Gas-liquid two-phase fluid feature recognition method, device and system and storage medium

Abstract

The invention relates to a gas-liquid two-phase fluid feature recognition method, device and system and a storage medium, and the method comprises the steps: obtaining a plurality of images of a gas-liquid two-phase fluid flowing in a pipeline, and obtaining an energy signal from the fluid; wherein the plurality of images and the energy signals are synchronously acquired fluid information; determining at least one flow pattern corresponding to the plurality of images by using a convolutional neural network; determining an energy signal corresponding to each flow pattern based on the plurality of images and the time information corresponding to the energy signals; and determining an energy characteristic corresponding to each flow pattern based on the energy signal corresponding to each flow pattern. Therefore, the flow pattern information corresponding to a large number of fluid images is identified by using the convolutional neural network, so that the identification efficiency, accuracy and objectivity of the fluid images can be improved; the flow pattern of the gas-liquid two-phase fluid and the synchronously collected energy signals are subjected to correlation analysis, energy feature recognition of the whole process of the gas-liquid two-phase fluid is achieved, and then design and improvement of a pipeline are assisted.

Description

Technical Field

[0001] The present invention relates to the field of metering technology, and in particular to a method, device, system and storage medium for identifying characteristics of a gas-liquid two-phase fluid. Background Art

[0002] Gas-liquid two-phase flow refers to the phenomenon that gas and liquid flow simultaneously in the same pipe or container. It is widely used in industrial fields such as power, petroleum, and chemical industry.

[0003] The flow pattern of gas-liquid two-phase fluid refers to the phenomenon that the gas-liquid interface is distributed in different geometric shapes or different structural forms, such as bubbly flow, stratified flow, wavy flow, etc. Different flow patterns of gas-liquid two-phase fluid are often accompanied by different energy changes, which has an important impact on the stability and safety of industrial production processes. Therefore, accurately identifying the characteristic information of gas-liquid two-phase fluid has become an urgent problem to be solved. Summary of the invention

[0004] One of the purposes of the present invention is to provide a method for identifying the characteristics of a gas-liquid two-phase fluid so as to solve the problem of identifying the characteristics of a gas-liquid two-phase fluid in the prior art; the second purpose is to provide a device for identifying the characteristics of a gas-liquid two-phase fluid; the third purpose is to provide a system for identifying the characteristics of a gas-liquid two-phase fluid; and the fourth purpose is to provide a computer-readable storage medium.

[0005] In order to achieve the above object, the technical solution adopted by the present invention is as follows:

[0006] A method for identifying characteristics of a gas-liquid two-phase fluid comprises: obtaining multiple images of a gas-liquid two-phase fluid flowing in a pipeline, and obtaining an energy signal from the fluid; wherein the multiple images and the energy signal are fluid information collected synchronously; using a convolutional neural network to determine at least one flow type corresponding to the multiple images; based on time information corresponding to the multiple images and the energy signal, determining an energy signal corresponding to each of the flow types; and based on the energy signal corresponding to each of the flow types, determining an energy characteristic corresponding to each of the flow types.

[0007] According to the above technical means, on the one hand, the convolutional neural network is used to identify the flow pattern information corresponding to a large number of fluid images, which can improve the efficiency, accuracy and objectivity of fluid image recognition; on the other hand, the correlation analysis of the flow pattern of the gas-liquid two-phase fluid and the synchronously collected energy signal realizes the energy characteristic recognition of the whole process of the gas-liquid two-phase fluid, so that based on the energy characteristics corresponding to various flow patterns, high-energy signals that have a greater impact on pipeline safety can be identified, thereby assisting the design and improvement of pipelines in engineering development.

[0008] Further, determining the energy signal corresponding to each of the flow types based on the time information corresponding to the multiple images and the energy signals includes: determining the flow type time information corresponding to each flow type based on the time information corresponding to the multiple images; determining the energy signal corresponding to each flow type based on the flow type time information and the time information corresponding to the energy signal.

[0009] According to the above technical means, the corresponding relationship between each flow pattern of the gas-liquid two-phase fluid and the energy signal can be accurately determined, and then the energy characteristics of each flow pattern can be determined.

[0010] Further, the energy signal is time domain data;

[0011] The energy characteristics corresponding to each flow type are determined based on the energy signal corresponding to each flow type, including: for each flow type, determining the frequency domain data corresponding to the energy signal based on the time domain data; and determining the energy characteristics corresponding to the flow type based on the frequency domain data.

[0012] According to the above technical means, the energy signal is converted from time domain data to frequency domain data, so that the energy information corresponding to the energy signal can be accurately determined by using the amplitude value in the frequency domain data.

[0013] Further, the energy signal includes at least one of a vibration signal and a noise signal;

[0014] The determining, based on the time domain data, frequency domain data corresponding to the flow type; and determining, based on the frequency domain data, energy features corresponding to the flow type, include at least one of the following:

[0015] Based on the time domain data of the vibration signal, determining the frequency domain data of the vibration signal; based on the frequency domain data of the vibration signal, determining the vibration energy characteristics corresponding to the flow pattern;

[0016] Based on the time domain data of the noise signal, the frequency domain data of the noise signal is determined; based on the frequency domain data of the noise signal, the noise energy feature corresponding to the flow type is determined.

[0017] According to the above technical means, the vibration energy characteristics and / or noise energy characteristics that have a greater impact on pipeline safety can be determined. Therefore, the design and improvement of pipelines can be effectively assisted based on the vibration energy characteristics and / or noise energy characteristics.

[0018] Furthermore, the convolutional neural network includes a feature extraction network and a classification network;

[0019] The method of using a convolutional neural network to determine at least one flow type corresponding to the multiple images includes:

[0020] Using the feature extraction network to perform feature extraction on the multiple images to obtain image features corresponding to the multiple images;

[0021] The at least one flow type is determined based on the image features using the classification network.

[0022] According to the above technical means, using a convolutional neural network to identify at least one flow type corresponding to multiple images can improve the recognition speed and accuracy of flow type information, that is, the convolutional neural network can be used to quickly identify flow type information in a large amount of image data.

[0023] Furthermore, before using the convolutional neural network to determine at least one flow type corresponding to the multiple images, the method further includes:

[0024] Acquire a plurality of sample images; wherein each of the sample images has corresponding sample flow type information;

[0025] Determining predicted flow type information corresponding to the plurality of sample images using an initial neural network;

[0026] Determining loss data based on the sample flow pattern information and the predicted flow pattern information corresponding to the plurality of sample images;

[0027] Based on the loss data, the initial neural network is updated to obtain the convolutional neural network.

[0028] According to the above technical means, by using a variety of sample images to train the initial network to obtain a trained convolutional neural network, the accuracy of the convolutional neural network in identifying the flow type corresponding to the image can be improved.

