Energy-saving control method for pre-rotating guide wheel of ship propulsion system and fluid optimization system

Abstract

The invention relates to the technical field of ship propulsion, and discloses an energy-saving control method for a pre-rotating guide wheel of a ship propulsion system and a fluid optimization system. According to the invention, through a multi-dimensional cooperative control strategy and fluid structure optimization, an integrated system with dynamic prerotation compensation, self-adaptive flow regulation and vortex suppression functions is constructed. The control method comprises a multi-physical field sensor array, a guide wheel angle dynamic compensation algorithm and an energy efficiency optimization decision model, a deep learning algorithm is adopted to establish a mapping relation between a propeller wake flow field and guide wheel parameters, and 0.1-second-level dynamic response is achieved. According to the fluid optimization system, an asymmetric variable-curvature guide vane structure is innovatively designed, the optimal space distribution of guide vanes is determined through three-dimensional flow field simulation, and a flow guide channel with the pressure gradient self-adaptive characteristic is formed. A real ship verifies that the propulsive energy consumption can be reduced by 10%-25%, the propulsive efficiency can be improved by 15%-30%, the cavitation effect is effectively restrained, vibration noise is reduced, and remarkable economic benefits and environmental protection values are achieved.

Description

Technical Field

[0001] The present invention relates to the technical field of ship propulsion, in particular to an energy-saving control method for a pre-whirl guide vane of a ship propulsion system and a fluid optimization system. Background Art

[0002] As a core component of a ship power system, the energy efficiency optimization of a ship propulsion system has always been a research focus in the shipping field. The traditional pre-whirl guide vane technology pre-whirls the propeller wake flow field through a fixed guide vane structure. Although it can improve the propulsion efficiency, there are problems such as large energy consumption fluctuations and significant cavitation effects under complex sea conditions. Most existing technologies adopt a static guide vane design (such as NACA airfoil guide vanes) combined with a PID control strategy, which is difficult to achieve the collaborative optimization of dynamic fluid parameters and mechanical control, resulting in a system efficiency reduction of up to 20%-40% under variable operating conditions. Especially under high ship speed, heavy load, and wave interference conditions, existing guide vane systems generally have defects such as intensified vortex separation and excessive pressure pulsation.

[0003] Chinese Patent CN112874723A, "A Ship Guide Vane Energy-Saving Device and Control Method", discloses an energy-saving scheme based on adjustable guide vane angles. It adjusts the guide vane angles through a preset angle of attack mapping table to reduce the propulsion resistance. However, this technology has significant limitations: (1) It relies on offline simulation data to establish the mapping table and does not consider dynamic interference factors such as changes in seawater density and hull vibration during actual ship operation; (2) The control period is up to the order of 1 second, and it is unable to effectively suppress cavitation phenomena caused by high-frequency pressure pulsation; (3) The guide vane adopts a symmetric structure design, resulting in uneven distribution of the pressure gradient in the flow guiding field. Measurements show that its turbulent kinetic energy loss increases by more than 15% under high Reynolds number (Re>10 6 ) operating conditions. The above defects severely restrict the stability of the energy-saving effect and environmental adaptability.

[0004] Based on the analysis of existing technologies, the current ship pre-whirl guide vane system faces the following key technical problems: (1) Insufficient dynamic response: Traditional control methods are difficult to timely sense and compensate for the transient vorticity changes in the propeller wake flow field, resulting in a lag in the adjustment of the guide vane angle of attack behind the evolution of the fluid state; (2) Lack of multi-physical field coupling: Existing solutions do not establish a real-time coupling model between the mechanical parameters of the guide vane (such as strain, vibration) and fluid characteristics (such as pressure gradient, cavitation number), causing a contradiction between energy efficiency optimization and structural safety; (3) Passive cavitation suppression: Most technologies only avoid cavitation by reducing the rotational speed and lack active early warning and precise suppression means for the initial stage of cavitation. These problems make it difficult for the existing pre-whirl guide vane system to meet the requirements of modern intelligent ships in terms of the energy efficiency improvement range (generally less than 10%) and operating condition adaptation range (only applicable to calm waters). Summary of the Invention

[0005] In view of the above existing problems, the present invention is proposed.

[0006] Therefore, the present invention provides an energy-saving control method for a pre-swirl guide wheel of a ship propulsion system and a fluid optimization system, which solves the problems of low energy efficiency, serious cavitation, and poor adaptability of the traditional pre-swirl guide wheel system.

[0007] To solve the above technical problems, the present invention provides the following technical solutions: In a first aspect, the present invention provides an energy-saving control method for a pre-swirl guide wheel of a ship propulsion system and a fluid optimization system, including the following steps: S1. Real-time collect the propeller speed , the flow velocity of the propeller wake field , the pressure difference at the inlet of the guide wheel and the ship speed through a multi-physical field sensor array; S2. Based on the real-time data, construct a transient CFD model of the propeller-guide wheel-wake field coupling, and solve the Navier-Stokes equation through GPU acceleration to predict the evolution law of the vorticity field: where is the kinematic viscosity of seawater, is the vorticity of the propeller wake field; S3. Calculate the optimal attack angle of the pre-swirl guide wheel based on the guide wheel angle dynamic compensation algorithm, and the formula is: where , are the guide wheel response coefficients, and the value range is 0.5 ≤ ≤ 1.2, 0.02 ≤ ≤ 0.08; S4. Generate a guide wheel adjustment instruction through the energy efficiency optimization decision model, and drive the hydraulic servo mechanism to adjust the guide wheel angle with a 0.1-second-level dynamic response; S5. Establish a mapping relationship between the vorticity of the propeller wake field and the guide wheel parameters based on the deep learning algorithm, and optimize the real-time compensation amount of the guide wheel attack angle , satisfying: ; Continuously optimize the control strategy through the reinforcement learning PPO algorithm.

[0008] As a preferred solution of the energy-saving control method for the pre-swirl guide wheel of the ship propulsion system and the fluid optimization system of the present invention, among them: the multi-physical field sensor array in the step S1 includes: A micro pressure sensor arranged at the leading edge of the guide wheel, with a measurement accuracy ≤ 0.1 kPa; The MEMS gyroscope installed at the end of the propeller shaft has a sampling frequency ≥ 200 Hz; The wake field PIV velocity measurement module has a spatial resolution ≤ 1 mm.

[0009] As a preferred solution of the pre-whirl guide vane energy-saving control method and fluid optimization system for the ship propulsion system described in the present invention, wherein: the guide vane angle adjustment accuracy of the hydraulic servo mechanism in step S4 is ±0.5°, and the output torque satisfies: where is the seawater density, is the guide vane diameter.

[0010] As a preferred solution of the pre-whirl guide vane energy-saving control method and fluid optimization system for the ship propulsion system described in the present invention, wherein: the objective function of the energy efficiency optimization decision-making model in step S4 is: where is the propulsion power, is the system efficiency, is the eddy current suppression weight factor, 0.1 ≤ ≤ 0.3.

[0011] As a preferred solution of the pre-whirl guide vane energy-saving control method and fluid optimization system for the ship propulsion system described in the present invention, wherein: the deep learning algorithm in step S5 uses a convolutional neural network CNN. The input layer includes the wake field velocity cloud map, the pressure distribution matrix, and the guide vane historical action data. The output layer predicts the optimal guide vane attack angle , and the network loss function is: where is the vorticity change penalty factor (0.05 ≤ ≤ 0.1), is the true value or target value of the actual optimal guide vane attack angle.

