Air-injected cavitation water wing drag reduction effect ai control method and system

Abstract

The application discloses a drag reduction effect AI control method and system of a natural-ventilation mixed cavitation hydrofoil, and relates to the technical field of ship engineering. A cavitation flow experiment is performed to obtain an input data set and an output data set; feature extraction is performed based on the input data set and the output data set to obtain a cavitation coverage area rate, a flow field instability characteristic parameter, an interface evolution stability index and a signal cavitation index, and the indexes are integrated into a response data set; a comprehensive data source set is constructed based on the input data set, the output data set and the response data set; and the influence level of a current ventilation behavior is determined based on the response data set, and an optimal design scheme is determined based on the influence level and the comprehensive data source set. The method overcomes the limitation of a traditional static black box model that only relies on an input-output relationship and lacks process feedback, and effectively distinguishes between a temporarily low-drag but soon-to-be unstable state and a truly stable and efficient operating state.

Description

Technical Field

[0001] This application relates to the field of marine engineering technology, specifically to an AI control method and system for drag reduction effect of a natural-ventilated hybrid cavitation hydrofoil. Background Technology

[0002] In the research of cavitation drag reduction technology, data-driven methods are gradually replacing traditional empirical design and high-cost numerical simulations, becoming an important means of parameter optimization. A typical technical approach involves using hydrofoil geometric parameters (such as chord length, thickness, and vent distribution), incoming flow conditions (flow velocity, angle of attack, and cavitation number), and ventilation parameters (airflow rate) as input features, and using the drag coefficient obtained from experiments or simulations as the output target. A predictive model based on neural networks, random forests, or Gaussian processes is then constructed. By training this model, a nonlinear mapping relationship between the input parameters and drag reduction performance is established. Combined with search strategies such as genetic algorithms and Bayesian optimization, the theoretically optimal parameter combination is automatically obtained. This type of method has a certain predictive capability under static operating conditions and is widely used in the design optimization of components such as hydrofoils and propellers.

[0003] Most existing data-driven methods employ static mapping black-box models, whose fundamental limitation lies in their inability to assess the dynamic evolution of the flow field during control. These models typically only establish a direct correlation between input parameters (such as ventilation rate and angle of attack) and final drag, ignoring the fact that cavitation development is a highly unsteady and time-varying process. For example, after applying a certain ventilation strategy, the system may be in a drag-reducing evolution stage where cavitation coverage gradually expands and drag continues to decrease; it may also have already shown signs of instability such as violent interface oscillations and vortex structure breakage; or it may have entered a low-drag and stable optimal steady state. Traditional models completely fail to see these key intermediate states. The core reason for this problem is that their input features often only contain static information such as geometry or boundary conditions, failing to incorporate dynamic response data that reflects the real-time evolution of the flow field. Information that could reveal the intrinsic behavior of the flow field, such as the evolution of cavitation morphology recorded by high-speed cameras, the velocity field measured by PIV, the pulsating signals captured by pressure sensors, and the sequence of drag changes over time, is largely ignored in existing modeling. Therefore, the model cannot obtain physical indicators such as cavitation coverage area, interface evolution stability, or flow field instability, making it difficult to determine the true nature of the current operating state. As a result, even if drag temporarily decreases, if this is accompanied by severe shedding or structural oscillations, the model may still misjudge it as effective optimization, lacking not only physical interpretability but also potential safety risks. Thus, the model cannot distinguish between temporarily low drag but impending instability and a truly stable and efficient operating state; the optimization process relies entirely on the statistical regularities of historical data, lacking interpretability of the physical process.

[0004] More seriously, this stateless optimization mechanism introduces significant safety risks. In actual operation, flow field disturbances, incoming flow fluctuations, or model extrapolation may cause the system to enter unexpected operating conditions. The black-box model cannot promptly identify intermediate states such as unstable levels or drag-reduction evolution levels, and will still blindly output the next set of parameters, even exacerbating flow field instability and causing structural vibration or cavitation damage. Furthermore, due to the lack of ability to determine the impact level of the current control behavior, the system cannot achieve dynamic termination (early stop), safe rollback, or strategy adjustment of candidate parameter combinations, resulting in an uncontrollable optimization process. Summary of the Invention

[0005] In view of this, the present disclosure provides an AI control method for drag reduction of a natural-ventilated hybrid cavitation hydrofoil, which at least partially solves the problems existing in the prior art.

[0006] AI control methods for drag reduction of natural-ventilated hybrid cavitation hydrofoils include:

[0007] Step 1: Perform the cavitation flow experiment to obtain the input and output datasets;

[0008] Step 2: Based on the input and output datasets, feature extraction is performed to obtain cavitation coverage area ratio, flow field instability characteristic parameters, interface evolution stability index, and signal cavitation index, which are then integrated into a response dataset.

[0009] Step 3: Based on the input dataset, output dataset, and response dataset, construct a comprehensive data source set;

[0010] Step 4: Based on the response dataset, determine the impact level of the current ventilation behavior, and based on the impact level and the comprehensive data source set, determine the optimal design scheme.

[0011] Furthermore, the input dataset includes: incoming flow velocity, static pressure of the working section, and angle of attack; the output dataset includes: a first image sequence, a second image, a third image, a fourth image, and a resistance timing signal.

[0012] Furthermore, feature extraction is performed based on the input and output datasets to obtain cavitation coverage area ratio, flow field instability characteristic parameters, interface evolution stability index, and signal cavitation index, which are then integrated into a response data set, including:

[0013] Step 21: Based on each frame of the first image in the first image sequence, extract the cavitation region and determine the cavitation projection area based on the cavitation region; determine the hydrofoil chord length and hydrofoil span based on the cavitation flow experiment, and obtain the cavitation coverage area ratio based on the cavitation projection area, hydrofoil chord length and hydrofoil span.

[0014] Step 22: Based on the second image, extract the vorticity modulus length, and determine the instability characteristic parameters of the flow field based on the vorticity modulus length;

[0015] Step 23: Determine the gas phase volume fraction gradient vector field based on the third and fourth images, obtain the absolute angle change based on the gas phase volume fraction gradient vector field, and determine the interface evolution stability index based on the absolute angle change.

[0016] Step 24: Perform time-frequency analysis on the drag time series signal to obtain the signal cavitation index and drag coefficient. Integrate the cavitation coverage area ratio, flow field instability characteristic parameters, interface evolution stability index and signal cavitation index into a response data set.

[0017] Furthermore, the product term of hydrofoil chord length and hydrofoil span is determined, and the ratio of cavitation projected area to the product term is marked as cavitation coverage area ratio.

[0018] Furthermore, the maximum value of the vorticity modulus is marked as a characteristic parameter of the flow field instability.

[0019] Furthermore, based on the input dataset, output dataset, and response dataset, a comprehensive data source set is constructed, including:

[0020] Step 31: Based on the current cavitation flow experiment, determine several experimental parameters, and determine a set of structural parameters based on the experimental parameters; wherein, the experimental parameters include: vent number, vent size, and distribution pattern;

[0021] Step 32: Label the input dataset as an environment parameter group;

[0022] Step 33: Determine the cavitation evolution intensity index based on the cavitation coverage area ratio, flow field instability characteristic parameters, interface evolution stability index, and signal cavitation index.

[0023] Step 34: Based on the structural parameter group, environmental parameter group, cavitation evolution intensity index, and response data group, obtain the comprehensive data source set.

