Method for measuring pressure pulsation amplitude of crown trimming on runner

Abstract

The present invention relates to the technical field of analysis of the coupling effect of mechanical vibration and fluid excitation, and in particular to a method for measuring the pressure pulsation amplitude of a crown modification on a runner. The pressure gradient distribution and cavitation phase change characteristic parameters on the flow channel surface are extracted through fluid dynamics simulation, and the targeted arrangement of the high-frequency pressure sensor array in the vortex core area, the critical position of boundary layer separation, and the mutation area of ​​the modification structure is guided. The ring encryption and axial equal spacing arrangement strategies are adopted to improve the signal acquisition accuracy. A dynamic error feedback mechanism for actual measurement and simulation is established. The proportion of abnormal frequency bands is verified by frequency domain kurtosis and envelope spectrum harmonic analysis, and a modification parameter combination with a pressure pulsation amplitude reduction that meets the standard and a stable frequency band distribution is screened. The present invention solves the measurement error problem caused by the frequency domain aliasing of multi-source high-frequency noise and effective signals, provides a high-precision quantitative basis for the optimization of runner modification, and is suitable for the field of dynamic characteristic testing of mechanical systems.

Description

Technical Field

[0001] The present invention relates to the technical field of mechanical vibration and fluid excitation coupling effect analysis, and in particular to a method for measuring the pressure pulsation amplitude of a crown trim on a runner. Background Art

[0002] During the crown modification process of a power plant's steam turbine runner, the pressure pulsation amplitude can be measured using a high-frequency pressure sensor array and a dynamic signal analyzer. The specific method is as follows: First, based on the geometric parameters of the runner crown and the fluid dynamics model, multiple pressure sensors are arranged along the flow path surface under specific operating conditions to simultaneously collect transient pressure signals corresponding to different modification schemes. A joint time-frequency domain analysis method is then used to extract the dominant frequency component and amplitude characteristics of the pressure pulsation. Combined with fluid-structure interaction simulation results, the modification structure's ability to suppress unsteady flows such as vortex cavitation is quantified. Finally, through amplitude spectrum comparison and energy fraction calculation, an optimized solution is selected that significantly reduces the pressure pulsation amplitude and stabilizes the frequency distribution.

[0003] The technical challenge in measuring the pressure pulsation amplitude of a power plant's steam turbine crown trim is the frequency domain aliasing of the multi-source high-frequency noise caused by the unsteady flow within the flow channel with the actual pressure pulsation signal. Because the flow field in the crown trim region is affected by rotation and boundary layer separation, the pressure pulsation signal includes both low-frequency components generated by cavitation vortices and high-frequency interference from the coupling of turbulent pulsations and mechanical vibrations. Conventional time-frequency analysis methods struggle to accurately separate the effective signal components from these overlapping frequency bands, resulting in increased amplitude measurement errors. Summary of the Invention

[0004] In response to the shortcomings of the existing technology, the present invention provides a method for measuring the pressure pulsation amplitude of the crown trim on the runner. The present invention solves the problem of amplitude measurement error caused by the frequency domain aliasing of multi-source high-frequency noise and effective signals in the pressure pulsation measurement of the crown trim on the runner, which makes the conventional time-frequency analysis method unable to accurately separate the effective components.

[0005] In order to solve the above technical problems, the specific technical solutions of the present invention are as follows:

[0006] The present invention provides a method for measuring the pressure pulsation amplitude of a runner crown trimming process, comprising:

[0007] Obtain the three-dimensional geometric data of the crown-shaped flow channel on the runner, and combine it with fluid dynamics simulation to generate the pressure gradient distribution and cavitation phase change characteristic parameters on the flow channel surface;

[0008] Based on the pressure gradient distribution and cavitation phase change characteristic parameters, a high-frequency pressure sensor array is arranged on the surface of the runner upper crown modified flow channel, and the arrangement position of the high-frequency pressure sensor array is dynamically adjusted in the vortex core area, the critical position of boundary layer separation and the modified structure mutation area;

[0009] Establish a trigger signal synchronized with the runner rotation cycle, collect transient pressure signals under all working conditions under different modification parameter configurations, and simultaneously record speed, flow rate and vibration parameters to generate a raw signal data set with time stamps;

[0010] The raw signal data set is periodically segmented, low-frequency offset is eliminated by baseline drift correction, mechanical vibration noise is filtered out by an adaptive band-stop filter, and turbulent pulsation noise is separated by wavelet packet decomposition and narrow bandpass filtering to generate a denoised cavitation characteristic signal;

[0011] A joint time-frequency domain analysis is performed on the cavitation characteristic signal to extract the amplitude characteristics of the main frequency component of the cavitation vortex. Combined with the amplitude spectrum comparison results after fluid-structure coupling simulation calibration, the modification parameter combination with a pressure pulsation amplitude reduction exceeding a first preset threshold and a frequency band distribution that meets the stability standard is screened.

[0012] Furthermore, the method for measuring the pressure pulsation amplitude of the crown modification of the runner described in the present invention, based on the pressure gradient distribution and cavitation phase change characteristic parameters, arranges a high-frequency pressure sensor array on the surface of the runner crown modification flow channel, and dynamically adjusts the arrangement position of the high-frequency pressure sensor array in the vortex core area, the critical position of boundary layer separation, and the modification structure mutation area, including:

[0013] Based on the pressure gradient distribution and cavitation phase change characteristic parameters, an annular layout is adopted in the vortex core area, where the sensor spacing is dynamically adjusted according to the simulation data of the local pressure gradient change rate, and the sensors are arranged at equal intervals along the axial direction of the flow channel at the critical position of boundary layer separation.

[0014] Furthermore, the method for measuring the pressure pulsation amplitude of the crown trimming of the runner described in the present invention, wherein the raw signal data set is periodically segmented, low-frequency offset is eliminated by baseline drift correction, mechanical vibration noise is filtered out by an adaptive band-stop filter, and turbulent pulsation noise is separated by wavelet packet decomposition and narrow-bandpass filtering to generate a denoised cavitation characteristic signal, includes:

[0015] Performing wavelet packet decomposition on the original signal data set to divide it into multiple frequency sub-bands and identifying high-frequency random noise sub-bands dominated by turbulent pulsation;

[0016] The high-frequency random noise sub-band is subjected to shrinkage processing using a soft threshold function, and a low-frequency sub-band related to the cavitation vortex band characteristics is retained;

[0017] The low-frequency sub-band is subjected to narrow bandpass filtering, and the periodic cavitation component is enhanced by synchronous superposition and averaging in the time domain.

[0018] Furthermore, the method for measuring the pressure pulsation amplitude of the crown-shaped runner of the present invention adopts an annular layout in the vortex core region based on the pressure gradient distribution and cavitation phase change characteristic parameters, wherein the sensor spacing is dynamically adjusted according to the simulation data of the local pressure gradient change rate, and is evenly spaced along the axial direction of the flow channel at the critical position of boundary layer separation, including:

[0019] Performing short-time Fourier transform on the denoised cavitation characteristic signal to generate a time-varying amplitude curve;

[0020] Based on the time-varying amplitude curve, a wavelet ridge tracking algorithm is used to identify the cavitation characteristic frequency in the frequency band overlap region;

[0021] The cavitation characteristic frequency is combined with the fluid-structure coupling simulation calibration result to generate a cavitation main frequency amplitude distribution map.

[0022] Furthermore, the method for measuring the pressure pulsation amplitude of the crown trim of the runner according to the present invention further includes: performing an error comparison between the extracted main frequency amplitude characteristics of the cavitation vortex and the predicted amplitude in the amplitude spectrum comparison result after the fluid-structure coupling simulation calibration;

[0023] If the error exceeds a second preset threshold, adjusting the cavitation phase change characteristic parameters and the boundary layer separation criterion of the fluid dynamics simulation model;

[0024] By iteratively optimizing the cavitation phase change characteristic parameters and the boundary layer separation criterion, the error between the simulation prediction amplitude and the measured amplitude is converged to within the second preset threshold.

[0025] Furthermore, the method for measuring the pressure pulsation amplitude of the crown shaping of the runner according to the present invention further comprises: statistically analyzing the historical distribution of the cavitation main frequency amplitude based on the original signal data set with time stamps and the denoised cavitation characteristic signal of different shaping parameters;

[0026] Setting a dynamic evaluation threshold based on the historical distribution and calculating the frequency domain energy ratio of the cavitation suppression effect of the modification parameters;

[0027] A modification parameter set having an amplitude reduction exceeding the dynamic evaluation threshold and a main frequency energy proportion higher than a third preset proportion is screened out.

[0028] Furthermore, the method for measuring the pressure pulsation amplitude of the crown shaping of a runner according to the present invention further includes:

[0029] Based on the cavitation main frequency amplitude distribution spectrum, the frequency domain kurtosis index of the denoised cavitation characteristic signal is verified, and the abnormal frequency band whose frequency domain kurtosis value exceeds a fourth preset threshold is calculated;

[0030] Simultaneously, envelope spectrum analysis is performed on the cavitation characteristic signal to extract the amplitude of the harmonic component in the envelope spectrum corresponding to the rotation period of the rotor;

[0031] According to the verification results that the proportion of the abnormal frequency band is less than 5% and the amplitude fluctuation of the harmonic component is lower than the fifth preset threshold, it is confirmed that the amplitude retention of the cavitation main frequency component is more than 90% and the high-frequency noise suppression effect meets the signal-to-noise ratio ≥ 15dB;

[0032] A test report including the cavitation main frequency amplitude distribution map, the timestamp data and the screened modification parameter set is generated, and a target modification parameter set is output.

