A method and system for diagnosing multi-modal vibration coupling of a water supply pump turbine shaft system
Patent Information
- Application Number
- CN202611208081.3
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-08-11
- Publication Date
- 2026-11-03
AI Technical Summary
Existing technologies cannot achieve synchronous acquisition and spatiotemporal registration of multimodal vibration signals of feedwater pump turbine shaft systems. They lack targeted coupling and decoupling algorithms, cannot adapt to the characteristics of variable operating conditions, resulting in distortion of multimodal vibration coupling feature extraction and misjudgment of faults, making it difficult to locate the root cause of the fault.
A distributed sensor array is used to synchronously acquire multimodal vibration signals. An improved adaptive variational mode decomposition algorithm is used for decoupling, a dedicated coupled fault feature map is constructed, and the diagnostic threshold is dynamically updated to achieve accurate decoupling of multimodal vibration coupling features and fault tracing.
It achieves spatiotemporal synchronous acquisition of multimodal vibration signals and accurate extraction of coupling features, improving fault diagnosis accuracy and identification coverage. It can accurately locate the root cause of faults and output graded early warnings under complex operating conditions, adapting to the full operating conditions of the unit.
Smart Images
Figure CN123030841A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of turbine fault diagnosis and maintenance technology, specifically a method and system for multimodal vibration coupling diagnosis of a feedwater pump turbine shaft system. Background Technology
[0002] The feedwater pump turbine is a core auxiliary equipment in thermal power generating units, undertaking the critical task of pressurizing and delivering boiler feedwater. Its shaft system is constantly under complex operating conditions of high-speed rotation, variable load fluctuations, and alternating steam flow disturbances, making it highly susceptible to multi-dimensional vibration phenomena such as radial, axial, and torsional vibrations. The vibrations of different modes do not exist independently but are strongly coupled and correlated. For example, shaft misalignment faults can simultaneously induce coupled radial and axial vibration disturbances, and rotor crack faults can cause superimposed interference from torsional and radial vibrations. The superposition of multi-mode vibration coupling can accelerate shaft bearing wear, rotor deformation, coupling failure, and other faults. In severe cases, it can lead to feedwater pump turbine shutdown, affecting the overall operational safety and economy of the unit.
[0003] Currently, most diagnostic technologies for shaft vibration in feedwater pump turbines employ a single radial vibration signal analysis mode, focusing only on the vibration amplitude, frequency, and phase. A few solutions analyze axial or torsional vibration signals separately, but all neglect the coupling correlation characteristics between multimodal vibrations. Existing technologies suffer from the following core defects: First, they cannot achieve synchronous acquisition and spatiotemporal registration of multimodal vibration signals, resulting in timing deviations between modes and distorted coupling feature extraction. Second, they lack targeted coupling decoupling algorithms, failing to isolate superimposed interference from multimodal vibrations and easily misdiagnosing coupled faults as single-mode conventional vibration faults. Third, fixed diagnostic thresholds cannot adapt to the variable operating characteristics of feedwater pump turbines, leading to extremely low accuracy in identifying coupled faults under low load and start-stop transition conditions. Fourth, they cannot trace the coupling correlation of multimodal vibrations, making it difficult to locate the root cause mode and coupling propagation path of the fault.
[0004] Existing publicly available multimodal vibration diagnosis schemes are mostly applicable to conventional rotating machinery and have not been adapted and optimized for the operating characteristics, shaft system structure features, and fault coupling laws of feedwater pump turbines. These schemes suffer from poor specificity, weak coupling feature recognition capabilities, and insufficient engineering applicability, failing to meet the current operational and maintenance needs for high-precision fault diagnosis and early warning for feedwater pump turbines in large-capacity, high-parameter thermal power units. Therefore, those skilled in the art provide a multimodal vibration coupling diagnosis method and system for feedwater pump turbine shaft systems to address the problems mentioned in the background. Summary of the Invention
[0005] The purpose of this invention is to provide a method and system for multimodal vibration coupling diagnosis of a feedwater pump turbine shaft system, so as to solve the problems mentioned in the background art.