[0024] Further, assess the impact level of the current ventilation behavior, including:

[0025] Step 41: Determine the change in cavitation evolution intensity index between the current cavitation shedding cycle and the previous cavitation shedding cycle. If the change in cavitation evolution intensity index is greater than +0.05, the impact level of the current ventilation behavior is determined to be unstable, the first image of the next cavitation shedding cycle is taken, and the process returns to step 2. If the change in cavitation evolution intensity index is less than -0.05, the impact level of the current ventilation behavior is determined to be drag reduction evolution, and the next step is executed.

[0026] Step 42: Retrieve similar historical cavitation shedding cycles from the historical database, and determine the rate of change of interface evolution stability index and the mean rate of change of interface evolution stability index based on similar historical cavitation shedding cycles.

[0027] Step 43: Determine the current rate of change of the interface evolution stability index based on the current interface evolution stability index;

[0028] Step 44: If the current rate of change of the interface evolution stability index is less than the average rate of change of the interface evolution stability index, and the drag coefficient is less than the first target value, then the impact level of the current ventilation behavior is determined to be effective; otherwise, take a new image of the first image of the next cavitation shedding cycle and return to step 2.

[0029] Furthermore, based on the impact level and the comprehensive data source set, the optimal design scheme is determined, which is an optimized combination of optimized ventilation volume and optimized angle of attack:

[0030] Initial ventilation rate determined based on cavitation flow experiments;

[0031] Determine the ventilation adjustment volume, and then determine the optimal ventilation volume based on the initial ventilation volume and the ventilation adjustment volume;

[0032] Determine the angle of attack to be adjusted, and then determine the optimal angle of attack based on the original angle of attack and the adjusted angle of attack.

[0033] The optimized angle of attack and optimized ventilation volume are used as candidate optimization combinations, and the next cavitation shedding cycle is run based on the candidate optimization combinations;

[0034] Based on each cavitation shedding cycle, acquire the input and output data and return to execute step 2;

[0035] Extract the drag coefficients for n consecutive cavitation shedding cycles, determine the change in drag coefficients between adjacent cavitation shedding cycles, and determine whether the drag coefficients show a continuous downward trend and whether the change in drag coefficients is within the target range. If so, mark the candidate optimization combination as the optimization combination; otherwise, end the process.

[0036] Further, determine the ventilation adjustment volume, and determine the optimal ventilation volume based on the initial ventilation volume and the ventilation adjustment volume; determine the adjustment angle of attack, and determine the optimal angle of attack based on the angle of attack and the adjustment angle of attack, including:

[0037] If the impact level is effective, mark the ventilation adjustment amount and the angle of attack adjustment as 0, mark the initial ventilation amount as the optimized ventilation amount, and mark the angle of attack as the optimized angle of attack;

[0038] If the impact level is drag reduction evolution level, mark the ventilation adjustment amount as +3, mark the adjusted angle of attack as +3°, mark the optimized ventilation amount as the initial ventilation amount +3, and mark the optimized angle of attack as the angle of attack +3°.

[0039] If the impact level is unstable, mark the ventilation adjustment amount as +15, mark the adjusted angle of attack as -0.5°, mark the optimized ventilation amount as the initial ventilation amount +15, and mark the optimized angle of attack as the angle of attack -0.5°;

[0040] If the impact level is transitional, mark the ventilation adjustment amount as +8, the angle of attack adjustment as +2°, the optimized ventilation amount as initial ventilation amount +8, and the optimized angle of attack as angle of attack +2°.

[0041] This invention also provides an AI control system for the drag reduction effect of a natural-ventilated hybrid cavitation hydrofoil, comprising the following modules:

[0042] The experimental module is used to perform cavitation flow experiments and obtain input and output datasets.

[0043] The feature extraction module is used to extract features based on the input and output datasets to obtain cavitation coverage area, flow field instability characteristic parameters, interface evolution stability index, and signal cavitation index, and integrate them into the response dataset.

[0044] The data collection module constructs a comprehensive data source set based on the input dataset, output dataset, and response dataset;

[0045] The optimization module determines the impact level of the current ventilation behavior based on the response dataset, and then determines the optimal design scheme based on the impact level and the comprehensive data source set.

[0046] This disclosure provides an AI control method and system for drag reduction of a natural-ventilated hybrid cavitation hydrofoil. By constructing a comprehensive dataset including input parameters and multi-source output responses, it achieves dynamic perception of the cavitation flow field evolution process and determines its impact level based on the flow field response characteristics of the current cycle, thereby generating a targeted optimized control scheme. This technical mechanism overcomes the limitations of traditional static black-box models that rely solely on input-output relationships and lack process feedback, effectively distinguishing between temporarily low drag but impending instability and truly stable and efficient operating states.

[0047] This application extracts dynamic features such as cavitation coverage area, interface evolution stability index, flow field instability characteristic parameters, and signal cavitation index, and combines them with current input parameters to form a comprehensive data source set, thereby determining the impact level of ventilation behavior in the current cycle. For example, when drag decreases briefly but is accompanied by violent interface oscillation, rapid increase in vorticity, and divergence in frequency domain energy distribution, the current impact level is determined to be unstable, identifying it as a temporary low drag but imminent instability. Conversely, if drag continues to decrease and cavitation coverage expands stably, the interface is smooth, and vorticity is controlled, it is determined to be in a drag-reducing evolution level or an effective level, confirming that it is in a truly stable and efficient evolution path, making optimization controllable and improving safety.

[0048] Based on the impact level of the current cycle, this application generates ventilation adjustment and angle of attack adjustment strategies for the next shedding cycle: immediate intervention for unstable stages, either by reverting or enhancing stabilizing excitation; and continuous advancement for drag-reducing evolution stages. The current parameter combination is only considered an optimal solution when multiple consecutive cycles reach the effective level or when drag is continuously and controllably reduced. Attached Figure Description

[0049] Figure 1 This is a flowchart of the drag reduction AI control method for the natural-ventilated hybrid cavitation hydrofoil provided in this embodiment;

[0050] Figure 2 This is a structural block diagram of the drag reduction effect AI control system for the natural-ventilated hybrid cavitation hydrofoil provided in this embodiment;

[0051] Figure 3 This is a schematic diagram of the structure of the electronic device provided in the embodiments of this application.

[0052] In the accompanying drawings, unless otherwise specified, the same reference numerals throughout the various drawings denote the same or similar parts or elements. These drawings are not necessarily drawn to scale. It should be understood that these drawings depict only some embodiments disclosed in this application and should not be construed as limiting the scope of this application. Detailed Implementation

[0053] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0054] like Figure 1 As shown, the drag reduction AI control method for the natural-ventilated hybrid cavitation hydrofoil provided in this application includes:

[0055] Step 1: Perform the cavitation flow experiment to obtain the input and output datasets.

[0056] The input dataset includes: incoming flow velocity, static pressure in the working section, and angle of attack; the output dataset includes: a first image sequence, a second image, a third image, a fourth image, and a resistance timing signal.