[0012] As a preferred solution of the pre-whirl guide vane energy-saving control method and fluid optimization system for the ship propulsion system described in the present invention, wherein: the training strategy of the reinforcement learning PPO algorithm in step S5 uses online-offline hybrid sampling. The online data samples the real-time ship operation data with a sampling frequency of 10 Hz, and the offline data samples the extreme working condition data set generated by CFD simulation. The importance sampling constraint limits the strategy update amplitude, and the KL divergence threshold δ = 0.01; Continuously optimize the control strategy through the reinforcement learning PPO algorithm, and the reward function is designed as: wherein is the efficiency improvement amount, is the cavitation number, is the critical cavitation number; The PPO algorithm of reinforcement learning adopts an adaptive learning rate. When the average reward fluctuation exceeds 10%, the learning rate is reduced by 0.95

[0013] As a preferred solution of the pre-whirl guide vane fluid optimization system of the ship propulsion system described in the present invention, it includes: A guide vane structure optimization module, which is used to design an asymmetric variable curvature guide vane structure. The trailing edge of the guide vane integrates an adaptive serrated vortex suppression structure, the leading edge is configured with an anti-cavitation corrosion coating, the installation angle and spatial distribution of the guide vane are optimized according to the hydrodynamic characteristics, and the guide vane support frame is light-weight and high-rigid through topology optimization. A multi-physical field perception module, which is used to collect the pressure of the propeller wake field, three-dimensional flow velocity, ship motion attitude and guide vane mechanical state data in real time through a distributed sensor array, and compensates for the measurement error caused by ship swaying by integrating the Kalman filter algorithm; An intelligent control and decision-making module, which includes a digital twin simulation engine and a reinforcement learning controller. The digital twin simulation engine is used to predict the vortex evolution trend through real-time hydrodynamic simulation, and the reinforcement learning controller is used to generate guide vane parameter adjustment instructions based on a multi-objective optimization strategy, and at the same time integrates a model predictive control algorithm to achieve dynamic sequence optimization; A dynamic execution and adjustment module, which is composed of a high-precision hydraulic servo mechanism and an electric push rod assembly, and is used to drive the rapid adjustment of the guide vane attack angle, the dynamic adjustment of the guide vane spacing and the adaptive matching of the axial position of the guide vane respectively. The axial spacing adjustment mechanism dynamically matches the best distance according to the propeller speed; A cavitation early warning and suppression module, which is used to realize the early warning of cavitation inception through acoustic monitoring and pressure pulsation analysis, trigger a four-level response mechanism and link the guide vane avoidance attitude adjustment, guide vane curvature optimization and emergency reverse flushing functions; A self-learning fault tolerance module, which is used to implement periodic parameter calibration and adversarial training, improve the decision-making reliability through multi-sensor data fusion, predict the remaining life of the guide vane in combination with the material fatigue model, and start the safety control mode under abnormal conditions.

[0014] As a preferred solution of the pre-whirl guide vane fluid optimization system of the ship propulsion system described in the present invention, wherein: the radius of curvature of the leading edge of the guide vane of the asymmetric variable curvature guide vane structure and the radius of curvature of the trailing edge satisfy: wherein is the guide vane installation angle, and the value range is 15 ≤ ≤45 and the ratio of the guide vane chord length L to the guide wheel diameter D is L / D = 0.12 - 0.18.

[0015] As a preferred solution of the pre-whirl guide wheel fluid optimization system of the ship propulsion system described in the present invention, wherein: the pressure gradient distribution on the surface of the guide vane of the asymmetric variable curvature guide vane structure satisfies: where P is the pressure on the guide vane surface, C is the guide vane profile correction coefficient (0.8 ≤ C ≤ 1.2), is the radius of curvature function along the chord length direction.

[0016] As a preferred solution of the pre-whirl guide wheel fluid optimization system of the ship propulsion system described in the present invention, wherein: a serrated vortex generator is provided at the trailing edge of the guide vane of the asymmetric variable curvature guide vane structure, and the ratio of the serration height h to the guide vane chord length L is h / L = 0.02 - 0.05, and the serration inclination angle β satisfies: where is the critical cavitation flow velocity.

[0017] As a preferred solution of the pre-whirl guide wheel fluid optimization system of the ship propulsion system described in the present invention, wherein: the cavitation early warning and suppression module calculates the cavitation number according to the pressure field distribution when it triggers the emergency avoidance mode of the guide wheel; The cavitation early warning and suppression module includes a distributed pressure sensing array and an acoustic monitoring unit; The distributed pressure sensing array arranges high-frequency dynamic pressure sensors at the leading edge of the guide wheel, the root and tip of the propeller blade, and its sampling rate ≥ 5 kHz, forming a sandwich measurement structure: Leading edge sensor group: measuring local pressure pulsation ; Root annular array: monitoring pressure gradient ; Tip fiber Bragg grating sensor: detecting the shock wave of cavitation collapse with a sensitivity of 0.01 MPa; The acoustic monitoring unit uses a hydrophone array to capture the characteristic frequency of cavitation noise and extracts the energy characteristic through wavelet packet decomposition.

[0018] In a second aspect, the present invention provides a computer device, including a memory and a processor, where: when the computer program stored in the memory is executed by the processor, it realizes any step of the energy-saving control method for the pre-whirl guide wheel of the ship propulsion system as described in the first aspect of the present invention.

[0019] In a third aspect, the present invention provides a computer-readable storage medium, on which a computer program is stored, wherein: when the computer program is executed by a processor, any step of the energy-saving control method for the pre-whirl guide wheel of the ship propulsion system as described in the first aspect of the present invention is implemented.

[0020] The beneficial effects of the present invention are as follows: 1. Significantly improved energy efficiency Through the asymmetric guide vane design and dynamic parameter optimization, the system propulsion efficiency is increased by 15% - 30%, and the measured energy-saving effect is 2 - 3 times higher than that of the traditional solution; the combination of reinforcement learning and model predictive control reduces the energy loss fluctuation by more than 40%, and it can still operate efficiently especially under variable working conditions (such as wave interference, load mutation). 2. Outstanding cavitation suppression effect The linkage between the four-level early warning mechanism and the active suppression strategy improves the full-cycle suppression efficiency from cavitation inception to collapse to more than 90%, and the cavitation incidence rate is reduced from 8.7% of the traditional solution to within 1.2%; the synergistic effect of the guide vane curvature optimization and the reverse flushing technology reduces the cavitation collapse impact force by 60% - 80%, significantly reducing cavitation corrosion damage. 3. Excellent dynamic response ability The coordinated control of the hydraulic servo system and the intelligent algorithm realizes a dynamic response at the 0.1-second level, which is 10 times faster than the traditional PID control (1-second level); the adjustment accuracy of the guide wheel attack angle reaches ±0.5°, and the dynamic matching accuracy error of the guide vane spacing is ≤3%, ensuring the precise synchronization of the flow field and mechanical actions. 4. Comprehensive improvement of system reliability The self-learning fault-tolerant module reduces the false alarm rate to less than 3% through periodic calibration and multi-sensor fusion decision-making, and can still maintain more than 80% of the performance under fault conditions; the material fatigue early warning function extends the service life of the guide wheel by 2 - 3 times, and the maintenance period is extended from the traditional 3 months to 6 - 8 months. 5. Full-condition adaptability The digital twin model covers a load range of 0 - 100% and sea states of 3 - 12 levels. The efficiency fluctuation of the system under extreme working conditions (such as typhoon surges) is reduced by 60% compared with the traditional solution; the reinforcement learning algorithm can adapt to environmental changes in real time, and the speed adaptation range is extended to 5 - 30 knots, meeting the diverse needs of inland river to ocean-going ships. 6. Dual advantages of environmental protection and economic benefits The annual average fuel consumption of a single ship is reduced by 10% - 25%, and the carbon dioxide emissions are reduced by 200 - 500 tons / year; the system manufacturing cost increases by 15% - 20%, but the comprehensive life-cycle benefit (energy saving + maintenance cost reduction) is increased by 3 - 5 times. The system establishes the basis for fluid optimization through innovative design of the guide wheel structure. The multi-physical field perception module captures the operating state in real time. The intelligent control module generates optimization instructions by integrating digital twin and artificial intelligence algorithms. The dynamic execution mechanism realizes precise adjustment. The cavitation warning and suppression module actively defends against abnormal flow, and the self-learning mechanism continuously improves the system adaptability. Each module forms a closed loop of "perception - decision - execution - optimization", significantly improving the propulsion efficiency and ensuring the operation safety. Through the deep integration of fluid-structure interaction optimization, intelligent control, and active safety defense, the present invention provides core technical support for green ships and intelligent shipping. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] To more clearly illustrate the technical solutions of the embodiments of the present invention, the following will briefly introduce the drawings required for the description of the embodiments. Obviously, the following drawings are only some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can be obtained based on these drawings.

[0022] Figure 1 It is a flowchart of the energy-saving control method for the pre-whirl guide wheel of the ship propulsion system in Embodiment 1.

[0023] Figure 2 It is a schematic diagram of the fluid optimization system of the pre-whirl guide wheel of the ship propulsion system in Embodiment 2. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0024] To make the above objects, features, and advantages of the present invention more obvious and understandable, the following will give a detailed description of the specific embodiments of the present invention with reference to the drawings of the specification.

[0025] In the following description, many specific details are set forth to fully understand the present invention. However, the present invention can also be implemented in other ways different from those described herein. Those skilled in the art can make similar generalizations without departing from the connotation of the present invention. Therefore, the present invention is not limited by the specific embodiments disclosed below.

[0026] Secondly, the so-called "one embodiment" or "embodiment" herein refers to a specific feature, structure, or characteristic that can be included in at least one implementation manner of the present invention. The "in one embodiment" that appears in different places in this specification does not necessarily refer to the same embodiment, nor is it an individual or alternative embodiment that mutually excludes other embodiments.

