An AI-based rotary vane flow self-distributing head and its automatic control method

Abstract

The present invention relates to the technical field of heat exchangers, and specifically to an AI-based rotary vane flow self-distributing head and an automatic control method thereof, wherein the self-distributing head comprises a primary distribution component and a secondary distribution component, wherein the primary distribution component comprises a head and a connecting pipe, and the secondary distribution component comprises a multi-rotary vane component and a fixing frame; the present invention combines the primary distribution component and the secondary distribution component, and provides a stronger flow balancing capability through the multi-rotary vane component of the secondary distribution component, and can realize active and passive distribution, and can be dynamically controlled through an AI algorithm, thereby further improving the operating efficiency of the heat exchanger.

Description

Technical Field

[0001] The present invention relates to the technical field of heat exchangers, and in particular to an AI-based rotary vane flow self-distributing head and an automatic control method thereof. Background Art

[0002] Heat exchangers are general process equipment for allocating energy between different logistics and completing heat transfer. They are widely used in a large number of industries such as power generation, chemical industry, power, metallurgy, etc., especially in power circulation systems with supercritical carbon dioxide as the working fluid. Heat exchangers play an important role in transferring and allocating energy between working fluids.

[0003] With the continuous improvement of science and technology, people pay more and more attention to the special application scenarios of power systems involving nuclear power plants, thermal power plants, and aircraft engines. Reducing the size of equipment, improving efficiency, and reducing equipment manufacturing and operating costs and natural resource consumption are one of the future development directions of heat exchangers.

[0004] At present, the heat exchangers used in conventional industrial fields mainly include shell and tube heat exchangers, sleeve heat exchangers, plate heat exchangers, plate-fin heat exchangers, etc. They cannot simultaneously meet the requirements of large heat exchange specific surface area, high welding strength, and small size. In recent years, with the improvement of industrial manufacturing level, microchannel heat exchangers with high-precision chemical etching and vacuum diffusion welding as the core process have gradually entered the application stage. Their microchannels are small in size, highly compact, and the welding method has no welding slag. The strength of the joints is close to the strength of the parent material, which has obvious advantages. However, in the actual testing of microchannel heat exchangers, it was found that due to the small and large number of channels of such heat exchangers, the working fluid flow rate is high under operating conditions. It has been observed many times that the blockage and wear degree of the heat exchanger in the conventional single semi-cylindrical head is related to the position. It can be seen that the local flow rate of the fluid at different positions is different. This uneven flow rate may even affect the heat transfer performance.

[0005] In view of this, the present invention proposes an AI-based rotary vane flow self-distributing head and an automatic control method, which provide a means to improve the uniformity of fluid flow distribution in a microchannel heat exchanger, optimize the operating parameters of the heat exchanger, and narrow the gap between design and actual operation. Summary of the invention

[0006] The technical problem to be solved by the present invention is to provide an AI-based rotary vane flow self-distributing head and its automatic control method in view of the characteristics of a large number of flow channels in a microchannel heat exchanger and a weak flow distribution capacity of a conventional head, thereby improving the uniformity of fluid flow distribution in the microchannel heat exchanger.

[0007] The present invention is achieved through the following technical solutions:

[0008] An AI-based rotary vane flow self-distributing head comprises: a primary distribution component and a secondary distribution component, wherein the secondary distribution component is fixedly arranged inside the primary distribution component;

[0009] The primary distribution assembly comprises: a sealing head and a connecting pipe, the sealing head is provided with a hole adapted to the connecting pipe, and one end of the connecting pipe is fixedly connected to the hole;

[0010] The secondary distribution component comprises: a multi-rotor blade component and a fixing frame, the multi-rotor blade component is fixedly arranged in the head by the fixing frame, and the secondary distribution component is arranged corresponding to the connecting pipe.

[0011] Optionally, the rotation axis of the multi-rotor blade assembly coincides with the central axis of the connecting pipe, the first end of the fixing frame is fixedly connected to the inner side surface of the head, and the second end of the fixing frame is fixedly connected to the multi-rotor blade assembly;

[0012] The fixing frame is provided with a flow channel hole for fluid to pass through.

[0013] Optionally, the multi-rotor blade assembly comprises: a single-rotor blade assembly, a rotating assembly and a perforated plate, the rotating assembly is fixedly connected to the fixing frame, the inner ends of the plurality of single-rotor blade assemblies are fixedly connected to the rotating end of the rotating assembly, and the plurality of single-rotor blade assemblies are evenly distributed in an annular shape with the rotating end as the axis;

[0014] The orifice plate is arranged between the rotating assembly and the connecting pipe, and the orifice plate is fixedly connected to the fixing frame, and the orifice plate is parallel to the rotating plane of the single rotating blade assembly.