[0057] In this embodiment, the cavitation flow experiment was conducted in a closed-loop circulating water tunnel. The experimental section measures 0.3 m (width) × 0.3 m (height) × 1.0 m (length). The incoming flow velocity can be continuously adjusted within the range of 0.5–18 m / s, the turbulence intensity is below 0.5%, and a pressure regulating box and degassing device are provided to precisely control the static pressure of the working section and remove dissolved gases. The experimental model is a NACA66 series hydrofoil with a chord length of 150 mm and a span of 300 mm. It is cantilevered in the middle of the experimental section by lateral support rods, 150 mm from the lower wall. The angle of attack can be precisely set within the range of 8°–18° via an external adjustment mechanism. The incoming flow velocity, working section static pressure, and angle of attack in the input dataset are all physical quantities measured in real time during the experiment: the incoming flow velocity is calibrated by an upstream standard Pitot tube and differential pressure sensor; the working section static pressure is directly acquired by a high-precision pressure sensor installed 5 times the chord length upstream of the model's leading edge; and the angle of attack is confirmed by encoder feedback. The hydrofoil's leading edge features a replaceable ventilation module containing multiple sets of ventilation holes with different diameters (1–3 mm) and arrangements (single hole, linear multi-hole array, and ring array). Specific experimental parameters include the hole number (used to identify the hole layout, e.g., 1# is a central single hole, 2#–6# are spanwise distributed multi-hole arrays), hole diameter (1.0 mm, 1.5 mm, or 2.0 mm), and distribution method (centralized, linear, or ring array). These parameters collectively constitute the structural parameter set and serve as part of the input characteristics. The ventilation volume is regulated by a mass flow controller (MFC) and synchronously recorded by a data acquisition system. The output dataset was acquired through a multiphysics synchronous measurement system: the first image sequence is a high-speed photographic image, taken from the side of the experimental section using a Phantom V2012 high-speed camera to capture the dynamic evolution of cavitation; the second image is the instantaneous velocity vector distribution of the flow field, obtained by using a dual-cavity Nd:YAG laser (wavelength 532 nm) in conjunction with a PIV sheet light source and a synchronous controller, combined with fluorescent tracer particles and cross-correlation algorithms to obtain the liquid phase velocity field; the third image is the gas phase concentration distribution, obtained by using an X-ray density imaging system (based on a synchrotron radiation source or a miniature X-ray source) to penetrate the middle section of the hydrofoil and inverting the gas phase volume fraction using the absorption intensity; the fourth image is a local density 3D reconstruction image, taken synchronously from orthogonal directions by two high-speed X-ray cameras, combined with a tomographic reconstruction algorithm to obtain the internal structure of the cavitation cloud; the drag time series signal is acquired in real time by a six-axis force sensor to collect the three-dimensional forces and torques acting on the hydrofoil.

[0058] Step 2: Based on the input and output datasets, feature extraction is performed to obtain cavitation coverage area ratio, flow field instability characteristic parameters, interface evolution stability index, and signal cavitation index, which are then integrated into a response dataset.

[0059] Step 21: Extract the cavitation region based on each frame of the first image in the first image sequence, and determine the cavitation projection area based on the cavitation region; determine the hydrofoil chord length and hydrofoil span based on the cavitation flow experiment, and obtain the cavitation coverage area ratio based on the cavitation projection area, hydrofoil chord length and hydrofoil span.

[0060] First, each frame of the first image in the first image sequence is preprocessed, including background subtraction, contrast enhancement, and median filtering, to eliminate uneven lighting and noise interference. Then, a gray-scale threshold-based segmentation method is used, combined with the characteristic that cavitation regions appear as high-brightness areas in the image (due to the significant differences in light reflection and scattering between the gas and liquid phases), followed by image binarization to initially identify the pixel set of the cavitation region. To extract the cavitation region more accurately, edge detection algorithms (such as the Canny operator) are used to assist in identifying the cavitation boundary, and morphological operations (such as closing operations) are combined with post-processing to fill the holes inside the cavitation region and remove isolated noise points, thereby obtaining a continuous and complete cavitation contour. In high-speed camera images, cavitation typically manifests as a transparent or semi-transparent gas layer attached to the leading edge of the hydrofoil, with its trailing edge often accompanied by periodically detached cavitation clouds or microbubble structures; the overall morphology changes continuously with the development stage of cavitation.

[0061] Based on this, all pixels marked as cavitation in the processed binary image are statistically analyzed, and multiplied by the actual physical area corresponding to a single pixel to obtain the cavitation projection area. Further, using the product of the hydrofoil chord length and span as a reference area, the ratio of the cavitation projection area to this reference area is defined as the cavitation coverage area ratio. The product of the hydrofoil chord length and span is often used as an approximation of its surface area, serving as a benchmark for measuring cavitation coverage. In the experiment, cavitation images were acquired through high-speed photography, and the projected area of ​​the cavitation region on a two-dimensional plane was extracted after image processing; this is the cavitation projection area. This area reflects the actual extent of gas phase adhering to the hydrofoil surface. Dividing the cavitation projection area by the product of the chord length and span yields the cavitation coverage area ratio. This ratio intuitively expresses the coverage proportion of the cavitation structure across the entire hydrofoil surface, providing a quantitative description of the degree of cavitation development. Compared to traditional methods that rely solely on single indicators such as drag coefficient or cavitation length to judge drag reduction effects, the cavitation coverage area ratio provides more direct and comprehensive information. For example, under certain operating conditions, although drag may decrease, if the cavitation coverage ratio remains low, it indicates that the gas phase has not fully covered the hydrofoil surface, and the drag reduction mechanism may not have been stably established. In this case, the low drag state may not be reliable or sustainable. Therefore, introducing the cavitation coverage ratio helps to more accurately identify the truly effective drag reduction state and avoid misjudgments due to over-reliance on a single indicator. A higher cavitation coverage ratio usually indicates that a large area of ​​low-density gas phase has formed on the hydrofoil surface, which helps to effectively isolate the water flow from the wall, thereby significantly reducing frictional drag. This indicator itself has a clear physical meaning, is calculated directly, and does not rely on specific equipment or complex calibration procedures, thus exhibiting good comparability across different experimental platforms or studies. More importantly, the cavitation coverage ratio can reflect the actual state of cavitation development during control, compensating for the shortcomings of relying solely on terminal indicators such as drag to capture intermediate evolutionary stages. With this indicator, different situations such as optimized state transition, stable drag reduction, insufficient coverage, and limited effect can be more precisely distinguished, providing a reliable basis for adjusting ventilation strategies. In complex and unsteady cavitation flow fields, this ability to determine the state based on physical observation is a key support for achieving safe, efficient and intelligent control.

[0062] Step 22: Based on the second image, extract the vorticity modulus and determine the instability characteristic parameters of the flow field based on the vorticity modulus.

[0063] Using the liquid-gas two-phase velocity field obtained in step 1, the instantaneous velocity vector distribution at each point in the flow field can be obtained. Based on this, the velocity field is spatially differentiated using the finite difference method to calculate the vorticity vector at each location, thus constructing the vorticity vector field for the entire observation area. The modulus of this vector field is then taken to form the vorticity modulus. In actual flow fields, high values ​​of the vorticity modulus are usually concentrated at the trailing edge of cavitation clouds, the shear layer, and the wake vortex region. These locations are precisely where shearing is intense and turbulent fluctuations are active, often indicating impending flow instability or drastic structural changes. To facilitate the quantification of the overall flow field instability, the maximum value of the vorticity modulus is selected as the flow field instability characteristic parameter. This parameter effectively reflects the strongest vortex intensity in the current flow field and is an important basis for judging whether the system is approaching an unstable state.