[0029] A gas-liquid two-phase fluid feature recognition device, comprising:

[0030] An information acquisition module, used to acquire multiple images of the gas-liquid two-phase fluid flowing in the pipeline, and to acquire an energy signal from the fluid; wherein the multiple images and the energy signal are synchronously acquired fluid information;

[0031] A flow type determination module, configured to determine at least one flow type corresponding to the plurality of images using a convolutional neural network;

[0032] An energy signal determination module, used to determine the energy signal corresponding to each of the flow types based on the multiple images and the time information corresponding to the energy signals;

[0033] The energy characteristic determination module is used to determine the energy characteristic corresponding to each of the flow types based on the energy signal corresponding to each of the flow types.

[0034] A gas-liquid two-phase fluid feature recognition system includes: a pipeline, a camera, a light source, an energy sensor and a control device; wherein:

[0035] The light source is arranged on one side of the pipe;

[0036] The camera is arranged at the other side of the pipeline to collect multiple images of the gas-liquid two-phase fluid flowing in the pipeline and transmit the multiple images to the control device;

[0037] The energy sensor is arranged on the outside of the pipeline to synchronously collect energy signals from the fluid during the process of the camera collecting the multiple images, and transmit the energy signals to the control device;

[0038] The control device is used to use a convolutional neural network to determine at least one flow type corresponding to the multiple images; based on the time information corresponding to the multiple images and the energy signals, determine the energy signal corresponding to each of the flow types; based on the energy signal corresponding to each of the flow types, determine the energy characteristics corresponding to each of the flow types.

[0039] Furthermore, the energy sensor includes at least one of the following: a vibration sensor; a sound sensor.

[0040] A computer-readable storage medium having a computer program stored thereon, characterized in that when the computer program is executed by a processor, the steps in the above-mentioned gas-liquid two-phase fluid feature identification method are implemented.

[0041] Beneficial effects of the present invention:

[0042] (1) The present invention correlates and analyzes the flow pattern of the gas-liquid two-phase flow and the synchronously collected energy signal, thereby realizing the energy feature recognition of the entire process of the gas-liquid two-phase flow. Therefore, based on the energy features corresponding to various flow patterns, high-energy signals with a greater impact on pipeline safety can be identified, thereby assisting the design and improvement of pipelines in engineering development;

[0043] (2) The present invention can determine the vibration energy characteristics and / or noise energy characteristics that have a greater impact on pipeline safety, and can then effectively assist in the design and improvement of pipelines based on the vibration energy characteristics and / or noise energy characteristics. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] Figure 1 A schematic diagram of the implementation process of a gas-liquid two-phase fluid feature recognition method provided by the present invention;

[0045] Figure 2 A schematic diagram of the implementation process of using a convolutional neural network to determine at least one flow type corresponding to multiple images in the gas-liquid two-phase fluid feature recognition method provided by the present invention;

[0046] Figure 3 A schematic diagram of an implementation flow of an embodiment of a method for identifying characteristics of a gas-liquid two-phase fluid provided by the present invention;

[0047] Figure 4 A schematic diagram showing the change of flow pattern information of a gas-liquid two-phase fluid over time in one embodiment of the gas-liquid two-phase fluid feature recognition method provided by the present invention;

[0048] Figure 5A A schematic diagram of noise energy characteristics corresponding to each flow type in an embodiment of a gas-liquid two-phase fluid characteristic recognition method provided by the present invention;

[0049] Figure 5B A schematic diagram of vibration energy characteristics corresponding to each flow pattern in an embodiment of a gas-liquid two-phase fluid characteristic identification method provided by the present invention;

[0050] Figure 6 A schematic diagram of the composition structure of a gas-liquid two-phase fluid feature recognition device provided by the present invention;

[0051] Figure 7 A structural block diagram of a gas-liquid two-phase fluid feature recognition system provided by the present invention. DETAILED DESCRIPTION

[0052] The following will describe the embodiments of the present invention with reference to the accompanying drawings and preferred embodiments. Those skilled in the art can easily understand other advantages and effects of the present invention from the contents disclosed in this specification. The present invention can also be implemented or applied through other different specific embodiments, and the details in this specification can also be modified or changed in various ways based on different viewpoints and applications without departing from the spirit of the present invention. It should be understood that the preferred embodiments are only for illustrating the present invention, not for limiting the scope of protection of the present invention.

[0053] It should be noted that the illustrations provided in the following embodiments are only schematic illustrations of the basic concept of the present invention, and thus the drawings only show components related to the present invention rather than being drawn according to the number, shape and size of components in actual implementation. In actual implementation, the type, quantity and proportion of each component may be changed arbitrarily, and the component layout may also be more complicated.

[0054] In the following description, reference is made to “some embodiments”, which describe a subset of all possible embodiments, but it will be understood that “some embodiments” may be the same subset or different subsets of all possible embodiments and may be combined with each other without conflict.

[0055] The terms "first / second / third" involved are merely used to distinguish similar objects and do not represent a specific ordering of the objects. It is understandable that "first / second / third" can be interchanged with a specific order or sequence where permitted so that the embodiments of the invention described herein can be implemented in an order other than that illustrated or described herein.

[0056] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as those commonly understood by those skilled in the art to which the present invention belongs. The terms used herein are only for the purpose of describing the present invention and are not intended to limit the present invention.

[0057] In order to better understand the gas-liquid two-phase fluid feature identification method provided by the present invention, the relevant technology is first described below.

[0058] As mentioned above, since the interface of the gas-liquid two-phase fluid is affected by the relative velocity and interaction force, the flow characteristics are complex and diverse, which makes it difficult to identify the flow pattern of the gas-liquid two-phase fluid, and the noise and vibration characteristics under various flow patterns cannot be accurately identified.

[0059] In the related art, a variety of flow pattern identification methods for gas-liquid two-phase fluids are provided. Specifically, the flow pattern identification methods for gas-liquid two-phase fluids are divided into direct methods and indirect methods, among which:

[0060] The direct method refers to determining the flow pattern by directly observing the flow pattern of the gas-liquid two-phase fluid. Common direct methods include visual inspection and ray attenuation methods. For example, in the visual inspection method, a large number of flow pattern images are first taken with a high-speed camera, and then a certain number of flow pattern images with obvious features are selected from the flow pattern images, and then the noise and vibration characteristics corresponding to these flow patterns are analyzed in turn;

[0061] The indirect method refers to inferring the flow pattern by analyzing and processing the sensor detection signal. In the indirect method, the flow pattern characteristics are first identified by measuring the difference characteristics of the time domain, frequency and pressure difference signals in the pipeline by the sensor, and then the noise and vibration characteristics corresponding to these flow patterns are analyzed.

[0062] From the above, it can be seen that when using the direct method and the indirect method to analyze the flow pattern, noise and vibration characteristics of the gas-liquid two-phase fluid, the operator is required to judge the flow pattern information corresponding to the current image or sensor signal based on work experience. Therefore, there are problems of strong subjectivity and strong dependence on the work experience of the operator, and it is difficult to comprehensively and objectively analyze the data characteristics of noise and vibration under various flow patterns.