[0033] Beneficial effects of the present invention:

[0034] The present invention extracts the pressure gradient distribution on the flow channel surface and the cavitation phase change characteristic parameters through fluid dynamics simulation, guides the targeted dynamic arrangement of the high-frequency pressure sensor array in the vortex core area, the critical position of boundary layer separation and the mutation area of ​​the modified structure, combines the annular encryption and axial equal spacing arrangement strategies to improve the spatial resolution and dynamic response capability of signal acquisition; adopts multi-stage signal processing technology, including baseline drift correction, adaptive band-stop filtering, wavelet packet decomposition and narrow bandpass filtering, to effectively separate turbulent pulsation noise and mechanical vibration interference, and enhance the extraction accuracy of cavitation main frequency characteristics; calibrates model parameters through fluid-solid coupling simulation, establishes a dynamic error feedback mechanism of measured data and simulation prediction, combines frequency domain kurtosis verification and envelope spectrum analysis, quantifies the cavitation suppression effect and screens the modification parameter combination that meets the stability standard, systematically solves the measurement error problem caused by the frequency domain aliasing of multi-source high-frequency noise and effective signals, provides a high-precision quantitative basis for the optimization of the runner modification, and meets the accuracy and reliability requirements of the mechanical system dynamic characteristics test field for the measurement of the cavitation vortex pressure pulsation amplitude. BRIEF DESCRIPTION OF THE DRAWINGS

[0035] In order to more clearly illustrate the technical solution of the present invention, the following is a brief introduction to the drawings required for use in the embodiments. Obviously, for ordinary technicians in this field, other drawings can be obtained based on the drawings without paying any creative labor.

[0036] Figure 1 This is a flow chart of the method for measuring the pressure pulsation amplitude of crown shaping on a runner provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0037] In order to make the purpose, technical solutions and advantages of the present invention clearer, the technical solutions of the present invention will be clearly and completely described below in conjunction with the specific embodiments of the present invention and the corresponding drawings. Obviously, the embodiments described are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention. The technical solutions provided by each embodiment of the present invention are described in detail below in conjunction with the drawings. In order to better understand the purpose of the present invention, the present invention is further described in detail below.

[0038] See also Figure 1 The present invention provides a method for measuring the pressure pulsation amplitude of the crown shaping of a runner, comprising:

[0039] Step 1: Obtain the three-dimensional geometric data of the crown-shaped flow channel on the runner, and generate the pressure gradient distribution and cavitation phase change characteristic parameters of the flow channel surface by combining fluid dynamics simulation;

[0040] When acquiring the 3D geometric data for the runner crown's modified flow channel, 3D laser scanning or computer-aided design (CAD) models are used to reconstruct the modified flow channel surface topology, extracting the channel curvature distribution, width variation trends, and local geometric mutation characteristics in the modified area. This 3D geometric data is aligned with the original design model using a point cloud registration algorithm to eliminate measurement errors and generate a high-precision flow channel surface mesh model, providing geometric input for subsequent fluid dynamics simulations.

[0041] When combining fluid dynamics simulation to generate the pressure gradient distribution on the flow channel surface, a numerical model is constructed based on the Reynolds-averaged Navier-Stokes equations, and the cavitation phase change model is used to describe the gas-liquid two-phase flow characteristics. During the simulation process, the inlet boundary conditions are set to the flow rate and speed parameters of the actual working conditions, the outlet boundary conditions are pressure free outflow, and the medium physical properties are set according to the properties of the working fluid. The pressure gradient distribution cloud map on the flow channel surface is obtained through transient calculation, the extreme points of the pressure fluctuation amplitude in the vortex core area, the critical position of boundary layer separation, and the modified structure mutation area are identified, and the cavitation phase change characteristic parameters are extracted. The cavitation phase change characteristic parameters include the cavitation initiation threshold, the cavitation volume fraction distribution, and the pressure oscillation amplitude in the vortex core area, which are output as a spatial coordinate correlation data set through the post-processing module to guide the targeted layout strategy of the sensor array.

[0042] The above steps establish the foundation for quantitative analysis of the flow characteristics of the modified flow channel through a progressive process of geometric data acquisition, numerical modeling, and simulation calculations. Three-dimensional geometric data provides precise geometric input for simulation, while fluid dynamics simulation reveals the dynamic characteristics of the flow field through parametric modeling. The logical connection between the two forms a complete mapping from physical structure to flow characteristics, meeting the technical requirements for the coordinated optimization of simulation and measured data in the field of dynamic characteristic testing of mechanical systems.

[0043] Step 2: Based on the pressure gradient distribution and cavitation phase change characteristic parameters, a high-frequency pressure sensor array is arranged on the surface of the runner upper crown modified flow channel, and the arrangement position of the high-frequency pressure sensor array is dynamically adjusted in the vortex core area, the critical position of boundary layer separation, and the modified structure mutation area;

[0044] Based on the pressure gradient distribution and cavitation phase change characteristic parameters, a circular layout strategy is adopted in the vortex core region. High-frequency pressure sensors are evenly arranged along the circumference of the vortex core center. The sensor spacing is dynamically adjusted based on simulation data of the local pressure gradient change rate. Specifically, in areas where the pressure gradient change rate exceeds a preset threshold, the sensor spacing is reduced to 5%-10% of the vortex core diameter to capture high-frequency pressure pulsation characteristics; in areas with flat gradients, the spacing is increased to reduce redundant data. The diameter of the circular layout is set based on the lateral expansion range of the vortex core predicted by simulation. The number of sensors is positively correlated with the pressure fluctuation amplitude, forming a high-density monitoring network for the dynamic characteristics of the vortex core.

[0045] At the critical point of boundary layer separation, an array of sensors is arranged equidistantly along the axial direction of the flow channel, with spacing set at 1% to 3% of the axial length of the flow channel. This spacing is determined based on the wavelength characteristics of the main frequency of the pressure pulsation in the separation zone, ensuring that the distance between adjacent sensors is no greater than one-quarter of the main frequency wavelength to avoid signal distortion caused by spatial aliasing. Using the coordinates of the flow separation point obtained through simulation, sensors are arranged equidistantly along the axial extension of the flow channel surface to monitor the unsteady flow signals caused by boundary layer separation in real time.

[0046] For the modified structural abrupt change region, sensors are densely deployed upstream and downstream of the geometric abrupt change point, based on the curvature radius of the geometric abrupt change point and the predicted pressure gradient abrupt change amplitude. The upstream sensor density is higher than the downstream sensor density, forming a gradient monitoring network to track the dynamic process of flow separation initiation and development. The spacing between sensors is dynamically adjusted based on the pressure fluctuation amplitude range in the abrupt change region. A larger amplitude range requires a smaller sensor spacing, thereby improving the spatial resolution of signal acquisition in the abrupt change region.

[0047] When dynamically adjusting the sensor layout position, the local difference between the measured pressure pulsation amplitude and the simulation prediction value is compared in real time to determine whether the flow field state deviates from the initial model. If the measured amplitude exceeds 15% of the simulation prediction value, the sensor position optimization program is triggered: the diameter of the annular layout of the vortex core area and the sensor spacing are recalculated, and the axial equal spacing parameters of the boundary layer separation area are updated. The adjustment command is sent to the sensor array through the wireless transmission module, and the micro-servo mechanism drives the sensor to move to the target coordinate to complete the online position calibration. This closed-loop feedback mechanism realizes the synchronous optimization of sensor layout and flow field evolution through the dynamic interaction of measured data and simulation model, avoids the signal acquisition blind spot caused by flow instability, and meets the measurement accuracy and real-time requirements of the dynamic characteristics test of the mechanical system.

[0048] Step 3: Establish a trigger signal synchronized with the runner rotation cycle, collect transient pressure signals under all working conditions under different shaping parameter configurations, and simultaneously record the speed, flow rate, and vibration parameters to generate a raw signal data set with a timestamp;

[0049] To establish a trigger signal synchronized with the rotor's rotational period, an encoder or Hall effect sensor monitors the rotor's rotational phase in real time, generating a pulse trigger signal precisely aligned with the rotational period. The rising edge of the trigger signal corresponds to a specific geometric marker on the rotor. A phase-locked loop circuit eliminates phase deviations caused by speed fluctuations, ensuring that the data acquisition system maintains a time base synchronization with the rotor's rotation. The trigger signal's frequency resolution is set to one thousandth of the rotational period to accommodate the synchronization accuracy requirements under varying speed conditions.