[0064] Step 23: Determine the gas phase volume fraction gradient vector field based on the third and fourth images, obtain the absolute angle change based on the gas phase volume fraction gradient vector field, and determine the interface evolution stability index based on the absolute angle change.

[0065] This embodiment integrates two measurement methods: laser-induced fluorescence (LIF) and X-ray density field. A calibration plate is used to spatially register the LIF image and the X-ray 3D reconstructed image. The LIF image reflects the instantaneous gas phase concentration distribution, while the X-ray data provides high-resolution local density information. Through rigid body transformation combined with affine correction, the two sets of images are aligned in a unified coordinate system, ensuring that the same physical location corresponds consistently in both images. After registration, at each time step, the local gas phase volume fraction is calculated based on the fused data, and a finite difference is performed spatially to obtain the gradient vector at that point. This gradient direction represents the direction of the most drastic change in gas phase concentration and can be approximated as the normal direction of the interface at that point, thus constructing the gradient vector field for the entire field. Principal component analysis (PCA) is performed on the entire gradient vector field at the current time, and the direction corresponding to the first principal component is extracted and denoted as the principal direction of interface evolution at that time. This direction represents the main trend of the current interface as a whole, whether it tends to expand, contract, or deflect. The above process is repeated over 10 consecutive time steps (t = 1 to t = 10) to obtain 10 principal direction angles. Then, the angle differences between adjacent principal directions are calculated sequentially, their absolute values ​​are taken, and the average is calculated. The result is defined as the interface evolution stability index. The smaller the index value, the more stable the interface evolution; if the value is large, it indicates frequent oscillations in the interface direction, and the flow field may tend to be unstable. This method provides a quantitative criterion for cavitation stability from the perspective of interface geometric evolution, avoiding the limitations of relying solely on macroscopic drag or empirical judgments.

[0066] Step 24: Perform time-frequency analysis on the drag time series signal to obtain the signal cavitation index and drag coefficient. Integrate the cavitation coverage area ratio, flow field instability characteristic parameters, interface evolution stability index and signal cavitation index into a response data set.

[0067] To investigate the dynamic behavior of hydrofoils in high-speed cavitation flow, we first employed a six-axis force sensor to monitor their three-dimensional force in real time. For the drag component in the incoming flow direction, we performed a series of preprocessing steps, including eliminating DC bias, applying a Hanning window to reduce spectral leakage, and using a db6 wavelet basis for wavelet denoising to remove high-frequency noise.

[0068] A time-frequency joint analysis was performed on the preprocessed drag time-series signal. On one hand, a power spectral density map was generated by calculating the short-time Fourier transform (STFT) to identify the main frequencies and their energy distribution. Particularly under mixed natural and ventilated cavitation conditions, significant energy peaks were observed in the 50-150 Hz range, reflecting the characteristic frequencies of periodic detachment of cavitation clouds. The energy integral value in this frequency band was defined as the signal cavitation index, used to measure the intensity and regularity of cavitation activity. On the other hand, the drag coefficient was obtained by normalizing the average drag signal over the entire time period and dividing it by the product of the dynamic pressure (i.e., 1 / 2ρU², where ρ is the fluid density and U is the inflow velocity) and the chord length and span. This method not only provides macroscopic performance parameters but also captures important dynamic characteristics reflecting flow field instability.

[0069] This embodiment integrates cavitation coverage area ratio, flow field instability characteristic parameters, interface evolution stability index, and signal cavitation index to construct a comprehensive response dataset. This dataset enables a shift from static result perception to dynamic state assessment, surpassing traditional methods that rely solely on single macroscopic indicators such as drag coefficient. The cavitation coverage area ratio quantifies the extent of gas phase expansion on the hydrofoil surface, directly impacting drag reduction potential; flow field instability characteristic parameters reveal instability indicators such as shear layer disturbances and enhanced turbulence; the interface evolution stability index, utilizing principal component analysis, demonstrates the directionality and continuity of the overall gas-liquid interface motion, helping to identify non-ideal states such as periodic oscillations or deflections; and the signal cavitation index provides time-frequency domain information on the regularity and energy intensity of cavitation shedding, aiding in determining whether the system is in a highly efficient and stable operating state. It can distinguish between a temporarily low-drag state that may be on the verge of instability and a truly stable and efficient operating state, thus providing a comprehensive and quantitative evaluation framework.

[0070] Step 3: Construct a comprehensive data source set based on the input dataset, output dataset, and response dataset.

[0071] Step 31: Based on the current cavitation flow experiment, determine several experimental parameters, and determine a set of structural parameters based on the experimental parameters; wherein, the experimental parameters include: vent number, vent size, and distribution pattern.

[0072] Step 32: Label the input dataset as an environment parameter group.

[0073] Step 33: Determine the cavitation evolution intensity index based on the cavitation coverage area ratio, flow field instability characteristic parameters, interface evolution stability index, and signal cavitation index.

[0074] The cavitation evolution intensity index is used to quantify the overall evolution intensity of cavitation from its inception to its development into a stable or unstable state.

[0075] The expression for the cavitation evolution intensity index is as follows:

[0076] ;

[0077] In the formula, CDI is the cavitation evolution intensity index, A is the cavitation coverage area ratio, and A c The baseline value for the cavitation area to be completely covered (the theoretical maximum area that can be covered under experimental conditions, i.e. the area when the entire leading edge region is completely covered by cavitation). This represents the maximum value of the vortex modulus. The critical vorticity threshold (the critical vorticity level that triggers significant nonlinear instability (such as cloud cavitation shedding)). The absolute average value of the angle change is denoted as F, which is the signal cavitation index, and F0 is the reference signal cavitation index. w1, w2, w3 and w4 are weighting coefficients, all of which range from 0 to 1.

[0078] Step 34: Based on the structural parameter group, environmental parameter group, cavitation evolution intensity index, and response data group, obtain the comprehensive data source set.

[0079] The structural parameter set reflects the physical configuration of the hydrofoil leading-edge ventilation device, determining the location, scale, and momentum distribution of the initial gas injection. These are key design variables influencing the cavitation initiation and development path. The environmental parameter set characterizes external flow conditions, directly affecting the cavitation number and boundary layer development state, and represents the boundary input for system operation. This embodiment also introduces a cavitation evolution intensity index as a core intermediate state variable, achieving comprehensive quantification of cavitation development degree, interface stability, and flow pulsation. Furthermore, the time-series information of each original feature is preserved, unifying these four types of data into a comprehensive data source set. This not only achieves end-to-end data connectivity from static input to dynamic response to state assessment, but more importantly, it breaks through the limitations of black-box mapping in traditional data-driven models, making the optimization process physically interpretable: it can identify whether a drag reduction effect is caused by a stable expanding cavitation layer or by a brief but intense shedding event; it can also determine whether a parameter combination, while effective in the short term, is approaching the instability boundary. This fusion mechanism provides sufficient basis for the accurate determination of subsequent impact levels.

[0080] Step 4: Based on the response dataset, determine the impact level of the current ventilation behavior, and based on the impact level and the comprehensive data source set, determine the optimal design scheme.

[0081] Step 41: Determine the change in cavitation evolution intensity index between the current cavitation shedding cycle and the previous cavitation shedding cycle. If the change in cavitation evolution intensity index is greater than +0.05, the impact level of the current ventilation behavior is determined to be unstable, the first image of the next cavitation shedding cycle is captured, and the process returns to step 2. If the change in cavitation evolution intensity index is less than -0.05, the impact level of the current ventilation behavior is determined to be drag reduction evolution level, and the next step is executed. If the change in cavitation evolution intensity index is greater than or equal to -0.05 and less than or equal to +0.05, the impact level of the current ventilation behavior is determined to be initial evolution level, and ventilation volume is not optimized.

