Method for testing vibration of unit gap based on transient numerical simulation

Abstract

The present application relates to the technical field of unit gap vibration test, and particularly relates to a unit gap vibration test method based on transient numerical simulation, which constructs a multi-degree-of-freedom dynamic model by acquiring unit structure parameters and operation parameters, adopts a genetic algorithm to globally optimize stiffness matrix and damping matrix parameters, and generates a high-precision dynamic model; integrates a thermal fluid coupling module in the model, dynamically corrects coupling weight factors in combination with a particle swarm optimization algorithm, generates a multi-physical-field non-steady-state coupling vibration response equation, and solves a Pareto optimal solution set to optimize gap parameters in combination with a multi-objective genetic algorithm. The present application realizes real-time evaluation of a vibration state and closed-loop output of a fault early warning signal. The present application solves the problem of insufficient characterization of non-steady-state multi-field coupling characteristics in the prior art, and provides high-precision technical support for online monitoring of rotating machinery.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of unit gap vibration test, and particularly relates to a unit gap vibration test method based on transient numerical simulation. BACKGROUND

[0002] In the process of power plant operation, the gap vibration of a unit (such as a rotating equipment such as a steam turbine and a generator) directly affects the safety and reliability of the equipment. The vibration test method based on transient numerical simulation can accurately analyze the transmission characteristics of vibration energy in the gap and the dynamic response law of key components by constructing a unit dynamic model and simulating transient excitation (such as start-stop, load mutation, etc.) under actual working conditions. This method can identify abnormal vibration modes caused by temperature gradients, fluid impact or mechanical imbalance, optimize gap design parameters and operation control strategies, thereby reducing the risk of unplanned shutdown and prolonging the service life of the equipment. Its application provides theoretical support and technical means for online monitoring, fault warning and preventive maintenance of power plant units, and significantly improves the operation reliability of large rotating machinery.

[0003] In the unit gap vibration test technology based on transient numerical simulation, the existing method has limitations in characterizing the non-steady coupling characteristics of dynamic excitation sources and actual operating conditions of the unit, especially under transient excitation (such as rapid start-stop or load mutation), it is difficult to accurately quantify the distortion of vibration energy transmission path caused by multi-physical field interaction, which may cause vibration response prediction deviation and affect the reliability of gap vibration test under complex working conditions. SUMMARY

[0004] To overcome the deficiencies of the prior art, the present application provides a unit gap vibration test method based on transient numerical simulation, which solves the problem of insufficient characterization of non-steady coupling characteristics of dynamic excitation sources and actual operating conditions of the unit in the existing transient numerical simulation method.

[0005] To solve the above technical problems, the specific technical solutions of the present application are as follows:

[0006] The unit gap vibration test method based on transient numerical simulation provided by the present application comprises:

[0007] Step S1, obtaining the structure parameters and operating parameters of the unit, wherein the structure parameters include blade gap and bearing support stiffness, and the operating parameters include rotating speed and temperature field distribution;

[0008] Step S2, constructing a multi-degree-of-freedom dynamic model of the unit based on the structure parameters and operating parameters, and performing global parameter optimization on the stiffness matrix and damping matrix of the multi-degree-of-freedom dynamic model by a genetic algorithm to generate an optimized dynamic model;

[0009] Step S3, integrating a thermal fluid coupling module in the optimized dynamic model, simulating a multi-field coupling vibration response under a non-steady boundary condition in combination with a transient excitation input function, and dynamically correcting a thermal fluid coupling weight factor through a particle swarm optimization algorithm to generate a multi-physical field non-steady coupling vibration response equation;

[0010] Step S4, extracting a vibration energy distribution spectrum of a gap region based on the multi-physical field non-steady coupling vibration response equation, identifying a distortion region of a vibration energy transmission path through a transmission path analysis, and real-time correcting a transmission path distortion index of the distortion region based on an adaptive fuzzy logic algorithm to generate a corrected vibration energy transmission characteristic;

[0011] Step S5, establishing a sensitivity mapping relationship between a gap geometric parameter and a vibration amplitude according to the corrected vibration energy transmission characteristic, and solving a Pareto optimal solution set of a gap design parameter and an operation control parameter through a multi-objective genetic algorithm to generate a collaborative optimization strategy;

[0012] Step S6, performing a bench test based on the collaborative optimization strategy, acquiring measured vibration data, and verifying a generalization of the optimized dynamic model through a time-frequency domain joint verification strategy to generate a verified dynamic model;

[0013] Step S7, embedding the verified dynamic model into an online monitoring system, real-time fusing vibration data collected by a sensor and a prediction result of the dynamic model through a Kalman filtering algorithm, dynamically updating model parameters, and outputting a real-time state evaluation and a fault early warning signal of a unit gap vibration.

[0014] Further, the unit gap vibration test method based on transient numerical simulation provided by the application, the step S2 comprises:

[0015] Filtering an initial population based on a historical model parameter library, and setting a population size of a genetic algorithm in combination with a structural freedom degree number of the multi-degree-of-freedom dynamic model;

[0016] Iteratively optimizing a stiffness matrix and a damping matrix of the multi-degree-of-freedom dynamic model through adaptive crossover probability and mutation probability, wherein the mutation probability is adjusted in stages according to a sensitivity analysis result of the stiffness matrix and the damping matrix;

[0017] When a continuous iterative descending amplitude of a mean square error between vibration energy predicted by the multi-degree-of-freedom dynamic model and preset experimental data in the historical model parameter library is less than a preset threshold, outputting an optimized dynamic model parameter.

[0018] Further, the unit gap vibration test method based on transient numerical simulation provided by the application, the step S3 comprises:

[0019] The particle swarm dimension is set according to the number of thermal fluid coupling fields, and the inertia weight of the particle swarm optimization is adjusted based on a nonlinear decreasing strategy;

[0020] The vibration energy transmission error of the multi-physical field unsteady coupling vibration response equation is taken as an optimization objective function, and the search space of the thermal fluid coupling weight factor is dynamically updated through individual learning factors and social learning factors;

[0021] When the thermal fluid coupling weight factor converges to a preset energy conservation threshold, a corrected multi-physical field unsteady coupling vibration response equation is output.

[0022] Further, the transient numerical simulation-based unit gap vibration test method, the step S4 comprises:

[0023] The membership function and fuzzy rule of the high-frequency resonance and low-frequency vortex distortion mode are defined based on a historical fault case library;

[0024] According to the dynamic relationship between the frequency domain phase shift and the amplitude attenuation ratio of the multi-physical field unsteady coupling vibration response equation, the transmission path distortion index of the distortion region is calculated;

[0025] The amplitude-frequency characteristics of the energy transmission path of the distortion region are corrected by the gradient descent method, and the correction is based on the real-time feedback of the membership function and fuzzy rule;

[0026] When the transmission path distortion index is lower than a preset safety threshold, the corrected vibration energy transmission characteristics are output.

[0027] Further, the transient numerical simulation-based unit gap vibration test method, the step S5 comprises: taking the minimum vibration amplitude and the highest operation stability as optimization objectives, combining the sensitivity mapping relationship and the penalty function method into the gap design threshold and the operation safety boundary constraint;

[0028] Non-dominated solution sets are generated by simulating binary crossover and polynomial mutation, and the Pareto front solution is screened based on crowding degree sorting;

[0029] When the Pareto front solution covers a preset engineering applicability range determined by the corrected vibration energy transmission characteristics, a cooperative optimization strategy is output.