[0015] Specifically, the single rotary blade assembly includes: a blade and a micro-motion assembly, the blade is connected to the rotating assembly through the micro-motion assembly, the central axis of the micro-motion assembly is perpendicular to the central axis of the rotating end, and the micro-motion assembly is used to adjust the angle of attack of the blade facing the fluid.

[0016] Optionally, the blades are sheet-like structures of spline curves, and a plurality of the blades are bent in the same direction.

[0017] Optionally, the rotating assembly includes: a rotating shaft and a rotating motor, the rotating motor is fixedly connected to the fixed frame, and the rotating shaft is coaxial with and fixedly connected to the torque output shaft of the rotating motor, the inner ends of the plurality of single-rotating blade assemblies are fixedly connected to the rotating shaft, and the plurality of single-rotating blade assemblies are evenly distributed in an annular shape with the rotating shaft as the axis.

[0018] Optionally, the micro-motion assembly includes: a micro-motion motor and a micro-motion shaft, the blade is connected to the rotating shaft through the micro-motion shaft and the micro-motion motor, the central axis of the micro-motion shaft is perpendicular to the central axis of the rotating shaft, and the micro-motion motor is used to adjust the angle of attack of the blade facing the fluid.

[0019] Furthermore, the self-distributing head further comprises a primary flow sensor, a secondary flow sensor and a control module, wherein the primary flow sensor is arranged on the circumferential inner side of the connecting pipe, and the secondary flow sensor is arranged in the through hole of the orifice plate;

[0020] The control module is electrically connected to the primary flow sensor, the secondary flow sensor, the rotary motor and the micro-motor.

[0021] An automatic control method of a rotary vane flow self-distributing head based on AI, based on the above-mentioned rotary vane flow self-distributing head based on AI, the automatic control method includes:

[0022] A head model with the same structure as the physical head was built in CFD, and the head model was simulated for fluid mechanics. The simulation input was different primary flow rates and different state parameters, and the simulation output was the velocity field in the head model. The simulated secondary flow rate at the secondary flow sensor; the primary flow rate is the flow rate at the primary flow sensor, and the state parameters include the rotation speed of the rotating component and the angle of attack of the blade to the fluid;

[0023] Run the physical head, adjust the flow rate and state parameters at the primary flow sensor, and obtain the The actual secondary flow collected by a secondary flow sensor;

[0024] Construct a convolutional neural network model, and conduct qualitative training on the convolutional neural network through primary flow, state parameters and simulated secondary flow; after completing the qualitative training, conduct quantitative training on the convolutional neural network through primary flow, state parameters and actual secondary flow to obtain the final flow prediction model;

[0025] Determine the primary flow in actual operation and obtain multiple sets of state parameters and their corresponding predicted secondary flow ;

[0026] Calculate the flow distribution coefficient ,in, , is the total number of secondary flow sensors, It is the average value of the secondary flow detected by all secondary flow sensors;

[0027] Screening flow distribution coefficient The state parameter corresponding to the minimum value;

[0028] The control module regulates the rotating component and the micro-motion component according to the state parameters;

[0029] Monitor primary flow in real time, and recalculate status parameters and make adjustments if changes occur.

[0030] Specifically, the method for constructing the traffic prediction model includes:

[0031] Build a convolutional neural network model, and use the primary flow signal, the rotation speed of the rotating component, and the angle of attack of the blade to the fluid as the model input; use the secondary flow signal as the model output;

[0032] The secondary flow sensor collects the actual secondary flow from the distribution head under the actual working state;

[0033] Conduct qualitative training on the positive and negative weights in the convolutional neural network model: increase the first variable of the three variables in the model input, keep the other variables unchanged, output the predicted secondary flow, and calculate the difference between the predicted secondary flow and the actual secondary flow. If the difference increases or decreases, assign all positive or negative signs to the weights a1 to ai;

[0034] Then increase the second variable of the model input, keep other variables unchanged, output the predicted secondary flow and calculate the difference between the predicted secondary flow and the actual secondary flow. If the difference still increases, keep the signs of all weights unchanged; if it decreases, change the signs of some weights.