[0050] When collecting transient pressure signals under all working conditions with different shaping parameter configurations, the high-frequency pressure sensor array is configured in a multi-channel synchronous sampling mode, and the sampling frequency is set to more than 5 times the estimated value of the cavitation main frequency to meet the signal fidelity requirements of the Nyquist sampling theorem. In response to the dynamic switching requirements of the shaping parameters, automated fixtures or electronic control adjustment devices are used to adjust the flow channel structure parameters in real time to achieve continuous collection of pressure signals for different shaping schemes within the full working condition coverage range. When synchronously recording the speed, flow rate and vibration acceleration parameters, the data streams of each sensor are associated through the same timestamp marking system. The speed parameter is directly obtained by the encoder, the flow parameter is measured in real time using an electromagnetic flowmeter, and the vibration parameter is collected by a piezoelectric acceleration sensor. The data synchronization error is controlled within the microsecond level.

[0051] To generate a timestamped raw signal dataset, transient pressure signals, speed, flow, and vibration parameters collected from multiple channels are encoded according to a unified time base, forming a multidimensional data matrix. The timestamps are generated using a high-precision clock chip with nanosecond resolution. Each data point is associated with a configuration identifier for the modified parameters, a label for the operating condition, and the spatial coordinates of the sensor. The dataset is stored in a structured binary format, enabling rapid retrieval and seamless integration with subsequent signal processing modules, providing a complete data foundation for noise separation and feature extraction.

[0052] The above steps achieve high-fidelity capture and precise correlation of transient signals across all operating conditions through the integrated design of trigger signal synchronization, multi-parameter collaborative acquisition, and data encoding. The uniformity of the trigger signal's time base ensures data phase alignment, while the multi-channel synchronous sampling mode prevents time drift between signals. The timestamp tagging and parameter association mechanism provides traceable context for subsequent analysis, meeting the systematic requirements for complex operating condition data acquisition in the field of dynamic characteristics testing of mechanical systems.

[0053] Step 4: performing periodic segmentation processing on the original signal data set, eliminating low-frequency offsets through baseline drift correction, filtering out mechanical vibration noise using an adaptive band-stop filter, and separating turbulent pulsation noise through wavelet packet decomposition and narrow bandpass filtering to generate a denoised cavitation characteristic signal;

[0054] When performing periodic segmentation on the original signal dataset, the continuous signal is divided into multiple equal-length segments based on the phase signal of the wheel's rotation period, eliminating period length deviations caused by speed fluctuations. This phase signal is acquired in real time by an encoder or Hall effect sensor. The segmented signal segments are time-stamped and strictly aligned with the rotation phase, forming a periodic segmented dataset, providing a consistent time-base input for subsequent signal processing.

[0055] When eliminating low-frequency offsets through baseline drift correction, a polynomial fitting algorithm is used to model the low-frequency trend term of each signal segment, and the least squares method is used to remove low-frequency interference introduced by sensor zero drift or ambient temperature changes. The corrected signal retains the high-frequency pressure pulsation component, preventing low-frequency offsets from interfering with cavitation feature analysis. When using an adaptive band-stop filter to filter mechanical vibration noise, the filter center frequency is dynamically set based on the natural frequency of the impeller support structure and vibration acceleration test data. The bandwidth is adaptively adjusted based on the frequency domain distribution of the vibration energy, effectively suppressing the mechanical vibration harmonic components coupled to the impeller rotation frequency in real time.

[0056] When using wavelet packet decomposition technology to divide the signal into multiple frequency subbands, the Daubechies wavelet basis function is selected for multi-level decomposition, and the number of decomposition levels is determined according to the estimated range of the cavitation main frequency. By calculating the energy proportion and variance characteristics of each subband, the high-frequency random noise subband dominated by turbulent pulsation is identified, and the soft threshold function is used to shrink the coefficients of the high-frequency subband to suppress random noise while retaining the low-frequency effective signal related to the cavitation vortex band. When performing narrow-band pass filtering on the retained low-frequency subband, the passband range is set in combination with the cavitation frequency prediction value of the fluid-structure coupling simulation, and the stopband attenuation is not less than 40dB, and the signal components near the cavitation main frequency are extracted. The filtered signal is enhanced by time-domain synchronous superposition and averaging. Based on the signal segments with phase alignment of the rotation period, multiple periodic signals of the same phase are superimposed and averaged to suppress non-periodic turbulent noise and generate a denoised cavitation characteristic signal.

[0057] The above steps, through a series of processing steps including periodic segmentation, baseline correction, band-stop filtering, wavelet decomposition, and narrow-bandpass filtering, form a complete denoising chain from the original signal to the cavitation signature signal. Periodic segmentation eliminates the effects of speed fluctuations, baseline correction and band-stop filtering suppress low-frequency offsets and mechanical vibration interference, respectively. Wavelet packet decomposition separates noise in the frequency domain, and narrow-bandpass filtering and time-domain averaging enhance the main cavitation frequency component. These steps are logically seamless, meeting the technical requirements for signal separation accuracy and fidelity in the field of dynamic characteristics testing of mechanical systems.

[0058] Step 5: Perform a joint time-frequency domain analysis on the cavitation characteristic signal to extract the amplitude characteristics of the main frequency component of the cavitation vortex. Combined with the amplitude spectrum comparison results after fluid-solid coupling simulation calibration, select a modification parameter combination whose pressure pulsation amplitude reduction exceeds a first preset threshold and whose frequency band distribution meets the stability standard.

[0059] When performing a joint time-frequency domain analysis on the cavitation characteristic signal, a short-time Fourier transform is used to generate a time-varying amplitude curve. Signal segments are intercepted through a Hamming window and the spectrum is calculated to reflect the characteristics of the cavitation main frequency amplitude evolving over time. Combined with the wavelet ridge tracking algorithm, the Morlet wavelet basis function is selected for continuous wavelet transform, and the local energy maximum points are detected on the time-frequency plane. The candidate points are connected to form a main ridge line through a path optimization algorithm, and the cavitation characteristic frequency trajectory and corresponding amplitude are extracted to solve the signal separation problem in the frequency band overlapping area. The frequency trajectory and amplitude data of the main ridge line are integrated to generate the amplitude feature set of the main frequency component of the cavitation vortex.

[0060] When comparing the amplitude spectrum after calibrating the fluid-structure interaction simulation, the measured cavitation main frequency amplitude is matched against the simulated predicted spectrum. Corresponding points with frequency deviations less than 1% are identified, and the relative amplitude error is calculated. The error data is then linked to the sensor layout coordinates on the flow channel surface to generate an error distribution heat map, locating regional deviations between the simulation model and the measured data. If the error exceeds a first preset threshold, a model parameter optimization program is initiated to adjust the cavitation phase change characteristic parameters (such as the cavitation bubble collapse rate) and the boundary layer separation criterion (such as the vorticity-pressure gradient coupling criterion). Iterative optimization is performed using the gradient descent method until the error converges to within the threshold.

[0061] When screening the modification parameter combinations that meet the pressure pulsation amplitude reduction standards, the threshold for the main frequency amplitude reduction is set to no less than 20%, and the frequency band distribution stability standard is that the main frequency energy accounts for more than 70% of the total frequency band energy and the amplitude fluctuation range of the secondary frequency component is less than 10%. A multi-objective optimization algorithm is used to perform Pareto frontier analysis on the modification parameters, and the conflict elimination scheme is combined with the matrix of coupling effects between parameters to retain the parameter set that coordinates the optimization of amplitude reduction and stability. The screening results are verified by frequency domain kurtosis to verify the abnormal frequency band proportion and envelope spectrum harmonic analysis to further confirm the signal fidelity. Finally, the output is a modification parameter combination that meets the cavitation suppression effect and pressure pulsation stability requirements, providing a quantitative basis for the optimization of the impeller structure. The above steps systematically solve the problem of amplitude measurement error under multi-source noise interference through a closed-loop process of time-frequency analysis, model calibration and multi-objective screening, and meet the accuracy and reliability requirements of modification parameter screening in the field of mechanical system dynamic characteristics testing.

[0062] The specific implementation of the method for measuring the pressure pulsation amplitude of the crown shaping of a runner provided by the present invention is as follows:

[0063] In step 1, when obtaining the three-dimensional geometric data of the crown-shaped flow channel on the runner, it is necessary to extract the topological structure and geometric parameters of the flow channel surface based on three-dimensional laser scanning or computer-aided design models, including curvature distribution, flow channel width changes and local geometric mutation characteristics of the modified area. Through fluid dynamics simulation, the Reynolds-averaged Navier-Stokes equation is combined with the cavitation model to simulate the unsteady flow characteristics in the flow channel under different working conditions. During the simulation process, the boundary conditions are set to the speed, flow rate and medium physical parameters of the actual operating conditions, and the pressure gradient distribution and cavitation phase change characteristic parameters on the flow channel surface are calculated. The cavitation phase change characteristic parameters include the cavitation starting threshold, the cavitation volume fraction distribution and the pressure fluctuation amplitude of the vortex core area. The simulation results are exported as pressure gradient cloud maps and spatial coordinate data of the cavitation phase change area through the post-processing module to guide the subsequent sensor layout.