[0030] Further, the transient numerical simulation-based unit gap vibration test method, the step S6 comprises:

[0031] Based on stratified sampling, K-fold verification data sets are divided, covering cold start, load rejection and peak regulation operation conditions;

[0032] The consistency of the frequency domain energy distribution of the simulation data of the optimized dynamic model generated through the spectral kurtosis difference index analysis step S2 and the measured data obtained by the current test bench experiment is consistent;

[0033] When the time domain mean square error and the frequency domain spectral kurtosis difference of the simulation data are both lower than the preset verification threshold, it is determined that the dynamic model passes the generalization verification.

[0034] Further, the transient numerical simulation-based unit gap vibration test method, the step S7 comprises:

[0035] According to the accuracy of the vibration displacement sensor and the temperature field infrared monitoring sensor, the process noise covariance matrix of the Kalman filtering algorithm is set, and the observation noise covariance is dynamically updated through the prediction residual of the verified dynamic model;

[0036] The vibration displacement sensor data, the temperature field infrared monitoring data and the output value of the verified dynamic model generated in step S6 are covariance weighted and fused;

[0037] When the parameter update period of the verified dynamic model is synchronized with the sampling period of the unit control system, the real-time state evaluation and fault warning signal of the unit gap vibration are output.

[0038] The present application has the following advantages:

[0039] The present application improves the accuracy and reliability of the unit gap vibration test by stage-by-stage parameter optimization and dynamic coupling correction. Based on the global optimization mechanism of genetic algorithm and particle swarm optimization algorithm, combined with historical model parameter library and sensitivity analysis, the stiffness matrix, damping matrix and thermal fluid coupling weight factor of the multi-degree-of-freedom dynamic model are optimized to accurately represent the influence of multi-field interaction on the vibration energy transmission path under non-steady-state boundary conditions. The self-adaptive fuzzy logic algorithm is used to correct the transmission path distortion index in real time, suppress the high-frequency resonance and low-frequency vortex distortion effect, and the multi-objective genetic algorithm is used to generate the Pareto front solution to optimize the gap design parameters and operation control strategy. The time-frequency domain joint verification strategy strengthens the generalization ability of the dynamic model under complex working conditions such as cold start and load rejection, and the Kalman filtering algorithm fuses the sensor data and model prediction value to realize the closed-loop output of vibration state real-time evaluation and fault warning signal. The above technical scheme effectively solves the problem of insufficient characterization of transient excitation source and multi-field coupling characteristics in the existing method, and provides high-precision technical support for online monitoring and maintenance of rotating machinery. BRIEF DESCRIPTION OF DRAWINGS

[0040] In order to more clearly illustrate the technical solutions of the present application, the following will briefly introduce the drawings needed to be used in the embodiments. Obviously, for those skilled in the art, other drawings can also be obtained without creative labor based on the drawings.

[0041] Figure 1 The flowchart of the unit gap vibration test method based on transient numerical simulation provided by the embodiments of the present application. DETAILED DESCRIPTION

[0042] In order to make the purpose, technical solutions and advantages of the present application more clear, the following will combine the specific embodiments of the present application and the corresponding drawings to clearly and completely describe the technical solutions of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, not all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor belong to the protection scope of the present application. The following will combine the drawings to specifically describe the technical solutions provided by the embodiments of the present application. In order to better understand the purpose of the present application, the following will further describe the present application in detail.

[0043] Please refer to Figure 1 The unit gap vibration test method based on transient numerical simulation provided by the present application comprises:

[0044] In step S1, the structure parameters of the unit are obtained, and the structure parameters comprise a blade gap and bearing support stiffness. The operation parameters comprise a rotating speed and a temperature field distribution.

[0045] In step S1, the structure parameters of the unit are obtained by measuring the blade gap through a high-precision three-dimensional scanning technology, and the dynamic change data of the bearing support stiffness is captured in real time by combining a laser displacement sensor to provide geometric constraints and mechanical property inputs for a multi-degree-of-freedom dynamic model construction. The measurement of the blade gap is aimed at the radial and axial gap distribution between the rotor and the stator, and a non-contact optical sensor is adopted to avoid mechanical interference. The bearing support stiffness is calibrated through a loading experiment and a frequency response function inversion method to reflect the dynamic support characteristics under different rotating speeds. In the operation parameters, the rotating speed is collected in real time by a Hall sensor or an optical encoder, the temperature field distribution is monitored by a distributed optical fiber temperature sensor network, the sensors are arranged along the axial and circumferential directions of the unit, and the thermal gradient changes in the high-temperature and high-pressure areas are covered. The temperature field data are calibrated by a thermal imager and a spatial interpolation algorithm to generate a continuous distribution map, and are transmitted to a data processing module synchronously with the rotating speed signal to form a time-space correlated operation parameter database. The above parameter acquisition scheme provides high-fidelity inputs for subsequent model construction and optimization.

[0046] In step S2, a multi-degree-of-freedom dynamic model of the unit is constructed based on the structural parameters and the operating parameters, and a global parameter optimization is performed on the stiffness matrix and the damping matrix of the multi-degree-of-freedom dynamic model by using a genetic algorithm, so as to generate an optimized dynamic model;

[0047] In step S2, when the multi-degree-of-freedom dynamic model of the unit is constructed based on the structural parameters and the operating parameters obtained in step S1, the number of degrees of freedom of the model is determined according to the support structure, the blade distribution and the bearing connection mode of the rotor of the unit, and each degree of freedom corresponds to a translation and rotation motion component of the rotor. The initial parameters of the stiffness matrix are constructed based on the measured data of the bearing support stiffness, and the initial values of the damping matrix are calibrated by a frequency domain attenuation experiment. The two matrices jointly represent the dynamic characteristics of the vibration energy transmission of the unit. The initial population of the genetic algorithm is selected from a historical model parameter library. The population size is positively correlated with the number of degrees of freedom of the model. The historical parameter combination with a high correlation degree with the geometric characteristics and the operating conditions of the current unit is selected by using a similarity matching algorithm, so as to improve the global optimization efficiency.

[0048] The adaptive crossover probability is dynamically adjusted according to a population diversity index. When the difference between the population individuals is lower than a preset threshold, the crossover probability is increased to enhance the search breadth. The mutation probability is set in stages according to the sensitivity analysis results of the stiffness matrix and the damping matrix. A low mutation probability is used for the matrix elements with a significant influence on the vibration energy transmission to maintain stability, and a high mutation probability is used for the secondary elements to expand the search space. The fitness function takes the mean square error between the predicted vibration energy of the model and the historical experimental data as an evaluation index. When the error reduction amplitude in continuous iterations is less than a preset convergence threshold, it is determined that the parameter optimization of the stiffness matrix and the damping matrix is completed, and the optimized dynamic model is output. The optimized model accurately represents the dynamic response law of the unit under non-steady state conditions by modifying the distribution characteristics of the stiffness and the damping, and provides a high-precision base model for subsequent multi-field coupling analysis.