[0006] To achieve the above objectives, the present invention provides the following technical solution:
[0007] A method for multimodal vibration coupling diagnosis of a feedwater pump turbine shaft system includes the following steps:
[0008] S1. Synchronous acquisition and preprocessing of multimodal vibration signals: Based on the distributed sensor array, the radial vibration signal, axial vibration signal, torsional vibration signal and real-time operating parameters of the feedwater pump turbine shaft system are synchronously acquired. All signals are preprocessed by time alignment, outlier removal and noise reduction filtering to obtain a standardized multimodal vibration dataset.
[0009] S2. Adaptive decoupling of multimodal vibration coupling components: An improved adaptive variational mode decomposition algorithm is used to iteratively decompose the standardized multimodal vibration dataset. The decomposition order and penalty factor are dynamically optimized according to the operating parameters of the feedwater pump turbine. The independent feature components and cross-coupled feature components of each mode vibration are separated to complete the decoupling of multimodal vibration coupling signals.
[0010] S3. Coupled Fault Feature Extraction and Map Construction: Extract the time domain and frequency domain features of each independent feature component and the coupling correlation features of the cross-coupled feature components respectively. Combine the operating parameters to construct the multimodal vibration coupling feature matrix of the feedwater pump turbine shaft system. Train and generate a dedicated coupled fault feature map based on the historical fault sample dataset.
[0011] S4. Adaptive Threshold Grading Diagnosis: The vibration coupling diagnosis threshold is dynamically updated based on the real-time load, speed, and steam pressure parameters of the unit. The coupling features extracted in real time are matched and compared with the coupling fault feature map to determine the shaft vibration fault type, fault coupling degree, and fault level.
[0012] S5. Fault Coupling Source Tracing and Early Warning Output: Based on feature matching results, analyze the coupling transmission path of multimodal vibration, locate the root cause mode and fault inducing factor, and output corresponding early warning information and operation and maintenance optimization suggestions according to the fault level.
[0013] As a further aspect of the present invention: In step S1, the distributed sensor array includes radial vibration sensors deployed on the shaft bearing housing, axial displacement sensors at the shaft end, and torsional vibration sensors at the coupling position. The signal sampling frequency is set to 1024Hz. The operating parameters include unit load, rotor speed, main steam pressure, and feedwater temperature. The preprocessing specifically involves: using the 3σ criterion to remove abnormal signal abrupt changes, using a wavelet threshold denoising algorithm to remove industrial site interference noise, using the real-time rotor speed as a unified time reference to complete the time sequence alignment of multimodal signals, and unifying the amplitude dimensions of all signals to obtain a standardized dataset.
[0014] As a further aspect of the present invention: In step S2, the parameter optimization rules of the improved adaptive variational mode decomposition algorithm are as follows: under the steady-state full-load condition of the unit, the decomposition order is set to 6-8; under the low-load and start-stop transition conditions of the unit, the decomposition order is set to 3-5. The penalty factor is dynamically iteratively optimized according to the real-time signal-to-noise ratio to select the optimal mode decomposition result.
[0015] As a further aspect of the present invention: In step S2, the parameter optimization rules of the improved adaptive variational mode decomposition algorithm are as follows: under the steady-state full-load operating condition of the unit, the decomposition order is set to 6-8; under the low-load and start-stop transition operating conditions of the unit, the decomposition order is set to 3-5. The penalty factor is dynamically iteratively optimized based on the real-time signal-to-noise ratio to select the optimal mode decomposition result.
[0016] As a further aspect of the present invention: In step S4, the method for dynamically updating the vibration coupling diagnostic threshold is as follows: a multi-dimensional correlation mapping model between operating parameters and diagnostic thresholds is established, and the vibration amplitude threshold and modal coupling coefficient threshold are adaptively adjusted according to the real-time operating range of the unit to achieve full-condition adaptive diagnosis. The fault levels are divided into 1-4 levels, with level 1 being no fault, level 2 being a minor single-mode fault, level 3 being a moderate dual-mode coupling fault, and level 4 being a severe tri-mode coupling fault.