[0064] In step 2, based on the pressure gradient distribution and cavitation phase change characteristic parameters obtained in step 1, a ring layout strategy is adopted in the vortex core area, and the high-frequency pressure sensor array is distributed circumferentially along the center of the vortex core. The sensor spacing is dynamically adjusted according to the simulation data of the local pressure gradient change rate. Specifically, smaller sensor spacing is used in areas with higher pressure gradient change rates to capture high-frequency pressure pulsations, while the spacing is increased in areas with flat gradients to reduce redundant data. At the critical position of boundary layer separation, sensors are arranged at equal intervals along the axial direction of the flow channel, and the spacing is set to 1% to 3% of the axial length of the flow channel to adapt to the flow instability caused by boundary layer separation. For the mutation area of ​​the modified structure, combined with the curvature radius of the geometric mutation point and the pressure fluctuation amplitude predicted by simulation, the sensor layout is encrypted upstream and downstream of the mutation area to form a gradient monitoring network.

[0065] In step 3, a trigger signal synchronized with the rotor rotation cycle is generated by an encoder or Hall sensor, so that the data acquisition system maintains phase synchronization with the rotor rotation. A high-frequency pressure sensor array is used to collect transient pressure signals under all operating conditions under different shaping parameter configurations, and the speed, flow rate, and vibration acceleration signals are recorded simultaneously. The data acquisition system is configured in multi-channel synchronous sampling mode, and the sampling frequency is set to more than 5 times the estimated cavitation main frequency to meet the Nyquist sampling theorem. The original signal data set is marked with a timestamp and associated with the corresponding shaping parameter configuration, speed, and flow rate conditions to form a multi-dimensional data matrix, which is convenient for subsequent signal processing and parameter comparison analysis.

[0066] In step 4, when the original signal data set is periodically segmented, the continuous signal is divided into multiple equal-length segments according to the rotation period of the impeller to eliminate the periodic deviation caused by speed fluctuations. A polynomial fitting algorithm is used to correct the baseline drift of each signal segment to remove low-frequency interference introduced by temperature drift or sensor zero offset. An adaptive band-stop filter is used to filter out mechanical vibration noise, where the band-stop frequency range is dynamically adjusted according to the natural frequency of the impeller support structure and vibration test data. The signal is divided into multiple frequency sub-bands by wavelet packet decomposition, and the high-frequency random noise sub-band dominated by turbulent pulsation is identified. The high-frequency sub-band is shrunk by a soft threshold function to suppress random noise. The low-frequency sub-band related to the cavitation vortex band characteristics is retained, and the signal components near the cavitation main frequency are extracted by a narrow bandpass filter. The repeatability characteristics of the periodic cavitation components are enhanced by combining time-domain synchronous superposition averaging to generate a denoised cavitation characteristic signal.

[0067] In step 5, when performing a joint time-frequency domain analysis on the denoised cavitation characteristic signal, a short-time Fourier transform is used to obtain the time-varying amplitude curve of the signal, and the wavelet ridge tracking algorithm is combined to identify the cavitation characteristic frequency in the frequency band overlap area. The amplitude characteristics of the main frequency component of the cavitation vortex are extracted through the amplitude spectrum calibrated by the fluid-solid coupling simulation, and a dynamic error comparison is performed with the simulation prediction results. When screening the modification parameter combination whose pressure pulsation amplitude reduction exceeds the first preset threshold, the main frequency amplitude reduction threshold is set to no less than 20%, and the frequency band distribution is required to meet the stability standard, that is, the main frequency energy accounts for more than 70% of the total frequency band energy and the amplitude fluctuation range of the secondary frequency component is less than 10%. By iteratively optimizing the cavitation phase change characteristic parameters and boundary layer separation criteria of the fluid dynamics model, the amplitude error between the simulation prediction and the measured data is converged to within 5%, and finally a modification parameter set that meets the cavitation suppression effect and pressure pulsation stability requirements is output.

[0068] The above steps, guided by simulation, create a closed-loop process for sensor placement, dynamic signal acquisition, multi-source noise separation, and cavitation feature extraction. This resolves measurement errors caused by frequency-domain aliasing of multi-source high-frequency noise and valid signals, providing a high-precision quantitative basis for runner optimization. Data transmission and feedback mechanisms form a closed-loop technology loop between each step, with simulation results guiding sensor placement, measured data feeding back into model calibration, and signal processing results driving parameter selection.

[0069] Specifically, the method for measuring the pressure pulsation amplitude of the crown modification on the runner described in the present invention, based on the pressure gradient distribution and cavitation phase change characteristic parameters, arranges a high-frequency pressure sensor array on the surface of the crown modification flow channel of the runner, and dynamically adjusts the arrangement position of the high-frequency pressure sensor array in the vortex core area, the critical position of boundary layer separation, and the modification structure mutation area, including:

[0070] Based on the pressure gradient distribution and cavitation phase change characteristic parameters, an annular layout is adopted in the vortex core area, where the sensor spacing is dynamically adjusted according to the simulation data of the local pressure gradient change rate, and the sensors are arranged at equal intervals along the axial direction of the flow channel at the critical position of boundary layer separation.

[0071] When arranging the high-frequency pressure sensor array, the spatial location of the vortex core area and the area with the extreme pressure fluctuation amplitude are first identified based on the pressure gradient distribution cloud map on the flow channel surface generated by fluid dynamics simulation. The annular layout of the vortex core area is specifically as follows: with the center of the vortex core as the center of the circle, multiple sensors are evenly distributed along the circumference. The annular diameter is determined based on the simulation data of the lateral expansion range of the vortex core. The number of sensors is positively correlated with the amplitude of the pressure gradient change in the vortex core area. For areas where the pressure gradient change rate exceeds the preset threshold, an adaptive encryption algorithm is used to reduce the spacing between the sensors in the annular layout. The spacing adjustment step is 5% to 10% of the vortex core diameter to capture the local high-frequency pressure pulsation characteristics.

[0072] At the critical position of boundary layer separation, combined with the simulation coordinates of the flow separation point on the flow channel surface, the sensor array is arranged at equal intervals along the axial direction of the flow channel. The basis for setting the equal spacing is 1% to 3% of the axial length of the flow channel in the boundary layer separation zone. The specific value is adjusted according to the wavelength characteristics of the main frequency of the pressure pulsation in the separation zone, so that the spacing between adjacent sensors is no more than one-quarter of the main frequency wavelength to avoid spatial aliasing. For the modified structure mutation area, based on the curvature radius of the geometric mutation point and the pressure gradient mutation amplitude, sensors are arranged upstream and downstream of the mutation point to form a gradient monitoring network. The upstream arrangement density is higher than the downstream to track the dynamic process of the initiation and development of flow separation.

[0073] When dynamically adjusting the sensor layout position, the difference between the pressure gradient change rate predicted by the simulation and the local amplitude of the measured signal is compared in real time to determine whether the flow state in the vortex core area or the boundary layer separation area deviates from the initial simulation model. If the measured pressure pulsation amplitude exceeds 15% of the simulation prediction value, the sensor position optimization program is triggered: the diameter of the annular layout of the vortex core area and the sensor spacing are recalculated according to the current pressure gradient distribution, and the axial equal spacing parameters of the critical position of boundary layer separation are updated at the same time. The adjusted sensor array receives the update instruction through the wireless transmission module, and the micro servo mechanism drives the sensor to move to the target coordinate to complete the online position calibration. This dynamic adjustment mechanism realizes the synchronous optimization of sensor layout and flow field evolution through closed-loop feedback, avoiding signal acquisition blind spots caused by flow instability.

[0074] Specifically, the method for measuring the pressure pulsation amplitude of the crown trim on the runner of the present invention comprises performing periodic segmentation processing on the original signal data set, eliminating low-frequency offsets through baseline drift correction, filtering out mechanical vibration noise using an adaptive band-stop filter, and separating turbulent pulsation noise through wavelet packet decomposition and narrow-bandpass filtering to generate a denoised cavitation characteristic signal, which includes:

[0075] Performing wavelet packet decomposition on the original signal data set to divide it into multiple frequency sub-bands and identifying high-frequency random noise sub-bands dominated by turbulent pulsation;

[0076] The high-frequency random noise sub-band is subjected to shrinkage processing using a soft threshold function, and a low-frequency sub-band related to the cavitation vortex band characteristics is retained;

[0077] The low-frequency sub-band is subjected to narrow bandpass filtering, and the periodic cavitation component is enhanced by synchronous superposition and averaging in the time domain.

[0078] During signal processing, the original signal dataset is first segmented periodically based on the rotor's rotation period. Using the rotor's rotation phase signal, acquired by an encoder or Hall effect sensor, the continuously acquired transient pressure signal is segmented into multiple signal segments synchronized with the rotation period, eliminating period length deviations caused by speed fluctuations. Each signal segment is timestamped and aligned with the corresponding rotation period phase, forming a periodic segmented dataset.

[0079] When performing baseline drift correction on a periodic segmented data set, a polynomial fitting algorithm is used to model the low-frequency trend term of each signal segment. Low-frequency offsets introduced by sensor zero drift or ambient temperature variations are eliminated using the least squares method. The corrected signal segments are then filtered through an adaptive band-stop filter to remove mechanical vibration noise. The center frequency of the band-stop filter is dynamically set based on the natural frequency of the rotor support structure and vibration test data, and the bandwidth is adaptively adjusted based on the frequency domain distribution of the vibration energy. The filter parameters are updated in real time by online monitoring of the vibration acceleration signal to suppress mechanical vibration interference coupled to the rotor rotation frequency.