[0035] Increase the third variable of the model input, keep other variables unchanged, repeat the above steps and adjust the signs of some weights until the difference between the predicted secondary flow and the actual secondary flow increases;

[0036] Get the sign of the weight;

[0037] Conduct quantitative training of weights in the convolutional neural network model: perform convolution calculation on the primary flow signal, the rotation speed of the rotating component, and the angle of attack of the blade on the fluid with the diagonal matrix convolution kernel, extract local features, and use them as the feature extraction layer C1;

[0038] Perform weighted calculation and summation on the feature extraction layer C1 to obtain a feature mapping layer P1 with a weight value of a1;

[0039] The feature mapping layer P1 is re-convolved with the diagonal matrix convolution kernel to obtain the feature extraction layer C2;

[0040] Perform weighted calculation and summation on the feature extraction layer C2 to obtain a feature mapping layer P2 with a weight value of a2;

[0041] Iterate the operation until the feature mapping layer Pi with weight ai is obtained;

[0042] Compare the predicted secondary flow output by the feature mapping layer Pi with the actual secondary flow. If the two are not equal, adjust the weights a1, a2, ..., ai through the optimization algorithm and re-perform the convolution calculation until the two are equal;

[0043] Obtain the traffic prediction model.

[0044] Compared with the prior art, the present invention has the following advantages and beneficial effects:

[0045] The present invention combines a primary distribution component and a secondary distribution component. The multi-rotor blade component of the secondary distribution component provides a stronger flow distribution capability, can realize active and passive distribution, and can be dynamically regulated by an AI algorithm, thereby further improving the operating efficiency of the heat exchanger. BRIEF DESCRIPTION OF THE DRAWINGS

[0046] The accompanying drawings illustrate exemplary embodiments of the present invention and, together with the description thereof, are used to explain the principles of the present invention. These drawings are included to provide a further understanding of the present invention, and the accompanying drawings are included in and constitute a part of this specification and do not constitute a limitation of the embodiments of the present invention.

[0047] Figure 1 It is a structural schematic diagram of an AI-based rotary vane flow self-distributing head according to the present invention.

[0048] Figure 2 It is a top view of an AI-based rotary vane flow self-distributing head according to the present invention.

[0049] Figure 3 is a side view of a multi-rotor blade assembly according to the present invention.

[0050] Figure 4 is a top view of the multi-rotor blade assembly according to the present invention.

[0051] Figure 5 is a side view of an orifice plate according to the present invention.

[0052] Figure 6 is a top view of the orifice plate according to the present invention.

[0053] Figure 7 It is a schematic diagram of the structure of a single rotary blade assembly according to the present invention.

[0054] Figure 8 It is a flow chart of an automatic control method of an AI-based rotary vane flow self-distributing head according to the present invention.

[0055] Figure markings: 1-head, 2-pipe connection, 3-multi-rotor blade assembly, 4-fixed frame, 21-primary flow sensor, 31-single-rotor blade assembly, 32-rotating shaft, 33-rotating motor, 34-orifice plate, 35-secondary flow sensor, 311-blade, 312-micro-motor, 313-micro-rotating shaft. DETAILED DESCRIPTION

[0056] To make the purpose, technical solution and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and implementation methods. It is understood that the specific implementation methods described herein are only used to explain the relevant content, rather than to limit the present invention.

[0057] It should also be noted that, for the convenience of description, only the parts related to the present invention are shown in the drawings.

[0058] In the absence of conflict, the embodiments and features of the embodiments of the present invention may be combined with each other. The present invention will be described in detail below with reference to the accompanying drawings and in combination with the embodiments.

[0059] Embodiment 1

[0060] like Figure 1 , Figure 2 , Figure 3 and Figure 4 As shown, this embodiment provides an AI-based rotary vane flow self-distributing head, including: a primary distribution component and a secondary distribution component, and the secondary distribution component is fixedly arranged inside the primary distribution component. The self-distributing head is suitable for heat exchange equipment with a large number of flow channels, including microchannel heat exchangers, plate-fin heat exchangers, shell and tube heat exchangers, etc.

[0061] The primary distribution component distributes the fluid flow through the pipe 2 and into the head 1 for the first time; the secondary distribution component distributes the fluid flow through the head 1 and into the heat exchanger for the second time.

[0062] The primary distribution component includes: a head 1 and a pipe 2. The head 1 is provided with a hole adapted to the pipe 2, and one end of the pipe 2 is fixedly connected to the hole. During heat exchange, the head 1 is fixed to the inlet section of the heat exchanger by welding. The materials used for the various components of the head 1 include various easily processed metal materials such as stainless steel, ferroalloy, and titanium alloy. The pipe 2 of the head 1 uses metal pipes in various shapes including circular and square, and the head 1 can use a three-dimensional shell with various geometric curves including semicircular, parabolic, and hyperbolic cross-sections. Different shapes are suitable for fluids with different flow rate ranges and can be improved according to actual operating conditions.

