Pressure pulsation suppression system of vane pump

Abstract

The invention discloses a vane pump pressure pulsation suppression system which is characterized in that a multi-dimensional working condition sensing module collects real-time working condition data, dynamic external correction data provided by an external variable introduction and correction module is combined, future pressure pulsation and cavitation states are synchronously predicted through an LSTM prediction control module, and a cooperative suppression instruction is generated. The instruction drives the cavitation suppression module to execute active cavitation suppression operation on the oil suction pipeline and drives the piezoelectric execution module to adjust the volume of the elastic cavity in the pump body so as to offset the pressure pulsation at the same time. And the closed-loop feedback optimization module calculates and feeds back errors according to the suppression effect, and dynamically corrects prediction model parameters and external variable correction weights. According to the invention, the cooperative inhibition of cavitation and pressure pulsation and the adaptive optimization of the overall performance of the system are realized.

Description

Technical Field

[0001] This invention relates to the fields of intelligent control and fluid machinery engineering technology, and in particular to a pressure pulsation suppression system for a vane pump. Background Technology

[0002] As a commonly used fluid power component in hydraulic systems, vane pumps inevitably generate pressure pulsations during operation. These pulsations not only cause pipeline vibration and structural noise but also accelerate the wear of seals and critical moving parts, reducing system stability and service life. In applications such as high-end hydraulic equipment and precision machine tools, where high fluid stability and dynamic response are required, pressure pulsations are particularly prominent and have become a significant factor restricting equipment performance improvement.

[0003] As vane pumps evolve towards higher speeds, higher power densities, and more complex load conditions, their operating states exhibit significant nonlinear and time-varying characteristics. Traditional pulsation suppression methods based on single operating parameters struggle to maintain stable performance under complex conditions. On one hand, vane pumps are affected not only by their own structural parameters and instantaneous operating conditions during actual operation but also by dynamic disturbances from various external factors such as environmental conditions and downstream load changes. The combined effect of these factors can easily lead to rapid changes in the system state, making the pulsation characteristics difficult to predict accurately. On the other hand, existing suppression strategies typically rely on reactive responses, lacking the ability to effectively predict future operating trends. Under conditions of sudden changes in operating conditions or high-frequency pulsation, they are prone to response lag or insufficient suppression.

[0004] Furthermore, during the operation of vane pumps, there is often a significant interaction between fluid cavitation and pressure pulsation. When a local pressure drop triggers cavitation, the formation and collapse of bubbles further exacerbate pressure fluctuations; conversely, increased pressure pulsation may induce or worsen cavitation, creating a vicious cycle that is detrimental to stable system operation. Under complex loads or high-speed conditions, suppressing only a single pressure pulsation is often insufficient to fundamentally improve the overall operating condition, and the suppression effect diminishes significantly with changes in operating conditions.

[0005] Therefore, how to fully perceive the multi-dimensional operating state of the vane pump, comprehensively consider the influence of external dynamic factors, make forward-looking predictions on pressure pulsation and its related unstable factors, and on this basis achieve effective regulation of the fluid state, so as to achieve more stable and precise operation control under complex working conditions, remains a technical problem that urgently needs to be solved in this field. Summary of the Invention

[0006] In view of the shortcomings of the existing technology, the purpose of this invention is to provide a pressure pulsation suppression system for vane pumps, which can realize the advanced prediction, coordinated suppression and adaptive optimization of pressure pulsation and cavitation phenomena, and significantly improve the operational stability and reliability of vane pumps under complex and variable operating conditions.

[0007] To achieve the above objectives, the present invention provides the following technical solution: a pressure pulsation suppression system for a vane pump, comprising: A multi-dimensional operating condition sensing module is used to collect real-time operating condition data of the vane pump, including basic operating condition data and cavitation monitoring data. The external variable introduction and correction module is used to collect dynamic external variables and process the dynamic external variables to generate external correction data with dynamic weights. The long short-term memory neural network prediction and control module is connected to the multi-dimensional working condition perception module and the external variable introduction and correction module, respectively. It is used to receive the real-time working condition data and the external correction data, and predict the future pressure pulsation and cavitation state based on the data, and generate a cooperative suppression command that includes piezoelectric drive command and cavitation suppression command. A cavitation-pulsation coordinated suppression module is connected to the long short-term memory neural network prediction and control module, and includes a cavitation suppression unit disposed in the oil suction port pipeline of the vane pump, for driving the cavitation suppression unit to perform active cavitation suppression operation according to the cavitation suppression command; A distributed piezoelectric-elastic cavity execution module is connected to the long short-term memory neural network predictive control module, and includes an elastic cavity formed in the inner wall of the stator of the vane pump and a corresponding piezoelectric ceramic actuator, for driving the piezoelectric ceramic actuator to adjust the volume of the elastic cavity to counteract pressure pulsation according to the piezoelectric drive command; The closed-loop feedback optimization module is connected to the multi-dimensional working condition perception module and the long short-term memory neural network prediction control module, respectively. It is used to calculate the pulsation suppression error and the cavitation suppression error based on the suppressed actual working condition data, and feed the pulsation suppression error and the cavitation suppression error back to the long short-term memory neural network prediction control module to dynamically correct the prediction model parameters and the correction weights of the external variable introduction and correction module.

[0008] Furthermore, the multi-dimensional working condition perception module includes: A data acquisition unit, used to acquire the real-time operating data, includes: The basic operating condition data acquisition subunit includes pressure sensors, vibration sensors, flow sensors, speed sensors, and viscosity sensors respectively installed at the inlet flange, outlet flange, stator inner wall circumferential direction, pump shaft end, and fluid inlet pipeline of the vane pump. The cavitation monitoring data acquisition subunit includes at least two ultrasonic cavitation sensors installed in the cavitation-prone area of ​​the inner wall of the stator of the vane pump. The data transmission unit is connected to the data acquisition unit and is used to sample the basic operating condition data at a frequency not lower than a first preset frequency, sample the cavitation monitoring data at a frequency not lower than a second preset frequency, and transmit the sampled data to the long short-term memory neural network prediction and control module, wherein the second preset frequency is higher than the first preset frequency.

[0009] Furthermore, the external variable introduction and correction module includes: The variable acquisition unit is used to acquire various types of dynamic external variables; The preprocessing unit, connected to the variable acquisition unit, is used to filter and normalize the acquired dynamic external variables to obtain standardized variable data. The weight dynamic allocation unit, connected to the preprocessing unit and the closed-loop feedback optimization module, is used to dynamically allocate correction weights to each of the standardized variable data based on the current working conditions and according to a preset weight allocation strategy, so as to form the external correction data. The weight dynamic allocation unit is further configured to adjust the preset weight allocation strategy when the pulsation suppression error or the cavitation suppression error exceeds a preset error threshold.

[0010] Furthermore, the long short-term memory neural network prediction control module includes: An input vector construction unit is used to construct an input vector with a preset dimension from the real-time operating data and the external correction data, wherein the input vector includes at least the current value of the basic operating data, the first derivative value calculated from the basic operating data, the current value of the external correction data, and the current value of the cavitation monitoring data. A collaborative instruction generation unit, connected to the input vector construction unit, is used to input the input vector into a pre-trained long short-term memory neural network model, wherein the model synchronously outputs the piezoelectric driving instruction and the cavitation suppression instruction.