[0080] The wavelet packet decomposition algorithm is used to divide the corrected signal into multiple frequency subbands. The wavelet basis function selects the Daubechies series with compact support characteristics, and the number of decomposition levels is determined according to the estimated range of the cavitation main frequency. By calculating the energy proportion and variance characteristics of each subband, the high-frequency random noise subband dominated by turbulent pulsation is identified. The high-frequency subband is shrunk using a soft threshold function. The threshold setting is based on the statistical characteristics of the noise subband. The low-frequency subband related to the cavitation vortex band characteristics is retained by shrinking the random noise coefficient. The low-frequency subband is extracted through a narrow bandpass filter to extract the signal components near the cavitation main frequency. The passband range of the filter is set in combination with the cavitation frequency prediction value of the fluid-structure interaction simulation, and the stopband attenuation is not less than 40dB.

[0081] When performing time-domain synchronous superposition and averaging on the filtered low-frequency subband, multiple periodic signals of the same phase are superimposed and averaged based on the phase-aligned signal segments of the rotational period to suppress non-periodic turbulent pulsation noise. The number of superpositions is dynamically adjusted based on the signal-to-noise ratio (SNR) improvement required. The averaged signal retains the repetitive characteristics of the periodic cavitation components, generating a denoised cavitation signature signal. Through the synergistic effects of frequency domain decomposition, threshold noise reduction, and time domain enhancement, these processing steps effectively separate cavitation vortex characteristics from multi-source interference signals, providing high-fidelity input data for subsequent time-frequency analysis.

[0082] The logical relationship between these steps is as follows: periodic segmentation eliminates the effects of speed fluctuations, baseline correction and band-stop filtering suppress low-frequency interference and mechanical vibration, wavelet packet decomposition separates frequency-domain noise, and narrow-bandpass filtering and time-domain averaging enhance cavitation characteristics. This technical solution uses multi-stage signal processing to gradually remove noise components, forming a complete denoising chain from the original signal to the cavitation signature signal, meeting the signal separation accuracy requirements for dynamic characteristics testing of mechanical systems.

[0083] Specifically, the method for measuring the pressure pulsation amplitude of the crown trim on the runner described in the present invention adopts an annular layout in the vortex core region based on the pressure gradient distribution and cavitation phase change characteristic parameters, wherein the sensor spacing is dynamically adjusted according to the simulation data of the local pressure gradient change rate, and is evenly spaced along the axial direction of the flow channel at the critical position of boundary layer separation, including:

[0084] Performing short-time Fourier transform on the denoised cavitation characteristic signal to generate a time-varying amplitude curve;

[0085] Based on the time-varying amplitude curve, a wavelet ridge tracking algorithm is used to identify the cavitation characteristic frequency in the frequency band overlap region;

[0086] The cavitation characteristic frequency is combined with the fluid-structure coupling simulation calibration result to generate a cavitation main frequency amplitude distribution map.

[0087] When performing a short-time Fourier transform on the denoised cavitation characteristic signal, a Hamming window is used as the time-domain window function. The window length is set to 2–3 times the estimated period of the cavitation main frequency to balance time-frequency resolution. The sliding step size of the window function is determined by the phase resolution of the rotor rotation period and is typically 1 / 4 of the window length. Signal segments are windowed segment by segment and the spectrum is calculated to generate a time-varying amplitude curve. The horizontal axis of the time-varying amplitude curve represents the time series, and the vertical axis represents the frequency component. The amplitude is represented by a color map, reflecting the temporal evolution of the amplitude of the cavitation characteristic frequency.

[0088] When applying the wavelet ridge tracking algorithm, the Morlet wavelet is selected as the basis function, with its center frequency matching the estimated range of the cavitation main frequency. The time-frequency energy distribution of the signal is calculated using a continuous wavelet transform, and local energy maxima are detected on the time-frequency plane to form a preliminary set of candidate ridge points. A path optimization algorithm is used to connect the candidate points, eliminating false ridges caused by noise or cross-terms while retaining the main ridge that reflects the dynamic characteristics of the cavitation vortex. The frequency trajectory of the main ridge is the cavitation characteristic frequency, and its corresponding amplitude is extracted through ridge energy integration to form a cavitation characteristic frequency-amplitude correlation dataset.

[0089] When combined with the fluid-structure interaction simulation calibration results, the measured cavitation characteristic frequencies were aligned in the frequency domain with the frequency-amplitude distribution predicted by the simulation. A frequency matching algorithm was used to identify corresponding points in the simulated and measured data where the frequency deviation was less than 1%, and the cavitation phase change parameters of the simulation model were then corrected for amplitude. The corrected model outputs the three-dimensional spatial distribution of the cavitation main frequency amplitude. Combined with the spatial coordinates of the sensor placement, a Kriging interpolation algorithm was used to generate a distribution map of the cavitation main frequency amplitude on the flow channel surface. Different colored areas in the map represent amplitude intensity, and the geometric contours of the flow channel are superimposed to visually demonstrate the spatial evolution of the cavitation vortex.

[0090] The above steps convert the denoised cavitation characteristic signal into an amplitude spectrum with spatial distribution characteristics through the progressive processing of time-frequency analysis, ridge extraction and data fusion. Short-time Fourier transform provides basic data for time-varying amplitude, wavelet ridge tracking achieves accurate separation of frequency band overlapping areas, and fluid-solid coupling simulation calibration establishes a dynamic correlation between actual measurement and simulation. The technical solution quantifies the spatiotemporal distribution characteristics of the cavitation main frequency amplitude through the collaborative processing of multi-source data, providing a visual basis for the optimization of the runner modification parameters. A data closed loop is formed between each step. For example, simulation calibration relies on measured ridge data, and spectrum generation feeds back to the iterative optimization of the sensor layout strategy, reflecting the systematicness and operability of the technical solution.

[0091] Specifically, the method for measuring the pressure pulsation amplitude of the crown trim of the runner according to the present invention further includes: performing an error comparison between the extracted amplitude characteristics of the main frequency of the cavitation vortex and the predicted amplitude in the amplitude spectrum comparison result after the fluid-structure coupling simulation calibration;

[0092] If the error exceeds a second preset threshold, adjusting the cavitation phase change characteristic parameters and the boundary layer separation criterion of the fluid dynamics simulation model;

[0093] By iteratively optimizing the cavitation phase change characteristic parameters and the boundary layer separation criterion, the error between the simulation prediction amplitude and the measured amplitude is converged to within the second preset threshold.

[0094] During the error comparison phase, the extracted amplitude characteristics of the cavitation vortex's main frequency are aligned with the calibrated amplitude spectrum from the fluid-structure interaction simulation using a frequency matching algorithm. This algorithm sets a matching tolerance based on a frequency deviation threshold between the measured and simulated data. An amplitude mapping relationship is established for frequencies with a frequency deviation of less than 1%, and the relative amplitude error of the corresponding frequency points is calculated. The error data is then spatially linked to the sensor locations on the flow channel surface, generating an error distribution heatmap that can be used to pinpoint regional deviations between the simulation model and the measured data.

[0095] When the error exceeds the second preset threshold, the parameter optimization program of the fluid dynamics simulation model is started. The adjustment of the characteristic parameters of the cavitation phase change includes the cavitation starting threshold, the cavitation bubble rupture rate and the phase change energy transfer coefficient. The parameters that significantly affect the amplitude error are determined by sensitivity analysis. The optimization of the boundary layer separation criterion is based on the dynamic correlation characteristics of the pressure gradient and the velocity vector in the separation zone, and the vortex-pressure gradient coupling criterion is used to update the triggering conditions of the separation criterion. During the parameter adjustment process, the objective function is optimized based on the gradient descent method. The objective function is defined as the root mean square error between the measured and simulated amplitudes, and the iterative step size is dynamically adjusted according to the error convergence speed.

[0096] During the iterative optimization process, the fluid-structure interaction simulation is re-executed after each parameter adjustment to generate an updated amplitude spectrum prediction result. A secondary error comparison is performed between the new predicted data and the measured data. If the error still exceeds the threshold, the parameter weight is adjusted according to the spatial gradient direction of the error distribution heat map, and the corresponding parameters in the high error area are corrected first. This process is cyclically executed through a closed-loop feedback mechanism until the amplitude error between the measurement and simulation converges to within the second preset threshold range. The convergence judgment condition is that the error change rate for three consecutive iterations is less than 0.5%, and the standard deviation of the amplitude error at each frequency point is required to drop below 20% of the initial value.

[0097] The above steps achieve dynamic calibration between the simulation model and the measured data through a collaborative mechanism of data mapping, parameter optimization, and closed-loop verification. The frequency matching algorithm establishes a data correlation benchmark, the parameter adjustment program drives the model to approximate the actual flow characteristics, and the iterative convergence mechanism ensures the stability of the optimization process. The technical solution improves the simulation model's prediction accuracy of the dynamic characteristics of cavitation vortex strips through the collaborative correction of multi-physical field parameters, providing a reliable quantitative basis for the screening of modification parameters. A positive feedback link is formed between each step. For example, error analysis guides the direction of parameter adjustment, and the optimized model feeds back to more accurate simulation predictions, reflecting the adaptability and systematic nature of the technical solution.