[0082] First, two complete cavitation shedding cycles are continuously monitored (e.g., each cycle lasts approximately 20 milliseconds, corresponding to a dominant frequency of 50 Hz), and the average value of the cavitation evolution intensity index within each shedding cycle is calculated. In actual operation, this embodiment uses changes in the cavitation evolution intensity index to determine the impact level of the current ventilation behavior. For example, if the index value in the previous cavitation shedding cycle was 0.72, and it rises to 0.78 in the current cycle, the change is +0.06. Since this value exceeds the preset positive threshold of +0.05, this embodiment classifies the current state as unstable. This judgment means that the cavitation structure is rapidly intensifying, often accompanied by violent shedding of cavitation clouds, large interface oscillations, or a significant increase in vorticity, posing a risk of flow field instability. To prevent further parameter adjustments in an unstable state from inducing structural resonance or exacerbating cavitation damage, this embodiment does not immediately execute the next parameter optimization step. Instead, it triggers high-speed photography to capture a complete sequence of dynamic cavitation images in the next cycle for further analysis of interface morphology and shedding characteristics. Only after confirming that the flow field has stabilized is the optimization process restarted. Conversely, if the cavitation evolution intensity index of the current cycle decreases from 0.68 to 0.62, a change of -0.06 (less than -0.05), it is determined to be a drag-reduction evolution stage. This indicates that the cavitation coverage is expanding in an orderly manner, the flow structure is stabilizing, and it has good drag-reduction potential. At this point, the system will execute the next parameter adjustment according to the strategy.

[0083] If the indicator changes only slightly, for example, from 0.70 to 0.71, the change is only +0.01, falling within the range of [-0.05, +0.05], then it is classified as the initial evolution stage. This embodiment is considered to be in a stable transition phase, and the ventilation volume is not adjusted for the time being to avoid control oscillations caused by frequent intervention.

[0084] The aforementioned ±0.05 threshold is not arbitrarily set, but is based on statistical analysis of a large amount of experimental data: when the index changes beyond this range, PIV and X-ray imaging results generally show significant changes in the flow field structure, such as a gas-liquid interface deflection angle exceeding 10°, or a peak vorticity increase exceeding 20%. Step 42: Based on the historical database, similar historical cavitation shedding cycles are retrieved, and the rate of change of the interface evolution stability index and the mean rate of change of the interface evolution stability index are determined based on these similar historical cavitation shedding cycles.

[0085] Assuming the experimental conditions corresponding to the current cavitation shedding cycle are: hydrofoil vent number 5#, vent diameter 1.5 mm, linear distribution (structural parameter group); incoming flow velocity 12.0 m / s, static pressure in the working section 98 kPa, angle of attack 14° (environmental parameter group); the calculated cavitation evolution intensity index CDI = 0.78.

[0086] Retrieve historical periods from the historical database that meet the following three conditions:

[0087] (1) The structural parameters are completely consistent with the current experimental conditions.

[0088] (2) The deviation of environmental parameters from the current environmental parameter set is within the preset range (such as incoming flow velocity ±0.3 m / s, working section static pressure ±3 kPa, angle of attack ±0.5°).

[0089] (3) The absolute value of the difference between the historical CDI value and the current CDI is less than the preset threshold (set to 0.03 after calibration).

[0090] Among the historical cycles that meet the above conditions, the three cycles with the smallest CDI differences are selected as similar samples. For example, the detected historical cycles A (CDI=0.76), B (CDI=0.77), and C (CDI=0.79) have absolute deviations of CDI from the current value of 0.02, 0.01, and 0.01, respectively, all less than 0.03, and their structures match the environmental parameters. For these three similar historical cycles, the interface evolution stability index sequence within each consecutive cycle is extracted. The rate of change of the index between adjacent cycles is calculated by combining the historical cycle step size, and the average value is obtained.

[0091] Step 43: Determine the current rate of change of the interface evolution stability index based on the current interface evolution stability index.

[0092] Based on the current interface evolution stability index and the interface evolution stability index of the previous shedding cycle, and combined with the cycle step size, the change rate of the interface evolution stability index is calculated.

[0093] Step 44: If the current rate of change of the interface evolution stability index is less than the average rate of change of the interface evolution stability index, and the drag coefficient is less than the first target value, then the impact level of the current ventilation behavior is determined to be effective; otherwise, take a new image of the first image of the next cavitation shedding cycle and return to step 2.

[0094] The first target value is the optimal drag coefficient that can be achieved by similar hydrofoils or aircraft under specific operating conditions (such as incoming flow velocity, angle of attack, etc.).

[0095] This embodiment marks the current interface evolution stability index change rate as less than the average interface evolution stability index change rate as a screening criterion for drag reduction potential in the current operating condition; and marks the drag coefficient as less than the first target value as a reference benchmark for the stability of the current operating condition. If the reference benchmark for the stability of the current operating condition is valid, it is determined that the interface evolution trend is under control and the flow field structure is continuously stable, belonging to a truly stable and efficient state; conversely, if the current interface evolution stability index change rate is greater than or equal to the average interface evolution stability index change rate, even if the drag coefficient is lower than the first target value, it is considered a high-risk state of temporarily low drag but imminent instability. This overcomes the limitation of traditional optimization methods that rely solely on instantaneous drag indices, leading to misjudgments. By introducing a dynamic trend reference mechanism based on historically similar operating conditions, it achieves interpretable assessment of the cavitation evolution process and early identification of instability precursors. On the one hand, it avoids misjudging a brief low-resistance state as a successful optimization and thus misadjusting parameters; on the other hand, it effectively identifies dangerous operating conditions where the interface is fluctuating rapidly and about to become unstable, even though the resistance is low, by trend comparison, triggering a reshoot and reassessment mechanism to prevent the system from entering the resonance or cavitation damage area.

[0096] In actual operation, even if the current drag coefficient is lower than the set first target value, the system will not determine that drag reduction is successful based solely on this. A comprehensive judgment must also be made in conjunction with the rate of change of the interface evolution stability index. For example, if this rate of change is greater than or equal to the average value under similar historical operating conditions, it indicates that although the drag is temporarily low, the gas-liquid interface may be accelerating its deflection, its oscillation is significantly intensified, or even showing signs of periodic rupture. This is a typical manifestation of temporarily low drag but impending instability. This state usually occurs when the ventilation volume is too large or the angle of attack is too high: although the cavitation coverage area is large, the overall flow field tends to be unstable due to enhanced backflow and turbulent vortex structure. In this case, the system will not classify it as an "effective level," but will trigger a high-speed camera to capture the cavitation dynamic image for the next cycle and return to step 2 to re-extract features and evaluate the state, to prevent mistakenly identifying a temporary performance improvement as successful optimization.

[0097] Conversely, if the drag coefficient is lower than the first target value, and the rate of change of the interface evolution stability index is less than the average of similar historical operating conditions, it indicates that during the continuous expansion of the cavitation structure, the interface movement is actually more stable than before, without any abnormal disturbances or violent oscillations. This situation suggests that not only has drag reduction been achieved, but the system is also in a more orderly operating state than historical stable operating conditions, which can be reliably identified as a truly stable and efficient optimization range, thus supporting subsequent parameter solidification or further fine-tuning.