[0011] Furthermore, the cooperative instruction generation unit includes: The joint prediction subunit is used to simultaneously predict, based on the input vector and through the long short-term memory neural network model, the pressure pulsation characteristic parameters in the first preset time period and the cavitation occurrence probability and location parameters in the second preset time period. The instruction mapping subunit, connected to the joint prediction subunit, is used to map the pressure pulsation characteristic parameters into piezoelectric drive commands for different elastic cavities according to a preset mapping strategy, and to map the cavitation occurrence probability and position parameters into cavitation suppression commands for different cavitation suppression units.

[0012] Furthermore, the cavitation-pulsation synergistic suppression module includes: The instruction parsing and distribution unit, connected to the long short-term memory neural network prediction and control module, is used to receive and parse the cavitation suppression instruction; The cavitation suppression unit, connected to the instruction parsing and distribution unit, includes: The solenoid valve control subunit is used to adjust the opening degree of the miniature electromagnetic proportional valve in the cavitation suppression unit according to the parsed instructions. The gas injection subunit is used to control the low-pressure gas supply component to inject inert gas at a preset pressure into the oil suction line according to the parsed instructions. The flow fine-tuning subunit is used to fine-tune the fluid flow rate of the oil suction port pipeline according to the parsed instructions.

[0013] Furthermore, the cavitation-pulsation synergistic suppression module also includes: The linkage control unit, connected to the instruction parsing and distribution unit, is used to generate a piezoelectric drive compensation instruction after the cavitation suppression unit is activated according to the instruction; The piezoelectric drive compensation command is used to instruct the long short-term memory neural network predictive control module or the distributed piezoelectric-elastic cavity execution module to adjust the drive parameters of the piezoelectric drive command or the piezoelectric ceramic actuator to compensate for pressure fluctuations that may be caused by performing the active cavitation suppression operation.

[0014] Furthermore, the distributed piezoelectric-elastic cavity actuation module includes: The structure execution unit includes: The cavity assembly includes a plurality of annular elastic cavities opened at a predetermined circumferential interval on the inner wall of the stator of the vane pump; A piezoelectric drive assembly includes stacked piezoelectric ceramic sheets disposed corresponding to each of the annular elastic cavities, wherein each of the piezoelectric ceramic sheets is fixed to the outer wall of the corresponding annular elastic cavity; A drive control unit, connecting the long short-term memory neural network predictive control module and the piezoelectric drive assembly, includes: A signal receiving and parsing subunit is used to receive and parse the piezoelectric drive command; The parameter compensation subunit, connected to the signal receiving and parsing subunit, is used to compensate and adjust the driving parameters of the parsed piezoelectric driving command according to the activation state of the cavitation suppression command. The high-voltage drive subunit, connected to the parameter compensation subunit and the piezoelectric drive assembly, is used to generate a high-voltage drive signal based on the compensated and adjusted drive parameters to control the expansion and contraction of the corresponding piezoelectric ceramic sheet.

[0015] Furthermore, the closed-loop feedback optimization module includes: The dual error calculation unit, connected to the multi-dimensional working condition sensing module, is used to calculate the pulsation suppression error and cavitation suppression error respectively based on the collected suppressed actual working condition data. The collaborative triggering unit, connected to the dual error calculation unit, is used to generate a model correction trigger signal when the pulsation suppression error exceeds a first preset error threshold or the cavitation suppression error exceeds a second preset error threshold. The model collaborative correction unit, connected to the collaborative triggering unit, is used to respond to the model correction triggering signal, package the pulsation suppression error, the cavitation suppression error, the current operating parameters and the current driving command into a collaborative correction data packet, and send it to the long short-term memory neural network prediction control module.

[0016] Furthermore, the closed-loop feedback optimization module also includes: Self-learning iterative units include: The data association storage subunit is used to continuously store the associated dataset consisting of the current operating condition parameters, the external correction data, the cavitation monitoring data, the collaborative suppression command, and the actual operating condition data after suppression; The offline model training subunit is connected to the data association storage subunit. When the size of the stored association dataset reaches a preset threshold, iterative training of the prediction model in the long short-term memory neural network prediction control module is initiated. During training, the weight allocation strategy of the external variable introduction and correction module and the collaborative logic of the cavitation-pulsation collaborative suppression module and the distributed piezoelectric-elastic cavity execution module are optimized simultaneously.

[0017] The beneficial effects of this invention are: 1. Strong synergistic suppression capability: By comprehensively sensing and predicting multi-dimensional operating data and dynamic external variables of the vane pump, it can make forward-looking judgments on pressure pulsation and cavitation state, and implement synergistic control based on the prediction results, effectively avoiding the problems of response lag or single action of traditional suppression methods, and significantly improving the overall stability of the system.

[0018] 2. High suppression accuracy and stability: Based on the cooperative suppression command generated by predictive control, it can actively adjust before pulsation and cavitation occur; through precise physical control of the fluid state, it reduces the amplitude of pulsation and suppresses cavitation inducing factors, so that the vane pump can maintain a low pressure fluctuation level and stable operation under complex working conditions.

[0019] 3. Good adaptability and robustness to operating conditions: By introducing external variables and performing adaptive correction, the impact of environmental changes, load fluctuations and other factors on prediction accuracy is reduced; combined with a closed-loop feedback mechanism, the model parameters are dynamically corrected, enabling the system to adapt to the operating requirements under different speed, flow and medium conditions.

[0020] 4. The coupling effect of cavitation and pulsation is effectively weakened: In response to the coupling relationship between cavitation and pressure pulsation, the probability of cavitation and its impact on pressure fluctuations are reduced simultaneously through synergistic suppression, thereby avoiding a vicious cycle and improving the long-term reliability and lifespan of the vane pump.

[0021] 5. Strong feasibility and wide applicability: This solution is based on conventional sensing, control and execution components, without the need for major modifications to the main structure of the vane pump. It is easy to integrate into existing hydraulic systems and is suitable for fluid transportation and hydraulic equipment applications with high stability and high precision requirements. Attached Figure Description

[0022] Figure 1 This is a schematic diagram of the pressure pulsation suppression system for the blade pump in this invention; Figure 2 This is a schematic diagram of the data acquisition unit in this invention; Figure 3 This is a schematic diagram of the structure of the cooperative instruction generation unit in this invention; Figure 4 This is a schematic diagram of the cavitation suppression unit in this invention; Figure 5 This is a schematic diagram of the drive control unit in this invention; Figure 6 This is a schematic diagram of the structure of the self-learning iterative unit in this invention.

[0023] Figure reference numerals: 1. Multi-dimensional working condition perception module; 11. Data acquisition unit; 111. Basic working condition data acquisition sub-unit; 112. Cavitation monitoring data acquisition sub-unit; 12. Data transmission unit; 2. External variable introduction and correction module; 21. Variable acquisition unit; 22. Preprocessing unit; 23. Weight dynamic allocation unit; 3. Long short-term memory neural network prediction and control module; 31. Input vector construction unit; 32. Cooperative instruction generation unit; 321. Joint prediction sub-unit; 322. Instruction mapping sub-unit; 4. Cavitation-pulsation cooperative suppression module; 41. Instruction parsing and distribution unit; 42. 421. Suppression Unit; 422. Solenoid Valve Control Subunit; 423. Gas Injection Subunit; 424. Flow Fine-tuning Subunit; 43. Linkage Control Unit; 5. Distributed Piezoelectric-Elastic Chamber Execution Module; 51. Structural Execution Unit; 52. Drive Control Unit; 521. Signal Reception and Analysis Subunit; 522. Parameter Compensation Subunit; 523. High-Pressure Drive Subunit; 6. Closed-Loop Feedback Optimization Module; 61. Dual Error Calculation Unit; 62. Collaborative Triggering Unit; 63. Model Collaborative Correction Unit; 64. Self-Learning Iteration Unit; 641. Data Association and Storage Subunit; 642. Offline Model Training Subunit. Detailed Implementation

[0024] The present invention will be further described in detail below with reference to the accompanying drawings and embodiments. Identical components are denoted by the same reference numerals. It should be noted that the terms "front," "rear," "left," "right," "upper," and "lower" used in the following description refer to directions in the accompanying drawings, and the terms "bottom surface," "top surface," "inner," and "outer" refer to directions toward or away from the geometric center of a specific component, respectively.