[0049] In step S3, a thermal fluid coupling module is integrated into the optimized dynamic model, a multi-field coupling vibration response under non-steady state boundary conditions is simulated by combining a transient excitation input function, and a particle swarm optimization algorithm is used to dynamically correct a thermal fluid coupling weight factor, so as to generate a multi-physical field non-steady state coupling vibration response equation.

[0050] In step S3, when the thermal fluid coupling module is integrated into the optimized dynamic model, a finite volume method is used to solve the interaction between the fluid domain and the structure domain. The parameters of the fluid domain are defined based on the temperature field distribution and the flow velocity characteristics in the operating parameters, and the parameters of the structure domain inherit the optimization results of the stiffness matrix and the damping matrix of the dynamic model. The transient excitation input function simulates non-steady state boundary conditions (such as thermal shock during cold start or pressure sudden change during load rejection). The time stepping method is used to couple the fluid pressure field, the temperature gradient field and the mechanical vibration field, so as to generate transient vibration response data under the interaction of multi-physical fields.

[0051] The dimension of the particle swarm optimization algorithm is set according to the number of thermal-fluid coupling fields, each coupling field corresponds to an independent dimension of the particle position vector, and the weight factor represents the distribution characteristics of the fluid-structure energy transfer coefficient. The inertia weight adopts a nonlinear decreasing strategy, setting a higher weight in the initial stage to enhance the global search ability, and decreasing exponentially with the increase of the number of iterations, gradually turning to local fine search. The optimization objective function is the vibration energy transfer error of the multi-physical field unsteady coupling vibration response equation, and the error value is quantified by calculating the interaction loss rate of the kinetic energy of the fluid domain and the strain energy of the structure domain.

[0052] The individual learning factor controls the tracking strength of the particle to its own historical optimal position, and the social learning factor guides the particle to approach the group optimal solution, and the two cooperate to dynamically update the search space direction and step length of the weight factor. When the iterative update of the weight factor enables the energy transfer error to be lower than the preset energy conservation threshold (satisfying the energy conservation law between physical fields), it is determined that the optimization process converges, and the corrected multi-physical field unsteady coupling vibration response equation is output. The equation accurately represents the nonlinear distortion characteristics of the vibration energy transfer path caused by fluid-structure interaction under transient excitation, providing high-precision input for subsequent transfer path analysis.

[0053] In step S4, based on the multi-physical field unsteady coupling vibration response equation, the vibration energy distribution spectrum of the gap region is extracted, the distortion region of the vibration energy transfer path is identified through transfer path analysis, and the transfer path distortion index of the distortion region is corrected in real time based on an adaptive fuzzy logic algorithm, and a corrected vibration energy transfer characteristic is generated;

[0054] In step S4, when extracting the vibration energy distribution spectrum of the gap region based on the multi-physical field unsteady coupling vibration response equation, a wavelet packet transform is performed on the vibration response data using a frequency domain decomposition technique to separate the energy density distribution of different frequency bands and generate an energy distribution spectrum containing time-frequency characteristics. The transfer path analysis identifies the transfer path of vibration energy between the blade gap and the bearing support structure through an energy flow tracking algorithm, locates the distortion region by calculating the path contribution degree, and the distortion region is characterized by abnormal aggregation of energy density or phase mutation in a specific frequency band.

[0055] The adaptive fuzzy logic algorithm is based on the statistical characteristics of the high-frequency resonance and low-frequency vortex distortion modes in the historical fault case library, defines a membership function to quantify the probability distribution of energy anomalies in different frequency bands, and establishes a dynamic mapping relationship between the distortion index and the phase shift and amplitude attenuation ratio through fuzzy rules associated with expert experience and experimental data. The frequency domain phase shift is extracted by Hilbert transform to obtain the instantaneous phase difference of the vibration waveform, and the amplitude attenuation ratio is calculated based on the ratio of the decay rate to the steady-state amplitude, and the two are combined to generate a comprehensive evaluation index of the transfer path distortion index.

[0056] The gradient descent method takes the distortion index as the optimization objective, combines the weight coefficients output by the membership function with the modified direction constrained by the fuzzy rule, and iteratively adjusts the amplitude-frequency characteristic parameters of the distortion region. During the correction process, the fuzzy rule provides physical constraints for the gradient descent to avoid parameter adjustment beyond the feasible region of the actual working condition, while the membership function dynamically allocates the correction priority of different frequency bands. When the distortion index of the energy transmission path is iteratively optimized and is lower than the preset safety threshold (set according to the statistical boundary of the historical safe working condition), the corrected vibration energy transmission characteristic is output. This characteristic represents the distribution of the energy transmission path after the distortion is suppressed, providing a dynamic benchmark for gap parameter optimization.

[0057] In step S5, a sensitivity mapping relationship between the gap geometric parameters and the vibration amplitude is established according to the corrected vibration energy transmission characteristic, and a multi-objective genetic algorithm is used to solve the Pareto optimal solution set of the gap design parameters and the operation control parameters, generating a collaborative optimization strategy.

[0058] In step S5, when establishing the sensitivity mapping relationship between the gap geometric parameters and the vibration amplitude based on the corrected vibration energy transmission characteristic, a parameter perturbation method is used to apply a small perturbation to the geometric parameters such as the blade gap and bearing support stiffness, and the response sensitivity value of the vibration amplitude to the parameter change is calculated through the dynamic model to generate a parameter sensitivity matrix. This matrix quantifies the influence weight of different design parameters on the vibration energy, providing a priority basis for multi-objective optimization. The optimization objectives are set as the minimum vibration amplitude and the highest running stability, and the gap design threshold (such as the minimum allowable gap value) and the running safety boundary (such as the maximum allowable vibration amplitude) are converted into constraint conditions through the penalty function method, and the solution set is forced to converge to the engineering feasible region through the penalty coefficient.

[0059] The multi-objective genetic algorithm simulates binary crossover operation to exchange parameter code fragments of parent individuals, maintaining population diversity; polynomial mutation operation introduces random disturbance to avoid premature convergence of the algorithm, generating a non-dominated solution set covering the trade-off relationship of multiple objectives. The crowding degree sorting algorithm selects the Pareto frontier solution according to the distribution density of the solution set in the objective space, and preferentially retains individuals with sparse distribution and wide coverage. The pre-set engineering applicability range is defined according to the corrected vibration energy transmission characteristic, covering the vibration amplitude tolerance interval and stability threshold under typical working conditions such as cold start and load rejection. When the Pareto frontier solution covers all boundary conditions, the collaborative optimization strategy is output, which includes the recommended value range of the gap geometric parameters and the adjustment priority sequence of the operation control parameters, such as preferentially reducing the gap parameters with the highest sensitivity to the vibration amplitude. The above process balances the conflicting demands of vibration suppression and running stability through the multi-objective optimization and constraint fusion mechanism, providing an implementable parameter combination scheme for the bench test.