[0063] In practice, more than one pipe 2 can be installed on the head 1 and maintained evenly distributed on the head 1. When the number of pipes 2 is greater than 1, a secondary flow distribution component needs to be added at the same time to be coaxially arranged with the pipe 2.

[0064] The secondary distribution component includes: a multi-rotor blade component 3 and a fixing frame 4. The multi-rotor blade component 3 is fixed in the head 1 through the fixing frame 4. The secondary distribution component is arranged corresponding to the connecting pipe 2, that is, one secondary distribution component corresponds to one connecting pipe 2.

[0065] The rotation axis 32 of the multi-rotor assembly 3 coincides with the central axis of the pipe 2, the first end of the fixing frame 4 is fixedly connected to the inner side of the head 1, and the second end of the fixing frame 4 is fixedly connected to the multi-rotor assembly 3;

[0066] The fixing frame 4 is provided with a flow channel hole for the fluid to pass through. Generally, the fixing frame 4 is fixed by fixing rods made of multiple metal rods, that is, one end of the fixing rod is fixedly connected to the inner side of the head 1, and the other end of the fixing rod is fixedly connected to the multi-rotor assembly 3. Of course, it can also be set as the entire structure, and then multiple through holes are set on the fixing frame 4, connected between the multi-rotor assembly 3 and the head 1 so that it does not shake.

[0067] The multi-vane assembly 3 includes: a single-vane assembly 31, a rotating assembly and an orifice plate 34. The rotating assembly is fixedly connected to the fixed frame 4. The inner ends of the multiple single-vane assemblies 31 are fixedly connected to the rotating end of the rotating assembly, and the multiple single-vane assemblies 31 are evenly distributed in a ring shape with the rotating end as the axis. Under normal circumstances, the multi-vane assembly 3 of the secondary flow distribution assembly is generally one layer, and the number of layers can also be increased to achieve a multi-layer coaxial arrangement to improve the flow distribution capability.

[0068] The orifice plate 34 is disposed between the rotating assembly and the connecting pipe 2 , and the orifice plate 34 is fixedly connected to the fixing frame 4 , and the orifice plate 34 is parallel to the rotating plane of the single rotary blade assembly 31 .

[0069] The orifice plate 34 is disposed between the rotating assembly and the pipe 2. It is a metal plate with a large number of tiny holes obtained by chemical etching, mechanical milling and other processing methods on a circular metal sheet, which is used to further control and regulate the flow of the fluid and ensure that the fluid can be evenly distributed. Figure 5 and Figure 6 The circular holes shown in the figure also include rectangular holes, triangular holes, strip holes and other structural shapes, and the size and position of the holes can be further optimized and improved. The orifice plate 34 is a metal plate with a large number of tiny holes obtained on a circular metal sheet by chemical etching, mechanical milling and other processing methods, and is arranged downstream of the multi-rotor blade assembly 3 to redistribute the flow of the fluid that has been evenly distributed twice, thereby improving the uniformity of the flow.

[0070] During operation, the fluid enters the head 1 through the pipe 2 and enters the secondary distribution component through the hole. After the fluid is distributed once, it is distributed twice by the multi-rotor blade component 3 to achieve uniform flow distribution.

[0071] No force may be applied to the rotating assembly, so that the multi-vane assembly 3 rotates naturally under the impact of the fluid, thereby achieving passive secondary distribution.

[0072] Embodiment 2

[0073] The rotating assembly in this embodiment includes: a rotating shaft 32 and a rotating motor 33. The rotating motor 33 is fixedly connected to the fixed frame 4, and the rotating shaft 32 is coaxial with the torque output shaft of the rotating motor 33 and is fixedly connected. The inner ends of multiple single-vane assemblies 31 are all fixedly connected to the rotating shaft 32, and the multiple single-vane assemblies 31 are evenly distributed in a ring shape with the rotating shaft 32 as the axis.

[0074] That is, on the basis of Example 1, a rotating motor 33 is provided to control the rotation of the rotating assembly, so that the rotation of the multi-blade assembly 3 can be regulated, and the multi-blade assembly 3 can be actively controlled according to the fluid to achieve active secondary distribution.

[0075] In addition to using motor drive, external effects such as magnetic field and electric field can also be used to achieve self-distribution of flow during the operation of the heat exchanger.

[0076] Embodiment 3

[0077] like Figure 7 As shown, the single rotary blade assembly 31 provided in this embodiment includes: a blade 311 and a micro-motion assembly, the blade 311 is connected to the rotating assembly through the micro-motion assembly, the central axis of the micro-motion assembly is perpendicular to the central axis of the rotating end, and the micro-motion assembly is used to adjust the angle of attack of the blade 311 facing the fluid.