[0025] Reference Figure 1 Example 1 discloses a pressure pulsation suppression system for a vane pump.

[0026] I. Overall System Structure; This embodiment provides a pressure pulsation and cavitation synergistic suppression system for a vane pump, comprising: a multi-dimensional operating condition sensing module 1, an external variable introduction and correction module 2, a long short-term memory neural network predictive control module 3, a cavitation-pulsation synergistic suppression module 4, a distributed piezoelectric-elastic cavity execution module 5, and a closed-loop feedback optimization module 6.

[0027] The above modules achieve data interaction and collaborative control through an industrial controller, which is used to predictively suppress pressure pulsation and cavitation phenomena in vane pumps under complex operating conditions.

[0028] II. Module and Hardware Selection Guidelines; (a) Multi-dimensional working condition perception module 1; This module is used to collect basic operating condition data and cavitation monitoring data during the operation of the vane pump in real time.

[0029] The basic operating condition data includes: pressure data: collected by PTX5072 pressure sensor; vibration data: collected by 356A15 vibration sensor; flow rate data: collected by FS400A flow sensor; rotational speed data: collected by HS350 rotational speed sensor; fluid viscosity data: collected by SV-10 viscosity sensor.

[0030] Cavitation monitoring data includes cavitation intensity signals collected by the V306 ultrasonic cavitation sensor, which are used to reflect the cavitation status in the oil suction zone and oil discharge transition zone of the vane pump.

[0031] All collected data is transmitted to the industrial controller after signal conditioning, and simultaneously sent to the long short-term memory neural network predictive control module 3.

[0032] Technical benefits: By sensing operating conditions from multiple sources and in multiple dimensions, it provides a complete and real-time data foundation for subsequent prediction and control, avoiding misjudgments of the state caused by relying on only a single parameter.

[0033] (ii) External Variable Introduction and Correction Module 2; This module is used to collect and process dynamic external variables that affect system stability during the operation of the vane pump, including the following four categories: Ambient temperature variable: acquired by PT100 temperature sensor, accuracy ±0.2℃; Downstream load variables: acquired by Rosemount 3051 pressure transmitter, accuracy ±0.1%FS; Power supply voltage variation: acquired by ZMPT101B voltage sensor, with an accuracy of ±0.5%; Operating disturbance variable: Load change rate parameter calculated internally by the control system.

[0034] This module performs filtering, normalization, and weight allocation on the aforementioned external variables to generate externally corrected data with dynamic weights, which is then transmitted to the Long Short-Term Memory Neural Network Prediction and Control Module 3.

[0035] Technical effect: By introducing external variables and performing adaptive correction, the impact of environmental changes, load fluctuations and other factors on prediction accuracy is effectively reduced, and the stability and robustness of prediction results are improved.

[0036] (iii) Long Short-Term Memory Neural Network Prediction and Control Module 3; This module uses an LSTM model deployed in industrial controllers to predict future operating conditions.

[0037] The model inputs include: historical operating condition data, real-time basic operating condition data, cavitation monitoring data, and external variable correction data.

[0038] The LSTM model simultaneously predicts the following: pressure pulsation characteristics, cavitation probability, and possible regions within the next 0.1 to 0.3 seconds.

[0039] In this embodiment, the target control value for pressure pulsation is explicitly set to 0.2 MPa, and the threshold for cavitation occurrence probability is set to 30%.

[0040] When the prediction result exceeds the above threshold, the LSTM model generates a cooperative suppression instruction containing piezoelectric drive instructions and cavitation suppression instructions, and sends them to the cavitation-pulsation cooperative suppression module 4 and the distributed piezoelectric-elastic cavity execution module 5, respectively.

[0041] Technical benefits: By using predictive control, intervention can be made in advance before pulsation and cavitation are significantly formed, eliminating the time lag problem of traditional ex-post response control.

[0042] III. Execution and Cooperative Inhibition Structures; (iv) Cavitation-Pulsation Synergistic Suppression Module 4; This module is located at the oil inlet pipe of the vane pump and includes a cavitation suppression unit 42, which is used to perform active cavitation suppression operation according to the cavitation suppression command.

[0043] The cavitation suppression unit 42 specifically includes: a miniature electromagnetic proportional valve (4V210-08 type); a low-pressure gas supply component (QPB-01 type, gas supply pressure 0.1~0.3MPa); an inert gas supply source (industrial grade nitrogen, purity ≥99.9%); and an oil suction flow fine adjustment mechanism.

[0044] When the probability of cavitation is predicted to exceed 30%, cavitation is suppressed in a coordinated manner through the following methods: adjusting the opening of the electromagnetic proportional valve; performing low-pressure air replenishment; and fine-tuning the oil suction flow distribution.

[0045] Technical effect: It weakens the conditions for cavitation formation in advance, avoids the collapse of cavitation bubbles from causing or amplifying pressure pulsations, and reduces instability factors from the source.

[0046] (v) Distributed piezoelectric-elastic cavity actuation module 5; The module includes: multiple elastic cavities disposed on the inner wall of the stator of the vane pump and piezoelectric ceramic actuators disposed corresponding to the elastic cavities.

[0047] The structural parameters of the elastic cavity are as follows: Quantity: 4 annular elastic cavities; Axial length: 50mm; Radial thickness: 6mm; Inner wall material: hydrogenated nitrile rubber (Shore A 75); Outer wall material: No. 45 steel.

[0048] The piezoelectric actuator uses PZT-8 laminated piezoelectric ceramic sheets and is controlled by an OPA548 high-voltage drive circuit.

[0049] When a piezoelectric drive command is received, each piezoelectric ceramic sheet expands and contracts precisely according to the predicted results, adjusting the corresponding elastic cavity volume to form a reverse pressure wave inside the pump that is opposite in phase to the original pressure pulsation, thus achieving precise cancellation.

[0050] Technical effect: By using a point-to-point volume adjustment method, high-precision physical cancellation of pressure pulsation is achieved, thereby improving the suppression effect.

[0051] IV. Closed-loop feedback optimization mechanism; (vi) Closed-loop feedback optimization module 6; This module is used to collect actual pressure pulsation data and cavitation suppression effect data after suppression, and to calculate pulsation suppression error and cavitation suppression error. The above error data is fed back to the Long Short-Term Memory Neural Network Predictive Control Module 3, which is used to dynamically correct LSTM model parameters, adjust external variable correction weights, and optimize piezoelectric drive commands and cavitation suppression commands.

[0052] Under stability test conditions (72 hours of continuous operation), the system meets the following requirements: prediction error ≤ 2%; cavitation probability ≤ 20%.

[0053] Technical effect: Through a closed-loop self-learning mechanism, the inhibition effect is continuously optimized and the long-term stable operation is achieved.