[0017] As a further aspect of the present invention: in step S5, the fault tracing specifically involves: analyzing the coupling transmission weights of each mode of vibration, identifying the modal transmission paths of vibration faults, and pinpointing the causes of rotor misalignment, bearing wear, rotor cracks, and steam flow excitation faults. The early warning output includes tiered audible and visual warnings and data pop-up prompts, while simultaneously matching and outputting dedicated maintenance suggestions for online monitoring optimization, shutdown maintenance, and component verification corresponding to the fault level.
[0018] A multimodal vibration coupling diagnostic system for a feedwater pump turbine shaft system includes a synchronous acquisition module, a signal preprocessing module, a coupling and decoupling module, a feature modeling module, an intelligent diagnostic module, a source tracing and early warning module, and a data storage module.
[0019] As a further aspect of the present invention: the synchronous acquisition module is used to synchronously acquire radial, axial, and torsional multimodal vibration signals of the feedwater pump turbine shaft system and real-time operating parameters of the unit; the signal preprocessing module is used to complete the time-series alignment, outlier removal, noise reduction filtering, and standardization of the multimodal signals; the coupling and decoupling module is used to complete the adaptive decoupling of independent components and cross-coupled components of multimodal vibration through an improved adaptive variational mode decomposition algorithm; the feature modeling module is used to extract the time-domain, frequency-domain, and coupling correlation features of multimodal vibration, and construct a coupling feature matrix and a dedicated fault feature map; the intelligent diagnosis module is used to dynamically update the diagnostic threshold based on the operating parameters to complete the graded diagnosis of fault type, coupling degree, and fault level; the source tracing and early warning module is used to analyze the vibration coupling transmission path, locate the root cause of the fault, and output graded early warning information and operation and maintenance optimization suggestions; the data storage module is used to archive and store real-time acquired data, preprocessed data, feature map model data, and historical diagnostic record data.
[0020] Compared with the prior art, the beneficial effects of the present invention are:
[0021] This invention employs a spatiotemporal synchronous acquisition architecture for multimodal vibration signals, solving the problems of temporal misalignment and distortion in coupling feature extraction inherent in traditional technologies, thus ensuring the accuracy of coupling feature analysis. It utilizes an adaptive variational mode decomposition algorithm based on operating conditions, dynamically optimizing decomposition parameters according to the varying operating characteristics of the feedwater pump turbine. This accurately decouples independent and cross-coupled components of multimodal vibration, completely resolving the technical challenges of multimodal vibration feature aliasing and misdiagnosis of coupling faults. Furthermore, it constructs a unique multimodal vibration coupling fault feature map specific to feedwater pump turbines, distinct from general rotating... The mechanical diagnostic model is specifically adapted to the fault coupling law and operating condition characteristics of the feedwater pump turbine shaft system, significantly improving the fault diagnosis accuracy under complex operating conditions. It sets an adaptive dynamic diagnostic threshold, abandoning the traditional fixed threshold diagnostic mode, and effectively adapts to all operating conditions such as unit start-up and shutdown, variable load, and steady-state operation, significantly improving the fault identification coverage and accuracy. It can realize the source tracing of multimodal vibration coupling transmission path, accurately locate the root cause of the fault, and output graded early warning and operation and maintenance suggestions, realizing the integration of fault diagnosis, source tracing, and operation and maintenance guidance, which has high engineering application value. Attached Figure Description
[0022] Figure 1 This is a schematic diagram of the system flow for a multimodal vibration coupling diagnosis method and system for a feedwater pump turbine shaft system. Detailed Implementation
[0023] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.
[0024] Please see Figure 1 In this embodiment of the invention, a method for multimodal vibration coupling diagnosis of a feedwater pump turbine shaft system includes the following steps:
[0025] S1. Synchronous acquisition and preprocessing of multimodal vibration signals: Based on the distributed sensor array, the radial vibration signal, axial vibration signal, torsional vibration signal and real-time operating parameters of the feedwater pump turbine shaft system are synchronously acquired. All signals are preprocessed by time alignment, outlier removal and noise reduction filtering to obtain a standardized multimodal vibration dataset.