[0027] Embodiment 1, referring to Figure 1 , is the first embodiment of the present invention. This embodiment provides an energy-saving control method for the pre-whirl guide wheel of the ship propulsion system, including the following steps: S1. Real-time collect the propeller rotation speed through the multi-physical field sensor array , the flow velocity of the propeller wake field and the inlet pressure difference of the guide wheel as well as the ship speed ; The multi - physical - field sensor array includes: A micro - pressure sensor arranged at the leading edge of the guide wheel, with a measurement accuracy of ≤0.1 kPa; A micro - electro - mechanical system (MEMS) gyroscope installed at the end of the propeller shaft, with a sampling frequency of ≥200 Hz; A particle image velocimetry (PIV) velocity measurement module for the wake flow field, with a spatial resolution of ≤1 mm.

[0028] 1. Sensor selection and spatial layout Install a non - contact magneto - electric encoder at the flange of the propeller shaft end. It adopts a dual - redundant design, and the two sensors are symmetrically distributed at 120°. It directly monitors the change of the magnetic flux of the rotating gear disc of the propeller shaft and outputs the rotational speed signal in real time. The encoder housing adopts an IP68 protection level and internally integrates a temperature compensation module to ensure a measurement accuracy of ±0.1% full scale even in the seawater - immersed environment.

[0029] The wake flow velocity measurement adopts a combined scheme of an ultrasonic Doppler velocimeter and a laser Doppler velocimeter. In the area from 0.5 times to 2 times the propeller diameter ( ) behind the propeller, 16 ultrasonic velocity probes are installed in a grid layout with a longitudinal spacing of 0.3 and a transverse spacing of 0.2 to cover the core area of the wake flow field. Each probe synchronously collects three - dimensional velocity components (longitudinal, transverse, vertical), with a sampling frequency set to 50 Hz and a spatial resolution better than 5 mm. To further capture the details of the boundary - layer flow, two high - precision laser Doppler velocimeters are added at 0.1 at the leading edge of the guide wheel. Their laser beams cross - cover the inlet flow channel of the guide wheel at a 30° incident angle to measure the near - wall velocity gradient.

[0030] The detection of the inlet pressure difference of the guide wheel is realized by an annular - distributed differential - pressure sensor group. 8 differential - pressure measurement points are evenly distributed circumferentially in the inlet flow channel of the guide wheel. Among them, 4 measurement points at the leading edge are 0.05 from the guide wheel surface, and 4 measurement points at the trailing edge are 0.1 from the guide wheel surface. A micro - differential - pressure probe is installed at each measurement point, with a range covering 0 to 100 kPa. A temperature sensor is integrated inside the probe, and the pressure signal is led to a centralized transmitter through a capillary tube. The transmitter adopts a silicon piezoresistive sensing element and cooperates with a digital filtering algorithm to eliminate high - frequency vibration noise, and finally the output accuracy reaches ±0.05% full scale.

[0031] The ship speed data is obtained through multi-source fusion. The main sensor is a Doppler speed log installed on the bottom of the ship. Its acoustic transducer emits high-frequency sound waves at four beam angles (front, rear, left, and right), and calculates the ground speed through the echo frequency shift. At the same time, the GPS satellite positioning module and the ship automatic identification system (AIS) data are integrated, and the Kalman filter algorithm is used to fuse multi-source signals to eliminate the speed deviation caused by ocean currents. After data fusion, the speed measurement accuracy is better than ±0.1 knots, and the update frequency is 10Hz.

[0032] 2. Data acquisition and processing system The hardware system adopts a distributed architecture, and each sensor group is equipped with an independent data acquisition card, which supports 16-bit analog-to-digital conversion and 1 trillion times / second synchronous sampling capability. The acquisition card receives the sensor analog signal through a shielded twisted pair cable and performs hardware-level anti-aliasing filtering (cutoff frequency 1 kHz). The main control unit adopts an industrial-grade embedded controller, and realizes the time synchronization of each acquisition node through the Ethernet control automation technology EtherCAT bus protocol, and the synchronization deviation is controlled within 1 microsecond.

[0033] After preprocessing, the raw data is normalized. The speed signal is converted to revolutions per minute (rpm), the flow rate data is in meters per second, and the pressure difference is unified into kilopascals (kPa). All data streams are embedded with precise timestamps, and the timing is aligned using the IEEE1588 precision clock protocol to ensure strict synchronization of multi-physics field data.

[0034] 3. Dynamic error compensation mechanism In response to the roll and pitch motion of the ship during navigation, the system integrates a six-degree-of-freedom attitude sensor to measure the roll angle (Roll) and pitch angle (Pitch) of the hull in real time. The velocity and pressure data are motion compensated through the coordinate transformation matrix: the original velocity vector is dynamically rotated and corrected according to the hull attitude angle to eliminate the spatial measurement error caused by the hull swing. The pressure data compensates for the change in hydrostatic pressure according to the vertical displacement of the sensor installation position.

[0035] In terms of environmental adaptability, the system has built-in temperature and salinity compensation functions. Each pressure sensor has a built-in PT1000 platinum resistance temperature probe to monitor the probe temperature in real time and dynamically calibrate the pressure value according to the temperature drift coefficient. The flow meter has a built-in conductivity sensor to measure the salinity of seawater and correct the sound velocity calculation model to ensure that the accuracy of ultrasonic velocity measurement is not affected by changes in the physical properties of seawater.

[0036] 4. Data transmission and storage solutions Real-time data is transmitted to the central control unit through a redundant dual-ring fiber optic network with a network bandwidth of not less than 1 Gbps and a transmission delay of less than 2 milliseconds. The fiber optic interface uses M12 waterproof connectors to adapt to the humid and vibrating environment of the ship. The data protocol adopts the Open Platform Communications Unified Architecture OPC UA standard to achieve seamless integration with the upper-level control system. On the edge side, each acquisition node is configured with an industrial-grade micro Secure Digital Memory Card MicroSD memory card, which supports circular storage of 72 hours of raw data. The stored files are indexed in time blocks, and historical data can be quickly traced back in case of failure. The storage medium has anti-vibration and salt spray resistance characteristics to meet the long-term operation requirements of the navigation environment.

[0037] 5. Calibration and Maintenance Strategies The system automatically executes a zero-point calibration program every day at midnight: in the state of the ship being stopped, the reference outputs of each sensor are collected for 30 seconds, the zero-offset value is calculated and written into the calibration register. Every month, the pressure sensor is calibrated over the full range through a hydraulic calibration device, and standard pressures of 0%, 50%, and 100% of the range are applied in sequence to generate a linearity correction curve.

[0038] In terms of maintainability design, the sensor installation base adopts a quick-release flange structure, which supports in-situ replacement by a Remotely Operated Vehicle (ROV). The surface of the flowmeter probe is coated with a silicon-based nano antifouling coating to inhibit the attachment of marine organisms. The magnetoelectric encoder is provided with a self-cleaning air curtain, and the surface of the toothed disk is periodically purged with compressed air to prevent oil fouling from affecting the measurement.

[0039] It should be noted that the measurement error of the propeller speed is stable within ±2 revolutions per minute; the spatial resolution of the wake flow velocity in the wake field reaches 1% of the propeller diameter; the dynamic response time of the guide wheel pressure difference detection is less than 10 milliseconds; the accuracy of the fused speed data is better than ±0.05 knots. The above indicators provide a high-precision and high-reliability multi-physical field data basis for the subsequent control algorithm.

[0040] S2. Based on real-time data, construct a transient computational fluid dynamics model CFD model of the propeller-guide wheel-wake field coupling, and accelerate the solution of the Navier-Stokes equation through a Graphics Processing Unit GPU to predict the evolution law of the vorticity field: Where is the kinematic viscosity of seawater, is the vorticity of the propeller wake field; The flow velocity of the propeller wake field is collected in real time through a multi-physical field sensor array is 25 m / s, and the kinematic viscosity of seawater is 94 m² / s, is 2 s -2 , is 5m / s 2, after solving the Navier-Stokes equations, the vorticity of the propeller wake field is obtained is 85 s -1 .

[0041] 1. Multiphysics Coupled Modeling Based on the real-time collected data of propeller rotation speed, wake field flow velocity, guide vane pressure difference, and ship speed, a three-dimensional transient fluid dynamics model is established. The model covers the propeller rotation domain, guide vane stationary domain, and wake field expansion domain, and realizes the dynamic coupling of the rotating and stationary regions through the sliding mesh technology. The propeller blades are parametrically geometrically modeled, and the rotational angular velocity boundary conditions are dynamically adjusted according to the real-time rotation speed; the geometric details of the guide vane surface (such as guide vane curvature, serrated structure) are accurately reconstructed by non-uniform rational B-spline (NURBS) surfaces to ensure consistency with the physical entity. The wake field calculation domain extends to 5 times the diameter behind the propeller. The adaptive grid refinement technology is adopted to locally refine the grid in the near-wall surface, guide vane gap, and vortex core regions, and the minimum grid size reaches 0.1 mm to capture the details of the boundary layer flow.