[0054] V. Explanation of overall working principle; This embodiment constructs a complete system state perception foundation by introducing multi-dimensional operating condition perception and adaptive external variables; then, it uses an LSTM model to predict the pressure pulsation and cavitation state in the near future, and drives the cavitation suppression and pressure pulsation cancellation execution structure synchronously according to the prediction results; finally, it continuously corrects the model and control parameters through closed-loop feedback to achieve high-precision and highly adaptive collaborative suppression control of the vane pump under complex operating conditions.

[0055] Example 2 is the second embodiment of the present invention.

[0056] I. Structure and working principle of multi-dimensional working condition sensing module 1; In this embodiment, the multi-dimensional working condition perception module 1 includes a data acquisition unit 11 and a data transmission unit 12.

[0057] Reference Figure 2 The data acquisition unit 11 includes: The basic operating condition data acquisition subunit 111 is installed at the inlet flange, outlet flange, stator inner wall circumferential direction, pump shaft end and fluid inlet pipeline of the vane pump, and is used to collect pressure, vibration, flow rate, speed and fluid viscosity data during the operation of the vane pump. The cavitation monitoring data acquisition subunit 112 is located in the cavitation-prone area on the inner wall of the vane pump stator, including the transition between the vane oil suction area and the oil discharge area, and is used to collect high-frequency acoustic signals caused by the generation and collapse of cavitation bubbles.

[0058] The data transmission unit 12 is connected to the data acquisition unit 11 and is used to sample basic operating condition data at a frequency not lower than a first preset frequency and to sample cavitation monitoring data at a frequency not lower than a second preset frequency, wherein the second preset frequency is higher than the first preset frequency; in this embodiment, the first preset frequency is 10kHz and the second preset frequency is 20kHz.

[0059] The sampled data is transmitted to the Long Short-Term Memory Neural Network Predictive Control Module 3 via industrial Ethernet, with a data transmission delay of no more than 1ms.

[0060] Technical effect: By acquiring and transmitting basic operating condition data and cavitation monitoring data at different frequencies, the system can accurately reflect the overall operating status of the vane pump and capture the high-frequency transient characteristics caused by cavitation, providing high-time-resolution data support for subsequent prediction.

[0061] II. Structure and working principle of external variable introduction and correction module 2; In this embodiment, the external variable introduction and correction module 2 includes: a variable acquisition unit 21, a preprocessing unit 22, and a weight dynamic allocation unit 23.

[0062] The variable acquisition unit 21 is used to acquire various types of dynamic external variables, including: ambient temperature variables, downstream load fluctuation variables, power supply voltage fluctuation variables, and fluid medium aging variables. The fluid medium aging variables are characterized by the rate of change of fluid viscosity.

[0063] The preprocessing unit 22 is connected to the variable acquisition unit 21 and is used to filter and normalize the acquired dynamic external variables. The filtering uses the Kalman filter algorithm to eliminate random noise interference, and the normalization process is used to map each variable to the 0-1 interval to form standardized variable data.

[0064] The weight dynamic allocation unit 23 is connected to the preprocessing unit 22 and the closed-loop feedback optimization module 6. It is used to dynamically allocate correction weights to standardized variable data based on the current working conditions and form external correction data. The value range of the correction weight is 0.05–0.25.

[0065] When the pulsation suppression error or cavitation suppression error fed back by the closed-loop feedback optimization module 6 exceeds the 2% error threshold, the weight dynamic allocation unit 23 adjusts the preset weight allocation strategy to enhance the corrective effect of the corresponding external variables on the prediction results.

[0066] Technical effect: By introducing external variables that are strongly correlated with the operating status of the vane pump and dynamically adjusting their correction weights according to the suppression effect, the impact of environmental changes, load changes and power supply fluctuations on prediction accuracy is effectively reduced, and the robustness of the system under complex working conditions is improved.

[0067] III. Structure and principle of Long Short-Term Memory Neural Network Prediction and Control Module 3; In this embodiment, the long short-term memory neural network prediction control module 3 includes an input vector construction unit 31 and a cooperative instruction generation unit 32.

[0068] (a) Input vector construction unit 31; The input vector construction unit 31 is used to introduce the real-time operating condition data and external variables output by the multi-dimensional operating condition perception module 1 and the external correction data output by the correction module 2 to construct an input vector with a preset dimension.

[0069] In this embodiment, the input vector is a 16-dimensional vector, which includes at least: the current value of the basic operating condition data, the first derivative value of the basic operating condition data, the current value of the external correction data, and the current value of the cavitation monitoring data.

[0070] By introducing first-order derivative information, the input vector can simultaneously represent the current state and the trend of change.

[0071] (ii) Cooperative instruction generation unit 32; The cooperative instruction generation unit 32 is connected to the input vector construction unit 31, as shown in the reference. Figure 3 It includes: The joint prediction subunit 321 is used to simultaneously predict the pressure pulsation characteristic parameters in the first preset time period and the cavitation occurrence probability and location parameters in the second preset time period based on the input vector and through a pre-trained long short-term memory neural network model; in this embodiment, the first preset time period and the second preset time period are both 0.1–0.3s; The instruction mapping subunit 322 is used to map the predicted pressure pulsation characteristic parameters into piezoelectric drive commands for different elastic cavities according to a preset mapping strategy, and to map the predicted cavitation occurrence probability and position parameters into cavitation suppression commands for different cavitation suppression units 42.

[0072] When the predicted pressure pulsation peak exceeds the target threshold of 0.2 MPa, or the predicted cavitation probability exceeds the 30% probability threshold, the cooperative instruction generation unit 32 outputs the corresponding cooperative suppression instruction.

[0073] Technical effect: By synchronously outputting piezoelectric drive commands and cavitation suppression commands under a unified prediction framework, the coordinated control of pressure pulsation and cavitation is achieved, avoiding the control lag or suppression mismatch problems caused by the independent suppression methods in traditional systems.

[0074] Construction and inference rules of LSTM predictive control models: (a) Input vector construction; Construct a 16-dimensional input vector from the following data: the current value of the basic operating condition data, the first derivative value of the basic operating condition data, the current value of the external correction data, and the current value of the cavitation monitoring data.

[0075] (ii) Long Short-Term Memory Neural Network Model Structure; The structure of a long short-term memory neural network model includes: The input layer is 16-dimensional; there are three hidden layers with 128, 64, and 32 neurons respectively; the output layer is 10-dimensional and is used to output the piezoelectric driving parameters corresponding to the four elastic cavities and the driving parameters of the two cavitation suppression units 42; the activation function is ReLU; and the optimizer is Adam.

[0076] (iii) Joint prediction and instruction mapping; Simultaneous prediction using a long short-term memory neural network model: Pressure pulsation characteristic parameters within the first preset time period of 0.1–0.3s in the future; The probability and location parameters of cavitation occurring in the second preset time period of 0.1–0.3s.

[0077] When the predicted peak pressure pulsation exceeds the target threshold of 0.2 MPa or the predicted cavitation probability exceeds the threshold of 30%, a cooperative suppression command is generated.