[0026] S2. Adaptive decoupling of multimodal vibration coupling components: An improved adaptive variational mode decomposition algorithm is used to iteratively decompose the standardized multimodal vibration dataset. The decomposition order and penalty factor are dynamically optimized according to the operating parameters of the feedwater pump turbine. The independent feature components and cross-coupled feature components of each mode vibration are separated to complete the decoupling of multimodal vibration coupling signals.
[0027] S3. Coupled Fault Feature Extraction and Map Construction: Extract the time domain and frequency domain features of each independent feature component and the coupling correlation features of the cross-coupled feature components respectively. Combine the operating parameters to construct the multimodal vibration coupling feature matrix of the feedwater pump turbine shaft system. Train and generate a dedicated coupled fault feature map based on the historical fault sample dataset.
[0028] S4. Adaptive Threshold Grading Diagnosis: The vibration coupling diagnosis threshold is dynamically updated based on the real-time load, speed, and steam pressure parameters of the unit. The coupling features extracted in real time are matched and compared with the coupling fault feature map to determine the shaft vibration fault type, fault coupling degree, and fault level.
[0029] S5. Fault Coupling Source Tracing and Early Warning Output: Based on feature matching results, analyze the coupling transmission path of multimodal vibration, locate the root cause mode and fault inducing factor, and output corresponding early warning information and operation and maintenance optimization suggestions according to the fault level.
[0030] In step S1, the distributed sensor array includes radial vibration sensors deployed on the shaft bearing housing, axial displacement sensors at the shaft end, and torsional vibration sensors at the coupling position. The signal sampling frequency is set to 1024Hz. The operating parameters include unit load, rotor speed, main steam pressure, and feedwater temperature. The preprocessing specifically involves: using the 3σ criterion to remove abnormal signal abrupt changes; using a wavelet threshold denoising algorithm to remove industrial site interference noise; using the real-time rotor speed as a unified time reference to complete multimodal signal time alignment; and unifying the amplitude dimensions of all signals to obtain a standardized dataset.
[0031] In step S2, the parameter optimization rules of the improved adaptive variational mode decomposition algorithm are as follows: under the steady-state full-load condition of the unit, the decomposition order is set to 6-8; under the low-load and start-stop transition conditions of the unit, the decomposition order is set to 3-5. The penalty factor is dynamically iteratively optimized according to the real-time signal-to-noise ratio to select the optimal mode decomposition result.
[0032] In step S3, the features of the independent feature components include peak value, effective value, average frequency, and phase difference. The coupling correlation features of the cross-coupled feature components include coupling correlation coefficient, feature superposition amplitude, and frequency offset. By training and fitting historical fault samples through a convolutional neural network, a coupled fault feature map adapted to the variable operating condition characteristics of the feedwater pump turbine is generated.
[0033] In step S4, the method for dynamically updating the vibration coupling diagnostic threshold is as follows: a multi-dimensional correlation mapping model between operating parameters and diagnostic thresholds is established, and the vibration amplitude threshold and modal coupling coefficient threshold are adaptively adjusted according to the real-time operating range of the unit to achieve full-condition adaptive diagnosis. The fault levels are divided into 1-4 levels, with level 1 being no fault, level 2 being a minor single-mode fault, level 3 being a moderate dual-mode coupling fault, and level 4 being a severe tri-mode coupling fault.
[0034] In step S5, the fault tracing specifically involves: analyzing the coupling transmission weights of each vibration mode, identifying the modal transmission paths of vibration faults, and pinpointing the causes of rotor misalignment, bearing wear, rotor cracks, and steam flow excitation. The early warning output includes tiered audible and visual warnings and data pop-up prompts, while simultaneously providing customized maintenance suggestions for online monitoring optimization, shutdown maintenance, and component verification corresponding to the fault level.
[0035] A multimodal vibration coupling diagnostic system for a feedwater pump turbine shaft system includes a synchronous acquisition module, a signal preprocessing module, a coupling and decoupling module, a feature modeling module, an intelligent diagnostic module, a source tracing and early warning module, and a data storage module.