[0060] Step S6, based on the collaborative optimization strategy, carry out bench test, obtain the measured vibration data, adopt time-frequency domain joint verification strategy to verify the generalization of the optimized dynamic model, and generate the verified dynamic model;

[0061] In step S6, when carrying out the bench test based on the collaborative optimization strategy, the K-fold verification data set is divided by the stratified sampling method, covering the cold start, load rejection and peak regulation operation conditions. The cold start condition simulates the transient vibration characteristics of the unit under low temperature initial state, the load rejection condition corresponds to the dynamic response caused by sudden load shedding, and the peak regulation condition reflects the vibration energy distribution difference under variable load rate. Stratified sampling ensures that each fold data set contains representative samples of all types of operating conditions, avoiding the bias of the verification results towards a single operating condition.

[0062] In the time-frequency domain joint verification strategy, the spectral kurtosis difference index extracts the multi-resolution frequency band energy of the simulation data and the bench test data through wavelet packet decomposition, calculates the Euclidean distance of the kurtosis value of each frequency band, and quantifies the consistency of the frequency domain energy distribution. The time domain mean square error is calculated by comparing the point-to-point difference square mean of the simulated vibration signal and the measured signal, to evaluate the prediction accuracy of the dynamic model in the time dimension. The preset verification threshold is set according to the error statistical distribution of the historical safe operation model. When the time domain mean square error and the frequency domain spectral kurtosis difference of the simulation data are both lower than the threshold, it is determined that the optimized dynamic model has cross-condition generalization ability, and the verified dynamic model is output.

[0063] The parameters of the verified model are fine-tuned through iterative calibration and residual feedback of the measured data, to strengthen the model's ability to capture nonlinear vibration characteristics under non-steady-state excitation, and provide high-confidence input for the online monitoring system. The above verification process improves the robustness and engineering applicability of the model under complex conditions through stratified data sampling and multi-dimensional index fusion.

[0064] Step S7, embed the verified dynamic model into the online monitoring system, real-time fuse the vibration data collected by the sensor and the prediction results of the dynamic model through Kalman filtering algorithm, dynamically update the model parameters and output the real-time state evaluation and fault warning signal of the unit gap vibration.

[0065] In step S7, when the verified dynamic model is embedded into the online monitoring system, the process noise covariance matrix of the Kalman filtering algorithm is set according to the measurement accuracy of the vibration displacement sensor and the temperature field infrared monitoring sensor. The micron-level resolution of the vibration displacement sensor corresponds to the noise parameter of the mechanical vibration dimension, and the temperature measurement error range of the temperature sensor affects the covariance component of the thermal field dimension. The observation noise covariance is updated dynamically based on the residual error of the sensor and the model prediction value. The residual error is the real-time difference between the sensor measured value and the model prediction value. The noise parameter is adjusted adaptively based on the statistical characteristics of the residual sequence, to enhance the tracking ability of the algorithm under non-steady-state conditions.

[0066] The vibration displacement sensor data and the temperature field infrared monitoring data are integrated with the dynamic model output value through a covariance weighted fusion algorithm, and the covariance weight is dynamically allocated according to the sensor measurement error range and the confidence degree of the model historical prediction residual. The integrated signal comprehensively reflects the interaction effect of mechanical vibration state and thermal field, and generates vibration amplitude, energy transmission path distortion index and thermal coupling strength index with high confidence. The model parameters are dynamically updated based on real-time data flow through the Kalman gain matrix, to correct the local deviation of the stiffness matrix and the damping matrix under non-steady state conditions, and to maintain the synchronization of the model prediction accuracy and the real-time working condition.

[0067] The parameter update period is aligned with the sampling pulse signal of the unit control system through a clock synchronization module, and when the model refresh frequency is consistent with the sampling rate of the control system, the real-time state evaluation result is output. The state evaluation result triggers a hierarchical alarm mechanism through a preset safety threshold, for example, when the vibration amplitude exceeds the running safety boundary or the distortion index reaches the preset fault threshold, a corresponding level of early warning signal is generated. The above technical scheme realizes the online monitoring and real-time early warning function of the unit gap vibration through multi-source data fusion and closed-loop parameter updating.

[0068] In the unit gap vibration test method based on transient numerical simulation provided by the application, step S1 acquires the structural parameters and operating parameters of the unit to provide basic data for subsequent modeling and optimization. The structural parameters include blade gap and bearing support stiffness, which are obtained through three-dimensional scanning or high-precision displacement sensors; the operating parameters include rotating speed and temperature field distribution, wherein the temperature field distribution is monitored in real time by distributed temperature sensors or infrared thermal imaging technology. Step S2 constructs a multi-degree-of-freedom dynamic model of the unit based on the above parameters, and performs global parameter optimization on the stiffness matrix and damping matrix of the model through a genetic algorithm. The initial population of the genetic algorithm is generated based on the historical model parameter library, the population size is matched with the number of model degrees of freedom, and the fitness function takes the mean square error between the predicted vibration energy of the model and the historical experimental data as the evaluation index. When the error reduction amplitude is lower than a preset threshold, the optimized dynamic model parameters are output.

[0069] Step S3 integrates a thermal fluid coupling module in the optimized dynamic model to simulate the multi-field coupled vibration response under non-steady state boundary conditions through a transient excitation input function. The thermal fluid coupling module solves the interaction between the fluid domain and the structure domain by using the finite volume method, the particle swarm optimization algorithm sets the particle dimension according to the number of thermal fields, the inertia weight is dynamically adjusted based on a nonlinear decreasing strategy, the target function is the vibration energy transmission error of the multi-physical field equation, and the search space of the weight factor is updated through the individual learning factor and the social learning factor until the weight factor meets the energy conservation threshold, and the corrected multi-physical field non-steady state coupled vibration response equation is generated.

[0070] Step S4 extracts the vibration energy distribution spectrum of the gap region based on the multi-physical field equation, and identifies the distortion region in the vibration energy transmission path by using the transfer path analysis method. The dynamic relationship between the frequency domain phase shift and the amplitude attenuation ratio is used to calculate the transmission path distortion index. The adaptive fuzzy logic algorithm defines the membership function and fuzzy rules according to the historical fault case library, and corrects the amplitude-frequency characteristics of the distortion region by the gradient descent method. When the distortion index is lower than the preset safety threshold, the corrected vibration energy transmission characteristics are output.

[0071] Step S5 establishes the sensitivity mapping relationship between the gap geometric parameters and the vibration amplitude according to the corrected vibration energy transmission characteristics. The sensitivity analysis quantifies the influence of the design parameters on the vibration amplitude by the parameter perturbation method. The multi-objective genetic algorithm takes the minimum vibration amplitude and the highest running stability as the target, combines the penalty function method to integrate the gap design threshold and the running safety boundary constraint, generates the non-dominated solution set through the simulated binary crossover and the polynomial mutation, selects the Pareto frontier solution covering the engineering applicability range based on the crowding degree sorting, and outputs the collaborative optimization strategy.

[0072] Step S6 carries out the bench test based on the collaborative optimization strategy, simulates the cold start, load rejection and peak regulation operation conditions, and divides the K-fold verification data set by hierarchical sampling. The time-frequency domain joint verification strategy uses wavelet transform to analyze the time domain mean square error, and uses the spectral kurtosis difference index to evaluate the consistency of the frequency domain energy distribution of the simulation data and the measured data. When both are lower than the preset verification threshold, it is determined that the dynamic model passes the generalization verification.