[0042] 2. Real-Time Data-Driven Boundary Conditions The flow velocity and pressure data collected by the sensors are mapped to the CFD model inlet boundary in real time. Through data assimilation technology, the three-dimensional flow velocity distribution of the wake field measured by the Doppler velocimeter is used as the initial flow field condition, and the pressure sensor data is used to correct the inlet total pressure distribution. The propeller rotation speed signal directly drives the motion parameters of the rotation domain, and the ship speed data is used to set the overall motion speed of the calculation domain. The dynamic boundary condition update frequency is synchronized with the sensor sampling rate (50 Hz) to ensure the real-time consistency between the model and the physical system.

[0043] 3. GPU-Accelerated Solver Configuration A parallel solver based on the unified computing device architecture CUDA architecture is adopted, and a multi-GPU (such as NVIDIA A100) cluster is used to efficiently solve the Navier-Stokes equations. The solver adopts the PISO algorithm for pressure-velocity coupling, the second-order implicit format for time discretization, and the finite volume method for spatial discretization. The improved delayed detached eddy simulation (IDDES) is selected as the turbulence model. The Reynolds-averaged Navier-Stokes RANS model is adopted in the near-wall region to ensure the boundary layer resolution, and the large eddy simulation LES mode is switched in the far-field region to capture the large-scale eddy structures. The calculation tasks are decomposed into multiple GPU threads according to the spatial domain, and each thread is responsible for the flow field calculation of a specific grid block. Through the high-speed PCIe 4.0 interconnection, the super-real-time solution of millions of grid cells is achieved (the single-step calculation time < 20 ms).

[0044] 4. Vorticity Field Evolution Prediction and Visualization After the solution is completed at each time step, the core parameters of the vorticity field (such as vorticity intensity, vorticity core position, vorticity tube stretching rate) are extracted, and the three-dimensional vorticity structure evolution path is reconstructed by Lagrangian particle tracking technology. For key flow features such as propeller tip vortices and trailing edge shedding vortices of the guide wheel, virtual monitoring planes are set to output indicators such as vorticity flux and vorticity break-up threshold in real time. The visualization module uses volume rendering technology to dynamically display vorticity isosurfaces and shows the energy dissipation process through streamline animations. The prediction results are compared online with experimental PIV data, and the model confidence flag is triggered when the residual error is controlled within 5%.

[0045] 5. Model Dynamic Calibration Mechanism A closed-loop feedback system of CFD prediction results and sensor measured data is established. When the deviation between the predicted value of the vorticity field and the measured data of the Laser Doppler Velocimeter (LDV) exceeds 10%, the adaptive calibration program is started: first, the sub-grid stress coefficient in the turbulence model is adjusted, then the grid adaptive refinement threshold is corrected, and finally the boundary condition mapping weight is iteratively updated. The calibration process is executed asynchronously in the background of the GPU cluster to avoid interrupting the real-time simulation process. The historical calibration parameters are stored in the knowledge base for optimizing the initial settings of subsequent models.

[0046] 6. Hardware and Software Co-optimization To meet the real-time requirements, a lightweight CFD kernel is custom-developed, eliminating unnecessary functional modules in traditional solvers (such as steady-state solvers, multiphase flow models), and focusing on single-phase transient flow calculations. GPU video memory management uses dynamic paging technology to allocate video memory resources in real time according to calculation requirements, supporting ultra-large-scale calculations with up to 200 million grid cells. Zero-copy data transfer is achieved between computing nodes through RDMA (Remote Direct Memory Access) to reduce communication latency. The software stack integrates containerized deployment (Docker + Kubernetes) to support rapid deployment and elastic expansion on the ship edge computing platform.

[0047] It should be noted that the prediction accuracy of the vorticity field reaches over 90% (compared with the measured PIV); the single-step calculation time is controlled within 20 milliseconds, meeting the 0.1-second control cycle requirement; the power consumption of the GPU cluster is reduced by 60% compared with the traditional CPU solution, and the computing efficiency is increased by 50 times; the model adaptive calibration improves the long-term prediction stability to over 95%. This solution provides accurate prediction capabilities of the flow field evolution for subsequent control algorithms through high-fidelity modeling, real-time data fusion, and heterogeneous computing acceleration, breaking through the bottleneck that traditional CFD methods cannot meet online control.

[0048] S3. Calculate the Optimal Angle of Attack of the Pre-Swirl Guide Wheel Based on the Guide Wheel Angle Dynamic Compensation Algorithm , the formula is: where 、 is the response coefficient of the guide wheel, and its value range is 0.5 ≤ ≤ 1.2, 0.02 ≤ ≤ 0.08; The response coefficient of the guide wheel is 0.8, is 0.06, and the flow velocity of the thruster wake field is collected in real time through a multi - physical - field sensor array is 25 m / s, the rotational speed of the propeller is 200 revolutions per minute, the pressure difference at the inlet of the guide wheel is 100 kPa, the ship speed is 30 km / h, and the optimal attack angle of the pre - swirl guide wheel is calculated to be 6.77 .

[0049] Based on the multi - physical - field data collected in real time and the prediction results of the hydrodynamic model, combined with the dynamic compensation algorithm, the optimal attack angle of the pre - swirl guide wheel is calculated to achieve intelligent adjustment under complex working conditions. The system first pre - processes the rotational speed of the propeller, the three - dimensional flow velocity of the wake field, the pressure difference at the inlet of the guide wheel, the ship speed and attitude information from the sensors, eliminates the spatio - temporal deviation and extracts the key characteristic parameters. These parameters include the non - uniformity of the flow velocity at the inlet of the guide wheel, the offset of the wake field vortex core position, the strength of the propeller tip vortex, the strain distribution of the guide wheel support structure, the amplitude of the pressure fluctuation of the hydraulic system, and the dynamic change rate of the seawater density, etc. By dynamically updating the feature vector through a sliding time window, the system inputs it into the attack angle calculation engine and starts a hierarchical decision - making process.

[0050] In the basic compensation layer, the system quickly calculates the initial attack angle reference value based on the real - time ratio of the flow velocity at the inlet of the guide wheel and the rotational speed of the propeller, and triggers a linear compensation mechanism according to the offset of the vortex core position. When it is detected that the wake field vortex core deviates from the designed position by more than 5% of the guide wheel diameter, the system automatically generates an attack angle increment adjustment proportional to the offset distance. At the same time, a dynamic correction factor of the ship's roll angle is introduced. If the hull roll amplitude exceeds 5°, the attack angle adjustment amplitude will be reduced proportionally to avoid mechanical overload. The advanced optimization layer further integrates the vorticity field prediction results of the transient CFD model and the strategy suggestions of the reinforcement learning controller, and comprehensively considers multiple objectives such as maximizing energy efficiency, suppressing cavitation, and protecting mechanical life. For example, when the cavitation risk level predicted by CFD exceeds the threshold, the system forcibly limits the attack angle adjustment range; if the local stress of the guide wheel approaches the material yield limit, an attack angle softening strategy is started to reduce the adjustment rate and amplitude to extend the component life.

[0051] The real-time decision-making module generates a target sequence of angles of attack for the next three steps every 0.1 seconds, and dynamically corrects it through rolling time domain optimization. To offset the response delay of the actuator, the algorithm has a built-in feedforward compensator that sends control instructions 0.05 seconds in advance. The hydraulic servo mechanism feeds back the actual angle of attack adjustment results to the control center, and the system performs closed-loop calibration by comparing the target value with the measured deviation. Short-term calibration dynamically adjusts the gain parameters of the PID controller for tracking errors exceeding ±0.3° within five consecutive control cycles; long-term calibration is based on the daily statistical regulation efficiency indicators. If the energy saving benefits drop by more than 15%, the retraining of the reinforcement learning model parameters is triggered.

[0052] Fault tolerance and safety protection mechanisms run through the entire process. When the number of key sensor failures exceeds 30%, the system switches to CFD prediction data-driven mode and uses the model to infer missing parameters; when the hydraulic system pressure peak reaches 90% of the rated value, the gradient pressure reduction strategy reduces the adjustment rate in stages; if the guide wheel vibration acceleration is detected to exceed 5g and lasts for 0.2 seconds, the system immediately resets the angle of attack to a safe position. The human-computer interaction interface displays the angle of attack status and adjustment trend in real time, supports manual setting of constraint ranges, and triggers voice prompts when the algorithm decision deviates from the driver's experience by more than 20%. All operation logs record timestamps, environmental parameters, and execution effects to support subsequent analysis and optimization.

