A pump jet flow field testing device and flow field loss compensation method

Abstract

This invention discloses a pump-jet flow field testing device and a flow field missing compensation method, including a high-frequency dual-pulse laser, an image processing system, a camera synchronizer, a high-speed camera, a cavitation water tank, a reflective prism, a pump-jet model, and a transparent conduit. The high-frequency dual-pulse laser is placed at the rear end of the pump-jet model within the cavitation water tank, and the laser emitted by the high-frequency dual-pulse laser illuminates the pump-jet model through the cavitation water tank. A reflective prism and a high-speed camera are installed on the laser plane of the pump-jet model. The image processing system uses data assimilation and a generative adversarial network-based model to effectively solve the problems of flow field information loss caused by complex structural occlusion, particle concentration loss, and wall reflection. This invention addresses the shortcomings of existing experimental devices for testing the internal flow field of pump-jet propellers, effectively compensates for missing flow field information in pump-jet propellers, and is of great significance for understanding the evolution of the internal flow field morphology of pump-jet propellers from an experimental mechanism perspective, as well as for subsequent noise prediction and flow field calculation verification of pump-jet propellers.

Description

Technical Field

[0001] This invention belongs to the field of macroscopic PIV experimental measurement technology, specifically relating to a pump-jet flow field testing device and a flow field missing compensation method. Background Technology

[0002] Pump-jet propulsion (PJP), as a novel propulsion mechanism for marine equipment, has attracted widespread attention due to its high efficiency, low noise, and high critical speed. However, the complex coupling disturbances between the stator and rotor of PJP generate highly complex near-wall turbulence, resulting in a distinctly unsteady, nonlinear, and multi-scale three-dimensional flow field. To explore the intrinsic evolution mechanism of macroscopic phenomena such as pump-jet forces, noise, and cavitation, a thorough understanding of the fine flow field inside the pump-jet is necessary. Non-contact measurement methods such as laser Doppler velocimetry (LDV), particle image velocimetry (PIV), and laser-induced fluorescence velocimetry (LIF) provide high-precision flow field data for the study of the mechanisms of complex structures. Among these, PIV is widely used for flow field measurement due to its unique flow field visualization capabilities. PIV obtains flow field data by continuously photographing uniformly distributed tracer particles illuminated by a dual-pulse laser in a very short time and then using a cross-correlation algorithm. However, in actual measurement, issues such as structural obstruction, uneven distribution of tracer particles, and wall reflections lead to missing flow field data. For example, during the experimental measurement of the internal flow field of a pump-jet propeller, limitations imposed by the cavitation water tank and interference from structures such as the blades and hub on the laser beam result in missing flow field data. Therefore, missing flow field data is a common bottleneck in experimental fluid mechanics. Furthermore, due to the inherent complexity of pump-jet propellers, publicly available literature, both domestically and internationally, is very limited. Current model test studies focus primarily on the overall macroscopic performance of the pump-jet propeller (such as thrust, torque, power, and efficiency). Moreover, no institution has yet measured the internal flow field characteristics of a pump-jet propeller. This lack of internal flow field testing methods and capabilities makes it difficult to obtain detailed flow information and flow evolution characteristics within the pump-jet, and even more difficult to explore the underlying mechanisms of changes in macroscopic phenomena such as stress, noise, and cavitation.

[0003] In summary, current experiments on flow field measurement for pump-jet propulsion systems are incomplete, and measurement schemes are immature, failing to effectively compensate for missing flow fields in pump-jet systems. Therefore, the development of a complete testing system and a method for compensating for missing flow fields is urgently needed. Summary of the Invention

[0004] The purpose of this invention is to provide a pump-jet flow field testing device and a flow field deficiency compensation method.

[0005] The objective of this invention is achieved through the following technical solution:

[0006] A method for compensating for missing flow field during pump jet flow field testing, the specific steps of which are as follows:

[0007] Step 1: The obtained experimental and numerical simulation data are passed through 10 convolutional layers to collect spatial feature information of the internal flow field of the pump jet;

[0008] Step 2: Use intrinsic orthogonal decomposition to reduce the flow field cost and obtain information on different modes of the flow field;

[0009] Step 3: Establish the mapping relationship between the observation space and the state space through the Voronoi operator, fuse the modal parameters with spatial feature information, and realize the assimilation of pump-jet flow field test data and numerical simulation data;

[0010] Step 4: Input the variationally assimilated flow field into the neural network to generate a network including deformable convolution, ATN attention transfer layer, convolutional layer and deconvolutional layer;

[0011] Step 5: Discriminator network, i.e., D network, distinguishes between the flow field information generated by the generator network and the fine flow field obtained by CFD calculation, so that the missing region generated by the generator G is close to the real flow field.