[0098] This embodiment follows a clear priority order for determining and responding to impact levels: Unstable Level > Drag Reduction Evolution Level > Effective Level. Once the conditions for determining the unstable level are met, such as an excessively high rate of change in the interface evolution stability index, a surge in vorticity, or violent oscillation of the cavitation structure, the current optimization process is immediately halted regardless of whether the current drag has decreased. A high-speed camera is then triggered to capture cavitation images for the next cycle, and the process re-enters the state assessment phase. This approach reflects the principle of safety first. Only after eliminating unstable states will the next level of judgment proceed. If the current operating condition meets the characteristics of the drag reduction evolution level (e.g., continuously expanding cavitation coverage area, decreasing drag, and stable interface evolution), it is considered to be in an active phase transitioning to the optimized operating condition. At this point, a preset parameter fine-tuning strategy is allowed to be executed, pushing the process closer to the ideal state. While the effective level is the final goal, it is ranked last in the judgment logic because it requires simultaneously meeting two strict conditions: first, the drag coefficient must be lower than the first target value; second, the rate of change of the interface evolution stability index must be better than the average level of similar historical operating conditions. This means that not only must the drag be low, but the flow field structure must also be more stable than before. Only when both of these requirements are met can the current parameter combination be considered a valid optimization solution.

[0099] The optimal design scheme is an optimized combination of optimized ventilation volume and optimized angle of attack.

[0100] The optimal design scheme in this application is implemented through a pre-trained neural network model. This model takes a comprehensive data source set as input, predicts the impact level of ventilation behavior, and automatically optimizes ventilation rate and angle of attack based on the prediction results, ultimately outputting an optimized combination. The training set of the model comes from a large amount of historical cavitation flow experimental data. Each sample consists of input features and labeled impact level tags. The input features are the comprehensive data source set constructed in this scheme, including: structural parameter set, environmental parameter set, cavitation evolution intensity index, and response data set. The impact level label (as a prediction label) is obtained through a combination of manual and rule-based labeling of the dynamic evolution behavior during the experiment. Specifically, based on the CDI change, the trend of the interface evolution stability index, and the drag coefficient performance in each cavitation shedding cycle, backtracking labeling is performed according to the judgment logic set in this scheme: if the CDI rises by more than +0.05, it is labeled as unstable; if the CDI falls by more than -0.05, it is labeled as drag reduction evolution; if the CDI change is in the range of [-0.05, +0.05], it is labeled as initial evolution; for stable low-drag states that meet the condition of drag coefficient being lower than the first target value and interface evolution trend being better than the historical average, it is labeled as effective. During training, supervised learning is used, and the cross-entropy loss function is used to optimize the neural network weights, ultimately obtaining a classification model that can accurately predict the impact level.

[0101] The initial ventilation rate was determined based on cavitation flow experiments.

[0102] In this embodiment, the initial ventilation rate relies on a pre-constructed recommended ventilation rate mapping table, which is calibrated based on a large amount of cavitation flow experimental data. The mapping table is constructed as follows: First, based on cavitation flow experiments, typical combinations of incoming flow velocity, static pressure in the working section, angle of attack, and vent structure are covered. For each set of experimental conditions, the ventilation rate is gradually increased, and the evolution of cavitation morphology in high-speed photographic images is observed. The critical point for the transition from natural cavitation to stable ventilated cavitation is recorded: that is, the minimum ventilation rate at which the cavitation cloud first continuously covers more than 70% of the leading edge without violent shedding, which is recorded as the baseline initial ventilation rate under this condition. For example, under the conditions of incoming flow velocity of 12 m / s, static pressure of 98 kPa, angle of attack of 14°, vent diameter of 1.5 mm, and linear distribution, the baseline value is measured to be 1.1 L / min. All experimental combinations and their corresponding baseline values ​​are compiled into a multidimensional lookup table. The dimensions include: cavitation number range (e.g., 0.6–0.8, 0.8–1.0, 1.0–1.2), angle of attack range (e.g., 10°–12°, 12°–14°, 14°–16°), and vent type (coded 1–6). In actual operation, the baseline ventilation rate is obtained by looking up the table based on the currently measured in-flow velocity, static pressure, angle of attack, and vent number.

[0103] Determine the ventilation adjustment volume, and based on the initial ventilation volume and the ventilation adjustment volume, determine the optimal ventilation volume; determine the adjustment angle of attack, and based on the angle of attack and the adjustment angle of attack, determine the optimal angle of attack.

[0104] If the predicted impact level is effective, mark the ventilation adjustment amount and the adjusted angle of attack as 0, mark the initial ventilation amount as the optimized ventilation amount, and mark the angle of attack as the optimized angle of attack.

[0105] The determination of the effective level has been verified by two stringent conditions: firstly, the drag coefficient is indeed lower than the first target value, indicating that the drag reduction effect has been achieved; secondly, the rate of change of the interface evolution stability index is lower than the average of similar historical operating conditions, indicating that the current gas-liquid interface evolution is more stable than before, without accelerated oscillation, periodic fracture, or other signs of instability. This shows that the system not only achieves low drag but is also in a structurally stable and orderly evolving state, possessing good repeatability and engineering applicability. In this case, further adjusting the ventilation rate or angle of attack may disturb the established equilibrium, for example, by intensifying cavitation cloud shedding, enhancing backflow, or even causing the flow field to slide into an unstable region. Therefore, the most reasonable approach is not further optimization, but to keep the current parameters unchanged and allow the system to continue operating under this condition. This "stability is optimal" strategy avoids sacrificing overall reliability for the sake of marginal performance improvement, reflecting the priority consideration of state robustness in closed-loop control.

[0106] If the impact level is drag reduction evolution level, mark the ventilation adjustment amount as +3, mark the adjusted angle of attack as +3°, mark the optimized ventilation amount as the initial ventilation amount +3, and mark the optimized angle of attack as the angle of attack +3°.

[0107] When the impact level is determined to be the drag reduction evolution stage, the ventilation adjustment amount is marked as +3 (unit: mL / min or standardized flow unit), the angle of attack adjustment is marked as +3°, and the optimized ventilation amount is determined as the initial ventilation amount +3, and the optimized angle of attack is determined as the current angle of attack +3°. The drag reduction evolution stage indicates that the current operating condition is developing in a favorable direction: the cavitation coverage area continues to expand, the flow field tends to stabilize, and the drag coefficient shows a decreasing trend, but it has not yet reached the optimal stable state and has the potential to further improve the drag reduction effect.

[0108] If the impact level is unstable, mark the ventilation adjustment amount as +15, the adjusted angle of attack as -0.5°, the optimized ventilation amount as the initial ventilation amount +15, and the optimized angle of attack as the angle of attack -0.5°.

[0109] When the current ventilation behavior is determined to be of an unstable level, a specific parameter adjustment strategy is implemented: increasing the initial ventilation rate by 15 mL / min while decreasing the current angle of attack by 0.5°. The unstable level is characterized by frequent and violent detachment of cavitation clouds, large-scale oscillations at the gas-liquid interface, a significant increase in vorticity, or a sudden increase in the intensity of the backflow. These phenomena indicate that the flow field has entered a highly unsteady state. If the ventilation rate is abruptly reduced at this point, the cavitation structure may collapse rapidly, causing the water flow to directly impact the hydrofoil surface, which would exacerbate pressure pulsations and flow disturbances. Conversely, moderately increasing the ventilation rate (e.g., +15 mL / min) can effectively replenish gas phase momentum, promoting the extension of the cavitation region to the forward edge and tending towards continuous attachment, thereby suppressing periodic detachment. Experiments have repeatedly observed that this gas-stabilized flow approach significantly improves interface stability; this phenomenon is called the ventilation stabilization effect.