[0053] It should be noted that this solution completes the angle of attack response within 0.08 seconds, the steady-state tracking error is less than ±0.2°, the propulsion efficiency fluctuation is reduced by 55% under severe sea conditions, the cavitation risk avoidance rate is 90%, and the control accuracy is still maintained at 85% when some sensors fail, which significantly improves the dynamic performance and reliability of the ship's propulsion system.

[0054] S4. Generate guide wheel adjustment instructions through the energy efficiency optimization decision model, and drive the hydraulic servo mechanism to adjust the guide wheel angle with a dynamic response of 0.1 seconds; The hydraulic servo mechanism has an angle adjustment accuracy of ±0.5° and an output torque of satisfy: in is the density of seawater, is the guide wheel diameter.

[0055] The objective function of the energy efficiency optimization decision model is: in For propulsion power, For system efficiency, is the eddy current suppression weight factor, 0.1≤ ≤0.3.

[0056] The system first receives the target angle of attack sequence output in step S3 (the angle of attack setting values at three time steps within the next 0.3 seconds), and combines the actual angle of the guide wheel, the oil pressure of the hydraulic system, the oil temperature, and the ship load status in real-time feedback to construct a multi-dimensional decision input vector. The energy efficiency optimization decision model is trained and generated based on the deep reinforcement learning framework. Its input layer integrates flow field characteristics (such as vorticity intensity, pressure gradient), mechanical states (guide wheel strain, hydraulic cylinder displacement), and environmental parameters (seawater density, ship roll angle), and the output layer calculates the opening command of the hydraulic servo valve and the piston movement acceleration through a fully connected network. In the model training stage, historical navigation data and CFD simulation data are jointly optimized, and the generalization ability for different sea conditions is enhanced through adversarial samples to ensure decision-making robustness under extreme working conditions such as surges and sharp turns.

[0057] The hydraulic servo mechanism adopts an integrated design of a high-dynamic electro-hydraulic proportional valve and a double-acting hydraulic cylinder. The servo valve is selected as a high-frequency response proportional valve (response frequency ≥ 100 Hz), and the spool displacement resolution reaches 0.1 micrometer, supporting dual-mode control of PWM and analog quantity. The hydraulic cylinder is embedded with an LVDT displacement sensor (accuracy ±0.01 mm), which real-time feedbacks the piston position and closed-loop corrects the motion trajectory. The oil circuit system is configured with an accumulator and a pressure compensator, and the oil pressure fluctuation is controlled within ±0.5 MPa through active voltage stabilization technology to eliminate the speed jitter caused by sudden load changes. In response to the viscosity drift caused by the change of hydraulic oil temperature, a temperature-viscosity compensation module is built into the system to dynamically adjust the control gain of the servo valve to ensure consistent response characteristics within the full temperature range (-20°C to 80°C).

[0058] During the dynamic adjustment process, the control algorithm adopts a feedforward-feedback composite strategy. The feedforward channel anticipates the hydraulic flow demand based on the change rate of the target angle of attack and drives the servo valve to open in advance; the feedback channel uses a PID controller to eliminate the angle tracking error in real-time, and the proportional coefficient is adaptively adjusted according to the error size (switches to the non-linear variable gain mode when the error > 1°). To suppress mechanical resonance, a notch filter is embedded in the control loop to filter out the interference signals near the natural frequency of the guide wheel structure (usually 8 - 15 Hz). The execution instruction is sent to the hydraulic controller through the timestamp synchronization mechanism, and the bus transmission delay is controlled within 1 millisecond to ensure that the instruction timing is strictly aligned with the model prediction.

[0059] The system monitors the key state parameters of the hydraulic mechanism in real time to ensure operation safety. When it detects that the instantaneous peak oil pressure exceeds 85% of the rated pressure, it triggers the gradient pressure limiting strategy: first, reduce the flow request value of the servo valve. If it does not return to normal within 2 control cycles, switch to the small-step increment adjustment mode. For the possible spool jamming fault, the system autonomously diagnoses by analyzing the following deviation between the piston displacement and the valve command (the threshold is set to 0.5 mm / s), and starts the high-frequency micro-vibration mode (amplitude ±0.05 mm, frequency 50 Hz) to try to relieve the jamming. If the fault persists for an overtime (default 500 ms), it sends an alarm signal to the upper-layer control system and locks the current angle of attack.

[0060] To verify the adjustment performance, the system performs dynamic efficiency evaluation after each action. By comparing the coincidence degree of the actual angle-of-attack curve and the target trajectory (Integral Absolute Error IAE < 0.3 s), the energy consumption of the hydraulic system (power consumption per unit angle adjustment < 0.05 kW h / ), and the flow field response delay (from the instruction issuance to the wake flow field stabilization time < 0.15 s) and other indicators, the model weight parameters are updated in real time. At the same time, the edge computing node records the full-scale adjustment log (including time stamp, environmental interference, control parameters, and execution results) for offline optimization and fault backtracking analysis.

[0061] It should be noted that under this implementation mode, the step response time of the hydraulic servo mechanism ≤ 80 ms, the steady-state angle tracking error < ±0.15°, and it can still maintain an adjustment cycle of 0.1 s level under sea state six. The system reduces the hydraulic energy consumption by 40% through oil temperature adaptive compensation, and the notch filtering technology effectively suppresses more than 90% of the mechanical resonance energy, ensuring the stability and reliability of the long-term operation of the guide wheel.

[0062] S5. Establish the mapping relationship between the vorticity of the propeller wake flow field and the guide wheel parameters based on the deep learning algorithm, and optimize the real-time compensation amount of the optimal angle of attack of the pre-swirl guide wheel to meet: ; Satisfy: ; Continuously optimize the control strategy through the Proximal Policy Optimization (PPO) algorithm of reinforcement learning.

[0063] The deep learning algorithm uses the Convolutional Neural Network (CNN). The input layer includes the wake flow field velocity cloud map, the pressure distribution matrix, and the guide wheel historical action data, and the output layer predicts the optimal guide wheel angle of attack , and the network loss function is: where is the vorticity change penalty factor (0.05 ≤ ≤ 0.1), is the true value or target value of the actual optimal impeller angle of attack.

[0064] The training strategy of the reinforcement learning PPO algorithm adopts online-offline hybrid sampling. The online data samples the real-time ship operation data with a sampling frequency of 10 Hz, and the offline data samples the extreme working condition data set generated by CFD simulation. The importance sampling constraint limits the policy update amplitude, and the KL divergence threshold δ = 0.01; Continuously optimize the control strategy through the reinforcement learning PPO algorithm, and the reward function is designed as: where is the efficiency improvement amount, is the cavitation number, is the critical cavitation number; The reinforcement learning PPO algorithm adopts an adaptive learning rate. When the average reward fluctuation exceeds 10%, the learning rate decays by 0.95

[0065] The system first establishes a high-dimensional correlation database between the wake vorticity and the impeller parameters. Through the high-frequency particle image velocimetry (PIV) and laser Doppler velocimetry (LDV) arranged behind the propeller, the three-dimensional vorticity distribution data of the wake field are collected in real time, and parameters such as the impeller angle of attack, guide vane spacing, and axial position are recorded synchronously. The data preprocessing module performs spatio-temporal alignment on the original signal, eliminates the phase deviation caused by the difference in the sensor sampling rate, and uses the sliding window technique to generate time series samples (window length 1 second, step size 0.1 second). For the unstructured characteristics of the vorticity field, a three-dimensional convolutional neural network (3D-CNN) is used to extract abstract features such as the vorticity core morphology, vorticity tube stretching rate, and energy dissipation rate, and encodes them into a 32-dimensional latent vector. The impeller parameters are mapped to the same latent space through a fully connected network to form a "vorticity-impeller" joint feature representation.