[0012] Step 6: Optimize the missing compensation weights of the G network; feed back the loss function value calculated by the D network to the G network to promote backpropagation of the network, and repeat steps 4-5 to optimize the missing compensation effect of the flow field.

[0013] Step 7: Obtain the final pump-jet experimental and numerically missing accurate compensated flow field.

[0014] Furthermore, step 2 specifically includes:

[0015] The flow field information obtained through convolution has a size of f. x ×f y The flow field information is flattened according to the spatial dimension to ensure that the flattened flow field data does not lose the original flow field data structure. The formula for the flattened flow field feature snapshot matrix is ​​as follows:

[0016]

[0017] Where p∈[1,f x ],q∈[1,f y ],x(r p,q ,t k () represents the physical quantity of the flow field. In the internal flow field of a pump, the velocity along the three-dimensional spatial direction is mainly selected, i.e., v. x v y v z ;

[0018] Furthermore, to reduce the amount of data processing required for flow field dimensionality reduction, the steady-state average component in the flow field space is subtracted, and only the portion of the pump-jet flow that deviates from the mean is modeled, as shown in the following formula:

[0019]

[0020] A set of orthogonal bases is selected to achieve the best approximation of the original pump jet flow field data. The orthogonal bases are represented as follows:

[0021]

[0022] Where n represents the order of the orthogonal basis, i.e. the order of the modes; the first 60 modes are selected as the main flow field information in the pump-jet flow field compensation process.

[0023] Furthermore, step 3 specifically includes:

[0024] Construct a Voronoi inverse operator to map the observation space to the state space and define an observation field. As a set of observation points for pump-jet flow field test data at a given location, the location of this point is defined as {(i t,k ,j t,k The expression for )} is as follows:

[0025] (i t,k ,j t,k )∈[1,...,N y ]×[1,...,N x ]

[0026] Where k represents the number of observation points during the pump-jet test, and the Voronoi operator unit R t,k Satisfy the following formula:

[0027]

[0028] Where d(·) represents the Euclidean distance, and its calculation formula is as follows:

[0029]

[0030] Therefore, the observation space is divided into multiple Voronoi operator units, satisfying the following formula:

[0031]

[0032] To obtain:

[0033]

[0034] Substitute the obtained observation space into:

[0035]

[0036] Ultimately, the assimilation and fusion of pump spray test data and simulation data will be achieved.

[0037] Furthermore, the generative network in step 4 incorporates six convolutional layers in its encoding section. These layers downsample the flow field features after data fusion to obtain low-dimensional feature information of the pump-jet internal flow field. During the first two downsampling layers, deformable convolutions are embedded, adaptively changing the sampling position of the convolution kernels. The features obtained from adjacent convolutional layers are then input into the ATN module. Simultaneously, the ATN module calculates the inner and outer cosine similarity of the missing flow field regions in the input feature map. The data expression for the ATN's working principle is as follows:

[0038]

[0039]

[0040] in, p represents the cosine similarity of the feature maps at the l-th layer. i This indicates that flow field feature values ​​are extracted from the outside of the i-th Mask in the feature map, p j Flow field feature values ​​extracted from inside the j-th Mask; ||p i ||2 and||p j ||2 represent the flow field characteristic values ​​p, respectively. i and p j The L2 norm; the cosine similarity index Convert to attention weights Subsequently, the flow field with missing adjacent low-level features This can be achieved by weighting the attention weights to obtain the weights of the missing regions, guiding the fusion between high-level and low-level features, and compensating for the missing flow field while making up for the information loss caused by convolutional pooling.

[0041] Furthermore, step 5 introduces an adversarial loss function during network training, as follows:

[0042]

[0043] Where E(*) represents the expected value of the distribution function, p d (x) represents the internal flow field distribution of the pump nozzle in the CFD simulation, p d (z) represents the flow field distribution generated by generator G, and D(*) represents the output of the discriminator, whose output value varies from 0 to 1.