[0110] Meanwhile, reducing the angle of attack by 0.5° can mitigate the adverse pressure gradient near the leading edge while maintaining lift performance, reducing the tendency for boundary layer separation and further decreasing the possibility of flow instability. These two adjustments work synergistically to quickly pull the system back from a high-risk state to a controllable range, creating a safe environment for subsequent optimization.

[0111] If the impact level is transitional, mark the ventilation adjustment amount as +8, the angle of attack adjustment as +2°, the optimized ventilation amount as initial ventilation amount +8, and the optimized angle of attack as angle of attack +2°.

[0112] The ventilation rate was increased by 8 mL / min from the initial value, and the angle of attack was increased by 2° from the current value. This adjustment was not arbitrary, but based on the calibration results of a large number of hydrofoil gas-liquid two-phase flow experiments: under typical inflow conditions, an increase of 8 mL / min in the ventilation rate is sufficient to enhance the gas phase momentum, enabling the cavitation cloud to extend more effectively to the leading edge, without inducing strong backflow or interface instability due to excessive air supply; while an increase of 2° in the angle of attack can moderately strengthen the low-pressure region at the leading edge without causing boundary layer separation, which helps the cavitation region to expand in the chord and spanwise directions and improves its adhesion stability.

[0113] The angle of attack and ventilation volume were selected as candidate optimization combinations, and the next cavitation shedding cycle was run based on these candidate optimization combinations.

[0114] Based on each cavitation shedding cycle, obtain the input and output data and return to execute step 2.

[0115] If the influence level of the ventilation behavior is effective for n consecutive cavitation shedding cycles, then the candidate optimization combination is marked as the optimization combination.

[0116] Extract the drag coefficients for n consecutive cavitation shedding cycles, determine the change in drag coefficients between adjacent cavitation shedding cycles, and determine whether the drag coefficients show a continuous downward trend and whether the change in drag coefficients is within the target range. If so, mark the candidate optimization combination as the optimization combination; otherwise, end the process.

[0117] The target interval is [-0.05, 0].

[0118] In this embodiment, the parameter combination consisting of optimized ventilation rate and optimized angle of attack is used as a candidate optimization combination, and n consecutive cavitation shedding cycles are run based on this combination. After each cycle, drag coefficient time-series data is collected in real time, and the drag coefficient sequence of the n consecutive cycles is extracted. The change in drag coefficient between adjacent cycles is calculated to determine whether it is within the preset target range [-0.05, 0], and the drag coefficient is analyzed simultaneously to see if it shows a monotonically decreasing trend. If both conditions are met, the candidate optimization combination is considered to achieve a stable and continuous drag reduction effect, and it is marked as an optimization combination and the optimization process ends. If the drag change in any cycle exceeds the target range, or the rate of change of the interface evolution stability index shows a continuous upward trend, indicating that there is a risk of flow field instability, the operation of the current candidate optimization combination is immediately terminated, and the control parameters are rolled back to the initial ventilation rate or the previous optimized combination that has been verified as stable. Subsequently, based on the data obtained from the rerun cycle after the rollback, the influence level determination is re-executed, and a new ventilation adjustment rate and angle of attack are generated according to the determination result to form the next candidate optimization combination and put it into operation. The above process is repeated until an optimal combination that meets the judgment criteria is successfully identified, or the preset maximum number of attempts is reached. If no combination that meets the criteria is found after reaching the maximum number of attempts, the system outputs the candidate combination that achieves a continuous decrease in the drag coefficient with the change within the target range as the design optimization scheme.

[0119] like Figure 2 As shown, the present invention also provides an AI control system for the drag reduction effect of a natural-ventilated hybrid cavitation hydrofoil, comprising the following modules:

[0120] The experimental module is used to perform cavitation flow experiments and obtain input and output datasets.

[0121] The feature extraction module is used to extract features based on the input and output datasets to obtain cavitation coverage area, flow field instability characteristic parameters, interface evolution stability index and signal cavitation index, and integrate them into the response dataset.

[0122] The data collection module constructs a comprehensive data source set based on the input dataset, output dataset, and response dataset;

[0123] The optimization module determines the impact level of the current ventilation behavior based on the response dataset, and then determines the optimal design scheme based on the impact level and the comprehensive data source set.

[0124] To implement the above embodiments, this application also proposes an electronic device. Please see [link to relevant documentation]. Figure 3 , Figure 3 This is a schematic diagram of the structure of the electronic device provided in the embodiments of this application. For example... Figure 3As shown, the electronic device 500 includes: a processor 501 and a memory 502 communicatively connected to the processor 501; the memory 502 stores computer-executable instructions; the processor 501 executes the computer-executable instructions stored in the memory to implement the method provided in the foregoing embodiments.

[0125] To implement the above embodiments, this application also proposes a computer-readable storage medium storing computer-executable instructions, which, when executed by a processor, are used to implement the methods provided in the foregoing embodiments.

[0126] In the description of this specification, the references to terms such as "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of this application. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples. Moreover, without contradiction, those skilled in the art can combine and integrate the different embodiments or examples described in this specification, as well as the features of those different embodiments or examples.

[0127] Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated. Thus, a feature defined as "first" or "second" may explicitly or implicitly include at least one of that feature. In the description of this application, "a plurality of" means two or more, unless otherwise explicitly specified.

[0128] Any process or method description in the flowchart or otherwise herein can be understood as representing a module, segment, or portion of code comprising one or more executable instructions for implementing a particular logical function or process. Furthermore, the scope of the preferred embodiments of this application includes additional implementations in which functions may be performed not in the order shown or discussed, including substantially simultaneously or in reverse order depending on the functionality involved.

[0129] The logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as a sequenced list of executable instructions for implementing logical functions, and can be embodied in any computer-readable medium for use by, or in conjunction with, an instruction execution system, apparatus or device (such as a computer-based system, a processor-included system or other system that can fetch and execute instructions from, an instruction execution system, apparatus or device).

[0130] It should be understood that various parts of this application can be implemented using hardware, software, firmware, or a combination thereof. In the above embodiments, multiple steps or methods can be implemented using software or firmware stored in memory and executed by a suitable instruction execution system. All or part of the steps of the methods in the above embodiments can be implemented by a program instructing related hardware, the program being stored in a computer-readable storage medium, which, when executed, includes one or a combination of the steps of the method embodiments.

[0131] Furthermore, the functional units in the various embodiments of this application can be integrated into a processing module, or each unit can exist physically separately, or two or more units can be integrated into a module. The integrated module can be implemented in hardware or as a software functional module. If the integrated module is implemented as a software functional module and sold or used as an independent product, it can also be stored in a computer-readable storage medium. This storage medium can be a read-only memory, a disk, or an optical disk, etc.