[0036] The synchronous acquisition module is used to synchronously acquire radial, axial, and torsional multimodal vibration signals of the feedwater pump turbine shaft system and real-time operating parameters of the unit; the signal preprocessing module is used to complete the time-series alignment, outlier removal, noise reduction filtering, and standardization of multimodal signals; the coupling and decoupling module is used to complete the adaptive decoupling of independent and cross-coupled components of multimodal vibration through an improved adaptive variational mode decomposition algorithm; the feature modeling module is used to extract time-domain, frequency-domain, and coupling correlation features of multimodal vibration, and construct a coupling feature matrix and a dedicated fault feature map; the intelligent diagnosis module is used to dynamically update the diagnostic threshold based on operating parameters to complete the graded diagnosis of fault type, coupling degree, and fault level; the source tracing and early warning module is used to analyze the vibration coupling transmission path, locate the root cause of the fault, and output graded early warning information and operation and maintenance optimization suggestions; the data storage module is used to archive and store real-time acquired data, preprocessed data, feature map model data, and historical diagnostic record data.
[0037] The above description is only a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any equivalent substitutions or modifications made by those skilled in the art within the scope of the technology disclosed in the present invention, based on the technical solution and inventive concept of the present invention, should be covered within the scope of protection of the present invention.
Claims
1. A method for multimodal vibration coupling diagnosis of a feedwater pump turbine shaft system, characterized in that, Includes the following steps: S1. Synchronous acquisition and preprocessing of multimodal vibration signals: Based on the distributed sensor array, the radial vibration signal, axial vibration signal, torsional vibration signal and real-time operating parameters of the feedwater pump turbine shaft system are synchronously acquired. All signals are preprocessed by time alignment, outlier removal and noise reduction filtering to obtain a standardized multimodal vibration dataset. S2. Adaptive decoupling of multimodal vibration coupling components: An improved adaptive variational mode decomposition algorithm is used to iteratively decompose the standardized multimodal vibration dataset. The decomposition order and penalty factor are dynamically optimized according to the operating parameters of the feedwater pump turbine. The independent feature components and cross-coupled feature components of each mode vibration are separated to complete the decoupling of multimodal vibration coupling signals. S3. Coupled Fault Feature Extraction and Map Construction: Extract the time domain and frequency domain features of each independent feature component and the coupling correlation features of the cross-coupled feature components respectively. Combine the operating parameters to construct the multimodal vibration coupling feature matrix of the feedwater pump turbine shaft system. Train and generate a dedicated coupled fault feature map based on the historical fault sample dataset. S4. Adaptive Threshold Grading Diagnosis: The vibration coupling diagnosis threshold is dynamically updated based on the real-time load, speed, and steam pressure parameters of the unit. The coupling features extracted in real time are matched and compared with the coupling fault feature map to determine the shaft vibration fault type, fault coupling degree, and fault level. S5. Fault Coupling Source Tracing and Early Warning Output: Based on feature matching results, analyze the coupling transmission path of multimodal vibration, locate the root cause mode and fault inducing factor, and output corresponding early warning information and operation and maintenance optimization suggestions according to the fault level.
2. The method for multimodal vibration coupling diagnosis of a feedwater pump turbine shaft system according to claim 1, characterized in that, In step S1, the distributed sensor array includes a radial vibration sensor installed on the shaft bearing housing, an axial displacement sensor at the shaft end, and a torsional vibration sensor at the coupling position. The signal sampling frequency is set to 1024Hz. The operating parameters include unit load, rotor speed, main steam pressure, and feedwater temperature.
3. The method for multimodal vibration coupling diagnosis of a feedwater pump turbine shaft system according to claim 1, characterized in that, In step S1, the preprocessing specifically involves: using the 3σ criterion to remove abnormal signal mutation values, using wavelet threshold denoising algorithm to remove industrial field interference noise, using the real-time rotor speed as a unified time reference to complete the timing alignment of multimodal signals, and unifying the amplitude dimensions of all signals to obtain a standardized dataset.