[0078] The micro-motion component can be used to adjust the angle of attack of the blade 311 on the fluid, further improving the secondary distribution efficiency. The blade 311 is controlled by the micro-motion component, and the angle range of rotation around the micro-motion component is -10°~10°, and a certain blade 311 can be adjusted separately to achieve flexible control of the multi-rotating blade component 3.

[0079] The blades 311 are spline-shaped sheet structures, and the multiple blades 311 are all bent in the same direction. Figure 7 The vertical equal-width blades 311 shown also include various structures such as inclined blades 311 in turbines and variable-width blades 311 in aircraft, and parameters such as the number of blades 311 and the length of blades 311 can also be changed for optimization.

[0080] In addition, the micro-motion component includes: a micro-motion motor 312 and a micro-motion shaft 313. The blade 311 is connected to the rotating shaft 32 through the micro-motion shaft 313 and the micro-motion motor 312. The central axis of the micro-motion shaft 313 is perpendicular to the central axis of the rotating shaft 32. The micro-motion motor 312 is used to adjust the angle of attack of the blade 311 facing the fluid.

[0081] In addition to using motor drive, external effects such as magnetic fields and electric fields can also be used to achieve self-distribution of flow during the operation of the heat exchanger instead of manual adjustment; the micro motor 312 used can adopt other types of driving mechanisms such as direct drive motors.

[0082] Embodiment 4

[0083] Finally, in order to realize the automatic control of the AI-based rotary vane flow self-distributing head, the self-distributing head further includes a primary flow sensor 21, a secondary flow sensor 35 and a control module. The primary flow sensor 21 is arranged on the circumferential inner side of the pipe 2, and the secondary flow sensor 35 is arranged in the through hole of the orifice plate 34.

[0084] The control module is electrically connected to the primary flow sensor 21 , the secondary flow sensor 35 , the rotary motor 33 and the micro-motor 312 .

[0085] The primary flow sensor 21 and the secondary flow sensor 35 can use ultrasonic measuring elements and other elements that can measure fluid flow, and the number of primary flow sensors 21 arranged in the connecting pipe 2 is not limited to 8 as shown in the figure, and can be adjusted according to actual conditions. The arrangement position of the secondary flow sensor 35 in the micropore can also be adjusted to ensure that there is at least one in one micropore.

[0086] The primary flow sensor 21 is a high-precision sensor capable of measuring local flow. It is evenly attached to the inner wall of the pipe 2 in a circumferential direction to measure the flow of the fluid passing through each position in the pipe 2 and obtain the inlet flow distribution.

[0087] Embodiment 5

[0088] like Figure 8 As shown, this embodiment provides an automatic control method of a rotary vane flow self-distributing head based on AI. Based on the above-mentioned rotary vane flow self-distributing head based on AI, the automatic control method includes:

[0089] S1. Model a head model with the same structure as the physical head in CFD, and perform fluid mechanics simulation on the head model; the simulation input is different primary flow rates and different state parameters, and the simulation output result is the velocity field in the head model. The primary flow rate is the flow rate at the primary flow sensor, and the state parameters include the rotation speed of the rotating component and the angle of attack of the blade to the fluid.

[0090] Using CFD computational fluid dynamics software, a model identical to the physical head is built for fluid dynamics simulation. The simulation input is different inlet flow rates (i.e., primary flow) and state parameters of the multi-rotor assembly. The simulation result is the velocity field in the entire head, which naturally includes the velocity at the position corresponding to the secondary flow sensor of the physical head. This value is a simulation result, not a measured result, but it is still input into the training set of the convolutional neural network for preliminary training, in order to establish a qualitative relationship between different inlet flow rates and multi-rotor assembly parameters and secondary flow, laying the foundation for subsequent quantitative relationships.

[0091] S2, operate the physical head, adjust the flow rate and state parameters at the primary flow sensor, and obtain the The actual secondary flow collected by a secondary flow sensor;

[0092] Operate the physical head and continuously adjust the inlet flow rate and multi-rotor assembly parameters. At this time, a large amount of measured secondary flow can be obtained through collection to establish a database for the next step of model training.