[0044] A pump-jet flow field testing device, characterized in that it includes a high-frequency dual-pulse laser, an image processing system, a camera synchronizer, a high-speed camera, a cavitation water cylinder, a reflective prism, a pump-jet model, and a transparent conduit;

[0045] The high-frequency dual-pulse laser is placed at the rear end of the pump-jet model in the cavitation water cylinder. The laser emitted by the high-frequency dual-pulse laser irradiates the pump-jet model through the cavitation water cylinder. The pump-jet model is fixed inside the cavitation water cylinder through a transparent conduit. A stator and rotor coupled to each other are installed inside the pump-jet model. A reflective prism is installed on the laser plane of the pump-jet model, and a high-speed camera is used to capture image information and collect particle images. The high-speed camera is connected to the image processing system through a camera synchronizer, which is used to control the high-speed camera to acquire images synchronously.

[0046] Furthermore, one side of the cavitation water tank is made of transparent glass, and the two high-speed cameras are placed on the transparent glass side of the cavitation water tank.

[0047] A computer-readable storage medium having a computer program / instructions stored thereon, which, when executed by a processor, implements the steps of a method for compensating for missing flow field in a pump-jet flow field test.

[0048] A computer program product includes a computer program / instructions that, when executed by a processor, implement the steps of a method for compensating for missing flow field in a pump-jet flow field test.

[0049] An electronic device, comprising:

[0050] Memory, used to store computer programs;

[0051] A processor is used to execute the computer program to implement the instruction tracing method described in the method for compensating for missing flow fields in pump-jet flow field testing.

[0052] The beneficial effects of this invention are as follows:

[0053] The purpose of this invention is to address the shortcomings in experimental methods for testing the internal flow field of pump-jet propulsion systems, which lack detailed exploration of the internal flow field. Furthermore, it overcomes the data gaps caused by structural obstruction, uneven distribution of tracer particles, and wall reflections. The invention proposes a pump-jet flow field testing device and a flow field gap compensation method, providing technical support for understanding the evolution of the internal flow field morphology in pump-jet propulsion systems from an experimental mechanism perspective, and for subsequent noise prediction and flow field calculation verification. Attached Figure Description

[0054] Figure 1 This is an overall schematic diagram of the present invention;

[0055] Figure 2 This is a side view of the pump spray structure of the present invention;

[0056] Figure 3 This is a diagram showing the internal components of the pump sprayer of the present invention;

[0057] Figure 4 This is a flowchart of the pump-jet missing flow field compensation algorithm of the present invention;

[0058] Figure 5 This is a diagram showing the compensation results for the missing flow field in the pump spray of the present invention. Detailed Implementation

[0059] The present invention will be further described below with reference to the accompanying drawings.

[0060] according to Figures 1 to 3 The present invention provides a pump-jet flow field testing device, comprising: a high-frequency dual-pulse laser 1, an image processing system 2, a camera synchronizer 3, a high-speed camera 4, a cavitation water cylinder 5, a reflective prism 6, a pump-jet model 7, and a transparent conduit 8.

[0061] The high-frequency dual-pulse laser 1 is placed at the rear end of the pump-jet model 7 inside the cavitation water tank 5. The high-frequency dual-pulse laser 1 emits a laser wavelength of 527nm and a pulse frequency of 0.2-10kHz, which illuminates the pump-jet model 7 through the cavitation water tank 5. The pump-jet model 7 contains a mutually coupled stator 9 and rotor 10. The stator 9 is made of 6061-T6 aluminum alloy to ensure structural strength. The rotor 10 and the transparent conduit 8 are both made of transparent polymethyl methacrylate (PMMA, plexiglass). When the laser passes through the pump-jet model 7, the mutual coupling of the stator 9 and rotor 10, as well as the obstruction of the structure, inevitably leads to a loss of flow field structure. To capture rich flow field information within the pump-jet, a reflective prism 6 is installed on the laser plane of the pump-jet model 7 to enhance image capture by the high-speed camera 4. One side of the cavitation water tank 5 is made of transparent glass to facilitate the high-speed camera 4 in capturing the flow field. Both high-speed cameras 4 are Phantom VEO 640. A CMOS high-speed camera with a maximum resolution of 2560×1600 pixels and a maximum full-frame shooting frequency of 1400fps is placed on one side of the transparent glass of the bubble tube 5 for acquiring particle images; the camera synchronizer 3 is connected to two high-speed cameras 4 and an image processing system 2 for controlling the high-speed cameras 4 to acquire images synchronously.