[0063] Based on this, the present invention provides a method for identifying characteristics of a gas-liquid two-phase fluid, which can be executed by a processor of a computer device. The computer device may refer to a server, a laptop, a tablet computer, a desktop computer, a smart TV, a set-top box, a mobile device (such as a mobile phone, a portable video player, a personal digital assistant, a dedicated messaging device) and other devices with data processing capabilities.

[0064] Figure 1 A schematic diagram of the implementation process of a gas-liquid two-phase fluid feature recognition method provided by the present invention is shown in FIG. Figure 1 As shown, the method includes the following steps S101 to S103:

[0065] Step S101, obtaining multiple images of a gas-liquid two-phase fluid flowing in a pipeline, and obtaining an energy signal from the fluid; wherein the multiple images and the energy signal are fluid information collected synchronously.

[0066] Here, the pipeline refers to a carrier pipeline used to carry gas-liquid two-phase fluid.

[0067] In some embodiments, the pipe may be a transparent quartz tube, so that the image acquisition device can acquire image information of the gas-liquid two-phase fluid in the quartz tube.

[0068] In some embodiments, the pipeline can be a quartz tube arranged horizontally. In this way, when determining the flow pattern characteristics of the gas-liquid two-phase fluid in the quartz tube, the influence of other factors such as gravity can be excluded.

[0069] In some embodiments, a light source is used to illuminate the gas-liquid two-phase fluid in the pipeline so that the gas-liquid two-phase fluid can be evenly illuminated.

[0070] In some embodiments, a horizontal light source is disposed on one side of the pipeline to illuminate the gas-liquid two-phase fluid in the pipeline.

[0071] In some embodiments, a high-speed camera is disposed on one side of the pipeline to collect multiple images of the gas-liquid two-phase fluid in the pipeline using the high-speed camera.

[0072] In some embodiments, the horizontal light source and the high-speed camera are respectively disposed on both sides of the pipeline, and the horizontal light source, the high-speed camera and the pipeline are on the same horizontal plane.

[0073] Energy signal refers to the signal generated when gas-liquid two-phase fluid flows in a pipeline, which can characterize the energy change information during the flow of gas-liquid two-phase fluid.

[0074] In some embodiments, the energy signal refers to a vibration signal, a noise signal, a heat signal, etc. generated when the gas-liquid two-phase fluid flows in the pipeline.

[0075] Here, the multiple images and energy signals are synchronously collected fluid information, that is, during the flow of gas-liquid two-phase fluid in the pipeline, multiple images and energy signals corresponding to the fluid are collected simultaneously, so that the collection time of the multiple images and energy signals is synchronized, and then the correspondence between the multiple images and the energy signals can be established according to the collection time.

[0076] Step S102: Determine at least one flow type corresponding to the multiple images using a convolutional neural network.

[0077] Here, Convolutional Neural Networks (CNN) refers to a deep learning model that can be used for classification and recognition tasks of image data.

[0078] In some embodiments, the convolutional neural network is formed by cross-stacked image input layer, convolution layer, excitation layer (ReLu), pooling layer and fully connected layer. Among them, the image input layer is used to receive the original image data and convert it into a format that the neural network can process; the convolution layer is used to extract information from the image, that is, to extract features from the image; the excitation layer is used to introduce nonlinear factors to enhance the expression ability of the neural network, and at the same time, reduce overfitting through its sparsity; the pooling layer is used to reduce the feature dimension of the image while retaining the main features in the image and extracting information useful for the classification task; the fully connected layer is used as a classifier to map the learned distributed feature representation to the sample label space.

[0079] In some embodiments, the convolutional neural network is a VGG network (Visual Geometry Group Network, VGG network) model. Here, the VGG network model is developed on the basis of CNN, and fully considers the strong scalability and generalization ability of deep learning, so it can be well applied to the field of image recognition.

[0080] In some embodiments, the convolutional neural network is a VGG11 network model including 8 convolutional layers and 3 fully connected layers, and uses a convolution kernel with a pixel size of 3*3 and a maximum pooling size of 2*2. In this way, by using a small-sized convolution kernel and a smaller maximum pooling size, the number of parameters can be reduced, the nonlinear characteristics of the network can be increased, and the convergence speed of the model training can be improved during the model training stage.

[0081] In some embodiments, a convolutional neural network is used to perform feature analysis on multiple images to determine at least one flow type corresponding to each image, such as bubbly flow, stratified flow, wavy flow, plug flow, mist flow, and liquid phase flow, etc.

[0082] In some embodiments, before processing the multiple images using a convolutional neural network, normalization and standardization are performed on the multiple images.

[0083] In some embodiments, performing normalization processing on multiple images includes: first, performing tensor quantization processing on each image to obtain corresponding tensor quantization data; then, dividing the original pixel value of each pixel by the sum of the original pixel values ​​of all pixels to obtain the normalized pixel value corresponding to each pixel; finally, based on the normalized pixel value corresponding to each pixel, obtaining the image data after normalization processing.

[0084] In some embodiments, performing normalization processing on multiple images includes: performing scaling processing on the image data after normalization processing to obtain image data in a 224*224 format.

[0085] Step S103: determining an energy signal corresponding to each of the flow types based on the multiple images and the time information corresponding to the energy signals.

[0086] Here, since multiple images and energy signals are fluid information collected synchronously and have time synchronization, the energy signal corresponding to each of the multiple images can be determined based on the time information; further, according to the flow type corresponding to each image, the energy signal corresponding to each flow type can be determined.

[0087] Step S104: determining the energy characteristics corresponding to each of the flow patterns based on the energy signal corresponding to each of the flow patterns.

[0088] Here, the energy signal corresponding to each flow type is subjected to characteristic analysis, thereby determining the energy characteristics corresponding to each flow type.

[0089] For example, when the energy signal is a vibration signal, the vibration intensity of the vibration signal can be analyzed to determine the vibration intensity characteristics corresponding to each flow type;

[0090] For another example, when the energy signal is a noise signal, the noise intensity of the noise signal may be analyzed to determine the noise intensity characteristics corresponding to each flow type;

[0091] For another example, when the energy signal is a heat signal, the temperature of the heat signal may be analyzed to determine the heat characteristics corresponding to each flow type.

[0092] In the above-mentioned gas-liquid two-phase fluid feature recognition method provided by the present invention, first, multiple images of the gas-liquid two-phase fluid flowing in the pipeline are obtained, and energy signals from the fluid are obtained, wherein the multiple images and the energy signals are synchronously collected fluid information; then, at least one flow type corresponding to the multiple images is determined using a convolutional neural network; thereafter, based on the time information corresponding to the multiple images and the energy signals, the energy signal corresponding to each flow type is determined; finally, based on the energy signal corresponding to each flow type, the energy feature corresponding to each flow type is determined. In this way, on the one hand, the flow type information corresponding to a large number of fluid images is identified using a convolutional neural network, which can improve the recognition efficiency, accuracy and objectivity of the fluid image; on the other hand, the flow type of the gas-liquid two-phase fluid and the synchronously collected energy signal are correlated and analyzed to realize the energy feature recognition of the entire process of the gas-liquid two-phase fluid, so that high-energy signals with a greater impact on pipeline safety can be identified based on the energy features corresponding to various flow types, thereby assisting the design and improvement of pipelines in engineering development.