[0093] S3, construct a convolutional neural network model, and perform qualitative training on the convolutional neural network through primary flow, state parameters and simulated secondary flow; after completing the qualitative training, perform quantitative training on the convolutional neural network through primary flow, state parameters and actual secondary flow to obtain the final flow prediction model;

[0094] Conduct convolutional neural network model training. You can use TensorFlow or PyTorch to build a convolutional neural network model. The inlet flow rate and multi-rotor assembly parameters of the operating equipment are used as the input of the training set, and the actual flow field data obtained by S2 is used as the output of the training set to jointly train the convolutional neural network. First, input the inlet flow rate and multi-rotor assembly parameters of the actual operating equipment, and perform convolution calculations on them with the diagonal matrix convolution kernel (that is, an N-order matrix with all 1s on the main diagonal and all 0s on the rest), extract local features, and use them as the feature extraction layer C1. Then, perform weighted calculations and summation on the C1 layer to obtain the feature mapping layer P1 (with a weight of a1). Subsequently, P1 is re-convolved with the diagonal matrix convolution kernel to obtain the feature extraction layer C2. Then, perform weighted calculations and summation on the C2 layer to obtain the feature mapping layer P2 (with a weight of a2). Repeat this process many times to obtain the final output result, which is compared with the actual flow field data. Compare, if the two do not match, modify the weights of a1, a2...ai until they match the actual flow field data , and finally establish the quantitative relationship between different inlet flow rates and different multi-rotor blade assembly parameters and simulation results, that is, the weights a1, a2…ai;

[0095] S4. Determine the primary flow in actual operation and obtain multiple sets of state parameters and their corresponding predicted secondary flow ;

[0096] After clarifying the weights a1, a2, ..., ai, you can Under the premise of adjusting the parameters of the multi-rotor components, predicting multiple groups , further preprocess the secondary flow sensor data, such as denoising, and calculate the flow field characteristics, that is, calculate the flow distribution coefficient ,in, , is the total number of secondary flow sensors, It is the average value of the secondary flow detected by all secondary flow sensors;

[0097] Screening flow distribution coefficient The state parameter corresponding to the minimum value;

[0098] The control module regulates the rotating component and the micro-motion component according to the state parameters;

[0099] S5. Monitor the primary flow in real time. If any change occurs, recalculate the state parameters and make adjustments.

[0100] The primary flow sensor has a total of , numbered , a measured flow rate is ,in ; The final primary flow rate is obtained by averaging the detection values ​​of all primary flow sensors.

[0101] The secondary flow sensor has a total of , numbered , a measured flow rate is ,in ;

[0102] Yes All The average value of

[0103] So the flow distribution coefficient That is, the difference between the measured result of each secondary flow sensor and the average value is squared, the sum is then divided by the total number That's it.

[0104] The traffic prediction model can be built using AI algorithms such as neural network algorithms, genetic algorithms, and annealing algorithms. This embodiment provides a method for building a model based on a neural network algorithm, including:

[0105] Build a convolutional neural network model, and use the primary flow signal, the rotation speed of the rotating component, and the angle of attack of the blade to the fluid as the model input; use the secondary flow signal as the model output;

[0106] The secondary flow sensor collects the actual secondary flow signal from the distribution head under the actual working state;

[0107] Increase the first variable among the three variables of the model input (i.e., primary flow, rotation speed of the rotating component, and angle of attack of the blade to the fluid), keep the other variables unchanged, output the predicted secondary flow, and calculate the difference between the predicted secondary flow and the actual secondary flow. If the difference increases or decreases, assign all weights a1 to ai positive (+) or negative (-).

[0108] Then increase the second variable of the model input and keep other variables unchanged, output the predicted secondary flow and calculate the difference between the predicted secondary flow and the actual secondary flow. If the difference still increases, keep the signs of all weights unchanged; if it decreases, change the signs of some weights. Similarly, increase the third variable of the model input and keep other variables unchanged, repeat the above steps and adjust the signs of some weights until the difference between the predicted secondary flow and the actual secondary flow increases.

[0109] The positive and negative signs of the weights finally obtained determine the qualitative relationship between the input and output, reducing the workload for the subsequent quantitative relationship determination.

[0110] The primary flow signal, the rotation speed of the rotating component, and the angle of attack of the blade on the fluid are convolved with the diagonal matrix convolution kernel to extract local features, which are used as the feature extraction layer C1;

[0111] Perform weighted calculation and summation on the feature extraction layer C1 to obtain a feature mapping layer P1 with a weight value of a1;

[0112] The feature mapping layer P1 is re-convolved with the diagonal matrix convolution kernel to obtain the feature extraction layer C2;

[0113] Perform weighted calculation and summation on the feature extraction layer C2 to obtain a feature mapping layer P2 with a weight value of a2;

[0114] Iterate the operation until the feature mapping layer Pi with weight ai is obtained;

[0115] Compare the predicted secondary flow output by the feature mapping layer Pi with the actual secondary flow. If the two are not equal, adjust the weights a1, a2, ..., ai through the optimization algorithm and re-perform the convolution calculation until the two are equal;

[0116] Obtain the traffic prediction model.