[0078] The core algorithm formula and working mechanism of the Long Short-Term Memory (LSTM) neural network model: In this embodiment, to achieve synergistic prediction of pressure pulsation and cavitation, the following integrated prediction equation is introduced: ; in, As a predictive indicator for the combined effects of pressure pulsation and cavitation, To predict the length of the time window, The morphological parameters are obtained statistically from basic working condition data and used as the independent variables of the Gamma function. Let Gamma be the function, and N be the number of cavitation monitoring channels. Let be the characteristic index corresponding to the i-th type of external correction variable and use it as the independent variable of the Riemann Zeta function. For Riemann Zeta function, It is a Bessel function of the first kind and Let the order be related to the spatial distribution of the i-th monitoring point. To correspond to the angular frequency of the monitored signal, This is the scale normalization factor for the corresponding signal. The time decay coefficient, This is the cavitation signal enhancement coefficient. Let be the error function. The disturbance strength factor is formed by combining the weights of external variables. This is an information filtering function in elliptic integral form, used to characterize nonlinear operating condition disturbances; For fluid density parameters, For system state diffusion parameters, This is a time integral variable (continuous time variable) used to characterize the transient process of the vane pump's operating state evolving over time within the prediction time window. This is a normalization operator used to map the integral result to a dimensionless interval.

[0079] Comprehensive forecast indicators The range of is (0, +∞), where <1 indicates that the predicted state is in a stable range, and the corresponding pressure fluctuations and cavitation risks are controllable; ≥1 indicates that the predicted state has entered the unstable region, requiring the triggering of a cooperative suppression instruction, and A larger value indicates a higher risk of coupling between the pressure pulsation amplitude and the probability of cavitation.

[0080] This formula achieves quantitative prediction of the risk of pressure pulsation and cavitation coupling by integral coupling and nonlinear mapping of multi-source operating condition data, external variable disturbance characteristics and high-frequency cavitation information within a unified time window. It provides continuous and computable physical constraint indicators for the output decision of the long short-term memory neural network model, thereby improving the prediction stability and generalization ability.

[0081] IV. Prediction and Cooperative Instruction Generation Mechanism of Long Short-Term Memory Neural Network Prediction Control Module 3; In this embodiment, the long short-term memory neural network prediction control module 3 does not make independent predictions for a single physical quantity. Instead, it makes synchronous predictions for two operating phenomena that are highly coupled in physical mechanism but have different manifestations, namely pressure pulsation state and cavitation state, based on a unified input vector and a shared temporal feature extraction structure.

[0082] Specifically, the long short-term memory neural network model jointly models historical operating data, real-time operating data, external correction data, and cavitation monitoring data. Under the same time series prediction framework, it outputs pressure pulsation characteristic parameters reflecting the time-varying characteristics of fluid pressure inside the pump body, as well as cavitation occurrence probability and location parameters reflecting the risk of local phase change in the oil suction area.

[0083] The two types of prediction results are not generated independently, but are derived from the same hidden state vector, thus naturally preserving the temporal correlation and coupling relationship between pressure fluctuations and cavitation state within the model.

[0084] Based on this, the collaborative instruction generation unit 32 does not send the prediction results into two independent control loops separately, but generates them synchronously within the same prediction period according to a unified mapping strategy: The piezoelectric drive command applied to the elastic cavity inside the stator of the vane pump is used to physically cancel out the predicted pressure pulsations. The cavitation suppression command applied to the cavitation suppression unit 42 of the oil suction line is used to intervene in advance against the predicted cavitation trend.

[0085] Therefore, the Long Short-Term Memory Neural Network Predictive Control Module 3 realizes a collaborative control architecture at the system level that unifies the prediction object space, time base, and command generation logic. This ensures that the regulation behavior for pressure pulsation and the suppression behavior for cavitation are consistent in time and coordinated in terms of control objectives, avoiding the response mismatch or mutual interference problems caused by the separation of prediction models or independent control loops in traditional technologies.

[0086] Technical effect: By incorporating pressure pulsation prediction and cavitation state prediction into the same predictive control framework, and synchronously generating drive commands for two types of actuators based on unified prediction results, the system can identify potential unstable evolution trends in advance under complex operating conditions, and simultaneously weaken pulsation amplification and cavitation inducing factors in a synergistic manner, thereby significantly improving the overall stability and control consistency of the vane pump during operation.

[0087] V. Overall collaborative working effect of this embodiment; Through the above structural configuration, this embodiment achieves a tight coupling between multi-dimensional operating condition perception, external variable adaptive correction, and long short-term memory neural network predictive control. This enables the system to detect pressure pulsation and cavitation development trends in advance under complex operating conditions and output targeted collaborative suppression commands, thereby significantly improving the stability and reliability of the vane pump during operation.

[0088] Example 3 is the third embodiment of the present invention. Based on the systems described in Examples 1 and 2, this embodiment further details the internal structure configuration, linkage relationship, and collaborative working principle of the cavitation-pulsation coordinated suppression module 4 and the distributed piezoelectric-elastic cavity execution module 5, so as to highlight the technical effect brought about by the integrated collaborative execution mechanism of "cavitation prediction-active suppression-pulsation cancellation".

[0089] I. Structure and working principle of cavitation-pulsation synergistic suppression module 4; In this embodiment, the cavitation-pulsation coordinated suppression module 4 includes: an instruction parsing and distribution unit 41, a cavitation suppression unit 42, and a linkage control unit 43.

[0090] (a) Instruction parsing and dispatching unit 41; The instruction parsing and distribution unit 41 is connected to the long short-term memory neural network prediction control module 3. It is used to receive the cavitation suppression instruction output by the module, parse the cavitation suppression instruction, extract the control parameters related to the cavitation occurrence probability, prediction position and suppression intensity, and distribute the parsed control signal to the cavitation suppression unit 42 and the linkage control unit 43.

[0091] In this embodiment, when the long short-term memory neural network prediction control module 3 predicts that the probability of cavitation occurring within the next 0.1 seconds is ≥30%, the instruction parsing and distribution unit 41 determines that the cavitation suppression is triggered.

[0092] Technical effect: By introducing a dedicated instruction parsing and distribution structure into the system, a clear and controllable interface layer is formed between the predictive control module and the specific execution unit, thereby improving the stability and scalability of the system control logic.

[0093] (ii) Cavitation suppression unit 42; Cavitation suppression unit 42 is connected to instruction parsing and distribution unit 41, as shown in the reference. Figure 4 It includes: a solenoid valve control subunit 421, a gas injection subunit 422, and a flow fine-tuning subunit 423, which are distributed in the oil suction port pipeline of the vane pump.

[0094] The solenoid valve control subunit 421 is used to adjust the opening of the miniature electromagnetic proportional valve according to the analyzed cavitation suppression command. Its adjustment range is 0-100% to stabilize the local pressure at the oil suction port and avoid cavitation induced by excessively low pressure.

[0095] The gas injection subunit 422 is used to control the low-pressure gas injection component to inject inert gas into the oil suction line according to the analyzed cavitation suppression command. The gas injection pressure is set to 0.1–0.3 MPa. By forming a microbubble buffer layer in the oil suction area, the growth and collapse of cavitation bubbles are suppressed.

[0096] The flow fine-tuning subunit 423 is used to fine-tune the fluid flow rate of the oil suction port pipeline according to the parsed cavitation suppression command. Its adjustment range is ±5% of the rated flow rate, so as to optimize the flow field distribution in the oil suction area and further reduce the probability of cavitation.

[0097] The three sub-units mentioned above operate in coordination within the same control cycle, rather than starting and stopping independently.

[0098] Technical effect: Through the synergistic effect of multiple means such as pressure regulation, gas buffering and flow field optimization, active suppression of cavitation is achieved, thereby reducing the amplification effect of cavitation on pressure pulsation from the source.