[0093] In some embodiments, the energy signal corresponding to each flow type is determined based on the multiple images and the time information corresponding to the energy signal, that is, the above step S103 can be implemented as the following steps S1031 to S1032:

[0094] Step S1031, based on the time information corresponding to the multiple images, determine the flow type time information corresponding to each flow type.

[0095] Here, the multiple images collected during the flow of the gas-liquid two-phase fluid have corresponding image collection time information. In this way, after determining at least one flow pattern corresponding to the multiple images, the time information of the multiple images corresponding to each flow pattern can be used as the flow pattern time information of the corresponding flow pattern.

[0096] For example, the flow type corresponding to the 1st to 10th images is stratified flow, and the flow type corresponding to the 50th to 60th images is mist flow. Then the image acquisition time of the 1st to 10th images can be used as the flow type time information corresponding to the stratified flow, and the image acquisition time of the 50th to 60th images can be used as the flow type time information corresponding to the mist flow.

[0097] Step S1032: determining the energy signal corresponding to each flow type based on the flow type time information and the time information corresponding to the energy signal.

[0098] Here, since the multiple images and the energy signals are fluid information collected synchronously, the energy signal corresponding to each flow pattern can be determined based on the flow pattern time information and the time information corresponding to the energy signal.

[0099] Continuing with the above example, after respectively determining the flow pattern time information corresponding to the stratified flow and the mist flow, the energy signal collected within the flow pattern time can be used as the energy signal of the corresponding flow pattern.

[0100] In the above embodiment, based on the acquisition time information of multiple images and energy signals, the energy information corresponding to each flow type is determined, and the corresponding relationship between each flow type and the energy signal of the gas-liquid two-phase fluid can be accurately determined, thereby determining the energy characteristics of each flow type.

[0101] In some embodiments, the energy signal is time domain data.

[0102] Here, time domain data refers to the change of energy signal over time. For example, when the energy signal is a vibration signal, time domain data refers to the change of vibration frequency of the vibration signal over time; for another example, when the energy signal is a noise signal, time domain data refers to the change of noise frequency of the noise signal over time.

[0103] In this way, the energy signature corresponding to each flow type is determined based on the energy signal corresponding to each flow type, that is, the above step S104 can be implemented as the following step S1041:

[0104] Step S1041, for each of the flow types, based on the time domain data, determine the frequency domain data corresponding to the energy signal; based on the frequency domain data, determine the energy feature corresponding to the flow type.

[0105] Here, frequency domain data refers to the change of the amplitude of the energy signal with frequency, that is, it is used to describe the characteristics of the energy signal in terms of frequency.

[0106] Since frequency domain data focuses on the frequency components contained in the signal and their intensity, and displays the signal amount corresponding to each given frequency within a frequency range, frequency domain data can more intuitively display the energy characteristics corresponding to the flow type.

[0107] For example, based on the frequency domain data, it is determined that an amplitude peak appears within the first frequency range, which means that the energy of the flow pattern appearing within the first frequency range is the highest.

[0108] In some embodiments, the time domain data corresponding to the energy signal is converted into frequency domain data using a Fourier Transform (FT) algorithm or a Fast Fourier Transform (FFT) algorithm.

[0109] In the above embodiment, the energy signal is converted from time domain data to frequency domain data, so that the energy information corresponding to the energy signal can be accurately determined by using the amplitude value in the frequency domain data.

[0110] In some embodiments, the energy signal includes at least one of a vibration signal and a noise signal.

[0111] Here, by arranging a vibration sensor on the pipeline, the vibration signal generated during the flow of the gas-liquid two-phase fluid can be collected. In some embodiments, the vibration sensor is arranged on the horizontal surface of the pipeline.

[0112] By setting a sound sensor on the pipeline or at a specified distance from the pipeline, the noise signal generated during the flow of the gas-liquid two-phase fluid can be collected. In some embodiments, the sound sensor is a microphone. In some embodiments, the sound sensor is set at a vertical distance of 20 cm to 30 cm from the pipeline.

[0113] In this way, the step S1041 of determining the frequency domain data corresponding to the flow type based on the time domain data; and determining the energy characteristics corresponding to the flow type based on the frequency domain data can be implemented as at least one of the following steps S1042 to S1043:

[0114] Step S1042: determining the frequency domain data of the vibration signal based on the time domain data of the vibration signal; and determining the vibration energy characteristics corresponding to the flow pattern based on the frequency domain data of the vibration signal.

[0115] Here, in the gas-liquid two-phase fluid flow process, the vibration signal collected by the vibration sensor is the time domain data of the vibration signal.

[0116] In some embodiments, the time domain data of the vibration signal is converted into frequency domain data using a Fourier transform algorithm or a fast Fourier transform algorithm, that is, converted into a signal of a vibration amplitude changing with a vibration frequency.

[0117] In this way, based on the vibration amplitude information displayed in the frequency domain signal of the vibration signal, the vibration energy information can be determined, thereby determining the vibration energy characteristics corresponding to the flow type.

[0118] Step S1043: determining the frequency domain data of the noise signal based on the time domain data of the noise signal; and determining the noise energy characteristics corresponding to the flow type based on the frequency domain data of the noise signal.

[0119] Here, in the gas-liquid two-phase fluid flow process, the noise signal collected by the sound sensor is the time domain data of the noise signal.

[0120] In some embodiments, the time domain data of the noise signal is converted into frequency domain data using a fast Fourier transform algorithm, that is, converted into a signal of the variation of the noise amplitude with the frequency.

[0121] In this way, based on the amplitude information displayed in the frequency domain signal of the noise signal, the noise energy information can be determined, thereby determining the noise energy characteristics corresponding to the flow type.

[0122] In the above embodiment, by converting the time domain data corresponding to the energy signal into frequency domain data, the energy characteristics corresponding to each flow type can be displayed more intuitively, thereby assisting the industrial design and improvement of the pipeline; in addition, by determining the vibration energy characteristics and / or noise energy characteristics that have a greater impact on pipeline safety, the design and improvement of the pipeline can be effectively assisted.

[0123] In some embodiments, the convolutional neural network includes a feature extraction network and a classification network.

[0124] Here, the feature extraction network is used to perform multi-layer feature extraction on the image; the classification network is used to classify the features in the image according to preset types based on the features extracted by the feature extraction network.

[0125] In some embodiments, the feature extraction network is a plurality of convolutional layers in a convolutional neural network; and the classification network is a fully connected layer in a convolutional neural network.

[0126] In this way, the method of determining at least one flow type corresponding to the multiple images by using a convolutional neural network, that is, the above step S102, can be implemented as the following steps S1021 to S1022:

[0127] Step S1021: Use the feature extraction network to perform feature extraction on the multiple images to obtain image features corresponding to the multiple images.