[0062] according to Figure 4The generative adversarial network-based model used in the image processing system 2 mainly consists of a pump-jet flow field data assimilation network, a generator network, and a discriminator network. It is primarily used to compensate for missing flow fields in the pump-jet system. The pump-jet flow field data assimilation network assimilates the flow field information processed through experimental measurements with the flow field information simulated by computational fluid dynamics under the same operating conditions. The pump-jet flow field data assimilation network internally consists of a Voronoi operator, 10 convolutional layers, and an intrinsic orthogonal decomposition (POD) layer. Its algorithm steps are as follows:

[0063] Step 1: The obtained experimental and numerical simulation data are passed through 10 convolutional layers to collect spatial characteristic information of the internal flow field of the pump jet.

[0064] Step 2: Utilize intrinsic orthogonal decomposition to reduce the flow field cost and obtain information on different flow modes. The flow field information obtained through convolution has a size of f. x ×f y The flow field information is flattened according to the spatial dimension to ensure that the flattened flow field data does not lose the original flow field data structure. Its flattened flow field feature snapshot matrix is ​​shown in formula (1):

[0065]

[0066] Where p∈[1,f x ],q∈[1,f y ],x(r p,q ,t k () represents the physical quantity of the flow field. In the internal flow field of a pump, the velocity along the three-dimensional spatial direction is mainly selected, i.e., v. x v y v z Furthermore, to reduce the amount of data processing required for flow field dimensionality reduction, the steady-state average component in the flow field space is usually subtracted, and only the part of the pump that deviates from the mean is modeled, as shown in Equation (2):

[0067]

[0068] A set of orthogonal bases is selected to achieve the best approximation of the original pump jet flow field data. The orthogonal bases are represented as follows:

[0069]

[0070] Here, n represents the order of the orthogonal basis, i.e., the order of the modes. In the pump-jet flow field compensation process, the first 60 modes are selected as the main flow field information.

[0071] Step 3: Establish a mapping relationship between the observation space and the state space using the Voronoi operator, fusing modal parameters with spatial feature information to assimilate pump-jet flow field test data and numerical simulation data. This is achieved by constructing an inverse Voronoi operator to map the observation space to the state space. Define an observation field. As a set of observation points for pump-jet flow field test data at a given location, the location of this point is defined as {(i t,k ,j t,k )}, its expression is:

[0072] (i t,k ,j t,k )∈[1,...,N y ]×[1,...,N x (4) Where k represents the number of observation points during the pump-jet test, and the Voronoi operator unit R t,k Satisfy the following formula:

[0073] (5) Where d(·) represents the Euclidean distance, and its calculation formula is as follows:

[0074]

[0075] Therefore, the observation space is divided into multiple Voronoi operator units, satisfying the following formula:

[0076]

[0077] To obtain:

[0078]

[0079] Substitute the obtained observation space into:

[0080]

[0081] Ultimately, the assimilation and fusion of pump spray test data and simulation data will be achieved.

[0082] Step 4: The variationally assimilated flow field is input into the neural network, where the generator network consists of deformable convolutions, ATN (Attention Transfer) layers, convolutional layers, and deconvolutional layers. In the encoding part of the generator network, six convolutional layers are designed to sample the flow field features after data fusion, thereby obtaining low-dimensional spatial feature information of the pump-jet internal flow field. Furthermore, considering that the shape of the pump-jet rotor changes with position during spatial sampling, and that traditional convolutions are limited by fixed kernel size and position, making it impossible to learn features of objects with different deformations at different positions, deformable convolutions are embedded in the first two downsampling layers to adaptively change the sampling position of the convolution kernel, thus adapting to the rotor shape change. The features obtained from adjacent convolutional layers are jointly input into the ATN module. Simultaneously, the ATN module calculates the inner and outer cosine similarity of the missing flow field regions in the input feature map. The data expression for the ATN working principle is shown below:

[0083]

[0084] in, p represents the cosine similarity of the feature maps at the l-th layer. i This indicates that flow field feature values ​​are extracted from the outside of the i-th Mask in the feature map, p j Flow field feature values ​​extracted from the j-th Mask. ||p i ||2 and||p j ||2 represent the flow field characteristic values ​​p, respectively. i and p j The L2 norm. The cosine similarity index. Convert to attention weights Subsequently, the flow field with missing adjacent low-level features This can be achieved by weighting the attention weights to obtain the weights of the missing regions, guiding the fusion between high-level and low-level features, and compensating for the missing flow field while making up for the information loss caused by convolutional pooling.