[0098] Specifically, the method for measuring the pressure pulsation amplitude of the crown shaping of the runner according to the present invention further includes: statistically analyzing the historical distribution of the cavitation main frequency amplitude based on the original signal data set with time stamps and the cavitation characteristic signal after denoising for different shaping parameters;

[0099] Setting a dynamic evaluation threshold based on the historical distribution and calculating the frequency domain energy ratio of the cavitation suppression effect of the modification parameters;

[0100] A modification parameter set having an amplitude reduction exceeding the dynamic evaluation threshold and a main frequency energy proportion higher than a third preset proportion is screened out.

[0101] To calculate the historical distribution of cavitation frequency amplitude, we first categorize and store the original signal datasets and denoised cavitation characteristic signals for different modification parameters according to their timestamps. For each modification parameter, we extract the time-domain peak and frequency-domain integrated energy of the cavitation frequency amplitude, constructing a historical amplitude database indexed by the modification parameter. We then use a sliding window statistical method to aggregate and analyze the historical data by time period, calculating the mean, variance, and extreme value distribution characteristics of the cavitation frequency amplitude for each parameter. This generates a trend curve showing how the cavitation amplitude changes with the modification parameter.

[0102] When setting dynamic evaluation thresholds based on historical distributions, an adaptive threshold generation algorithm is employed. Using the historical amplitude mean as a benchmark, the threshold range is dynamically adjusted in conjunction with variance data. Specifically, the dynamic evaluation threshold is determined by a weighted average of 70% of the baseline amplitude and the historical maximum amplitude reduction value. Weighting coefficients are dynamically assigned based on the operating condition coverage of the modified parameters. For new modified parameter combinations, an interpolation method is used to combine historical distribution data of neighboring parameters to predict the initial threshold, and the threshold accuracy is gradually refined through an online learning mechanism.

[0103] To calculate the frequency domain energy fraction, power spectral density analysis is performed on the denoised cavitation signature signal. The energy integral within the cavitation main frequency bandwidth is extracted and then compared to the total energy across the entire frequency band. The main frequency bandwidth is set based on the cavitation frequency predictions from fluid-structure interaction simulations, and the bandwidth should not exceed ±5% of the predicted main frequency. The calculated frequency domain energy fraction is then compared with historical data from the same parameters in a database to generate a quantitative indicator of the cavitation suppression effectiveness under various modification parameters.

[0104] When selecting the modification parameter set, the amplitude reduction of the cavitation main frequency must exceed the dynamic evaluation threshold by 20%, and the main frequency energy ratio must be higher than 65% of the third preset ratio. A multi-objective optimization algorithm is used to perform a Pareto frontier analysis on the modification parameters, selecting parameter combinations that meet both the amplitude reduction and energy ratio requirements. During the screening process, conflicting modification solutions are eliminated by combining the coupling effect matrix between the parameters. Parameter sets with significant synergistic effects and satisfactory stability are retained to form the final optimized solution library.

[0105] The above steps achieve a quantitative evaluation of the cavitation suppression effect of the modification parameters through a progressive process of historical data modeling, dynamic threshold generation, and multi-objective screening. Historical distribution statistics provide a data benchmark, dynamic thresholds avoid the limitations of static evaluation, energy proportion calculations quantify frequency domain characteristics, and multi-objective screening integrates performance and stability requirements. The technical solution provides a traceable decision-making basis for the optimization of runner modification parameters through a data-driven evaluation system, meeting the requirements for scientific and reliable evaluation methods in the field of dynamic characteristic testing of mechanical systems. A data closed loop is formed between each step. For example, the screening results are fed back to the historical database to update the statistical model, and the dynamic threshold is continuously optimized as data accumulates, reflecting the adaptive iterative capability of the technical solution.

[0106] Specifically, the method for measuring the pressure pulsation amplitude of the crown shaping of a runner according to the present invention further includes:

[0107] Based on the cavitation main frequency amplitude distribution spectrum, the frequency domain kurtosis index of the denoised cavitation characteristic signal is verified, and the abnormal frequency band whose frequency domain kurtosis value exceeds a fourth preset threshold is calculated;

[0108] Simultaneously, envelope spectrum analysis is performed on the cavitation characteristic signal to extract the amplitude of the harmonic component in the envelope spectrum corresponding to the rotation period of the rotor;

[0109] According to the verification results that the proportion of the abnormal frequency band is less than 5% and the amplitude fluctuation of the harmonic component is lower than the fifth preset threshold, it is confirmed that the amplitude retention of the cavitation main frequency component is more than 90% and the high-frequency noise suppression effect meets the signal-to-noise ratio ≥ 15dB;

[0110] A test report including the cavitation main frequency amplitude distribution map, the timestamp data and the screened modification parameter set is generated, and a target modification parameter set is output.

[0111] In the frequency domain kurtosis index verification stage, the frequency domain kurtosis calculation is performed on the cavitation characteristic signal after denoising based on the cavitation main frequency amplitude distribution spectrum. The frequency domain kurtosis quantifies the energy distribution steepness of the signal spectrum through the fourth-order moment statistical method, calculates the kurtosis value of each frequency point and generates a kurtosis distribution curve. The identification basis of abnormal frequency bands is a set of continuous frequency points whose kurtosis values ​​exceed the fourth preset threshold. The threshold setting refers to the kurtosis mean of normal working conditions in the historical data plus two standard deviations to exclude instantaneous spike interference caused by random noise. The proportion of abnormal frequency bands is determined by calculating the ratio of the number of abnormal frequency points to the total number of analyzed frequency points, which is used to evaluate the residual degree of non-stationary interference in the signal.

[0112] In envelope spectrum analysis, the cavitation characteristic signal is Hilbert transformed to generate an analytical signal. Its envelope waveform is calculated and Fourier transformed to obtain the envelope spectrum. Frequency components corresponding to the rotor's fundamental frequency and its integer multiple harmonics are extracted from the envelope spectrum, and the amplitude fluctuation range of each harmonic is quantified. The amplitude fluctuation range of the harmonic components is calculated by calculating the ratio of the standard deviation to the mean amplitude of the same harmonic over different rotation periods. This fluctuation range is used to characterize the intensity of periodic interference caused by mechanical vibration or assembly clearance.

[0113] To verify the amplitude retention of the cavitation main frequency component, frequency domain kurtosis and envelope spectrum analysis results are combined. If the abnormal frequency band accounts for less than 5%, it indicates that the high-frequency noise suppression effect in the signal is satisfactory. If the amplitude fluctuation of the harmonic component is lower than the fifth preset threshold (usually set at 15% of the harmonic mean), it is determined that the periodic interference has been effectively filtered out. The power spectral density is integrated to calculate the energy proportion within the cavitation main frequency bandwidth. Combined with the signal-to-noise ratio evaluation formula, it is verified that the high-frequency noise suppression effect meets the signal-to-noise ratio requirement of ≥15dB. This combined verification of these indicators confirms that the amplitude retention of the cavitation main frequency in the denoised signal exceeds 90%, and the residual noise is within a controllable range.

[0114] When generating a test report, the cavitation main frequency amplitude distribution map, timestamp data, and the selected modification parameter set are integrated. The map uses three-dimensional visualization technology to superimpose the flow channel geometry contour, annotate the amplitude intensity distribution with different color levels, and mark the critical position of separation between the vortex core area and the boundary layer. The timestamp data contains synchronous recording information of the speed, flow rate, and vibration parameters, which are associated with the corresponding signal segments in time series. The modification parameter set is stored in matrix form, including parameter name, optimization amplitude, and cavitation suppression effect score. The target modification parameter set that conforms to the industry standard format is finally output for direct call by the runner structure optimization.

[0115] The above steps achieve dual verification of signal quality and modification effect through a progressive process of kurtosis verification, envelope analysis and comprehensive evaluation. Frequency domain kurtosis detects non-stationary interference residues, envelope spectrum analysis suppresses periodic noise, and comprehensive evaluation indicators quantify the fidelity of cavitation characteristics and the level of noise suppression. The technical solution forms a complete link from signal processing to optimization decision-making through multi-dimensional data fusion and standardized output, meeting the requirements of data reliability and application guidance in the field of dynamic testing of mechanical systems. The logical closed loop between each step is reflected in the verification results driving parameter screening and optimization, and the optimized parameters are fed back to the historical database through test reports to support subsequent iterative upgrades.

[0116] The technical features of the technical solution of the present invention are explained as follows:

[0117] Runner crown modified flow channel: refers to the fluid channel after the turbine runner crown has undergone structural modification. Its geometric shape is obtained through three-dimensional laser scanning or computer-aided design model, including curvature distribution, flow channel width change and local geometric mutation characteristics of the modified area, which is used to optimize flow stability and suppress cavitation.

[0118] Pressure gradient distribution: The flow channel surface pressure change rate data generated by fluid dynamics simulation reflects the pressure differences in different areas of the flow channel when the fluid flows. It is used to identify high-dynamic pressure fluctuation areas such as vortex core areas and boundary layer separation areas, and guide the targeted placement of sensors.

[0119] Cavitation phase transition characteristic parameters: Quantitative indicators that describe the cavitation phase transition process of fluid in low-pressure areas, including the cavitation initiation threshold (critical pressure that triggers cavitation), cavitation volume fraction (the proportion of cavitation per unit volume), and pressure fluctuation amplitude in the vortex core area, which are used for simulation model calibration and sensor layout optimization.