[0128] Here, the image input layer in the convolutional neural network is used to vectorize each image to obtain a vectorized representation corresponding to each image; then, the feature extraction network (i.e., multiple convolutional layers) in the convolutional neural network is used to extract features from the vectorized representation corresponding to each image to obtain image features corresponding to each image.

[0129] Step S1022: using the classification network to determine the at least one flow type based on the image features.

[0130] Here, after extracting features from multiple images, a classification network (i.e., a fully connected layer) is used to determine the flow type information corresponding to each image based on the image features.

[0131] In some embodiments, the convolutional neural network also includes an activation function layer.

[0132] In some embodiments, the activation function used by the activation function layer is a softmax function.

[0133] In this way, the classification network, that is, the fully connected layer outputs a one-dimensional vector containing multiple logical values ​​(logits), each logical value in the one-dimensional vector represents the original prediction value of the model for the corresponding flow type; the fully connected layer inputs the one-dimensional vector into the softmax activation function layer; using the softmax activation function, each logical value in the one-dimensional vector is converted into a probability distribution with a sum of 1; based on the probability distribution, the flow type prediction result for the current image is determined and output.

[0134] The mathematical expression of the softmax function is shown in the following formula (1):

[0135]

[0136] Among them, V i represents the i-th logical value in the one-dimensional vector input to the softmax function, which is equivalent to the original predicted value of the i-th flow type in the present invention; C represents the number of logical values ​​in the one-dimensional vector, which is equivalent to the number of categories of the flow type to be determined in the present invention; S i Represents the probability corresponding to the i-th flow type.

[0137] In the above embodiment, using a convolutional neural network to identify at least one flow type corresponding to multiple images can improve the recognition speed and accuracy of flow type information, that is, using a convolutional neural network can quickly identify flow type information in a large amount of image data.

[0138] Next, combine Figure 2 , the process of using convolutional neural network for image processing is explained, where the convolutional neural network is a VGG network model, the convolution kernel size is 3*3, the maximum pooling size is 2*2, and the activation function used is the softmax function.

[0139] like Figure 2 As shown, an image 201 with a pixel size of 224*224 is input into the first convolutional layer 202 in the VGG network model;

[0140] After the convolution operation of the first convolution layer 202, 64 feature images 203 with a pixel size of 224*224 are generated;

[0141] Input 64 feature images with a pixel size of 224*224 into the excitation layer 204 to introduce nonlinear features, thereby obtaining 64 feature images 205 with a pixel size of 224*224 containing nonlinear features;

[0142] 64 feature images 205 with a pixel size of 224*224 are input into the maximum pooling layer 206, and after the dimension reduction processing of the maximum pooling layer 206, 128 feature images 207 with a pixel size of 112*112 are obtained;

[0143] By analogy, after three times of convolution processing, excitation processing and maximum pooling processing, 512 feature images 208 with a pixel size of 7*7 are obtained;

[0144] 512 feature images 208 with a pixel size of 7*7 are input into a fully connected layer 209 for classification processing to obtain a one-dimensional feature vector 210 containing 6 logical values;

[0145] The one-dimensional feature vector 210 is input into the Softmax layer 211 to convert each logical value in the one-dimensional vector into a probability distribution with a sum of 1, which corresponds to the probability that the flow type in the current image belongs to bubbly flow, stratified flow, wavy flow, plug flow, mist flow and liquid phase flow.

[0146] In some embodiments, before using the convolutional neural network to determine at least one flow type corresponding to the multiple images, that is, before the above step S102, the gas-liquid two-phase fluid feature recognition method provided by the present invention further includes the following steps S105 to S108:

[0147] Step S105, obtaining a plurality of sample images; wherein each of the sample images has corresponding sample flow type information.

[0148] Here, the sample image refers to an image collected for a gas-liquid two-phase fluid and labeled with sample flow pattern information.

[0149] In some embodiments, a plurality of sample images are divided into a training set and a test set, wherein the training set is used to train the initial neural network, and the test set is used to test the trained convolutional neural network. For example, in one embodiment, 64 images representing bubbly flow, stratified flow, wavy flow, plug flow, mist flow, and liquid phase flow are obtained and evenly divided into two groups, respectively, as a training set and a test set.

[0150] In this way, by dividing the sample images into training sets and test sets, the subsequent test conclusions of the trained convolutional neural network can be made more accurate.

[0151] In some embodiments, multiple sample images can also be divided into a training set, a validation set and a test set; wherein, the roles of the training set and the test set are as described above; the validation set is used to verify the generalization ability of the model, and by evaluating the performance of the model under different hyperparameters, the best hyperparameter combination is selected to avoid overfitting of the model on the training set and improve the generalization ability of the model.

[0152] Step S106: using the initial neural network, determine the predicted flow type information corresponding to the multiple sample images.

[0153] Here, a plurality of sample images are input into the initial neural network, so as to use the initial neural network to predict the predicted flow type information corresponding to each image.

[0154] In some embodiments, the initial neural network includes a feature extraction network and a classification network. Thus, first, the feature extraction network in the initial neural network is used to extract image features corresponding to each image; then, the classification network in the convolutional neural network is used to classify the extracted image features to determine the predicted flow type information corresponding to each image.

[0155] Step S107: determining loss data based on the sample flow type information and predicted flow type information corresponding to the plurality of sample images.

[0156] Here, the loss data is used to measure the error between each predicted flow type information and the corresponding sample flow type information.

[0157] Here, using a specified loss function, loss data of sample flow type information and predicted flow type information of multiple sample images are calculated.

[0158] In some embodiments, the loss function is a cross-entropy loss function (Cross-Entropy Loss) or a multi-label cross-entropy loss function (Multi-Label Cross-Entropy Loss). The mathematical expression of the cross-entropy loss function is shown in the following formula (2):

[0159]

[0160] in, represents loss data; y represents sample flow type information; Represents the predicted flow type information; q represents the probability distribution predicted by the model.

[0161] Step S108: Based on the loss data, update the initial neural network to obtain the convolutional neural network.

[0162] Here, the initial neural network is updated based on the loss data corresponding to each sample image, so that the predicted flow type information generated by the updated initial neural network is closer to the corresponding sample flow type information.

[0163] In some embodiments, by calculating the loss data between the predicted flow type information and the corresponding sample flow type information and updating the initial neural network multiple times until the calculated loss data is reduced to a specified loss threshold, the neural network after the last update is determined as the corresponding convolutional neural network.

[0164] In some embodiments, an adaptive optimizer (Adaptive Moment Estimation, Adam optimizer) is used to update model parameters so that the loss function converges quickly, and the learning rate is adaptively adjusted to increase the model training speed.

[0165] In some embodiments, for the multiple sets of hyperparameter combinations obtained after model training, a validation set is used to verify the performance of the model under different hyperparameter combinations to select the best hyperparameter combination, thereby avoiding overfitting of the model on the training set and improving the generalization ability of the model.

[0166] In some embodiments, the trained convolutional neural network is tested using a test set to determine network performance.