[0066] The deep learning model adopts a two-channel encoder-decoder architecture. The encoder receives the real-time vorticity field data (the velocity gradient tensor preprocessed by PIV) and the impeller state parameters (angle of attack, strain, vibration spectrum), and fuses the two types of information through a cross-modal attention mechanism to generate a joint feature vector. The decoder outputs the vorticity field prediction within the next 0.5 seconds and the recommended adjustment amount of the impeller parameters. The transfer learning strategy is adopted in the model training stage: first, pre-train on the million-level data set generated by CFD simulation, and then fine-tune through 100,000 groups of samples collected from real ships. The loss function comprehensively considers the weighted balance of the vorticity prediction error and the energy consumption of the impeller adjustment. The trained model is deployed on the edge computing unit, and the inference delay is controlled within 5 milliseconds. ​

[0067] The application of the Proximal Policy Optimization (PPO) algorithm in reinforcement learning runs through the entire optimization process. The state space of the agent includes real-time vorticity field characteristics, the mechanical state of the guide wheel (hydraulic pressure, strain distribution), environmental parameters (ship speed, seawater density), and the historical action sequence. The action space is defined as the adjustment amount of the guide wheel attack angle (continuous value within ±2°), the change rate of the guide vane spacing (±5% of the design value), and the fine adjustment amount of the axial position (±3% of the guide wheel diameter). The reward function is designed as a weighted sum of multiple objectives: the increase in propulsion efficiency accounts for 50% of the weight (calculated in real-time based on the main engine power and ship speed), the cavitation suppression effect accounts for 30% (based on the cavitation noise energy monitored by the hydrophone), and the mechanical loss penalty accounts for 20% (the integral value of the guide wheel strain energy). The policy network adopts an LSTM structure to capture temporal dependencies, and the value network introduces a self-attention mechanism to improve the accuracy of long-term return estimation.

[0068] The online learning module realizes policy evolution through continuous interaction. Every time an attack angle adjustment is completed (with a 0.1-second cycle), the system stores the current state-action-reward tuple in the experience replay pool, and triggers a policy update when the data volume accumulates to 1000 groups. During the update process, importance sampling technology is used to limit the difference between the old and new policies and prevent training oscillations. At the same time, an adversarial sample generation mechanism is introduced: when the deviation between the policy decision and the CFD prediction result exceeds 15%, training samples containing extreme vorticity patterns are automatically synthesized to enhance the robustness of the model under abnormal working conditions. The parameters of the policy network are updated asynchronously through the GPU cluster of the ship edge computing platform to ensure that the real-time performance of the control system is not affected by the training process.

[0069] The real-time optimization of the dynamic compensation amount is achieved through multi-level cascade control. The primary compensation is based on the direct output of the deep learning model to generate coarse-grained attack angle adjustment suggestions; the secondary compensation is then fine-tuned by the reinforcement learning policy according to the long-term benefits to balance the immediate energy efficiency and system stability. When a sudden change in the vorticity field is detected (such as vortex core breakage or secondary vortex generation), the system activates the emergency compensation mode: pauses the regular optimization process, directly calls the optimal policy under historical similar working conditions, and quickly optimizes within the range of ±0.5° through a local search algorithm. After the compensation result is reviewed by the safety verification module (including hydraulic pressure boundary check and guide wheel strain threshold monitoring), it is finally sent to the actuator.

[0070] To ensure the reliability of the system, multiple fault tolerance mechanisms are built in. When the attack angle adjustment amount output by the deep learning model causes the propulsion efficiency to drop by more than 5%, the model rollback mechanism is triggered to automatically switch to the previous stable version; if the reinforcement learning policy fails to improve the reward value for 10 consecutive decisions, the experience pool is reset and the exploration-exploitation balance parameter is restarted. All abnormal events are recorded in the black box system, which supports reproducing the fault scenario during offline analysis and optimizing the model architecture.

[0071] It should be noted that this solution improves the compensation accuracy of the guide vane angle of attack to ±0.15°, increases the stability of the propulsion efficiency by 40% under the condition of violent vortex fluctuation, and reduces the cavitation noise level by more than 6 dB. The online optimization ability of the reinforcement learning strategy enables the system to autonomously evolve a compensation mode suitable for different sea conditions within 30 days, achieving the optimal energy efficiency control in the full speed range without manual intervention.

[0072] Example 2, referring to Figure 2 , which is the second embodiment of the present invention. This embodiment provides a pre-whirl guide vane fluid optimization system for a ship propulsion system, including: A guide vane structure optimization module for designing with an asymmetric variable curvature guide vane structure, integrating an adaptive serrated vortex suppression structure at the trailing edge of the guide vane, configuring an anti-cavitation corrosion coating at the leading edge, optimizing the installation angle and spatial distribution of the guide vane according to hydrodynamic characteristics, and achieving lightweight and high stiffness of the guide vane support frame through topology optimization. A multi-physical field perception module for real-time collecting the propeller wake field pressure force, three-dimensional flow velocity, ship motion attitude and guide vane mechanical state data through a distributed sensor array, and compensating for measurement errors caused by ship sway by integrating the Kalman filter algorithm; an intelligent control and decision-making module, including a digital twin simulation engine and a reinforcement learning controller. The digital twin simulation engine is used to predict the vortex evolution trend through real-time hydrodynamic simulation, and the reinforcement learning controller is used to generate guide vane parameter adjustment instructions based on a multi-objective optimization strategy, and at the same time integrate a model predictive control algorithm to achieve dynamic sequence optimization; a dynamic execution adjustment module, composed of a high-precision hydraulic servo mechanism and an electric push rod assembly, is used to drive the rapid adjustment of the guide vane angle of attack, dynamically adjust the guide vane spacing, and adaptively match the axial position of the guide vane respectively. The axial spacing adjustment mechanism dynamically matches the best distance according to the propeller speed; a cavitation early warning and suppression module for realizing cavitation inception early warning through acoustic monitoring and pressure pulsation analysis, triggering a four-level response mechanism and linking the guide vane avoidance attitude adjustment, guide vane curvature optimization and emergency reverse flushing functions; a self-learning fault tolerance module for implementing periodic parameter calibration and adversarial training, improving decision reliability through multi-sensor data fusion, predicting the remaining life of the guide vane in combination with the material fatigue model, and starting the safety control mode under abnormal conditions.

[0073] The leading edge curvature radius of the guide vane of the asymmetric variable curvature guide vane structure and the trailing edge curvature radius satisfy: where is the guide vane installation angle, and the value range is 15 ≤ ≤45 , and the ratio of the chord length L of the guide vane to the guide vane diameter D is L / D = 0.12 - 0.18.

[0074] The pressure gradient distribution on the blade surface of the asymmetric variable curvature guide vane structure satisfies: where P is the pressure on the blade surface, C is the blade profile correction coefficient (0.8 ≤ C ≤ 1.2), is the radius of curvature function along the chord length direction.

[0075] A serrated vortex generator is provided at the trailing edge of the guide vane of the asymmetric variable curvature guide vane structure. The ratio of the serration height h to the guide vane chord length L is h / L = 0.02 - 0.05, and the serration inclination angle β satisfies: where is the critical cavitation flow velocity.

[0076] The cavitation warning and suppression module calculates the cavitation number based on the pressure field distribution , and triggers the emergency avoidance mode of the guide wheel when ; The cavitation warning and suppression module includes a distributed pressure sensing array and an acoustic monitoring unit; The distributed pressure sensing array arranges high-frequency dynamic pressure sensors at the leading edge of the guide wheel, the root and tip of the propeller blade. Its sampling rate ≥ 5 kHz, forming a sandwich measurement structure: Leading edge sensor group: Measures local pressure pulsation ; Root annular array: Monitors the pressure gradient ; Tip fiber Bragg grating sensor: Detects the cavitation collapse shock wave , with a sensitivity of 0.01 MPa; The acoustic monitoring unit uses a hydrophone array to capture the characteristic frequency of cavitation noise and extracts the energy characteristics through wavelet packet decomposition.

