of a preset time to obtain a plurality of segmented signals;

[0100] The sixth module is used to construct the circle system of each segmented signal to finitely cross the visible graph to obtain the visible graph set;

[0101] The seventh module is used to train the mass flow detection model based on the visual graph to obtain the target mass flow detection model.

[0102] Based on the above embodiments, the first feature extraction layer includes a convolution layer and a pooling layer connected in sequence.

[0103] Based on the above embodiments, the second feature extraction layer includes multiple feature extraction branches, each of the feature extraction branches includes multiple feature extraction sub-branches connected in sequence, and the feature extraction sub-branches include convolutional layers and pooling layers connected in sequence.

[0104] Based on the above embodiments, the size of the convolution kernel in the first feature extraction layer is 16, the size of the convolution kernel of the last feature extraction sub-branch in each feature extraction branch of the second feature extraction layer is 64, and the size of the convolution kernel of the remaining feature extraction sub-branches is 32.

[0105] Based on the above embodiments, the fourth module 400 may include:

[0106] A verification unit is used to train the mass flow detection model based on the visual set to obtain a target mass flow detection model, and during the training process, verify the accuracy of the target mass flow detection model through a ten-fold cross validation method.

[0107] The following is a liquid-solid two-phase flow mass flow measurement device provided by an embodiment of the present invention, which may include:

[0108] An eighth module is used to obtain a measurement signal of a liquid-solid two-phase flow to be detected;

[0109] A ninth module is used to obtain a mass flow prediction value based on the liquid-solid two-phase flow measurement signal to be detected and a target mass flow detection model;

[0110] Wherein, the target mass flow detection model is a model trained by the mass flow detection model training device.

[0111] Based on the above embodiments, the present invention further provides an electronic device, which may include a memory and a processor, wherein the memory stores a computer program, and when the processor calls the computer program in the memory, the steps provided in the above embodiments may be implemented. Of course, the device may also include various necessary network interfaces, power supplies, and other components.

[0112] The present invention also provides a computer-readable storage medium on which a computer program is stored. When the computer program is executed by an execution terminal or a processor, the method provided in the embodiment of the present invention can be implemented; the storage medium may include: a USB flash drive, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk, and other media that can store program codes.

[0113] In this article, relational terms such as first and second, etc. are used only to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "comprise", "include" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, the elements defined by the statement "comprise a ..." do not exclude the presence of other identical elements in the process, method, article or device including the elements.

Claims

1. A mass flow detection model training method, characterized in that: include: Dividing a liquid-solid two-phase flow measurement signal set with a mass flow label into a plurality of signal subsets; wherein the liquid-solid two-phase flow measurement signals in each of the signal subsets have the same mass flow label; Constructing a circle system finite-travel visual graph of each of the liquid-solid two-phase flow measurement signals to obtain a visual graph set corresponding to the signal subset; Construct a mass flow detection model; Training the mass flow detection model based on the visual graph to obtain a target mass flow detection model; The mass flow detection model includes a first feature extraction layer, a second feature extraction layer, a Flatten layer, a Concatenate layer and a fully connected layer connected in sequence; The first feature extraction layer is used to extract the features of the circle system finitely crossing the visible graph in the visible graph set to obtain a visible graph feature set; The second feature extraction layer is used to extract the visible image features in the visible image feature set by grouping to obtain a plurality of group feature sets; The Flatten layer is used to convert the grouping features in the grouping feature set into one-dimensional grouping features; The Concatenate layer is used to concatenate all the one-dimensional grouping features to obtain fused features; The fully connected layer is used to receive the fusion features and output the mass flow prediction value; the activation function of the fully connected layer is the sigmoid function.

2. The mass flow detection model training method according to claim 1, characterized in that: Also includes: Segmenting the liquid-solid two-phase flow measurement signal by a sliding window of a preset time to obtain a plurality of segmented signals; Constructing the circle system of each segmented signal to finitely cross the visible graph to obtain the visible graph set; The mass flow detection model is trained based on the visual graph to obtain the target mass flow detection model.

3. The mass flow detection model training method according to claim 1 is characterized in that: The first feature extraction layer includes a convolutional layer and a pooling layer connected in sequence.

4. The mass flow detection model training method according to claim 3 is characterized in that: The second feature extraction layer includes multiple feature extraction branches, each of which includes multiple feature extraction sub-branches connected in sequence, and the feature extraction sub-branches include convolutional layers and pooling layers connected in sequence.

5. The mass flow detection model training method according to claim 4 is characterized in that: The size of the convolution kernel in the first feature extraction layer is 16, the size of the convolution kernel of the last feature extraction sub-branch in each feature extraction branch of the second feature extraction layer is 64, and the size of the convolution kernel of the remaining feature extraction sub-branches is 32.

6. The mass flow detection model training method according to claim 1, characterized in that: The step of training the mass flow detection model based on the visual graph to obtain a target mass flow detection model includes: The mass flow detection model is trained based on the visual set to obtain a target mass flow detection model. During the training process, the accuracy of the target mass flow detection model is verified by a ten-fold cross validation method.

7. A liquid-solid two-phase flow mass flow detection method, characterized in that: include: Acquire a liquid-solid two-phase flow measurement signal to be detected; Acquire a mass flow prediction value based on the liquid-solid two-phase flow measurement signal to be detected and a target mass flow detection model; Wherein, the target mass flow detection model is a model trained according to the mass flow detection model training method according to any one of claims 1 to 6.

8. A detection model training device, characterized in that: include: The first module is used to divide the liquid-solid two-phase flow measurement signal set with a mass flow label into a plurality of signal subsets; wherein the liquid-solid two-phase flow measurement signal in each of the signal subsets has the same mass flow label; The second module is used to construct a circle system finite traversal visual graph of each of the liquid-solid two-phase flow measurement signals to obtain a visual graph set corresponding to the signal subset; The third module is used to build a mass flow detection model; A fourth module is used to train the mass flow detection model based on the visual graph to obtain a target mass flow detection model; The mass flow detection model includes a first feature extraction layer, a second feature extraction layer, a Flatten layer, a Concatenate layer and a fully connected layer connected in sequence; The first feature extraction layer is used to extract the features of the circle system finitely crossing the visible graph in the visible graph set to obtain a visible graph feature set; The second feature extraction layer is used to extract the visible image features in the visible image feature set by grouping to obtain a plurality of group feature sets; The Flatten layer is used to convert the grouping features in the grouping feature set into one-dimensional grouping features; The Concatenate layer is used to concatenate all the one-dimensional grouping features to obtain fused features; The fully connected layer is used to receive the fusion features and output the mass flow prediction value; the activation function of the fully connected layer is the sigmoid function.

9. An electronic device, characterized in that: include: Memory, for storing computer programs; A processor, used to implement the mass flow detection model training method as described in any one of claims 1 to 6 and / or the liquid-solid two-phase flow mass flow detection method as described in claim 7 when executing the computer program.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores computer-executable instructions, and when the computer-executable instructions are executed by the processor, the mass flow detection model training method according to any one of claims 1 to 6 and / or the liquid-solid two-phase flow mass flow detection method according to claim 7 are implemented.

Images