[0117] In the description of this specification, the description with reference to the terms "one embodiment / method", "some embodiments / methods", "example", "specific example", or "some examples" etc. means that the specific features, structures, materials or characteristics described in conjunction with the embodiment / method or example are included in at least one embodiment / method or example of the present application. In this specification, the schematic representations of the above terms do not necessarily refer to the same embodiment / method or example. Moreover, the specific features, structures, materials or characteristics described may be combined in any one or more embodiments / methods or examples in a suitable manner. In addition, those skilled in the art may combine and combine the different embodiments / methods or examples described in this specification and the features of the different embodiments / methods or examples, unless they are contradictory.

[0118] In addition, the terms "first" and "second" are used for descriptive purposes only and should not be understood as indicating or implying relative importance or implicitly indicating the number of the indicated technical features. Therefore, the features defined as "first" and "second" may explicitly or implicitly include at least one of the features. In the description of this application, the meaning of "plurality" is at least two, such as two, three, etc., unless otherwise clearly and specifically defined.

[0119] It should be understood by those skilled in the art that the above embodiments are only for the purpose of clearly illustrating the present invention, and are not intended to limit the scope of the present invention. For those skilled in the art, other changes or modifications may be made based on the above invention, and these changes or modifications are still within the scope of the present invention.

Claims

1. A rotary vane flow self-distributing head based on AI, characterized in that: include: A primary distribution component and a secondary distribution component, wherein the secondary distribution component is fixedly arranged inside the primary distribution component; The primary distribution assembly comprises: a sealing head (1) and a connecting pipe (2), wherein the sealing head (1) is provided with a hole adapted to the connecting pipe (2), and one end of the connecting pipe (2) is fixedly connected to the hole; The secondary distribution component comprises: a multi-rotating blade component (3) and a fixing frame (4); the multi-rotating blade component (3) is fixedly arranged in the sealing head (1) via the fixing frame (4); and the secondary distribution component is arranged corresponding to the connecting pipe (2); The rotation axis (32) of the multi-rotor blade assembly (3) coincides with the central axis of the connecting pipe (2); the first end of the fixing frame (4) is fixedly connected to the inner side surface of the sealing head (1); and the second end of the fixing frame (4) is fixedly connected to the multi-rotor blade assembly (3); and a flow channel hole for fluid to pass through is provided on the fixing frame (4); The multi-rotor blade assembly (3) comprises: a single-rotor blade assembly (31), a rotating assembly and a perforated plate (34); the rotating assembly is fixedly connected to the fixed frame (4); the inner ends of the plurality of single-rotor blade assemblies (31) are fixedly connected to the rotating end of the rotating assembly; and the plurality of single-rotor blade assemblies (31) are evenly distributed in a ring shape with the rotating end as the axis; The orifice plate (34) is arranged between the rotating component and the connecting pipe (2), and the orifice plate (34) is fixedly connected to the fixing frame (4). The orifice plate (34) is parallel to the rotating plane of the single rotating blade component (31), and the orifice plate is a plate with a large number of tiny holes.

2. According to the AI-based rotary vane flow self-distributing head according to claim 1, it is characterized in that: The single rotary blade assembly (31) comprises: a blade (311) and a micro-motion assembly, wherein the blade (311) is connected to the rotary assembly via the micro-motion assembly, the central axis of the micro-motion assembly is perpendicular to the central axis of the rotary end, and the micro-motion assembly is used to adjust the angle of attack of the blade (311) facing the fluid.

3. The AI-based rotary vane flow self-distributing head according to claim 2 is characterized in that: The blades (311) are spline-curve sheet structures, and a plurality of blades (311) are bent in the same direction.

4. The AI-based rotary vane flow self-distributing head according to claim 2 is characterized in that: The rotating assembly comprises: a rotating shaft (32) and a rotating motor (33); the rotating motor (33) is fixedly connected to the fixed frame (4); the rotating shaft (32) is coaxial with and fixedly connected to the torque output shaft of the rotating motor (33); the inner ends of a plurality of the single-rotating blade assemblies (31) are all fixedly connected to the rotating shaft (32); and the plurality of the single-rotating blade assemblies (31) are evenly distributed in a ring shape with the rotating shaft (32) as the axis.