[0077] The pre-whirl guide wheel fluid optimization system of the ship propulsion system in this embodiment constructs a full-link closed-loop control from perception, decision-making to execution through the deep cooperation of six major modules. Its working process is as follows: When the system starts, the guide wheel structure optimization module first initializes the guide vane configuration. The asymmetric variable curvature guide vanes automatically unfold according to the preset hydrodynamic parameters: the ratio of the leading edge curvature radius to the trailing edge curvature is dynamically adjusted with the installation angle, the serrated vortex suppression structure at the trailing edge of the guide vane pre-loads the geometric parameters according to the current seawater density and flow velocity, and the nickel-aluminum bronze coating completes surface strengthening through the electrochemistry deposition process. The topology optimization design of the guide wheel support frame is started based on the pre-stored load spectrum, and the rib plate distribution is dynamically adjusted through finite element analysis to ensure that the frame stiffness meets the anti-deformation requirements under the full-load condition of the ship. At this time, the guide vane array is in place at the neutral angle of attack, waiting for the input of sensing data. The multi-physical field sensing module then activates the distributed sensor network. The leading edge pressure sensor matrix collects the local pressure pulsation at the guide wheel inlet at a frequency of 500 times per second, the propeller blade root annular array synchronously monitors the circumferential pressure gradient, and the fiber optic sensor captures the microsecond-level impact signal of the tip cavitation collapse in real time. The six-degree-of-freedom attitude sensor continuously outputs the ship's roll, pitch and heave data, and the MEMS gyroscope records the three-dimensional vibration spectrum of the propeller shafting. All the original data is dynamically denoised by the Kalman filter, and the flow velocity measurement deviation caused by the ship's roll is compensated in real time through coordinate transformation. The preprocessed data stream is transmitted to the control center through the fiber optic ring network with a bandwidth of 1 Gbps, forming a real-time state map containing 40-dimensional feature vectors such as pressure, flow velocity, attitude, and vibration. After receiving the state map, the intelligent control and decision-making module immediately starts the digital twin simulation engine. Based on the real-time sensor data, a three-dimensional flow field model is reconstructed, and the GPU-accelerated solver completes the prediction of the vorticity field evolution in the next 0.5 seconds within 20 milliseconds. The reinforcement learning controller synchronously analyzes the historical action sequence and the current environmental characteristics, and generates joint optimization instructions for the angle of attack adjustment, guide vane spacing adjustment and axial position matching through the policy network. The model predictive control algorithm rolls and optimizes the next three-step control sequence to balance the immediate energy efficiency and the long-term mechanical loss. For example, when it is predicted that the tip vortex intensity is about to exceed the standard, the algorithm issues a preliminary instruction to increase the guide vane spacing by 5% 0.2 seconds in advance. The dynamic execution and adjustment module drives the hydraulic servo mechanism and the electric push rod according to the control instructions. The hydraulic servo valve responds to the angle of attack adjustment requirement within 5 milliseconds, the double-acting hydraulic cylinder drives the guide wheel to rotate with an accuracy of 0.01 mm, and the displacement sensor real-time feeds back the actual angle and closes the loop to correct the trajectory. At the same time, the electric push rod assembly dynamically adjusts the axial position of the guide wheel according to the change of the propeller speed. When the speed increases to 90% of the rated value, the axial spacing automatically shortens to the minimum design distance to enhance the flow guiding effect. The parameters such as oil pressure, current and temperature during the execution process are real-time transmitted back to the decision-making module to form a control closed loop. The cavitation warning and suppression module monitors the flow field risks throughout the process. The hydrophone array continuously analyzes the noise characteristics of the 20-100kHz frequency band. When the specific spectrum pattern of the initial cavitation is detected, the system immediately initiates a four-level response: at the observation level, the guide wheel angle of attack is slightly adjusted and the flow field changes are recorded; after entering the warning level, the guide vane spacing is automatically expanded by 8% to reduce the local flow velocity; if the number of cavitations continues to be lower than the critical value, the danger level triggers the guide wheel avoidance attitude adjustment, the angle of attack is quickly deflected to a safe position, and the propeller is linked to reduce speed by 10%; once the shock wave signal of cavitation collapse is detected, the emergency level immediately cuts off the power and starts the reverse high-pressure water jet to clear the cavitation group within 0.5 seconds. The self-learning fault-tolerant module operates continuously in the background. Every morning, the microbubble injection device starts automatically, and the critical parameters of the cavitation model are calibrated through cross-validation of high-speed video and acoustic monitoring. The multi-sensor data fusion engine calculates the confidence weight of each sensor. When the pressure array fails partially, the system automatically increases the proportion of CFD prediction data in the decision-making. The material fatigue model dynamically predicts the remaining life of the guide wheel based on the stress cycle data of the strain sensor. When the cumulative damage reaches 80%, the system limits the angle of attack adjustment to 60% of the normal range to extend the service life. Under abnormal operating conditions (such as the hydraulic oil temperature exceeds 70°C), the safety control mode forces a reduction in the adjustment frequency and starts the backup air cooling system. The six modules are closely linked through the closed loop of "perception-decision-execution-optimization". For example, when a ship encounters a sudden lateral current that causes a deviation in heading, the attitude sensor detects abnormal roll, and the multi-physics field perception module immediately increases the velocity sampling rate; the digital twin engine predicts the distortion trend of the wake field, and the reinforcement learning controller generates an angle of attack compensation sequence; the hydraulic mechanism completes a 5° angle of attack adjustment within 0.1 seconds, and the synchronous electric push rod moves the guide wheel position backward to balance the flow field; the cavitation module temporarily activates the enhanced monitoring mode after detecting a sudden drop in pressure; the self-learning fault-tolerant module records the characteristics of this event for optimizing the response strategy for similar working conditions in the future. The entire system continuously iterates at a refresh rate of 10Hz to ensure that the ship maintains optimal propulsion efficiency and safety margin under various sea conditions. It should be noted that this embodiment can still control the fluctuation of propulsion efficiency within ±3% under level 6 sea conditions, reduce the cavitation-related failure rate to 0.5 times / thousand hours, and extend the life of the guide wheel mechanism to 2.3 times that of the traditional design, realizing the deep integration of fluid optimization and intelligent control.

[0078] This embodiment also provides a computer device, which is applicable to the energy-saving control method for a pre-rotation guide wheel of a ship propulsion system and a fluid optimization system, including: a memory and a processor; the memory is used to store computer-executable instructions, and the processor is used to execute computer-executable instructions to implement the energy-saving control method for a pre-rotation guide wheel of a ship propulsion system and a fluid optimization system as proposed in the above embodiments.

[0079] The computer device may be a terminal, which includes a processor, a memory, a communication interface, a display screen, and an input device connected via a system bus. Among them, the processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs in the non-volatile storage medium. The communication interface of the computer device is used to communicate with external terminals in a wired or wireless manner, and the wireless manner can be achieved through WIFI, carrier networks, NFC (Near Field Communication), or other technologies. The display screen of the computer device can be a liquid crystal display screen or an electronic ink display screen. The input device of the computer device can be a touch layer covering the display screen, or a button, trackball, or touchpad provided on the housing of the computer device, or an external keyboard, touchpad, or mouse, etc.

[0080] This embodiment also provides a storage medium, on which a computer program is stored. When the program is executed by a processor, it implements the energy-saving control method for the pre-whirl guide wheel of the ship propulsion system and the fluid optimization system as proposed in the above embodiment; the storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (Static Random Access Memory, abbreviated as SRAM), electrically erasable programmable read-only memory (Electrically Erasable Programmable Read-Only Memory, abbreviated as EEPROM), erasable programmable read-only memory (Erasable Programmable Read Only Memory, abbreviated as EPROM), programmable read-only memory (Programmable Red-Only Memory, abbreviated as PROM), read-only memory (Read-Only Memory, abbreviated as ROM), magnetic memory, flash memory, magnetic disk, or optical disc.

[0081] In summary, through the asymmetric guide vane design and dynamic parameter optimization of the present invention, the system propulsion efficiency is increased by 15% - 30%, and the measured energy-saving effect is 2 - 3 times higher than that of the traditional solution; the combination of reinforcement learning and model predictive control reduces the energy loss fluctuation by more than 40%, and it can still maintain high-efficiency operation especially under variable working conditions (such as wave interference and load mutation). The linkage between the four-level early warning mechanism and the active suppression strategy improves the full-cycle suppression efficiency from the initial formation to the collapse of cavitation to more than 90%, and the cavitation incidence rate is reduced from 8.7% of the traditional solution to within 1.2%; the synergistic effect of the guide vane curvature optimization and the reverse flushing technology reduces the cavitation collapse impact force by 60% - 80%, significantly reducing cavitation corrosion damage. The coordinated control of the hydraulic servo system and the intelligent algorithm achieves a 0.1-second-level dynamic response, which is 10 times faster than the traditional PID control (1-second-level); the adjustment accuracy of the guide vane attack angle reaches ±0.5°, and the dynamic matching accuracy error of the guide vane spacing is ≤3%, ensuring the precise synchronization of the flow field and mechanical actions. The self-learning fault tolerance module reduces the false alarm rate to below 3% through periodic calibration and multi-sensor fusion decision-making, and can still maintain more than 80% of its performance under fault conditions; the material fatigue early warning function extends the service life of the guide wheel by 2 - 3 times, and the maintenance cycle is extended from the traditional 3 months to 6 - 8 months. The digital twin model covers the load range of 0 - 100% and sea conditions of 3 - 12 levels, and the efficiency fluctuation of the system under extreme working conditions (such as typhoon surges) is reduced by 60% compared with the traditional solution; the reinforcement learning algorithm can adapt to environmental changes in real time, and the speed adaptation range is extended to 5 - 30 knots, meeting the diverse needs of inland to ocean-going ships. The annual average fuel consumption of a single ship is reduced by 10% - 25%, and the carbon dioxide emissions are reduced by 200 - 500 tons / year; the system manufacturing cost increases by 15% - 20%, but the comprehensive life-cycle benefit (energy saving + reduced maintenance cost) is increased by 3 - 5 times. This system establishes a fluid optimization foundation through the innovative design of the guide wheel structure, the multi-physical field perception module captures the operating state in real time, the intelligent control module generates optimization instructions by integrating the digital twin and artificial intelligence algorithms, the dynamic actuator realizes precise adjustment, and the cavitation early warning and suppression module actively defends against abnormal flow. The self-learning mechanism continuously improves the system adaptability. Each module forms a closed loop of "perception - decision - execution - optimization", significantly improving the propulsion efficiency and ensuring the operation safety. Through the deep integration of fluid-structure interaction optimization, intelligent control and active safety defense, the present invention provides core technical support for green ships and intelligent shipping.