[0073] Step S7 embeds the verified dynamic model into the online monitoring system. The Kalman filter algorithm sets the process noise covariance matrix according to the sensor accuracy, and dynamically updates the observation noise covariance based on the model prediction residual. The vibration displacement sensor data and the temperature field infrared monitoring data are integrated with the model output value set by the covariance weighted fusion algorithm, realizing real-time updating of the model parameters. When the model updating period and the sampling period of the unit control system are synchronized, the real-time state evaluation and fault warning signal are output, completing the closed-loop monitoring and control.

[0074] The above steps form a complete unit gap vibration test and online monitoring system through progressive technical means such as parameter optimization, multi-field coupled modeling, path correction and closed-loop verification.

[0075] Specifically, the unit gap vibration test method based on transient numerical simulation provided by the present application comprises the following steps:

[0076] The initial population is screened based on the historical model parameter library, and the population size of the genetic algorithm is set according to the number of structural degrees of freedom of the multi-degree-of-freedom dynamic model;

[0077] The stiffness matrix and the damping matrix of the multi-degree-of-freedom dynamic model are iteratively optimized through adaptive crossover probability and mutation probability, wherein the mutation probability is adjusted in stages according to the sensitivity analysis results of the stiffness matrix and the damping matrix.

[0078] When the continuous iterative descending amplitude of the mean square error of the vibration energy predicted by the multi-degree-of-freedom dynamic model and the preset experimental data in the historical model parameter library is less than a preset threshold, the optimized dynamic model parameters are output.

[0079] In step S2, when the initial population is screened based on the historical model parameter library, the historical model parameter library contains verified stiffness matrix and damping matrix parameter combinations under different working conditions, and the historical parameters with the highest correlation degree with the current unit structure parameters are selected as the initial population of the genetic algorithm through a similarity matching algorithm. The population size is determined according to the number of structural degrees of freedom of the multi-degree-of-freedom dynamic model, and a certain proportion of individual numbers in the population correspond to each degree of freedom to ensure the completeness of the parameter search space. In the adjustment strategy of the adaptive crossover probability and the mutation probability, the crossover probability is dynamically adjusted based on the population diversity index, and the mutation probability is set in stages according to the sensitivity analysis results of the stiffness matrix and the damping matrix. The matrix elements with high sensitivity correspond to a lower mutation probability to maintain stability, and the elements with low sensitivity increase the mutation probability to expand the search range. In the iterative optimization process, the fitness function takes the mean square error of the vibration energy predicted by the multi-degree-of-freedom dynamic model and the preset experimental data in the historical model parameter library as the evaluation index. When the descending amplitude of the mean square error in continuous iteration is lower than the preset threshold, it is determined that the parameters converge, and the optimized dynamic model parameters are output. The above process improves the global optimization efficiency through the historical data driving and dynamic parameter adjustment mechanism.

[0080] Specifically, the unit gap vibration test method based on transient numerical simulation provided by the application comprises the following steps:

[0081] The particle swarm dimension is set according to the number of thermal fluid coupling fields, and the inertia weight of the particle swarm optimization is adjusted based on a nonlinear decreasing strategy;

[0082] The vibration energy transmission error of the multi-physical field unsteady coupling vibration response equation is taken as an optimization objective function, and the search space of the thermal fluid coupling weight factor is dynamically updated through individual learning factors and social learning factors;

[0083] When the thermal fluid coupling weight factor converges to a preset energy conservation threshold, the corrected multi-physical field unsteady coupling vibration response equation is output.

[0084] In step S3, the number of thermal fluid coupling fields determines the dimension setting of the particle swarm optimization algorithm, each coupling field corresponds to an independent optimization dimension in the particle swarm, so that the particle position vector can completely characterize the weight factor distribution under the interaction of multiple fields. The adjustment of the inertia weight adopts a nonlinear decreasing strategy, a higher inertia weight is set in the initial stage to enhance the global search ability, and the inertia weight decreases exponentially with the increase of the iteration number, gradually turning to local fine search, balancing the convergence speed and accuracy of the algorithm.

[0085] The individual learning factor and the social learning factor control the tracking strength of the particles to their own historical optimal position and the group optimal position respectively, and guide the particles to gather in the area with the minimum vibration energy transmission error by dynamically updating the search space of the thermal fluid coupling weight factor. The optimization objective function is constructed based on the vibration energy transmission error of the multi-physical field non-steady coupling vibration response equation, and the error is quantified by the energy interaction loss rate between the fluid domain and the structure domain. When the iterative update of the weight factor makes the energy transmission error meet the preset energy conservation threshold, it is determined that the particle swarm optimization process converges, and the corrected multi-physical field non-steady coupling vibration response equation is output.

[0086] Specifically, the unit gap vibration test method based on transient numerical simulation provided by the application comprises the following steps:

[0087] The membership function and fuzzy rule of the high-frequency resonance and low-frequency vortex distortion mode are defined based on the historical fault case library;

[0088] According to the dynamic relationship between the frequency domain phase shift and the amplitude attenuation ratio of the multi-physical field non-steady coupling vibration response equation, the transmission path distortion index of the distortion region is calculated;

[0089] The amplitude-frequency characteristics of the energy transmission path of the distortion region are corrected by the gradient descent method, and the correction is based on the real-time feedback of the membership function and fuzzy rule;

[0090] When the transmission path distortion index is lower than the preset safety threshold, the corrected vibration energy transmission characteristics are output.

[0091] In step S4, the historical fault case library defines the typical characteristics of the high-frequency resonance and low-frequency vortex distortion mode by integrating the vibration abnormal data recorded by the unit under various operating conditions. The membership function is constructed based on the distribution probability of vibration energy in a specific frequency band, and the fuzzy rule is associated with expert experience and historical data to quantify the judgment logic of the transmission path abnormality under different distortion modes. The frequency domain phase shift is extracted by the Fourier transform of the multi-physical field non-steady coupling vibration response equation, reflecting the phase lag characteristics in the vibration energy transmission process; the amplitude attenuation ratio is calculated according to the attenuation rate of the peak energy in the frequency domain response spectrum, and the two are dynamically related to form a comprehensive evaluation index of the transmission path distortion index.

[0092] The gradient descent method takes the distortion index as the optimization objective function, combines the real-time feedback signal output by the membership function and the fuzzy rule, and adjusts the amplitude-frequency characteristic parameters of the energy transmission path. During the correction process, the fuzzy rule provides a constraint condition for the gradient direction to avoid the local optimal solution deviating from the actual physical characteristics, and the membership function dynamically weights the correction weight of different frequency bands to improve the pertinence of path correction. When the transmission path distortion index is lower than the preset safety threshold after iterative optimization, it is determined that the transmission path is in a stable state, and the corrected vibration energy transmission characteristic is output. The preset safety threshold is set according to the statistical characteristics of the safe operation condition in the historical fault case library, which meets the safety boundary requirements of the actual operation of the unit.

[0093] The above technical scheme accurately identifies and suppresses the distortion effect of the vibration energy transmission path through fault mode quantization, dynamic index calculation and adaptive correction mechanism.