5. The AI-based rotary vane flow self-distributing head according to claim 4 is characterized in that: The micro-motion assembly comprises: a micro-motion motor (312) and a micro-motion shaft (313); the blade (311) is connected to the rotating shaft (32) via the micro-motion shaft (313) and the micro-motion motor (312); the central axis of the micro-motion shaft (313) is perpendicular to the central axis of the rotating shaft (32); and the micro-motion motor (312) is used to adjust the angle of attack of the blade (311) facing the fluid.

6. The AI-based rotary vane flow self-distributing head according to claim 5 is characterized in that: It also includes a primary flow sensor (21), a secondary flow sensor (35) and a control module, wherein the primary flow sensor (21) is arranged on the circumferential inner side of the connecting pipe (2), and the secondary flow sensor (35) is arranged in the through hole of the orifice plate (34); The control module is electrically connected to the primary flow sensor (21), the secondary flow sensor (35), the rotating motor (33) and the micro-motor (312).

7. An automatic control method for a rotary vane flow self-distributing head based on AI, characterized in that: Based on the AI-based rotary vane flow self-distributing head according to claim 6, the automatic control method includes: A head model with the same structure as the physical head is built in CFD, and the head model is simulated for fluid mechanics; the simulation input is different primary flow rates and different state parameters, and the simulation output result is the velocity field in the head model, and the simulated secondary flow rate at the jth secondary flow sensor is obtained; the primary flow rate is the flow rate at the primary flow sensor, and the state parameters include the rotation speed of the rotating component and the angle of attack of the blade to the fluid; Operate the physical head, adjust the flow rate and state parameters at the primary flow sensor, and obtain the actual secondary flow rate collected by the jth secondary flow sensor; Construct a convolutional neural network model, and conduct qualitative training on the convolutional neural network through primary flow, state parameters and simulated secondary flow; after completing the qualitative training, conduct quantitative training on the convolutional neural network through primary flow, state parameters and actual secondary flow to obtain the final flow prediction model; Determine the primary flow in actual operation and obtain multiple sets of state parameters and their corresponding predicted secondary flow Q2(j); Calculate the flow distribution coefficient Where, j = 1, 2, 3, ..., n, n is the total number of secondary flow sensors, Q A It is the average value of the secondary flow detected by all secondary flow sensors; Filter the state parameter corresponding to the smallest value of the flow distribution coefficient S; The control module regulates the rotating component and the micro-motion component according to the state parameters; Monitor primary flow in real time, and recalculate status parameters and make adjustments if changes occur.

8. The automatic control method of the AI-based rotary vane flow self-distributing head according to claim 7 is characterized in that: The construction method of the traffic prediction model includes: Build a convolutional neural network model, and use the primary flow rate, the rotation speed of the rotating component, and the angle of attack of the blade to the fluid as the model input; use the secondary flow rate as the model output; The secondary flow sensor collects the actual secondary flow from the distribution head under the actual working state; Conduct qualitative training on the positive and negative weights in the convolutional neural network model: increase the first variable of the three variables in the model input, keep the other variables unchanged, output the predicted secondary flow, and calculate the difference between the predicted secondary flow and the actual secondary flow. If the difference increases or decreases, assign all positive or negative signs to the weights a1 to ai; Then increase the second variable of the model input, keep other variables unchanged, output the predicted secondary flow and calculate the difference between the predicted secondary flow and the actual secondary flow. If the difference still increases, keep the signs of all weights unchanged; if it decreases, change the signs of some weights. Increase the third variable of the model input, keep other variables unchanged, repeat the above steps and adjust the signs of some weights until the difference between the predicted secondary flow and the actual secondary flow increases; Get the sign of the weight; Conduct quantitative training of weights in the convolutional neural network model: perform convolution calculation on the primary flow signal, the rotation speed of the rotating component, and the angle of attack of the blade on the fluid with the diagonal matrix convolution kernel, extract local features, and use them as the feature extraction layer C1; Perform weighted calculation and summation on the feature extraction layer C1 to obtain a feature mapping layer P1 with a weight value of a1; The feature mapping layer P1 is re-convolved with the diagonal matrix convolution kernel to obtain the feature extraction layer C2; Perform weighted calculation and summation on the feature extraction layer C2 to obtain a feature mapping layer P2 with a weight value of a2; Iterate the operation until the feature mapping layer Pi with weight ai is obtained; Compare the predicted secondary flow output by the feature mapping layer Pi with the actual secondary flow. If the two are not equal, adjust the weights a1, a2, ..., ai according to the qualitative properties of the weights through the optimization algorithm, and re-perform the convolution calculation until the two are equal; Obtain the traffic prediction model.

Images