[0099] (iii) Linkage control unit 43; The linkage control unit 43 is connected to the instruction parsing and distribution unit 41, and is used to generate a piezoelectric drive compensation instruction after the cavitation suppression unit 42 is started according to the instruction.

[0100] The piezoelectric drive compensation command is used to instruct the long short-term memory neural network predictive control module 3 or the distributed piezoelectric-elastic cavity execution module 5 to compensate and adjust the drive parameters of the original piezoelectric drive command in order to offset the local pressure fluctuations that may be caused by the execution of active cavitation suppression operation.

[0101] Technical effect: By introducing a linkage control mechanism between cavitation suppression and pressure pulsation cancellation, new pressure disturbances are avoided due to the cavitation suppression operation itself, thus achieving stable control at the system level.

[0102] II. Structure and working principle of distributed piezoelectric-elastic cavity actuation module 5; In this embodiment, the distributed piezoelectric-elastic cavity execution module 5 includes a structural execution unit 51 and a drive control unit 52.

[0103] (a) Structural execution unit 51; The structure execution unit 51 includes: The cavity assembly includes a plurality of annular elastic cavities arranged at a predetermined circumferential interval on the inner wall of the stator of the vane pump; in this embodiment, the annular elastic cavities are evenly distributed along the circumference, the included angle between adjacent cavities is 90°, the axial length of each elastic cavity is 1 / 4 of the stator length, and the radial thickness is 5–8 mm. A piezoelectric drive assembly includes stacked piezoelectric ceramic sheets corresponding to each annular elastic cavity. The piezoelectric ceramic sheets are fixed to the outer wall of the corresponding elastic cavity and are used to drive the volume of the elastic cavity to undergo minute changes.

[0104] (ii) Drive control unit 52; The drive control unit 52 connects the long short-term memory neural network predictive control module 3 and the piezoelectric drive assembly, as shown in the reference. Figure 5 It includes: The signal receiving and parsing subunit 521 is used to receive and parse piezoelectric drive commands; The parameter compensation subunit 522 is used to compensate and adjust the driving parameters of the parsed piezoelectric driving command according to the activation state of the cavitation suppression command, so that the expansion amplitude and response rate of the piezoelectric ceramic are coordinated with the cavitation suppression action. The high-voltage drive subunit 523 is used to generate a high-voltage drive signal based on the compensated and adjusted drive parameters, control the expansion and contraction of the corresponding piezoelectric ceramic sheet, and realize the dynamic adjustment of the elastic cavity volume, with a volume adjustment range of 0–5%.

[0105] Technical effect: Through the coordinated work of distributed elastic cavities and piezoelectric ceramics, the pressure pulsation inside the pump body can be regionalized and precisely canceled, and the cancellation strategy can be adjusted in real time according to the cavitation suppression state.

[0106] III. Collaborative Working Mechanism and Comprehensive Technical Effects; In this embodiment, the long short-term memory neural network prediction control module 3 predicts the cavitation trend in advance. When the predicted cavitation probability reaches or exceeds the 30% threshold, the cavitation-pulsation collaborative suppression module 4 first activates the cavitation suppression unit 42 to actively suppress cavitation. At the same time, the linkage control unit 43 generates a piezoelectric drive compensation command, which causes the distributed piezoelectric-elastic cavity execution module 5 to synchronously adjust the volume change amplitude of the elastic cavity to offset the additional pressure fluctuations that may be caused during the cavitation suppression process.

[0107] Through the aforementioned collaborative mechanism, the system no longer passively cancels out existing pressure pulsations, but achieves closed-loop collaborative control of "cavitation prediction - active suppression - pulsation cancellation", thereby breaking the vicious cycle of mutual amplification of cavitation and pressure pulsations and significantly improving the operational stability and reliability of the vane pump under complex operating conditions.

[0108] Example 4 is the fourth embodiment of the present invention. Based on the systems described in Examples 1 to 3, this embodiment further details the internal structure of the closed-loop feedback optimization module 6, its dual-error driving logic, and the collaborative correction mechanism with the long short-term memory neural network prediction control module 3, emphasizing the online correction and offline self-learning iterative capabilities of the prediction model during actual operation.

[0109] I. Structural composition of closed-loop feedback optimization module 6; In this embodiment, the closed-loop feedback optimization module 6 includes: a dual error calculation unit 61, a collaborative triggering unit 62, a model collaborative correction unit 63, and a self-learning iteration unit 64.

[0110] This module maintains data connections with the multi-dimensional operating condition perception module 1 and the long short-term memory neural network predictive control module 3, respectively, and is used to dynamically optimize the predictive model and collaborative control strategy during system operation.

[0111] Description of the collaborative optimization mechanism of closed-loop feedback optimization module 6: In this embodiment, the closed-loop feedback optimization module 6 is not only used to correct errors for a single control target, but also serves as a system-level collaborative optimization hub for the entire blade pump pressure pulsation suppression system. By introducing a dual-error joint feedback mechanism, it achieves synchronous and adaptive optimization of the prediction model parameters and the correction weights of external variables.

[0112] Unlike existing technologies that rely solely on pressure pulsation as a single indicator for closed-loop regulation, in this embodiment, the closed-loop feedback optimization module 6 simultaneously calculates and incorporates pulsation suppression error and cavitation suppression error, treating them as optimization constraints with equal importance.

[0113] Specifically, the system synchronously obtains the pulsation suppression error, which characterizes the vibration of the pump body structure and the fluid instability, and the cavitation suppression error, which characterizes the degree of fluid phase instability and its impact on subsequent pulsation evolution, within each control cycle.

[0114] The two types of errors are not independent of each other, but form a composite error signal space through the closed-loop feedback optimization module 6, which together reflect the degree of deviation of the current system operating state from the comprehensive goal of "low pulsation-low altitude".

[0115] In this embodiment, when any error exceeds the corresponding preset threshold, the model collaborative correction unit 63 does not only correct the long short-term memory neural network prediction control module 3 for a single output dimension, but instead: The pulsation suppression error and cavitation suppression error are simultaneously introduced into the error backpropagation path of the prediction model; During the weight update process, the prediction model must make a coordinated trade-off between "reducing the prediction bias of pressure pulsation" and "reducing the prediction bias of cavitation state".

[0116] Therefore, the prediction model no longer simply pursues the minimization of pressure pulsations, but forms an optimization trajectory in the model parameter space constrained by two objectives, fundamentally avoiding the adverse situation of inducing or aggravating cavitation by suppressing pulsations through extreme piezoelectric driving means.

[0117] Furthermore, the closed-loop feedback optimization module 6 also synchronously feeds back the dual error results to the external variable introduction and correction module 2, which is used to dynamically correct the correction weights of various external variables.

[0118] In this embodiment: when cavitation suppression error is dominant, the system automatically increases the correction weight of external variables that are highly correlated with the fluid state (such as ambient temperature and fluid medium aging variables); When pulsation suppression error dominates, the system correspondingly increases the correction weight of external variables related to load disturbances and power fluctuations.

[0119] In this way, the external variable introduction and correction module is no longer a "static compensation link" independent of the control target, but is incorporated into the overall performance optimization closed loop driven by dual errors.