[0167] Next, combine Figure 3 , an embodiment of the gas-liquid two-phase fluid feature recognition method provided by the present invention is described. Figure 3 As shown, this embodiment includes the following steps S301 to S307:

[0168] Step S301, acquiring multiple images of gas-liquid two-phase fluid and noise signals and vibration signals; then, executing steps S302 and S306;

[0169] Here, a gas-liquid two-phase fluid is passed into a quartz tube, and a light source is set on one side of the quartz tube and a high-speed camera is set on the other side to collect multiple images of the fluid; at the same time, a sound sensor is used to collect noise information from the fluid, and a vibration sensor is used to collect vibration signals from the fluid.

[0170] Step S302, performing preprocessing on the multiple images; then, executing step S303;

[0171] Here, normalization processing and standardization processing are performed on a plurality of images.

[0172] Step S303, using the VGG network model to identify at least one type of flow type information corresponding to the multiple images; then, executing step S304;

[0173] Step S304, determining the flow pattern time information corresponding to each flow pattern; then, executing step S305;

[0174] Here, the flow type time information corresponding to each flow type is determined based on the image acquisition time of at least one image corresponding to each flow type.

[0175] Step S305, determining the noise signal and vibration signal corresponding to each flow pattern; then, executing step S307;

[0176] Here, based on the flow pattern time information corresponding to each flow pattern and the time information corresponding to the noise signal and the vibration signal, the noise signal and the vibration signal corresponding to each flow pattern are determined.

[0177] Step S306, pre-processing the noise signal and the vibration signal; then, executing step S307;

[0178] Here, preprocessing is performed on the noise signal and the vibration signal, which may include denoising, filtering and / or smoothing to improve the accuracy of signal analysis.

[0179] Step S307: Perform FFT processing on the noise signal and vibration signal corresponding to each flow type to identify the energy characteristics corresponding to each flow type.

[0180] Here, FFT processing is performed on the time domain data corresponding to the noise signal and the time domain data corresponding to the vibration signal to obtain the corresponding frequency domain data, and then the energy characteristics corresponding to each flow type are determined based on the frequency domain data.

[0181] Figure 4 FIG. 2 is a schematic diagram showing the flow pattern information of the gas-liquid two-phase fluid changing with time in the above embodiment. Figure 4 As shown, during 0-3 seconds, the flow pattern of the gas-liquid two-phase fluid is liquid phase flow; during 3-5 seconds, bubbly flow, liquid phase flow, plug flow and stratified flow exist simultaneously; during 5-7.5 seconds, bubbly flow, plug flow and stratified flow exist simultaneously; during 7.5-10.0 seconds, bubbly flow and stratified flow exist simultaneously; during 10.0-12.5 seconds, mist flow and a small amount of wavy flow exist; during 12.5-15.0 seconds, wavy flow and a small amount of mist flow exist; during 15.0 seconds to 20.0 seconds, only wavy flow exists.

[0182] Figure 5A and Figure 5B A schematic diagram of energy characteristics corresponding to various flow patterns of the gas-liquid two-phase fluid in the above embodiment is shown.

[0183] like Figure 5A As shown, FFT processing is performed on the time domain data of the noise signal corresponding to each flow type to obtain an energy characteristic diagram based on the noise signal corresponding to each flow type.

[0184] Figure 5A The grayscale images with different depths are used to show the energy characteristics of different flow types based on noise signals, that is, the intensity of wavy flow, stratified flow, plug flow, mist flow, liquid phase flow and bubble flow changes with frequency. Figure 5AAs shown in the figure, when the frequency is between 1000-8000Hz, the energy of bubble flow, liquid phase flow, plug flow and stratified flow is low, while the mist flow and wave flow have energy peaks at 4000Hz and 5500Hz. Therefore, it can be seen from the energy characteristic diagram that when the noise frequency is between 4000Hz and 5500Hz, the mist flow and wave flow are more likely to cause damage to the pipeline.

[0185] Figure 5B The energy characteristics of different flow types based on vibration signals are shown in grayscale images of different depths, that is, the energy of wavy flow, stratified flow, plug flow, mist flow, liquid phase flow and bubble flow changes with frequency. Figure 5B As shown in the figure, when the frequency is between 1000-2500Hz, the energy of each flow type is low, while the energy peaks of wavy flow, stratified flow, mist flow and bubble flow appear between 2500Hz and 6000Hz, and the energy peak values ​​of wavy flow and stratified flow are higher. Therefore, it can be seen from the energy characteristic diagram that when the vibration frequency is between 2500Hz and 6000Hz, wavy flow and stratified flow are more likely to cause damage to the pipeline.

[0186] In the above embodiment, the energy characteristics of each flow pattern are determined by identifying various flow patterns and corresponding vibration and noise signals during the flow of gas-liquid two-phase fluid in the pipeline. In this way, key frequency characteristics can be identified from the energy characteristics of each flow pattern, such as the vibration frequency and / or noise frequency, harmonic frequency, resonance frequency, etc. corresponding to the energy peak, and then the pipeline design and improvement can be assisted based on the identified key frequencies. At the same time, in the subsequent pipeline design process, the vibration signal and / or noise signal during the flow of gas-liquid two-phase fluid can be separately collected, and then based on the collected vibration signal and / or noise signal, and the energy characteristic diagram of each flow pattern determined in advance, the improvement plan of the pipeline design can be determined.

[0187] Based on the aforementioned method embodiment, the present invention also provides a gas-liquid two-phase fluid feature identification device, and each unit included in the device, and each module included in each unit, can be implemented by a processor in a computer device; of course, it can also be implemented by a specific logic circuit; in the implementation process, the processor can be a central processing unit (CPU), a microprocessor (MPU), a digital signal processor (DSP) or a field programmable gate array (FPGA), etc.

[0188] Figure 6A schematic diagram of the composition structure of a gas-liquid two-phase fluid feature recognition device provided by the present invention is shown in FIG. Figure 6 As shown, the gas-liquid two-phase fluid feature recognition device 600 includes: an information acquisition module 610, a flow pattern determination module 620, an energy signal determination module 630 and an energy feature determination module 640, wherein:

[0189] The information acquisition module 610 is used to acquire multiple images of the gas-liquid two-phase fluid flowing in the pipeline, and acquire an energy signal from the fluid; wherein the multiple images and the energy signal are synchronously acquired fluid information;

[0190] A flow type determination module 620, configured to determine at least one flow type corresponding to the plurality of images using a convolutional neural network;

[0191] An energy signal determination module 630, configured to determine an energy signal corresponding to each of the flow types based on the multiple images and time information corresponding to the energy signals;

[0192] The energy feature determination module 640 is used to determine the energy feature corresponding to each of the flow types based on the energy signal corresponding to each of the flow types.

[0193] In some embodiments, the energy signal determination module 630 is configured to:

[0194] Determine flow pattern time information corresponding to each flow pattern based on time information corresponding to the multiple images;

[0195] Based on the flow pattern time information and the time information corresponding to the energy signal, an energy signal corresponding to each flow pattern is determined.