[0120] High-frequency pressure sensor array: Multiple high-frequency response pressure sensors are arranged according to a specific spatial strategy. They adopt a ring layout in the vortex core area, are arranged at equal intervals along the flow channel axis in the boundary layer separation area, and are densely arranged upstream and downstream in the modified mutation area to capture transient pressure pulsation signals and improve spatial resolution.

[0121] Vortex core area: The core area of ​​the rotating vortex formed in the fluid flow has a significantly higher pressure gradient change rate than the surrounding area. After positioning through the simulated pressure cloud map, the ring sensor layout is used to dynamically adjust the spacing to capture high-frequency cavitation pulsation signals.

[0122] The critical position of boundary layer separation is the starting point where the fluid separates from the wall due to a sudden change in flow velocity on the channel surface. After the coordinates are determined through simulation, sensors are arranged along the axial direction of the channel with a spacing set at 1% to 3% of the channel length to monitor the unsteady flow caused by separation.

[0123] Modified structure mutation area: The local area where the geometric shape of the impeller crown mutates after modification (such as the curvature mutation point). Based on the pressure gradient mutation amplitude predicted by simulation, sensors are encrypted upstream and downstream to form a gradient monitoring network to track the initiation and development process of flow separation.

[0124] Baseline drift correction: A polynomial fitting algorithm is used to eliminate low-frequency trend items in the signal introduced by sensor zero drift or ambient temperature changes, retaining the high-frequency components of the actual pressure pulsation and preventing low-frequency interference from affecting subsequent analysis.

[0125] Adaptive band-stop filter: A filter that dynamically adjusts the stopband range based on the natural frequency of the rotor support structure and the measured vibration data. It is used to filter out specific frequency band noise coupled with mechanical vibration and suppress the interference of rotation frequency harmonics.

[0126] Wavelet packet decomposition: A technique that decomposes the signal into multiple frequency sub-bands based on the Daubechies wavelet basis function. High-frequency noise sub-bands dominated by turbulent pulsations are identified through energy ratio analysis. A soft threshold function is used to shrink the random noise coefficients, retaining the low-frequency effective signals associated with cavitation vortex bands.

[0127] Narrow bandpass filtering: The passband is set based on the cavitation main frequency range predicted by fluid-structure interaction simulation, and the stopband attenuation is not less than 40dB. It is used to extract signal components near the cavitation characteristic frequency and suppress out-of-band noise interference.

[0128] Time-domain synchronous superposition averaging: Based on the signal segments aligned with the phase of the rotor rotation period, multiple periodic signals of the same phase are superimposed and averaged to enhance the repeatability of the periodic cavitation components and suppress non-periodic turbulence noise.

[0129] Short-time Fourier transform: A Hamming window is used to intercept signal segments and calculate the spectrum to generate a time-varying amplitude curve, which reflects the characteristics of the cavitation main frequency amplitude evolving over time and provides a time-frequency analysis basis for wavelet ridge tracing.

[0130] Wavelet ridge tracking algorithm: Based on the continuous wavelet transform of Morlet wavelet, it detects local energy maximum points on the time-frequency plane, connects candidate points through path optimization to form the main ridge line, extracts the cavitation characteristic frequency trajectory and corresponding amplitude, and solves the signal separation problem in the frequency band overlapping area.

[0131] Fluid-structure coupling simulation calibration: The measured cavitation main frequency amplitude is compared with the simulation predicted spectrum, the error is calculated through the frequency point matching algorithm, and the cavitation phase change parameters (such as cavitation bubble collapse rate) and boundary layer separation criterion (vorticity-pressure gradient coupling criterion) are iteratively adjusted to make the simulation model close to the actual flow characteristics.

[0132] Dynamic evaluation threshold: Based on the weighted calculation of the average cavitation main frequency amplitude and the maximum amplitude reduction value in historical data, it is used to screen the modification parameter combinations with an amplitude reduction of more than 20% and a main frequency energy ratio higher than 65%, avoiding evaluation bias caused by static thresholds.

[0133] Frequency domain kurtosis verification: The steepness of the spectral energy distribution is quantified through fourth-order moment statistics. Abnormal frequency bands with kurtosis values ​​exceeding twice the standard deviation of the historical mean are identified. The residual non-stationary noise in the signal is evaluated. The proportion of abnormal frequency bands must be less than 5% to confirm the noise suppression effect.

[0134] Envelope spectrum analysis: Hilbert transform the signal to generate an envelope waveform, and extract the amplitude fluctuations of the wheel's rotation fundamental frequency and its harmonic components. The harmonic amplitude fluctuations are required to be less than 15% of the mean value to verify the filtering effect of periodic mechanical vibration interference.

[0135] Cavitation main frequency amplitude distribution map: This map is generated by fusing sensor measured data with simulation results using the Kriging interpolation algorithm. The amplitude intensity is annotated with color scales and the flow path geometry is superimposed to intuitively display the spatial evolution of the cavitation vortex band, providing a visual basis for optimizing shaping parameters.

[0136] The specific implementation of the present invention is as follows: In the measurement of the pressure pulsation amplitude of the crown modification on the runner, the three-dimensional geometric data of the flow channel is first obtained based on the three-dimensional laser scanning or computer-aided design model, including the curvature distribution, flow channel width and local mutation characteristics of the modification area. Through fluid dynamics simulation, the Reynolds-averaged Navier-Stokes equation is combined with the cavitation model to simulate the flow field characteristics under actual working conditions, and the pressure gradient distribution and cavitation phase change characteristic parameters on the flow channel surface are generated. The cavitation phase change characteristic parameters include the cavitation starting threshold, the cavitation volume fraction distribution and the pressure fluctuation amplitude of the vortex core area. The simulation results are exported as pressure gradient cloud maps and cavitation area coordinate data to guide subsequent sensor layout.

[0137] Based on simulation results, a circular layout strategy was adopted in the vortex core region, with high-frequency pressure sensors evenly distributed circumferentially around the vortex core center. The sensor spacing was dynamically adjusted based on the local pressure gradient rate. When the pressure gradient rate exceeded a preset threshold, the spacing was reduced to 5%-10% of the vortex core diameter. At the critical boundary layer separation point, sensors were evenly spaced along the flow channel at intervals of 1%-3% of the flow channel length, with spacing no greater than one-quarter of the wavelength of the pressure pulsation main frequency. In the modified structural abrupt region, sensors were densely distributed upstream and downstream based on the curvature radius and pressure gradient amplitude of the geometric abrupt point, forming a gradient monitoring network. An encoder or Hall effect sensor generated a trigger signal synchronized with the rotor rotation, and transient pressure signals were collected using a multi-channel synchronous sampling mode. The sampling frequency was set to at least five times the estimated cavitation main frequency. Speed, flow rate, and vibration acceleration parameters were simultaneously recorded to generate a timestamped raw signal dataset.

[0138] The original signal data set is periodically segmented and divided into equal-length signal segments according to the rotation period of the impeller to eliminate the influence of speed fluctuations. A polynomial fitting algorithm is used to correct the baseline drift and remove the low-frequency interference introduced by the sensor zero offset and temperature drift. An adaptive band-stop filter is used to filter out mechanical vibration noise. Its center frequency is dynamically adjusted according to the natural frequency of the impeller support structure, and the bandwidth is adaptively optimized based on the vibration energy distribution. The signal is divided into multiple frequency sub-bands through wavelet packet decomposition, and the high-frequency random noise sub-band dominated by turbulent pulsation is identified. The soft threshold function is used for shrinkage processing to suppress the noise, and the low-frequency sub-band related to the cavitation vortex band characteristics is retained. The low-frequency sub-band is narrow-band filtered, and the passband range is set in combination with the cavitation frequency prediction value of the fluid-solid coupling simulation. The stopband attenuation is not less than 40dB, and the periodic cavitation component is enhanced by synchronous superposition averaging in the time domain to generate a denoised cavitation characteristic signal.

[0139] A short-time Fourier transform (SFT) is performed on the denoised signal to generate a time-varying amplitude curve. A continuous wavelet transform (CT) is then performed using the Morlet wavelet basis function. A path optimization algorithm is used to extract wavelet ridges and identify cavitation characteristic frequencies. Combined with the fluid-structure interaction simulation calibration results, the measured frequencies are aligned with the simulation predictions, and a Kriging interpolation algorithm is used to generate a cavitation main frequency amplitude distribution map. A frequency matching algorithm is used to calculate the amplitude error between the measured and simulated values. If the error exceeds a second preset threshold, the cavitation phase transition characteristic parameters and boundary layer separation criteria are adjusted, and iterative optimization is performed until the error converges. The historical distribution of cavitation main frequency amplitudes for different modification parameters is statistically analyzed. Based on a dynamic evaluation threshold, parameter combinations with amplitude reductions exceeding 20% ​​and a main frequency energy fraction exceeding 65% are selected. Finally, frequency domain kurtosis is used to verify that the abnormal frequency band fraction is less than 5%, and envelope spectrum analysis shows that the harmonic amplitude fluctuation is less than 15%, confirming that the signal fidelity and noise suppression meet the requirements. A test report is then output, including the amplitude distribution map, timestamp data, and the optimized parameter set, providing a quantitative basis for runner modification.