[0085] Step 5: The discriminator network (D network) guides the generation of G to approximate the real flow field as closely as possible by distinguishing between the flow field information generated by the generator network and the fine flow field calculated via CFD. An adversarial loss function is introduced during training, as shown below:

[0086]

[0087] Where E(*) represents the expected value of the distribution function, p d (x) represents the internal flow field distribution of the pump nozzle in the CFD simulation, p d(z) represents the flow field distribution generated by generator G, and D(*) represents the output of the discriminator, whose output value varies from 0 to 1.

[0088] Step 6: Optimize the missing compensation weights of the G network. Feed the loss function value calculated by the D network back to the G network to promote backpropagation of the network. Repeat steps 4-5 to optimize the missing compensation effect of the flow field.

[0089] Step 7: Obtain the final pump-jet experimental and numerically missing accurate compensated flow field. Compared to the standard flow field calculated by CFD, its root mean square error is less than 3%. Figure 5 The diagram shown is a compensation result for the missing flow field in the pump jet. Figure 5 (a) shows the pump-jet flow field after data assimilation, and 5(c) shows the pump-jet flow field after data loss. Figure 4 The algorithm flow described in the text can achieve compensation for missing flow fields in pump-jet systems. From... Figure 5 (b) The compensation results show that the missing pump jet flow field can be effectively compensated after processing by the method of the present invention.

[0090] The principle or method of use of this invention is as follows: During the experiment, a high-frequency dual-pulse laser 1 and a high-speed camera 4 are turned on. The high-frequency dual-pulse laser 1 emits a laser with a wavelength of 527nm and a pulse frequency of 0.2-10kHz. After passing through the pump-jet model 7, the laser irradiates the flow field to be measured inside the pump-jet. The laser excites the tracer particles in the cavitation water cylinder 5, which emit reflected light. However, due to the mutual coupling between the rotor 10 and the stator 9, the flow field information is missing. At the same time, to make the flow field information more comprehensive, a reflective prism 6 is placed behind the pump-jet model 7. The two beams of light are captured by the high-speed camera 4. The acquired image information is transmitted to the image processing system 2 via Ethernet through the camera synchronizer 3. Finally, through the self-developed pump-jet flow field missing compensation algorithm, a precise measurement of the flow field inside the pump-jet is achieved.

[0091] The above description is merely a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any equivalent substitutions or modifications made by those skilled in the art within the technical scope disclosed in the present invention, based on the technical solution and inventive concept of the present invention, should be included within the scope of protection of the present invention. The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention. For those skilled in the art, the present invention can have various modifications and variations. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A method for compensating for missing flow field during pump jet flow field testing, characterized in that: The specific steps are as follows: Step 1: The obtained experimental and numerical simulation data are passed through 10 convolutional layers to collect spatial feature information of the internal flow field of the pump jet; Step 2: Use intrinsic orthogonal decomposition to reduce the flow field cost and obtain information on different modes of the flow field; Step 3: Establish the mapping relationship between the observation space and the state space through the Voronoi operator, fuse the modal parameters with spatial feature information, and realize the assimilation of pump-jet flow field test data and numerical simulation data; Step 4: Input the variationally assimilated flow field into the neural network to generate a network including deformable convolution, ATN attention transfer layer, convolutional layer and deconvolutional layer; Step 5: Discriminator network, i.e., D network, distinguishes between the flow field information generated by the generator network and the fine flow field obtained by CFD calculation, so that the missing region generated by the generator G is close to the real flow field. Step 6: Optimize the missing compensation weights of the G network; feed back the loss function value calculated by the D network to the G network to promote backpropagation of the network, and repeat steps 4-5 to optimize the missing compensation effect of the flow field. Step 7: Obtain the final pump-jet experimental and numerically missing accurate compensated flow field.

2. The method for compensating for missing flow field in pump jet flow field testing according to claim 1, characterized in that: Step 2 specifically involves: The flow field information obtained through convolution has a size of f. x ×f y The flow field information is flattened according to the spatial dimension to ensure that the flattened flow field data does not lose the original flow field data structure. The formula for the flattened flow field feature snapshot matrix is ​​as follows: Where p∈[1,f x ],q∈[1,f y ],x(r p,q ,t k ) represents the physical quantity of the flow field. In the internal flow field of a pump-jet pump, the velocity along the three-dimensional spatial direction is mainly selected, i.e., v. x v y v z ; Furthermore, to reduce the amount of data processing required for flow field dimensionality reduction, the steady-state average component in the flow field space is subtracted, and only the portion of the pump-jet flow that deviates from the mean is modeled, as shown in the following formula: A set of orthogonal bases is selected to achieve the best approximation of the original pump jet flow field data. The orthogonal bases are represented as follows: Where n represents the order of the orthogonal basis, i.e. the order of the modes; the first 60 modes are selected as the main flow field information in the pump-jet flow field compensation process.