[0094] Specifically, the gap vibration test method of the unit based on transient numerical simulation, the step S5 comprises: taking the minimum vibration amplitude and the highest running stability as the optimization target, combining the sensitivity mapping relationship and the penalty function method into the gap design threshold and the running safety boundary constraint;

[0095] Non-dominated solution set is generated by simulating binary crossover and polynomial mutation, and Pareto frontier solution is screened based on crowding degree sorting;

[0096] When the Pareto frontier solution covers the preset engineering applicability range determined by the corrected vibration energy transmission characteristic, the cooperative optimization strategy is output.

[0097] In step S5, the minimum vibration amplitude and the highest running stability are taken as the core indexes of multi-objective optimization, and the dynamic correlation characteristics of the gap geometric parameters and the vibration amplitude are quantified through the sensitivity mapping relationship. The sensitivity analysis adopts the parameter perturbation method, applies a small perturbation to the gap design parameters and calculates the change rate of the vibration response, generates the parameter sensitivity matrix, and provides a basis for optimization weight distribution. The penalty function method converts the gap design threshold and the running safety boundary constraint into the penalty term of the objective function, increases the penalty coefficient when the parameter combination exceeds the allowed range, and forces the solution set to converge to the feasible region.

[0098] Simulated binary crossover and polynomial mutation operation generates non-dominated solution set in multi-objective genetic algorithm, simulated binary crossover exchanges parameter code fragments of parent individuals by probability, maintains population diversity; polynomial mutation introduces local randomness based on probability disturbance, avoids premature convergence of algorithm. In the non-dominated solution set screening stage, the crowding degree sorting algorithm is used to calculate the density index of individuals in the solution set, and individuals with sparse distribution in the target space are preferentially retained to form the Pareto frontier solution which covers a wide range and is uniformly distributed.

[0099] The preset engineering applicability range is set according to the corrected vibration energy transmission characteristics, and covers the vibration amplitude and stability tolerance interval under the typical operating condition of the unit. The Pareto frontier solution needs to cover all boundary conditions of the range, and when the optimal solution set satisfies the coverage verification, the cooperative optimization strategy is output, and the strategy includes the recommended value range of the gap design parameter and the adjustment priority sequence of the operation control parameter. The above process balances the conflicting demands of vibration suppression and operation stability through a multi-objective optimization and constraint fusion mechanism.

[0100] Specifically, the transient numerical simulation-based unit gap vibration test method provided by the application comprises the following steps:

[0101] The K-fold verification data set is divided based on stratified sampling, covering cold start, load rejection and peak regulation operating conditions;

[0102] The frequency energy distribution consistency of the simulation data of the optimized dynamic model generated in step S2 and the measured data obtained by the current bench test is analyzed through the spectral kurtosis difference index;

[0103] When the time domain mean square error and the frequency domain spectral kurtosis difference of the simulation data are both lower than the preset verification threshold, it is determined that the dynamic model passes the generalization verification.

[0104] In step S6, the stratified sampling method divides the data set according to the typical characteristics of the unit operating condition, and cold start, load rejection and peak regulation are respectively taken as independent subsets. The K-fold verification data set is constructed by proportionally extracting samples, ensuring that each fold of data covers all types of operating conditions, and avoiding verification bias. K-fold cross-validation evaluates the generalization ability of the dynamic model under different data distributions by rotating the training set and the verification set, and improves the statistical significance of the verification result.

[0105] The spectral kurtosis difference index is used to quantify the consistency of the simulation data and the measured data of the bench test in the frequency energy distribution. The spectral kurtosis represents the energy aggregation characteristics by calculating the kurtosis value of the signal in a specific frequency band, and the difference index is constructed based on the Euclidean distance of the kurtosis values of each frequency band. The frequency energy distribution consistency analysis combines the wavelet packet decomposition technology to extract multi-resolution frequency band energy, compares the energy proportion of the simulation data and the measured data in each frequency band, and verifies the capturing ability of the dynamic model for the nonlinear vibration characteristics under transient excitation.

[0106] The time domain mean square error reflects the prediction accuracy of the dynamic model in the time dimension, and is obtained by calculating the point-to-point difference square mean of the simulated vibration signal and the measured signal; the frequency domain spectral kurtosis difference degree evaluates the restoration degree of the model to the complex vibration mode from the perspective of frequency domain energy distribution. The preset verification threshold is set according to the error distribution of the safe operation model in the historical verification case, when both indicators are lower than the threshold, it is determined that the dynamic model meets the engineering application requirements in the time-frequency domain characteristics, and the model parameters passed the generalization verification are output. The above process verifies the data in layers and evaluates the indicators in multiple dimensions, and strengthens the robustness of the model under complex working conditions.

[0107] Specifically, the transient numerical simulation-based unit gap vibration test method provided by the application comprises the following steps:

[0108] According to the accuracy of the vibration displacement sensor and the temperature field infrared monitoring sensor, the process noise covariance matrix of the Kalman filtering algorithm is set, and the observation noise covariance is dynamically updated according to the prediction residual of the verified dynamic model;

[0109] The vibration displacement sensor data, the temperature field infrared monitoring data and the output value of the verified dynamic model generated in step S6 are fused by covariance weighting;

[0110] When the parameter update period of the verified dynamic model is synchronized with the sampling period of the unit control system, the real-time state evaluation and fault warning signal of the unit gap vibration are output.

[0111] In step S7, the process noise covariance matrix of the Kalman filtering algorithm is set according to the measurement accuracy of the vibration displacement sensor and the temperature field infrared monitoring sensor, wherein the accuracy of the vibration displacement sensor determines the covariance component of the process noise in the mechanical vibration dimension, and the accuracy of the temperature sensor affects the noise parameter distribution in the thermal field dimension. The observation noise covariance is dynamically updated according to the prediction residual of the verified dynamic model, the residual is the real-time difference between the model prediction value and the sensor measured value, and the noise covariance matrix is adaptively adjusted based on the statistical characteristics of the residual sequence, thereby improving the tracking ability of the filtering algorithm to the non-steady state working condition.

[0112] The vibration displacement sensor data and the temperature field infrared monitoring data are integrated with the output value of the verified dynamic model through the covariance weighting fusion algorithm, and the covariance weighting weight is dynamically allocated according to the real-time confidence of each data source. The confidence is determined by the sensor measurement error range and the historical statistical results of the model prediction residual, and the fused signal comprehensively reflects the mechanical vibration state and the interactive effect of the thermal field, thereby generating a high-confidence unit gap vibration feature vector.

[0113] The parameter updating period of the verified dynamic model is aligned with the sampling period of the unit control system through a clock synchronization module, the model parameters are refreshed in real time based on the sensor data stream, and when the model updating period is synchronized with the control system sampling pulse signal, the real-time state evaluation result is output. The state evaluation result includes vibration amplitude, energy transmission path distortion index and thermal coupling strength index, and the fault warning signal is dynamically generated by comparing the evaluation result with the preset safety threshold, triggering the hierarchical alarm mechanism. The above technical scheme realizes online monitoring and real-time warning of the unit gap vibration through multi-source data fusion and closed-loop parameter updating.

[0114] Structural parameters and operating parameters: Structural parameters include blade clearance (physical gap between blade and casing, affecting airflow dynamics) and bearing support stiffness (bearing support stiffness of rotor system, determining vibration transmission efficiency), which are measured by three-dimensional scanning or high-precision displacement sensors.