[0140] The above implementation method systematically solves the problem of frequency domain aliasing of multi-source noise and effective signals through simulation-guided closed-loop processes of sensor layout, multi-level signal processing and dynamic model calibration. Technical linkage is formed between each step through data feedback. For example, sensor layout optimization relies on feedback from measured signals, and model parameter calibration drives higher-precision modification screening, meeting the requirements for measurement accuracy and reliability in the field of dynamic characteristics testing of mechanical systems.

[0141] The present invention solves the problem of amplitude measurement error caused by frequency domain aliasing of multi-source high-frequency noise and effective signal in the measurement of crown shaping pressure pulsation on the runner through the following technical solutions:

[0142] First, based on fluid dynamics simulation, the pressure gradient distribution and cavitation phase change characteristic parameters on the flow channel surface are extracted to guide the targeted placement of the high-frequency pressure sensor array in the vortex core area, the critical position of boundary layer separation, and the modified structure mutation area. A circular layout strategy is adopted in the vortex core area, and the sensor spacing is dynamically adjusted according to the rate of change of the local pressure gradient. The boundary layer separation area is arranged at equal intervals along the axial direction of the flow channel, and the modified structure mutation area is arranged in a denser manner upstream and downstream to form a gradient monitoring network. This layout strategy uses closed-loop optimization based on simulation data and measured feedback to accurately locate high-frequency noise sources and effective signal-sensitive areas, thereby improving the spatial resolution of signal acquisition.

[0143] Secondly, the acquired transient pressure signals undergo multi-stage signal processing to separate frequency domain aliasing components. Baseline drift correction eliminates low-frequency offsets, an adaptive band-stop filter suppresses mechanical vibration noise, and wavelet packet decomposition and narrow-bandpass filtering separate turbulent pulsation interference. The retained cavitation characteristic signals undergo synchronous time-domain superposition and averaging to enhance periodic components. The amplitude characteristics of the cavitation main frequency are extracted by combining short-time Fourier transform and wavelet ridge tracking algorithms. This processing chain effectively removes the frequency band overlap between high-frequency noise and the main frequency of the cavitation vortex band through frequency domain decomposition, threshold noise reduction, and joint time-frequency analysis.

[0144] Finally, the model parameters were calibrated through fluid-structure coupling simulation, and a dynamic error feedback mechanism was established between the measured data and the simulation predictions. The cavitation main frequency amplitude characteristics were compared with the simulated spectrum, and the cavitation phase change characteristic parameters and boundary layer separation criteria were iteratively optimized to converge the error to the preset threshold. Frequency domain kurtosis verification and envelope spectrum analysis were combined to confirm signal fidelity and noise suppression effects, and a cavitation main frequency amplitude distribution map and modification parameter set were generated. This solution quantifies the cavitation suppression effect through multi-source data fusion and closed-loop calibration, providing a high-precision basis for runner modification optimization and meeting the technical requirements of mechanical system dynamic characteristics testing.

Claims

1. A method for measuring the pressure pulsation amplitude of a crown shaping wheel, characterized in that: include: Obtain the three-dimensional geometric data of the crown-shaped flow channel on the runner, and combine it with fluid dynamics simulation to generate the pressure gradient distribution and cavitation phase change characteristic parameters on the flow channel surface; Based on the pressure gradient distribution and cavitation phase change characteristic parameters, a high-frequency pressure sensor array is arranged on the surface of the runner upper crown modified flow channel, and the arrangement position of the high-frequency pressure sensor array is dynamically adjusted in the vortex core area, the critical position of boundary layer separation and the modified structure mutation area; Establish a trigger signal synchronized with the runner rotation cycle, collect transient pressure signals under all working conditions under different modification parameter configurations, and simultaneously record speed, flow rate and vibration parameters to generate a raw signal data set with time stamps; The raw signal data set is periodically segmented, low-frequency offset is eliminated by baseline drift correction, mechanical vibration noise is filtered out by an adaptive band-stop filter, and turbulent pulsation noise is separated by wavelet packet decomposition and narrow bandpass filtering to generate a denoised cavitation characteristic signal; A joint time-frequency domain analysis is performed on the cavitation characteristic signal to extract the amplitude characteristics of the main frequency component of the cavitation vortex. Combined with the amplitude spectrum comparison results after fluid-structure coupling simulation calibration, the modification parameter combination with a pressure pulsation amplitude reduction exceeding a first preset threshold and a frequency band distribution that meets the stability standard is screened.

2. The method for measuring the pressure pulsation amplitude of the crown shaping of a runner according to claim 1, characterized in that: The method of arranging a high-frequency pressure sensor array on the surface of the runner upper crown modified flow channel based on the pressure gradient distribution and the cavitation phase change characteristic parameters, and dynamically adjusting the arrangement position of the high-frequency pressure sensor array in the vortex core area, the critical position of boundary layer separation, and the modified structure mutation area includes: Based on the pressure gradient distribution and cavitation phase change characteristic parameters, an annular layout is adopted in the vortex core area, where the sensor spacing is dynamically adjusted according to the simulation data of the local pressure gradient change rate, and the sensors are arranged at equal intervals along the axial direction of the flow channel at the critical position of boundary layer separation.

3. The method for measuring the pressure pulsation amplitude of the crown shaping of a runner according to claim 1, characterized in that: The process of performing periodic segmentation processing on the original signal data set, eliminating low-frequency offset by baseline drift correction, filtering out mechanical vibration noise by using an adaptive band-stop filter, and separating turbulent pulsation noise by wavelet packet decomposition and narrow-bandpass filtering to generate a denoised cavitation characteristic signal includes: Performing wavelet packet decomposition on the original signal data set to divide it into multiple frequency sub-bands and identifying high-frequency random noise sub-bands dominated by turbulent pulsation; The high-frequency random noise sub-band is subjected to shrinkage processing using a soft threshold function, and a low-frequency sub-band related to the cavitation vortex band characteristics is retained; The low-frequency sub-band is subjected to narrow bandpass filtering, and the periodic cavitation component is enhanced by synchronous superposition and averaging in the time domain.

4. The method for measuring the pressure pulsation amplitude during crown shaping of a runner according to claim 1, characterized in that: Based on the pressure gradient distribution and cavitation phase change characteristic parameters, an annular layout is adopted in the vortex core region, wherein the sensor spacing is dynamically adjusted according to the simulation data of the local pressure gradient change rate, and the sensors are arranged at equal intervals along the axial direction of the flow channel at the critical position of boundary layer separation, including: Performing short-time Fourier transform on the denoised cavitation characteristic signal to generate a time-varying amplitude curve; Based on the time-varying amplitude curve, a wavelet ridge tracking algorithm is used to identify the cavitation characteristic frequency in the frequency band overlap region; The cavitation characteristic frequency is combined with the fluid-structure coupling simulation calibration result to generate a cavitation main frequency amplitude distribution map.

5. The method for measuring the pressure pulsation amplitude during crown shaping of a runner according to claim 4, characterized in that: Also includes: Perform an error comparison between the extracted cavitation vortex main frequency amplitude characteristics and the predicted amplitude in the amplitude spectrum comparison result after the fluid-structure coupling simulation calibration; If the error exceeds a second preset threshold, adjusting the cavitation phase change characteristic parameters and the boundary layer separation criterion of the fluid dynamics simulation model; By iteratively optimizing the cavitation phase change characteristic parameters and the boundary layer separation criterion, the error between the simulation prediction amplitude and the measured amplitude is converged to within the second preset threshold.

6. The method for measuring the pressure pulsation amplitude during crown shaping of a runner according to claim 5, characterized in that: Also includes: According to the original signal data set with time stamps and the cavitation characteristic signal after denoising with different shaping parameters, the historical distribution of the cavitation main frequency amplitude is statistically analyzed; Setting a dynamic evaluation threshold based on the historical distribution and calculating the frequency domain energy ratio of the cavitation suppression effect of the modification parameters; A modification parameter set having an amplitude reduction exceeding the dynamic evaluation threshold and a main frequency energy proportion higher than a third preset proportion is screened out.

7. The method for measuring the pressure pulsation amplitude during crown shaping of a runner according to claim 6, characterized in that: Also includes: Based on the cavitation main frequency amplitude distribution spectrum, the frequency domain kurtosis index of the denoised cavitation characteristic signal is verified, and the abnormal frequency band whose frequency domain kurtosis value exceeds a fourth preset threshold is calculated; Simultaneously, envelope spectrum analysis is performed on the cavitation characteristic signal to extract the amplitude of the harmonic component in the envelope spectrum corresponding to the rotation period of the rotor; According to the verification results that the proportion of the abnormal frequency band is less than 5% and the amplitude fluctuation of the harmonic component is lower than the fifth preset threshold, it is confirmed that the amplitude retention of the cavitation main frequency component is more than 90% and the high-frequency noise suppression effect meets the signal-to-noise ratio ≥ 15dB; A test report including the cavitation main frequency amplitude distribution map, the timestamp data and the screened modification parameter set is generated, and a target modification parameter set is output.

Images