[0082] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit them. Although the present invention has been described in detail with reference to the preferred embodiments, those of ordinary skill in the art should understand that the technical solutions of the present invention can be modified or equivalently replaced without departing from the spirit and scope of the technical solutions of the present invention, and they should all be covered within the scope of the claims of the present invention.

Claims

1. A method for energy-saving control of a pre-rotation guide wheel of a ship propulsion system, characterized in that: The following steps are involved: S1. Real-time acquisition of propeller speed through a multi-physics sensor array , propeller wake flow velocity , guide wheel inlet pressure difference and ship speed ; S2. Based on real-time data, a transient CFD model of propeller-steering wheel-wake field coupling is constructed, and the Navier-Stokes equations are solved by GPU acceleration to predict the evolution law of the vortex field: in is the kinematic viscosity of seawater, is the vorticity of propeller wake field; S3. Calculate the optimal attack angle of the pre-spin guide wheel based on the guide wheel angle dynamic compensation algorithm , the formula is: in , is the guide wheel response coefficient, the value range is 0.5≤ ≤1.2,0.02≤ ≤0.08; S4. Generate guide wheel adjustment instructions through the energy efficiency optimization decision model, and drive the hydraulic servo mechanism to adjust the guide wheel angle with a dynamic response of 0.1 seconds; S5. Establishing the vorticity of propeller wake field based on deep learning algorithm Mapping relationship with guide wheel parameters to optimize guide wheel attack angle Real-time compensation ,satisfy: ； The control strategy is continuously optimized through the reinforcement learning PPO algorithm.

2. The energy-saving control method for pre-rotation guide wheel of a ship propulsion system according to claim 1, characterized in that: The multi-physics field sensor array in step S1 includes: The micro pressure sensor is arranged at the front edge of the guide wheel, with a measurement accuracy of ≤0.1kPa; MEMS gyroscope installed at the propeller shaft end, sampling frequency ≥ 200Hz; Wake field PIV velocity measurement module, spatial resolution ≤1mm.

3. The energy-saving control method for pre-rotation guide wheel of a ship propulsion system according to claim 1, characterized in that: The hydraulic servo mechanism in step S4 has an angle adjustment accuracy of ±0.5° and an output torque of satisfy: in is the density of seawater, is the guide wheel diameter.

4. The energy-saving control method for pre-rotation guide wheel of a ship propulsion system according to claim 1, characterized in that: The objective function of the energy efficiency optimization decision model in step S4 is: in For propulsion power, For system efficiency, is the eddy current suppression weight factor, 0.1≤ ≤0.

3.

5. The energy-saving control method for pre-rotation guide wheel of a ship propulsion system according to claim 1, characterized in that: The deep learning algorithm in step S5 adopts a convolutional neural network CNN, the input layer includes the wake field velocity cloud map, the pressure distribution matrix and the historical action data of the guide wheel, and the output layer predicts the optimal guide wheel attack angle , the network loss function is: in is the penalty factor for vorticity change (0.05≤ ≤0.1), is the true value or target value of the actual optimal guide wheel attack angle.

6. The energy-saving control method for pre-rotation guide wheel of a ship propulsion system according to claim 1, characterized in that: The reinforcement learning PPO algorithm training strategy in step S5 adopts online-offline hybrid sampling, the online data sampling is the real-time ship operation data with a sampling frequency of 10 Hz, and the offline data sampling is the extreme working condition data set generated by CFD simulation. The importance sampling constraint limits the strategy update amplitude, and the KL divergence threshold δ=0.01; The control strategy is continuously optimized through the reinforcement learning PPO algorithm, and the reward function is designed as: in To improve efficiency, is the cavitation number, is the critical cavitation number; The reinforcement learning PPO algorithm uses an adaptive learning rate. When the average reward fluctuation exceeds 10%, Learning Rate according to 0.95 attenuation.

7. A fluid optimization system for a pre-swirl guide wheel of a ship propulsion system, which is implemented based on the energy-saving control method for a pre-swirl guide wheel of a ship propulsion system according to any one of claims 1 to 6, characterized in that: include: The guide wheel structure optimization module is used to adopt an asymmetric variable curvature guide vane structure design. The guide vane trailing edge is integrated with an adaptive sawtooth vortex suppression structure, and the leading edge is equipped with an anti-cavitation corrosion coating. The guide vane installation angle and spatial distribution are optimized according to the fluid dynamics characteristics. The guide wheel support frame is lightweight and high in stiffness through topological optimization; The multi-physics field sensing module is used to collect propeller wake pressure, three-dimensional flow velocity, ship motion posture and guide wheel mechanical state data in real time through a distributed sensor array, and integrate the Kalman filter algorithm to compensate for the measurement error caused by the ship's swaying; An intelligent control and decision-making module, comprising a digital twin simulation engine and a reinforcement learning controller. The digital twin simulation engine is used to predict the vortex evolution trend through real-time fluid dynamics simulation. The reinforcement learning controller is used to generate guide wheel parameter adjustment instructions based on a multi-objective optimization strategy and integrate a model predictive control algorithm to achieve dynamic sequence optimization. The dynamic execution adjustment module is composed of a high-precision hydraulic servo mechanism and an electric push rod assembly, which are used to drive the guide wheel to quickly adjust the angle of attack, dynamically adjust the guide vane spacing, and adaptively match the axial position of the guide wheel. The axial spacing adjustment mechanism dynamically matches the optimal distance according to the propeller speed; Cavitation warning and suppression module, which is used to achieve cavitation initiation warning through acoustic monitoring and pressure pulsation analysis, trigger the four-level response mechanism and link the guide wheel avoidance posture adjustment, guide vane curvature optimization and emergency reverse flushing function; The self-learning fault-tolerant module is used to implement periodic parameter calibration and adversarial training, improve decision reliability through multi-sensor data fusion, predict the remaining life of the guide wheel in combination with the material fatigue model, and start the safety control mode under abnormal working conditions.

8. The pre-swirl guide wheel fluid optimization system for a ship propulsion system according to claim 7, characterized in that: The curvature radius of the guide vane leading edge of the asymmetric variable curvature guide vane structure Trailing edge curvature radius satisfy: in is the guide vane installation angle, the value range is 15 ≤ ≤45 , and the ratio of the guide vane chord length L to the guide wheel diameter D is L / D=0.12~0.

18.

9. The pre-swirl guide wheel fluid optimization system for a ship propulsion system according to claim 7, characterized in that: The pressure gradient distribution on the guide vane surface of the asymmetric variable curvature guide vane structure satisfies: Where P is the surface pressure of the guide vane, C is the correction coefficient of the guide vane profile (0.8≤C≤1.2), is a function of the radius of curvature along the chord length.

10. The pre-swirl guide wheel fluid optimization system for a ship propulsion system according to claim 7, characterized in that: The trailing edge of the guide vane of the asymmetric variable curvature guide vane structure is provided with a sawtooth vortex generator, the ratio of the sawtooth height h to the guide vane chord length L is h / L=0.02-0.05, and the sawtooth inclination angle β satisfies: in is the critical cavitation velocity.

11. The pre-swirl guide wheel fluid optimization system for a ship propulsion system according to claim 7, characterized in that: The cavitation warning suppression module calculates the cavitation number according to the pressure field distribution ,when The guide wheel emergency avoidance mode is triggered; The cavitation warning suppression module includes a distributed pressure sensing array and an acoustic monitoring unit; The distributed pressure sensing array arranges high-frequency dynamic pressure sensors at the leading edge of the guide wheel, the root and the tip of the propeller blade, with a sampling rate of ≥5kHz, forming a sandwich measurement structure: Leading edge sensor group: measuring local pressure pulsations ; Blade root ring array: monitoring pressure gradients ; Blade tip fiber Bragg grating sensor: Detecting cavitation collapse shock waves , sensitivity reaches 0.01MPa; The acoustic monitoring unit uses a hydrophone array to capture the characteristic frequency of cavitation noise and extracts energy characteristics through wavelet packet decomposition.

Images