[0120] Based on the above structure, this embodiment constructs a system-level collaborative optimization mechanism that differs from existing technologies, including: In the feedforward stage, the long short-term memory neural network prediction and control module 3 uniformly predicts the pressure pulsation and cavitation state and generates a cooperative suppression command. During the execution phase, the cavitation-pulsation coordinated suppression module 4 and the distributed piezoelectric-elastic cavity execution module 5 work together to implement the process. During the feedback phase, the closed-loop feedback optimization module 6 simultaneously corrects the prediction model and the external variable correction weights based on the joint evaluation results of the two errors. During the iteration phase, data is continuously accumulated and the overall collaborative logic is optimized offline through the self-learning iteration unit 64.

[0121] Thus, "cavitation-pulse coordination" is no longer just a one-time predictive control action, but evolves into a continuously converging, globally balanced, and multi-objective constrained adaptive optimization closed loop through a dual error feedback mechanism.

[0122] By introducing the aforementioned system-level collaborative optimization mechanism based on dual error feedback, this embodiment can achieve the following during long-term operation: Avoid system performance imbalance caused by a single control objective dominating; Simultaneously maintain pressure pulsation and cavitation state within their respective safe thresholds; Significantly improves the stability and robustness of the system under complex operating conditions, sudden load changes, and cavitation-prone conditions.

[0123] II. Dual Error Calculation and Trigger Judgment Mechanism; (a) Dual-error calculation unit 61; The dual-error calculation unit 61 is connected to the multi-dimensional working condition sensing module 1, and is used to receive the suppressed actual working condition data and calculate them respectively: Pulsation suppression error: The actual pressure pulsation amplitude collected by the distributed pressure sensor is compared with the preset pulsation target value for calculation; in this embodiment, the pulsation target value is set to pulsation amplitude ≤ 0.3 MPa; Cavitation suppression error: The actual cavitation occurrence probability obtained from the ultrasonic cavitation sensor is compared with the preset cavitation target value for calculation; in this embodiment, the cavitation target value is set to a cavitation occurrence probability ≤ 20%.

[0124] Through the above comparison, quantitative error indices reflecting the effects of pressure pulsation suppression and cavitation suppression were obtained respectively.

[0125] (ii) Cooperative triggering unit 62; The collaborative triggering unit 62 is connected to the dual error calculation unit 61 and is used to determine whether the prediction model needs to be corrected. When the pulsation suppression error exceeds the first preset error threshold of 2%; Or the cavitation suppression error exceeds the second preset error threshold by 10%; When any of the above conditions are met, the collaborative triggering unit 62 generates a model correction trigger signal.

[0126] Technical effect: By setting dual error criteria, the model is corrected by relying on a single control index, thus making the prediction model sensitive to both pressure pulsation and cavitation instability phenomena.

[0127] III. Online closed-loop correction mechanism of model collaborative correction unit 63; (a) Collaborative correction of data packet construction; In this embodiment, after responding to the model correction trigger signal, the model collaborative correction unit 63 encapsulates the following information to form a collaborative correction data packet: current pulsation suppression error, current cavitation suppression error, current operating condition parameters (corresponding to the basic operating condition part of the input vector in Embodiment 2), current external correction data, current piezoelectric drive command, and current cavitation suppression command.

[0128] The collaborative correction data packet is sent to the Long Short-Term Memory Neural Network Prediction Control Module 3 to drive the online correction of model parameters.

[0129] (ii) Online weight correction logic based on the model structure of Example 2; After receiving the collaborative correction data packet, the Long Short-Term Memory Neural Network Prediction and Control Module 3 does not rebuild the model structure, but rather adjusts the model weights online with small steps based on the Long Short-Term Memory Neural Network established in Example 2.

[0130] Specifically, the model incorporates the double error information from the collaboratively corrected data packet into the comprehensive prediction index described in Example 2. In the middle, make The system responds to the current prediction bias and updates the model weights using an online gradient descent algorithm with a learning rate of 0.001.

[0131] In this way, the model can gradually converge the prediction results to the target range that satisfies the pulsation amplitude ≤ 0.3 MPa and the cavitation probability ≤ 20% without destroying the original time-series feature extraction capability.

[0132] Technical effect: By using both pulsation suppression error and cavitation suppression error as driving factors for model correction, the predictive model can achieve collaborative adaptive correction of multi-physical coupling errors, avoiding control bias caused by single-objective optimization.

[0133] IV. Offline optimization mechanism of self-learning iterative unit 64; In this embodiment, refer to Figure 6 The self-learning iteration unit 64 includes: a data association storage subunit 641 and an offline model training subunit 642.

[0134] (a) Data association storage subunit 641; The data association storage subunit 641 is used to continuously store the association dataset generated during the operation of the system. The association dataset includes at least: current operating parameters, external correction data, cavitation monitoring data, prediction results, collaborative suppression instructions, and actual operating data after suppression.

[0135] By storing the above data with timestamp association, a historical database covering multiple operating conditions and load conditions can be built.

[0136] (ii) Offline model training subunit 642; When the cumulative size of the associated dataset reaches a preset threshold of 1000 sets, the offline model training subunit 642 starts iterative training of the prediction model in the long short-term memory neural network prediction control module 3.

[0137] During this iterative training process: the weight parameters of the LSTM model described in Example 2 are optimized as a whole; The weight allocation strategy of the external variable introduction and correction module 2 is optimized synchronously; the linkage control logic between the cavitation-pulsation collaborative suppression module 4 and the distributed piezoelectric-elastic cavity execution module 5 is adjusted synchronously.

[0138] After the iterative training is completed, the updated model parameters are loaded into the online prediction and control module to continue running.

[0139] Technical effect: By combining online minor corrections with offline deep iterations, the system has the ability to respond quickly to changes in operating conditions, and can continuously improve its adaptability to complex operating conditions and prediction accuracy in long-term operation.

[0140] V. Summary of the overall technical effects of Example 4; Through the above-mentioned closed-loop feedback optimization structure, this embodiment realizes a predictive model collaborative correction mechanism with dual errors as the core driving factors, enabling the vane pump pressure pulsation and cavitation collaborative suppression system to continuously correct prediction deviations and optimize control strategies during actual operation, thereby maintaining a stable and efficient suppression effect in complex and variable operating conditions over a long period of time.

[0141] The above are merely preferred embodiments of the present invention. The scope of protection of the present invention is not limited to the above embodiments. All technical solutions falling within the scope of the present invention's concept are within the scope of protection of the present invention. It should be noted that for those skilled in the art, any improvements and modifications made without departing from the principle of the present invention should also be considered within the scope of protection of the present invention.

Claims

1. A pressure pulsation suppression system for a vane pump, characterized in that, include: A multi-dimensional operating condition sensing module is used to collect real-time operating condition data of the vane pump, including basic operating condition data and cavitation monitoring data. The external variable introduction and correction module is used to collect dynamic external variables and process the dynamic external variables to generate external correction data with dynamic weights. The long short-term memory neural network prediction and control module is connected to the multi-dimensional working condition perception module and the external variable introduction and correction module, respectively. It is used to receive the real-time working condition data and the external correction data, and predict the future pressure pulsation and cavitation state based on the data, and generate a cooperative suppression command that includes piezoelectric drive command and cavitation suppression command. A cavitation-pulsation coordinated suppression module is connected to the long short-term memory neural network prediction and control module, and includes a cavitation suppression unit disposed in the oil suction port pipeline of the vane pump, for driving the cavitation suppression unit to perform active cavitation suppression operation according to the cavitation suppression command; A distributed piezoelectric-elastic cavity execution module is connected to the long short-term memory neural network predictive control module, and includes an elastic cavity formed in the inner wall of the stator of the vane pump and a corresponding piezoelectric ceramic actuator, for driving the piezoelectric ceramic actuator to adjust the volume of the elastic cavity to counteract pressure pulsation according to the piezoelectric drive command; The closed-loop feedback optimization module is connected to the multi-dimensional working condition perception module and the long short-term memory neural network prediction control module, respectively. It is used to calculate the pulsation suppression error and the cavitation suppression error based on the suppressed actual working condition data, and feed the pulsation suppression error and the cavitation suppression error back to the long short-term memory neural network prediction control module to dynamically correct the prediction model parameters and the correction weights of the external variable introduction and correction module.