[0196] In some embodiments, the energy signal is time domain data;

[0197] Thus, the energy characteristic determination module 640 is used to:

[0198] For each of the flow patterns, based on the time domain data, the frequency domain data corresponding to the energy signal is determined; based on the frequency domain data, the energy feature corresponding to the flow pattern is determined.

[0199] In some embodiments, the energy signal includes at least one of a vibration signal and a noise signal;

[0200] Thus, the energy feature determination module 640 is configured to perform at least one of the following:

[0201] Based on the time domain data of the vibration signal, determining the frequency domain data of the vibration signal; based on the frequency domain data of the vibration signal, determining the vibration energy characteristics corresponding to the flow pattern;

[0202] Based on the time domain data of the noise signal, the frequency domain data of the noise signal is determined; based on the frequency domain data of the noise signal, the noise energy feature corresponding to the flow type is determined.

[0203] In some embodiments, the convolutional neural network includes a feature extraction network and a classification network;

[0204] Thus, the flow pattern determination module 620 is used to:

[0205] Using the feature extraction network to perform feature extraction on the multiple images to obtain image features corresponding to the multiple images;

[0206] The at least one flow type is determined based on the image features using the classification network.

[0207] In some embodiments, the apparatus 600 further includes a model training module 650; the model training module 650 is configured to:

[0208] Acquire a plurality of sample images; wherein each of the sample images has corresponding sample flow type information;

[0209] Determining predicted flow type information corresponding to the plurality of sample images using an initial neural network;

[0210] Determining loss data based on the sample flow pattern information and the predicted flow pattern information corresponding to the plurality of sample images;

[0211] Based on the loss data, the initial neural network is updated to obtain the convolutional neural network.

[0212] The description of the above device embodiment is similar to the description of the above method embodiment, and has similar beneficial effects as the method embodiment. In some embodiments, the functions or modules included in the device provided by the embodiments of the present disclosure can be used to execute the method described in the above method embodiment. For technical details not disclosed in the device embodiment of the present invention, please refer to the description of the method embodiment of the present invention for understanding.

[0213] On the other hand, based on the above-mentioned embodiments, the present invention provides a gas-liquid two-phase fluid feature recognition system. Figure 7 As shown, the gas-liquid two-phase fluid feature recognition system 700 includes: a pipeline 710, a light source 720, a camera 730, an energy sensor and a control device 760; wherein,

[0214] The light source 720 is disposed on one side of the pipe 710;

[0215] The camera 730 is disposed at the other side of the pipeline 710 to collect multiple images of the gas-liquid two-phase fluid flowing in the pipeline 710 and transmit the multiple images to the control device 760;

[0216] The energy sensor is disposed outside the pipe 710 to synchronously collect energy signals from the fluid during the process of the camera 730 collecting the multiple images, and transmit the energy signals to the control device 760;

[0217] The control device 760 is used to determine at least one flow type corresponding to the multiple images using a convolutional neural network; determine the energy signal corresponding to each of the flow types based on the multiple images and the time information corresponding to the energy signals; and determine the energy characteristics corresponding to each of the flow types based on the energy signals corresponding to each of the flow types.

[0218] In some embodiments, the pipe 710 is a horizontally arranged quartz pipe, and a gas-liquid two-phase fluid can flow in the pipe 710 .

[0219] In some embodiments, the light source 720 is disposed horizontally to illuminate the gas-liquid two-phase fluid in the pipe 710 .

[0220] In some embodiments, camera 730 is a high-speed camera.

[0221] In some embodiments, the energy sensor includes at least one of the following: a vibration sensor 740 ; a sound sensor 750 .

[0222] In some embodiments, the vibration sensor 740 is disposed on the surface of the pipe 710 and is communicatively connected to the control device 760 to transmit the collected vibration signal to the control device 760 .

[0223] In some embodiments, the sound sensor 750 is set at a preset distance from the pipeline 710 to collect noise signals generated by the flow of the liquid two-phase fluid in the pipeline 710; at the same time, the sound sensor 750 is communicated with the control device 760 to transmit the collected images to the control device 760.

[0224] In some embodiments, sound sensor 750 is a microphone.

[0225] In some embodiments, the data processing device 760 determines the specific implementation methods of various flow types and energy characteristic diagrams corresponding to each flow type that appear during the flow of the gas-liquid two-phase fluid in the pipeline 710. Please refer to the detailed description of the above method embodiments and will not be repeated here.

[0226] The description of the above system embodiment is similar to the description of the above method embodiment, and has similar beneficial effects as the method embodiment. In some embodiments, the functions or devices included in the system provided by the embodiment of the present disclosure can be used to execute the method described in the above method embodiment. For technical details not disclosed in the system embodiment of the present invention, please refer to the description of the method embodiment of the present invention for understanding.

[0227] It should be noted that, in the embodiment of the present invention, if the above-mentioned gas-liquid two-phase fluid feature identification method is implemented in the form of a software function module and sold or used as an independent product, it can also be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the embodiment of the present invention can be essentially or partly reflected in the form of a software product that contributes to the relevant technology. The software product is stored in a storage medium, including several instructions to enable a computer device (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the methods described in each embodiment of the present invention. The aforementioned storage medium includes: various media that can store program codes, such as a U disk, a mobile hard disk, a read-only memory (ROM), a magnetic disk or an optical disk. In this way, the embodiment of the present invention is not limited to any specific hardware, software or firmware, or any combination of hardware, software, and firmware.

[0228] An embodiment of the present invention provides a computer device, including a memory and a processor, wherein the memory stores a computer program that can be run on the processor, and the processor implements part or all of the steps in the above method when executing the program.

[0229] The embodiment of the present invention provides a computer-readable storage medium on which a computer program is stored, and when the computer program is executed by a processor, some or all of the steps in the above method are implemented. The computer-readable storage medium may be transient or non-transient.

[0230] An embodiment of the present invention provides a computer program, including computer-readable codes. When the computer-readable codes are run in a computer device, a processor in the computer device executes some or all of the steps for implementing the above method.

[0231] An embodiment of the present invention provides a computer program product, the computer program product includes a non-transitory computer-readable storage medium storing a computer program, and when the computer program is read and executed by a computer, some or all of the steps in the above method are implemented. The computer program product can be implemented specifically by hardware, software or a combination thereof. In some embodiments, the computer program product is specifically embodied as a computer storage medium, and in other embodiments, the computer program product is specifically embodied as a software product, such as a software development kit (SDK), etc.

[0232] It should be noted here that the description of the various embodiments above tends to emphasize the differences between the various embodiments, and the same or similar aspects can be referenced to each other. The description of the above device, storage medium, computer program and computer program product embodiments is similar to the description of the above method embodiment, and has similar beneficial effects as the method embodiment. For technical details not disclosed in the embodiments of the device, storage medium, computer program and computer program product of the present invention, please refer to the description of the method embodiment of the present invention for understanding.