[0115] Operating parameters include rotational speed (rotor rotation frequency, affecting vibration main frequency) and temperature field distribution (clearance change caused by thermal expansion, monitored in real time by distributed temperature sensors or infrared thermal imaging technology).

[0116] Multi-degree-of-freedom dynamic model: A mathematical model based on the structural characteristics of the unit, which describes the vibration response of the system through multiple degrees of freedom (such as translation, rotation, etc.). Stiffness matrix represents the ability of the system to resist deformation, and damping matrix reflects the vibration energy dissipation characteristics, both of which are globally optimized by genetic algorithm. Genetic algorithm uses historical model parameter library to select initial population, adjusts parameters through crossover and mutation operations, and fitness function takes the mean square error of model predicted vibration energy and experimental data as evaluation index, and sensitivity analysis adjusts mutation probability in stages to balance search efficiency and stability.

[0117] Thermal fluid coupling module: Simulates the interaction between fluid field (such as steam or gas) and structural field, and transient excitation input function simulates non-steady-state boundary conditions (such as start-stop, load sudden change). Particle swarm optimization algorithm sets the dimension according to the number of coupled fields, and adjusts the inertia weight according to the non-linear decreasing strategy (large-scale search in the early stage, fine convergence in the later stage), taking vibration energy transmission error as the objective function, dynamically correcting the weight factor through individual learning factor (particle's own experience) and social learning factor (group optimal solution), until the energy conservation threshold (energy interaction loss rate meets the standard).

[0118] Distortion index of transmission path: based on the frequency domain characteristics extracted from the multi-physical field unsteady coupling vibration response equation, the distortion index is dynamically calculated by the phase shift (phase lag of vibration waveform) and amplitude attenuation ratio (amplitude attenuation rate in energy transmission process). The adaptive fuzzy logic algorithm defines the membership function (quantifies the abnormal probability of energy in different frequency bands) and fuzzy rules (such as high-frequency resonance trigger conditions) through the historical fault case library, and combines the gradient descent method to correct the amplitude-frequency characteristics and suppress the abnormal energy aggregation in the distortion region.

[0119] Multi-objective genetic algorithm and Pareto frontier solution: the sensitivity mapping relationship quantifies the influence weight of gap geometric parameters on vibration amplitude through parameter perturbation method. The penalty function method converts design thresholds (such as minimum allowable value of gap) and safety boundaries (such as maximum vibration amplitude) into constraint conditions. Simulated binary crossover (parameter coding segment exchange) and polynomial mutation (local random disturbance) generate non-dominated solution set, and the crowding degree sorting filters the uniformly distributed Pareto frontier solution, covering the engineering application range (such as stability tolerance interval under peak regulation condition) defined by the corrected energy characteristics.

[0120] Time-frequency domain joint verification strategy: hierarchical sampling divides the K-fold data set, covering typical working conditions such as cold start (low temperature initial state) and load rejection (sudden load shedding). The spectral kurtosis difference is calculated by wavelet packet decomposition to extract multi-band energy distribution, and the difference between simulation data and bench test data is calculated. Combined with the time domain mean square error (point-to-point signal difference), the model generalization is verified.

[0121] Kalman filter and real-time monitoring: the process noise covariance matrix is set according to the sensor accuracy (such as micron-level error of displacement sensor), and the observation noise covariance is dynamically updated based on the model prediction residual (difference between measured and predicted values). Covariance weighted fusion allocates weights according to sensor confidence (error range) and historical model residual, fuses vibration displacement, temperature field data and model output values. The parameter update period is synchronized through the clock module and the control system, and the vibration amplitude, distortion index and warning signal are output in real time, triggering the hierarchical alarm mechanism.

[0122] The specific embodiments of the present application are based on the actual application scene of the unit gap vibration test, combined with the dynamic characteristic demand of rotating equipment such as steam turbine and generator under the cold start, load rejection and peak regulation operation condition, a phased optimization and verification system is constructed. First, the structural parameters such as unit blade gap and bearing support stiffness are collected by three-dimensional scanning or high-precision displacement sensor, and the real-time monitoring of rotating speed and temperature field distribution is combined with distributed temperature sensor. Based on the historical model parameter library screening, the initial population matched with the current unit structure is selected, the population size of genetic algorithm is set according to the number of structural degrees of freedom of multi-degree-of-freedom dynamic model, the mutation probability is adjusted by adaptive crossover probability and sensitivity classification, the stiffness matrix and damping matrix parameters are optimized, and when the mean square error of model prediction vibration energy and historical experimental data decreases by less than a preset threshold, the optimized dynamic model is output.

[0123] When the thermal fluid coupling module is integrated in the dynamic model, the particle swarm optimization algorithm sets the dimension according to the number of thermal field, the inertia weight is adjusted according to the nonlinear decreasing strategy, and the target function is the vibration energy transfer error of the multi-physical field non-steady coupling vibration response equation. The weight factor search space is dynamically updated by individual learning factor and social learning factor, and when the energy transfer error meets the preset energy conservation threshold, the corrected multi-physical field equation is generated. Based on the equation, the vibration energy distribution spectrum of the gap region is extracted, the membership function and fuzzy rule of high-frequency resonance and low-frequency vortex distortion mode in the historical fault case library are combined, the transmission path distortion index is calculated, the amplitude-frequency characteristic is corrected under the constraint of fuzzy rule by gradient descent method, and the corrected vibration energy transfer characteristic is output.

[0124] According to the corrected characteristic, the sensitivity mapping relationship between gap geometric parameters and vibration amplitude is established, the multi-objective genetic algorithm takes the minimum vibration amplitude and the highest operation stability as the target, the design threshold and operation safety constraint are combined into the penalty function method, and the Pareto frontier solution covering the engineering applicability range is generated. The K-fold verification data set covering the cold start, load rejection and peak regulation conditions is divided by hierarchical sampling, the spectral kurtosis difference index is used to analyze the consistency of simulation data and bench test data, and when the time domain mean square error and frequency domain difference are lower than the preset threshold, it is determined that the model passes the generalization verification. The verified model is embedded in the online monitoring system, the Kalman filter algorithm sets the noise covariance matrix according to the sensor accuracy, dynamically fuses the vibration displacement sensor, temperature field infrared monitoring data and model prediction value, and the covariance weighted weight is dynamically allocated based on the confidence. The parameter update period is synchronized with the unit control system, and the real-time state evaluation and fault warning signal is output, forming a closed-loop monitoring and optimization system under non-steady state condition.

[0125] The application significantly improves the interactive representation ability of dynamic excitation source and non-steady state working condition by phased parameter optimization and multi-physical field coupling modeling. First, the global optimization of the stiffness matrix and damping matrix of the multi-degree-of-freedom dynamic model is carried out based on genetic algorithm, the initial population is selected combined with historical parameter library, the mutation probability is adjusted by sensitivity classification, and the accuracy and stability of the dynamic model are optimized. Second, the thermal fluid coupling module and transient excitation input function are introduced, the particle swarm optimization algorithm is used to dynamically correct the coupling weight factor, the multi-physical field non-steady state coupling vibration response equation is generated based on the energy conservation threshold, and the vibration energy transmission characteristics under fluid-structure interaction are accurately quantified.