3. The method for compensating for missing flow field in pump jet flow field testing according to claim 1, characterized in that: Step 3 specifically involves: Construct a Voronoi inverse operator to map the observation space to the state space and define an observation field. As a set of observation points for pump-jet flow field test data at a given location, the location of this point is defined as {(i t,k ,j t,k The expression for )} is as follows: (i t,k ,j t,k )∈[1,...,N y ]×[1,...,N x ] Where k represents the number of observation points during the pump-jet test, and the Voronoi operator unit R t,k Satisfy the following formula: Where d(·) represents the Euclidean distance, and its calculation formula is as follows: Therefore, the observation space is divided into multiple Voronoi operator units, satisfying the following formula: To obtain: Substitute the obtained observation space into: Ultimately, the assimilation and fusion of pump spray test data and simulation data will be achieved.

4. The method for compensating for missing flow field in pump jet flow field testing according to claim 1, characterized in that: The generative network in step 4 employs a six-layer convolutional architecture in its encoding section. These layers downsample the fused flow field features to obtain low-dimensional features of the pump-jet internal flow field. During the first two downsampling layers, deformable convolutions are embedded, adaptively changing the kernel sampling position. Features obtained from adjacent convolutional layers are then input into the ATN module. Simultaneously, the ATN module calculates the inner and outer cosine similarity of missing flow field regions in the input feature map. The data expression for the ATN's working principle is as follows: in, p represents the cosine similarity of the feature maps at the l-th layer. i This indicates that flow field feature values ​​are extracted from the outside of the i-th Mask in the feature map, p j Flow field feature values ​​extracted from inside the j-th Mask; ||p i ||2 and||p j ||2 represent the flow field characteristic values ​​p, respectively. i and p j The L2 norm; the cosine similarity index Convert to attention weights Subsequently, the flow field with missing adjacent low-level features This can be achieved by weighting the attention weights to obtain the weights of the missing regions, guiding the fusion between high-level and low-level features, and compensating for the missing flow field while making up for the information loss caused by convolutional pooling.

5. The method for compensating for missing flow field in pump jet flow field testing according to claim 1, characterized in that: Step 5 introduces an adversarial loss function during network training, as follows: Where E(*) represents the expected value of the distribution function, p d (x) represents the internal flow field distribution of the pump nozzle in the CFD simulation, p d (z) represents the flow field distribution generated by generator G, and D(*) represents the output of the discriminator, whose output value varies from 0 to 1.

6. A pump-jet flow field testing device used in the pump-jet flow field testing flow field missing compensation method according to any one of claims 1-5, characterized in that: Includes a high-frequency dual-pulse laser (1), an image processing system (2), a camera synchronizer (3), a high-speed camera (4), a cavitation water tank (5), a reflective prism (6), a pump-jet model (7), and a transparent conduit (8); The high-frequency dual-pulse laser (1) is placed at the rear end of the pump-jet model (7) in the cavitation water cylinder (5). The laser emitted by the high-frequency dual-pulse laser (1) irradiates the pump-jet model (7) through the cavitation water cylinder (5). The pump-jet model (7) is fixed inside the cavitation water cylinder (5) through a transparent conduit (8). The pump-jet model (7) is equipped with a mutually coupled stator (9) and rotor (10). A reflective prism (6) is installed on the laser plane of the pump-jet model (7), and a high-speed camera (4) is used to capture image information and collect particle images. The high-speed camera (4) is connected to the image processing system (2) through a camera synchronizer (3). The camera synchronizer (3) is used to control the high-speed camera (4) to collect images synchronously.

7. The pump jet flow field testing device according to claim 6, characterized in that: One side of the cavitation water tube (5) is made of transparent glass, and two high-speed cameras (4) are placed on the transparent glass side of the cavitation water tube (5).

8. A computer-readable storage medium having a computer program / instructions stored thereon, characterized in that: When the computer program / instructions are executed by the processor, they implement the steps of the method according to any one of claims 1 to 5.

9. A computer program product, comprising a computer program / instructions, characterized in that: When the computer program / instructions are executed by the processor, they implement the steps of the method according to any one of claims 1 to 5.

10. An electronic device, characterized in that, include: Memory, used to store computer programs; A processor for executing the computer program to implement the method as claimed in any one of claims 1 to 5.

Images