[0233] It should be understood that "one embodiment" or "an embodiment" mentioned throughout the specification means that specific features, structures or characteristics related to the embodiment are included in at least one embodiment of the present invention. Therefore, "in one embodiment" or "in an embodiment" appearing throughout the specification does not necessarily refer to the same embodiment. In addition, these specific features, structures or characteristics can be combined in one or more embodiments in any suitable manner. It should be understood that in various embodiments of the present invention, the size of the serial numbers of the above-mentioned steps / processes does not mean the order of execution. The execution order of each step / process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiment of the present invention. The serial numbers of the above-mentioned embodiments of the present invention are only for description and do not represent the advantages and disadvantages of the embodiments.

[0234] It should be noted that, in this article, the terms "include", "comprises" or any other variations 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, an element defined by the sentence "comprises a ..." does not exclude the existence of other identical elements in the process, method, article or device including the element.

[0235] In the several embodiments provided by the present invention, it should be understood that the disclosed devices and methods can be implemented in other ways. The device embodiments described above are only schematic. For example, the division of the units is only a logical function division. There may be other division methods in actual implementation, such as: multiple units or components can be combined, or can be integrated into another system, or some features can be ignored or not executed. In addition, the coupling, direct coupling, or communication connection between the components shown or discussed can be through some interfaces, and the indirect coupling or communication connection of the devices or units can be electrical, mechanical or other forms.

[0236] The units described above as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units; they may be located in one place or distributed on multiple network units; some or all of the units may be selected according to actual needs to achieve the purpose of the present embodiment.

[0237] In addition, all functional units in the embodiments of the present invention may be integrated into one processing unit, or each unit may be separately used as a unit, or two or more units may be integrated into one unit; the above-mentioned integrated units may be implemented in the form of hardware or in the form of hardware plus software functional units.

[0238] A person skilled in the art can understand that all or part of the steps of implementing the above method embodiment can be completed by hardware related to program instructions, and the aforementioned program can be stored in a computer-readable storage medium. When the program is executed, it executes the steps of the above method embodiment; and the aforementioned storage medium includes: mobile storage devices, read-only memories (ROM), magnetic disks or optical disks, etc., various media that can store program codes.

[0239] Alternatively, if the above-mentioned integrated unit of the present invention is implemented in the form of a software function module and sold or used as an independent product, it can also be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention can be essentially or partly embodied in the form of a software product that contributes to the relevant technology. The computer software product is stored in a storage medium and includes several instructions for a computer device (which can be a personal computer, server, or network device, etc.) to execute all or part of the methods described in each embodiment of the present invention. The aforementioned storage medium includes: various media that can store program codes, such as mobile storage devices, ROMs, magnetic disks, or optical disks.

[0240] The above embodiments are only preferred embodiments for fully illustrating the present invention, and the protection scope of the present invention is not limited thereto. Any equivalent substitution or change made by a person skilled in the art based on the present invention is within the protection scope of the present invention.

Claims

1. A method for identifying characteristics of a gas-liquid two-phase fluid, comprising: Acquire multiple images of a gas-liquid two-phase fluid flowing in a pipeline, and acquire an energy signal from the fluid; wherein the multiple images and the energy signal are fluid information acquired synchronously; Determine at least one flow type corresponding to the plurality of images using a convolutional neural network; Determine an energy signal corresponding to each of the flow patterns based on the multiple images and time information corresponding to the energy signals; Based on the energy signal corresponding to each of the flow types, the energy feature corresponding to each of the flow types is determined.

2. The method according to claim 1, wherein: The step of determining the energy signal corresponding to each of the flow types based on the multiple images and the time information corresponding to the energy signals includes: Determine flow pattern time information corresponding to each flow pattern based on time information corresponding to the multiple images; Based on the flow pattern time information and the time information corresponding to the energy signal, an energy signal corresponding to each flow pattern is determined.

3. The method according to claim 1 or 2, wherein: The energy signal is time domain data; The step of determining the energy feature corresponding to each of the flow patterns based on the energy signal corresponding to each of the flow patterns comprises: For each of the flow patterns, based on the time domain data, determining frequency domain data corresponding to the energy signal; Based on the frequency domain data, an energy feature corresponding to the flow pattern is determined.

4. The method according to claim 3, wherein: The energy signal includes at least one of a vibration signal and a noise signal; The determining, based on the time domain data, frequency domain data corresponding to the flow type; and determining, based on the frequency domain data, energy features corresponding to the flow type, include at least one of the following: Based on the time domain data of the vibration signal, determining the frequency domain data of the vibration signal; based on the frequency domain data of the vibration signal, determining the vibration energy characteristics corresponding to the flow pattern; Determine frequency domain data of the noise signal based on the time domain data of the noise signal; Based on the frequency domain data of the noise signal, a noise energy feature corresponding to the flow type is determined.

5. The method according to any one of claims 1 to 4, wherein: The convolutional neural network includes a feature extraction network and a classification network; The method of using a convolutional neural network to determine at least one flow type corresponding to the multiple images includes: Using the feature extraction network to perform feature extraction on the multiple images to obtain image features corresponding to the multiple images; The at least one flow type is determined based on the image features using the classification network.

6. The method according to any one of claims 1 to 5, wherein: Before determining at least one flow type corresponding to the multiple images using a convolutional neural network, the method further includes: Acquire multiple sample images; wherein each of the sample images has corresponding sample flow type information; Determining predicted flow type information corresponding to the plurality of sample images using an initial neural network; Determining loss data based on the sample flow pattern information and the predicted flow pattern information corresponding to the plurality of sample images; Based on the loss data, the initial neural network is updated to obtain the convolutional neural network.

7. A gas-liquid two-phase fluid feature recognition device, comprising: An information acquisition module, used to acquire multiple images of the gas-liquid two-phase fluid flowing in the pipeline, and to acquire an energy signal from the fluid; wherein the multiple images and the energy signal are synchronously acquired fluid information; A flow type determination module, configured to determine at least one flow type corresponding to the plurality of images using a convolutional neural network; An energy signal determination module, used to determine the energy signal corresponding to each of the flow types based on the multiple images and the time information corresponding to the energy signals; The energy characteristic determination module is used to determine the energy characteristic corresponding to each of the flow types based on the energy signal corresponding to each of the flow types.

8. A gas-liquid two-phase fluid feature recognition system, comprising: pipelines, cameras, light sources, energy sensors and control equipment; among which, The light source is arranged on one side of the pipe; The camera is arranged at the other side of the pipeline to collect multiple images of the gas-liquid two-phase fluid flowing in the pipeline and transmit the multiple images to the control device; The energy sensor is arranged on the outside of the pipeline to synchronously collect energy signals from the fluid during the process of the camera collecting the multiple images, and transmit the energy signals to the control device; The control device is used to use a convolutional neural network to determine at least one flow type corresponding to the multiple images; based on the time information corresponding to the multiple images and the energy signals, determine the energy signal corresponding to each of the flow types; based on the energy signal corresponding to each of the flow types, determine the energy characteristics corresponding to each of the flow types.

9. The system according to claim 8, wherein: The energy sensor includes at least one of the following: a vibration sensor; a sound sensor.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps in the method according to any one of claims 1 to 6 are implemented.

Images