2. The pressure pulsation suppression system for a vane pump according to claim 1, characterized in that: The multi-dimensional working condition perception module includes: A data acquisition unit, used to acquire the real-time operating data, includes: The basic operating condition data acquisition subunit includes pressure sensors, vibration sensors, flow sensors, speed sensors, and viscosity sensors respectively installed at the inlet flange, outlet flange, stator inner wall circumferential direction, pump shaft end, and fluid inlet pipeline of the vane pump. The cavitation monitoring data acquisition subunit includes at least two ultrasonic cavitation sensors installed in the cavitation-prone area of ​​the inner wall of the stator of the vane pump. The data transmission unit is connected to the data acquisition unit and is used to sample the basic operating condition data at a frequency not lower than a first preset frequency, sample the cavitation monitoring data at a frequency not lower than a second preset frequency, and transmit the sampled data to the long short-term memory neural network prediction and control module, wherein the second preset frequency is higher than the first preset frequency.

3. The pressure pulsation suppression system for a vane pump according to claim 1, characterized in that: The external variable introduction and correction module includes: The variable acquisition unit is used to acquire various types of dynamic external variables; The preprocessing unit, connected to the variable acquisition unit, is used to filter and normalize the acquired dynamic external variables to obtain standardized variable data. The weight dynamic allocation unit, connected to the preprocessing unit and the closed-loop feedback optimization module, is used to dynamically allocate correction weights to each of the standardized variable data based on the current working conditions and according to a preset weight allocation strategy, so as to form the external correction data. The weight dynamic allocation unit is further configured to adjust the preset weight allocation strategy when the pulsation suppression error or the cavitation suppression error exceeds a preset error threshold.

4. The pressure pulsation suppression system for a vane pump according to claim 1, characterized in that: The long short-term memory neural network prediction and control module includes: An input vector construction unit is used to construct an input vector with a preset dimension from the real-time operating data and the external correction data, wherein the input vector includes at least the current value of the basic operating data, the first derivative value calculated from the basic operating data, the current value of the external correction data, and the current value of the cavitation monitoring data. A collaborative instruction generation unit, connected to the input vector construction unit, is used to input the input vector into a pre-trained long short-term memory neural network model, wherein the model synchronously outputs the piezoelectric driving instruction and the cavitation suppression instruction.

5. The pressure pulsation suppression system for a vane pump according to claim 4, characterized in that: The cooperative instruction generation unit includes: The joint prediction subunit is used to simultaneously predict, based on the input vector and through the long short-term memory neural network model, the pressure pulsation characteristic parameters in the first preset time period and the cavitation occurrence probability and location parameters in the second preset time period. The instruction mapping subunit, connected to the joint prediction subunit, is used to map the pressure pulsation characteristic parameters into piezoelectric drive commands for different elastic cavities according to a preset mapping strategy, and to map the cavitation occurrence probability and position parameters into cavitation suppression commands for different cavitation suppression units.

6. The pressure pulsation suppression system for a vane pump according to claim 1, characterized in that: The cavitation-pulsation synergistic suppression module includes: The instruction parsing and distribution unit, connected to the long short-term memory neural network prediction and control module, is used to receive and parse the cavitation suppression instruction; The cavitation suppression unit, connected to the instruction parsing and distribution unit, includes: The solenoid valve control subunit is used to adjust the opening degree of the miniature electromagnetic proportional valve in the cavitation suppression unit according to the parsed instructions. The gas injection subunit is used to control the low-pressure gas supply component to inject inert gas at a preset pressure into the oil suction line according to the parsed instructions. The flow fine-tuning subunit is used to fine-tune the fluid flow rate of the oil suction port pipeline according to the parsed instructions.

7. The pressure pulsation suppression system for a vane pump according to claim 6, characterized in that: The cavitation-pulsation synergistic suppression module also includes: The linkage control unit, connected to the instruction parsing and distribution unit, is used to generate a piezoelectric drive compensation instruction after the cavitation suppression unit is activated according to the instruction; The piezoelectric drive compensation command is used to instruct the long short-term memory neural network predictive control module or the distributed piezoelectric-elastic cavity execution module to adjust the drive parameters of the piezoelectric drive command or the piezoelectric ceramic actuator to compensate for pressure fluctuations that may be caused by performing the active cavitation suppression operation.

8. The pressure pulsation suppression system for a vane pump according to claim 1, characterized in that: The distributed piezoelectric-elastic cavity actuation module includes: The structure execution unit includes: The cavity assembly includes a plurality of annular elastic cavities opened at a predetermined circumferential interval on the inner wall of the stator of the vane pump; A piezoelectric drive assembly includes stacked piezoelectric ceramic sheets disposed corresponding to each of the annular elastic cavities, wherein each of the piezoelectric ceramic sheets is fixed to the outer wall of the corresponding annular elastic cavity; A drive control unit, connecting the long short-term memory neural network predictive control module and the piezoelectric drive assembly, includes: A signal receiving and parsing subunit is used to receive and parse the piezoelectric drive command; The parameter compensation subunit, connected to the signal receiving and parsing subunit, is used to compensate and adjust the driving parameters of the parsed piezoelectric driving command according to the activation state of the cavitation suppression command. The high-voltage drive subunit, connected to the parameter compensation subunit and the piezoelectric drive assembly, is used to generate a high-voltage drive signal based on the compensated and adjusted drive parameters to control the expansion and contraction of the corresponding piezoelectric ceramic sheet.

9. The pressure pulsation suppression system for a vane pump according to claim 1, characterized in that: The closed-loop feedback optimization module includes: The dual error calculation unit, connected to the multi-dimensional working condition sensing module, is used to calculate the pulsation suppression error and cavitation suppression error respectively based on the collected suppressed actual working condition data. The collaborative triggering unit, connected to the dual error calculation unit, is used to generate a model correction trigger signal when the pulsation suppression error exceeds a first preset error threshold or the cavitation suppression error exceeds a second preset error threshold. The model collaborative correction unit, connected to the collaborative triggering unit, is used to respond to the model correction triggering signal, package the pulsation suppression error, the cavitation suppression error, the current operating parameters and the current driving command into a collaborative correction data packet, and send it to the long short-term memory neural network prediction control module.

10. The pressure pulsation suppression system for a vane pump according to claim 9, characterized in that: The closed-loop feedback optimization module also includes: Self-learning iterative units include: The data association storage subunit is used to continuously store the associated dataset consisting of the current operating condition parameters, the external correction data, the cavitation monitoring data, the collaborative suppression command, and the actual operating condition data after suppression; The offline model training subunit is connected to the data association storage subunit. When the size of the stored association dataset reaches a preset threshold, iterative training of the prediction model in the long short-term memory neural network prediction control module is initiated. During training, the weight allocation strategy of the external variable introduction and correction module and the collaborative logic of the cavitation-pulsation collaborative suppression module and the distributed piezoelectric-elastic cavity execution module are optimized simultaneously.

Images