[0126] For the problem of vibration energy transmission path distortion, the transmission path distortion index is corrected in real time by adaptive fuzzy logic algorithm. Based on the historical fault case library, the membership function and fuzzy rule of high frequency resonance and low frequency vortex distortion mode are defined, the distortion index is calculated combined with the dynamic relationship of frequency domain phase shift and amplitude attenuation ratio, the amplitude frequency characteristics are corrected under the constraint of fuzzy rule by using gradient descent method, and the path distortion caused by non-steady state excitation is suppressed. The corrected vibration energy transmission characteristics provide dynamic reference for gap parameter optimization, and strengthen the closed loop control ability of multi-physical field coupling effect.

[0127] Through time-frequency domain joint verification and online monitoring closed loop fusion, the engineering applicability of the model is ensured. The bench test adopts hierarchical sampling to divide K-fold verification data set, the spectral kurtosis difference index is used to evaluate the consistency of frequency domain energy distribution of simulation data and measured data, and the model generalization is verified combined with time domain mean square error. The verified dynamic model is embedded in the online monitoring system, the Kalman filter algorithm dynamically fuses sensor data and model prediction value, the covariance weighted weight is distributed according to the confidence, and the vibration state is real-time evaluated and fault warning is realized by synchronous updating period, forming a closed loop monitoring system under non-steady state working condition.

Claims

1. A method for testing a turbine gap vibration based on transient numerical simulation, characterized in that, The method comprises the following steps: Step S1, obtaining the structural parameters and operating parameters of the unit, wherein the structural parameters include blade clearance and bearing support stiffness, and the operating parameters include rotating speed and temperature field distribution; Step S2, constructing a multi-degree-of-freedom dynamic model of the unit based on the structural parameters and operating parameters, performing global parameter optimization on the stiffness matrix and damping matrix of the multi-degree-of-freedom dynamic model through a genetic algorithm, and generating an optimized dynamic model; Step S3, integrating a thermal fluid coupling module into the optimized dynamic model, simulating the multi-field coupling vibration response under the non-steady-state boundary condition in combination with a transient excitation input function, and dynamically correcting the thermal fluid coupling weight factor through a particle swarm optimization algorithm to generate a multi-physical field non-steady-state coupling vibration response equation; Step S4, extracting the vibration energy distribution spectrum of the clearance area based on the multi-physical field non-steady-state coupling vibration response equation, identifying the distortion area of the vibration energy transmission path through a transmission path analysis, and real-time correcting the transmission path distortion index of the distortion area based on an adaptive fuzzy logic algorithm to generate a corrected vibration energy transmission characteristic; Step S5, establishing a sensitivity mapping relationship between the clearance geometric parameters and the vibration amplitude according to the corrected vibration energy transmission characteristic, performing a Pareto optimal solution set solving on the clearance design parameters and operating control parameters through a multi-objective genetic algorithm, and generating a collaborative optimization strategy; Step S6, performing a bench test based on the collaborative optimization strategy, obtaining measured vibration data, performing a generalization verification on the optimized dynamic model through a time-frequency domain joint verification strategy, and generating a verified dynamic model; Step S7, embedding the verified dynamic model into an online monitoring system, real-time fusing the vibration data collected by a sensor and the prediction result of the dynamic model through a Kalman filtering algorithm, dynamically updating the model parameters, and outputting a real-time state evaluation and fault warning signal of the unit clearance vibration.

2. The transient numerical simulation based test method for turbine rotor blade tip rub according to claim 1, wherein, The step S2 comprises: screening an initial population based on a historical model parameter library, and setting the population size of the genetic algorithm in combination with the structural degree-of-freedom number of the multi-degree-of-freedom dynamic model; iteratively optimizing the stiffness matrix and damping matrix of the multi-degree-of-freedom dynamic model through adaptive crossover probability and mutation probability, wherein the mutation probability is adjusted in stages according to the sensitivity analysis result of the stiffness matrix and the damping matrix; when the continuous iterative descending amplitude of the mean square error between the vibration energy predicted by the multi-degree-of-freedom dynamic model and the preset experimental data in the historical model parameter library is less than a preset threshold, outputting the optimized dynamic model parameters.

3. The transient numerical simulation based method of testing for rotor rub in turbomachinery as defined in Claim 2, wherein, The step S3 comprises: setting the particle swarm dimension according to the number of thermal fluid coupling fields, and adjusting the inertia weight of the particle swarm optimization based on a nonlinear decreasing strategy; taking the vibration energy transmission error of the multi-physical field non-steady-state coupling vibration response equation as an optimization objective function, and dynamically updating the search space of the thermal fluid coupling weight factor through individual learning factor and social learning factor; when the thermal fluid coupling weight factor converges to a preset energy conservation threshold, outputting the corrected multi-physical field non-steady-state coupling vibration response equation.

4. The transient numerical simulation based method of testing for rotor rub in turbomachinery as defined in Claim 3, wherein, The step S4 comprises: The membership function and fuzzy rule of the high-frequency resonance and low-frequency vortex distortion mode are defined based on a historical fault case library; According to the dynamic relationship between the frequency domain phase shift and the amplitude attenuation ratio of the multi-physical field non-steady-state coupling vibration response equation, the transfer path distortion index of the distortion region is calculated; The amplitude-frequency characteristics of the energy transfer path of the distortion region are corrected by the gradient descent method, and the correction is based on the real-time feedback of the membership function and fuzzy rule; When the transfer path distortion index is lower than the preset safety threshold, the corrected vibration energy transfer characteristics are output.

5. The transient numerical simulation based method of testing for rotor rub in turbomachinery as defined in Claim 4 wherein, The step S5 includes: taking the minimum vibration amplitude and the highest operation stability as optimization objectives, combining the sensitivity mapping relationship and the penalty function method into the gap design threshold and the operation safety boundary constraint; Non-dominated solution sets are generated by simulating binary crossover and polynomial mutation, and the Pareto front solutions are screened based on the crowding degree sorting; When the Pareto front solutions cover the preset engineering applicability range determined by the corrected vibration energy transfer characteristics, the cooperative optimization strategy is output.

6. The transient numerical simulation based method of testing for rotor rub in turbomachinery as defined in Claim 5, wherein, The step S6 includes: The K-fold verification data set is divided based on stratified sampling, covering cold start, load rejection and peak regulation operation conditions; The frequency energy distribution consistency of the simulation data of the optimized dynamic model generated in step S2 and the measured data obtained by the current bench test is analyzed by the spectral kurtosis difference index; When the time domain mean square error and the frequency domain spectral kurtosis difference of the simulation data are both lower than the preset verification threshold, it is determined that the dynamic model passes the generalization verification.

7. The transient numerical simulation based method of testing for rotor rub in turbomachinery as defined in Claim 6 wherein, The step S7 includes: According to the accuracy of the vibration displacement sensor and the temperature field infrared monitoring sensor, the process noise covariance matrix of the Kalman filter algorithm is set, and the observation noise covariance is dynamically updated by the prediction residual of the verified dynamic model; The vibration displacement sensor data, the temperature field infrared monitoring data and the output value of the verified dynamic model generated in step S6 are covariance weighted and fused; When the parameter update period of the verified dynamic model is synchronized with the sampling period of the unit control system, the real-time state evaluation and fault warning signal of the unit gap vibration are output.

Images