[0132] The above are merely specific embodiments of this application, but the scope of protection of this application is not limited thereto. Any person skilled in the art can easily conceive of various variations or substitutions within the technical scope disclosed in this application, and these should all be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. An AI control method for drag reduction effect of a natural-ventilated hybrid cavitation hydrofoil, characterized in that, include: Step 1: Perform the cavitation flow experiment to obtain the input and output datasets; The input dataset includes: incoming flow velocity, static pressure in the working section, and angle of attack; the output dataset includes: a first image sequence, a second image, a third image, a fourth image, and a resistance timing signal; Step 2: Based on the input and output datasets, feature extraction is performed to obtain cavitation coverage area ratio, flow field instability characteristic parameters, interface evolution stability index, and signal cavitation index, which are then integrated into a response dataset, including: Step 21: Based on each frame of the first image in the first image sequence, extract the cavitation region and determine the cavitation projection area based on the cavitation region; determine the hydrofoil chord length and hydrofoil span based on the cavitation flow experiment, and obtain the cavitation coverage area ratio based on the cavitation projection area, hydrofoil chord length and hydrofoil span. Step 22: Based on the second image, extract the vorticity modulus and determine the instability characteristic parameters of the flow field based on the vorticity modulus. Step 23: Determine the gas phase volume fraction gradient vector field based on the third and fourth images, obtain the absolute angle change based on the gas phase volume fraction gradient vector field, and determine the interface evolution stability index based on the absolute angle change. Step 24: Perform time-frequency analysis on the drag time series signal to obtain the signal cavitation index and drag coefficient, and integrate the cavitation coverage area ratio, flow field instability characteristic parameters, interface evolution stability index and signal cavitation index into a response dataset. Step 3: Based on the input dataset, output dataset, and response dataset, construct a comprehensive data source set; Step 4: Based on the response dataset, determine the impact level of the current ventilation behavior, and based on the impact level and the comprehensive data source set, determine the optimal design scheme.

2. The drag reduction AI control method for the natural-ventilated hybrid cavitation hydrofoil according to claim 1, characterized in that: Determine the product term of hydrofoil chord length and hydrofoil span, and label the ratio of cavitation projected area to the product term as cavitation coverage area ratio.

3. The drag reduction AI control method for the natural-ventilated hybrid cavitation hydrofoil according to claim 1, characterized in that: The maximum value of the vorticity modulus is marked as the characteristic parameter of the flow field instability.

4. The drag reduction AI control method for the natural-ventilated hybrid cavitation hydrofoil according to claim 1, characterized in that: Based on the input dataset, output dataset, and response dataset, a comprehensive data source collection is constructed, including: Step 31: Based on the current cavitation flow experiment, determine several experimental parameters, and determine a set of structural parameters based on the experimental parameters; wherein, the experimental parameters include: vent number, vent size, and distribution pattern; Step 32: Label the input dataset as an environment parameter group; Step 33: Determine the cavitation evolution intensity index based on the cavitation coverage area ratio, flow field instability characteristic parameters, interface evolution stability index, and signal cavitation index. Step 34: Based on the structural parameter set, environmental parameter set, cavitation evolution intensity index, and response dataset, obtain the comprehensive data source set.

5. The drag reduction AI control method for the natural-ventilated hybrid cavitation hydrofoil according to claim 4, characterized in that: Assess the impact level of current ventilation behavior, including: Step 41: Determine the change in cavitation evolution intensity index between the current cavitation shedding cycle and the previous cavitation shedding cycle. If the change in cavitation evolution intensity index is greater than 0.05, the impact level of the current ventilation behavior is determined to be unstable, the first image of the next cavitation shedding cycle is taken, and the process returns to step 2. If the change in cavitation evolution intensity index is less than 0.05, the impact level of the current ventilation behavior is determined to be drag reduction evolution level, and the next step is executed. Step 42: Retrieve similar historical cavitation shedding cycles from the historical database, and determine the rate of change of interface evolution stability index and the mean rate of change of interface evolution stability index based on similar historical cavitation shedding cycles. Step 43: Determine the current rate of change of the interface evolution stability index based on the current interface evolution stability index; Step 44: If the current rate of change of the interface evolution stability index is less than the average rate of change of the interface evolution stability index, and the drag coefficient is less than the first target value, then the impact level of the current ventilation behavior is determined to be effective; otherwise, take a new image of the first image of the next cavitation shedding cycle and return to step 2.

6. The drag reduction AI control method for the natural-ventilated hybrid cavitation hydrofoil according to claim 5, characterized in that: Based on the impact level and the comprehensive data source set, the optimal design scheme is determined. The optimal design scheme is an optimized combination of optimized ventilation volume and optimized angle of attack. Initial ventilation rate determined based on cavitation flow experiments; Determine the ventilation adjustment volume, and then determine the optimal ventilation volume based on the initial ventilation volume and the ventilation adjustment volume; Determine the angle of attack to be adjusted, and then determine the optimal angle of attack based on the original angle of attack and the adjusted angle of attack. The optimized angle of attack and optimized ventilation volume are used as candidate optimization combinations, and the next cavitation shedding cycle is run based on the candidate optimization combinations; Based on each cavitation shedding cycle, acquire the input and output data and return to execute step 2; Extract the drag coefficients for n consecutive cavitation shedding cycles, determine the change in drag coefficients between adjacent cavitation shedding cycles, and determine whether the drag coefficients show a continuous downward trend and whether the change in drag coefficients is within the target range. If so, mark the candidate optimization combination as the optimization combination; otherwise, end the process.

7. The drag reduction AI control method for the natural-ventilated hybrid cavitation hydrofoil according to claim 6, characterized in that: Determine the ventilation adjustment volume, and then determine the optimal ventilation volume based on the initial ventilation volume and the ventilation adjustment volume; Determine the angle of attack to be adjusted, and then determine the optimized angle of attack based on the original angle of attack and the adjusted angle of attack, including: If the impact level is effective, mark the ventilation adjustment amount and the angle of attack adjustment as 0, mark the initial ventilation amount as the optimized ventilation amount, and mark the angle of attack as the optimized angle of attack; If the impact level is drag reduction evolution level, mark the ventilation adjustment amount as +3, mark the adjusted angle of attack as +3°, mark the optimized ventilation amount as the initial ventilation amount +3, and mark the optimized angle of attack as the angle of attack +3°. If the impact level is unstable, mark the ventilation adjustment amount as +15, mark the adjusted angle of attack as -0.5°, mark the optimized ventilation amount as the initial ventilation amount +15, and mark the optimized angle of attack as the angle of attack -0.5°; If the impact level is transitional, mark the ventilation adjustment amount as +8, the angle of attack adjustment as +2°, the optimized ventilation amount as initial ventilation amount +8, and the optimized angle of attack as angle of attack +2°.

8. A drag reduction AI control system for a natural-ventilated hybrid cavitation hydrofoil, used to execute the drag reduction AI control method for the natural-ventilated hybrid cavitation hydrofoil according to any one of claims 1-7, characterized in that, Includes the following modules: The experimental module is used to perform cavitation flow experiments and obtain input and output datasets. The feature extraction module is used to extract features based on the input and output datasets to obtain cavitation coverage area, flow field instability characteristic parameters, interface evolution stability index and signal cavitation index, and integrate them into the response dataset. The data collection module constructs a comprehensive data source set based on the input dataset, output dataset, and response dataset; The optimization module determines the impact level of the current ventilation behavior based on the response dataset, and then determines the optimal design scheme based on the impact level and the comprehensive data source set.

Images