4. The method for multimodal vibration coupling diagnosis of a feedwater pump turbine shaft system according to claim 1, characterized in that, In step S2, the parameter optimization rules of the improved adaptive variational mode decomposition algorithm are as follows: under the steady-state full-load condition of the unit, the decomposition order is set to 6-8; under the low-load and start-stop transition conditions of the unit, the decomposition order is set to 3-5. The penalty factor is dynamically iteratively optimized according to the real-time signal-to-noise ratio to select the optimal mode decomposition result.
5. The method for multimodal vibration coupling diagnosis of a feedwater pump turbine shaft system according to claim 1, characterized in that, In step S3, the features of the independent feature components include peak value, effective value, average frequency, and phase difference. The coupling correlation features of the cross-coupled feature components include coupling correlation coefficient, feature superposition amplitude, and frequency offset. By training and fitting historical fault samples through a convolutional neural network, a coupled fault feature map adapted to the variable operating condition characteristics of the feedwater pump turbine is generated.
6. The method for multimodal vibration coupling diagnosis of a feedwater pump turbine shaft system according to claim 1, characterized in that, In step S4, the method of dynamically updating the vibration coupling diagnostic threshold is as follows: establish a multi-dimensional correlation mapping model between operating parameters and diagnostic thresholds, and adaptively adjust the vibration amplitude threshold and modal coupling coefficient threshold according to the real-time operating range of the unit to achieve full-condition adaptive diagnosis.
7. The method for multimodal vibration coupling diagnosis of a feedwater pump turbine shaft system according to claim 1, characterized in that, In step S4, the fault level is divided into 1-4 levels, with level 1 being no fault, level 2 being a single-mode minor fault, level 3 being a dual-mode moderate coupling fault, and level 4 being a tri-mode severe coupling fault.
8. The method for multimodal vibration coupling diagnosis of a feedwater pump turbine shaft system according to claim 1, characterized in that, In step S5, the fault tracing specifically involves: analyzing the coupling transmission weights of each mode of vibration, sorting out the mode transmission paths of vibration faults, and locating the causes of rotor misalignment, bearing wear, rotor cracks, and steam flow excitation faults.
9. The method for multimodal vibration coupling diagnosis of a feedwater pump turbine shaft system according to claim 1, characterized in that, In step S5, the early warning output includes tiered audible and visual early warnings and data pop-up prompts, while also matching and outputting dedicated operation and maintenance suggestions for online monitoring optimization, shutdown maintenance, and component verification corresponding to the fault level.
10. A multi-modal vibration coupling diagnostic system for a feedwater pump turbine shaft system, characterized in that, The method for implementing the multimodal vibration coupling diagnosis method for the feedwater pump turbine shaft system as described in any one of claims 1-9 includes a synchronous acquisition module, a signal preprocessing module, a coupling decoupling module, a feature modeling module, an intelligent diagnosis module, a source tracing and early warning module, and a data storage module. The synchronous acquisition module is used to synchronously acquire radial, axial, and torsional multimodal vibration signals of the feedwater pump turbine shaft system and real-time operating parameters of the unit. The signal preprocessing module is used to perform time alignment, outlier removal, noise reduction filtering, and standardization of multimodal signals. The coupling and decoupling module is used to complete the adaptive decoupling of independent components and cross-coupled components of multimodal vibration through an improved adaptive variational mode decomposition algorithm. The feature modeling module is used to extract time-domain, frequency-domain, and coupled correlation features of multimodal vibration, and to construct a coupled feature matrix and a dedicated fault feature map. The intelligent diagnostic module is used to dynamically update the diagnostic threshold based on operating parameters to complete the hierarchical diagnosis of fault type, coupling degree and fault level. The source tracing and early warning module is used to analyze the vibration coupling transmission path, locate the root cause of the fault, and output graded early warning information and operation and maintenance optimization suggestions. The data storage module is used to archive and store real-time acquired data, preprocessed data, feature map model data, and historical diagnostic record data.