Method and system for evaluating vibration fault risk of steam turbine

Abstract

The invention relates to the technical field of equipment fault diagnosis, and discloses a steam turbine vibration fault risk evaluation method and system.The method comprises the following steps that real-time data such as vibration, temperature, pressure and rotating speed of a steam turbine are collected through multiple sensors; the sensors comprise an accelerometer, a displacement sensor, a temperature sensor, a pressure sensor and a rotating speed sensor; the collected multi-source data is preprocessed, including de-noising, time synchronization and data standardization, to eliminate environmental noise and time difference between the data. The technical scheme based on multi-dimensional data acquisition and fusion is adopted, vibration sensors, temperature sensors, pressure sensors, rotating speed sensors and the like are arranged at key parts of a steam turbine, the running state of equipment is obtained in real time, multi-dimensional fusion and feature extraction are carried out on data, and the effects of comprehensive and real-time monitoring and early prediction of vibration faults of the steam turbine are achieved.

Description

Technical Field

[0001] The present invention relates to the technical field of equipment fault diagnosis, and in particular to a method and system for evaluating the risk of vibration fault of a steam turbine. Background Art

[0002] With the advancement of industrialization, steam turbines, as important power machinery equipment, are widely used in the fields of electricity, petrochemicals, metallurgy, etc., playing a key role in energy conversion and driving. As a high-load, high-speed rotating equipment, the operating status and fault diagnosis of steam turbines directly affect the safety, stability and economy of production. In order to ensure the long-term operation of steam turbine equipment and timely discover and deal with potential faults, traditional fault diagnosis and maintenance methods usually rely on regular maintenance and manual judgment. These methods often cannot monitor the operating status of the equipment in real time, nor can they make full use of the large amount of data generated during the operation of the equipment, resulting in inaccurate predictions of potential faults, and even unnecessary downtime and maintenance costs due to overly conservative or overly delayed maintenance decisions.

[0003] At present, traditional steam turbine fault diagnosis technology is mainly divided into two categories: one is the manual inspection and regular maintenance method based on experience, and the other is the traditional monitoring method based on model. Although the former is simple to operate, it relies on personnel experience and is difficult to achieve comprehensive monitoring of equipment. It cannot detect sudden failures in time, nor can it provide early warning of potential equipment degradation trends. Although the latter can identify some fault signs by regularly monitoring key parameters such as vibration, temperature, and pressure of the equipment, most of these traditional monitoring methods lack in-depth mining of the correlation between data and fail to make full use of the multi-dimensional data of equipment operation, resulting in their evaluation results often not accurately reflecting the true health status of the equipment. In addition, existing risk assessment and fault prediction methods usually rely on simple mathematical models or empirical rules, and cannot flexibly adjust the prediction results according to real-time data, which in turn affects the accuracy of the prediction and the rationality of maintenance decisions.

[0004] Existing technologies have some key deficiencies in dynamic monitoring, fault prediction, risk assessment and decision support. First, existing technologies mostly rely on single or limited sensor data and fail to comprehensively evaluate the operating status of equipment through multi-dimensional data fusion. Second, existing fault prediction and risk assessment methods lack dynamics and cannot respond to changes in the equipment operation process in a timely manner. They cannot fully consider factors such as the real-time status of the equipment, load fluctuations and environmental changes. Summary of the invention

[0005] In view of the shortcomings of the prior art, the present invention provides a method and system for evaluating the risk of turbine vibration failure, which solves the problem that the prior art mostly relies on single or limited sensor data and fails to comprehensively evaluate the operating status of the equipment through multi-dimensional data fusion.

[0006] To achieve the above objectives, the present invention is implemented through the following technical solutions: A method for evaluating the risk of vibration failure of a steam turbine comprises the following steps:

[0007] Collecting real-time data such as vibration, temperature, pressure and speed of the steam turbine through a variety of sensors, including accelerometers, displacement sensors, temperature sensors, pressure sensors and speed sensors;

[0008] Preprocess the collected multi-source data, including denoising, time synchronization and data standardization, to eliminate environmental noise and time differences between data;

[0009] Perform multi-dimensional fusion on the pre-processed data, and use weighted average method or other data fusion algorithms to combine the data from different sensors into a unified comprehensive health assessment value;

[0010] Based on multi-source data and fusion results, dynamic risk assessment is performed using Bayesian networks or fuzzy logic models to assess the current failure risk of the turbine and adjust the assessment results based on real-time data.

[0011] Preferably, the evaluation method further comprises:

[0012] Use machine learning algorithms or deep learning algorithms (including support vector machines (SVMs) and long short-term memory networks (LSTMs)) to extract features and predict faults from vibration signals, and predict the remaining service life (RUL) of the equipment;

[0013] Based on the risk assessment results, maintenance recommendations are generated and multi-objective optimization algorithms (including particle swarm optimization PSO) are used to provide optimal maintenance decisions to reduce downtime and maintenance costs.

[0014] Preferably, the data preprocessing further includes using Kalman filtering or wavelet transform method to remove noise in the sensor signal to ensure data quality;

[0015] Multi-dimensional data fusion uses a weighted average method, which calculates the comprehensive health assessment value through the following formula:

[0016]

[0017] Among them, w i is the weight coefficient of the i-th sensor, X i is the data of the i-th sensor.

[0018] Preferably, the dynamic risk assessment is modeled using a Bayesian network, and the joint probability distribution of the Bayesian network is:

[0019]

[0020] Among them, X1, X2, …, X n is the relevant variable, Pa(X i ) is the variable X i The parent node of .

[0021] Preferably, the fault prediction uses a long short-term memory network (LSTM) to model the time series data of the vibration signal, and predicts the remaining service life of the turbine based on the model.

[0022] Preferably, the multi-objective optimization algorithm is a particle swarm optimization PSO algorithm, and the particle swarm optimization is calculated by the following formula:

[0023]

[0024]

[0025] in, is the position of the particle, is the velocity of the particle, ω is the inertia weight, c1, c2 are acceleration constants, and r1, r2 are random factors.

[0026] A steam turbine vibration failure risk assessment system, comprising:

[0027] Data acquisition unit: multiple sensor modules are used to collect data such as vibration, temperature, pressure and speed of the steam turbine, and transmit the data to the data processing unit;

[0028] Data processing unit: including preprocessing module, data fusion module, risk assessment module, fault prediction module and decision support module, used to process, analyze, evaluate and predict the collected data;

[0029] Risk assessment module: used to assess the failure risk of steam turbines through Bayesian network or fuzzy logic model based on preprocessed and fused data;

[0030] Fault prediction module: used to predict the fault risk and remaining service life UL of the steam turbine through support vector machine or LSTM algorithm;

[0031] Decision support module: Generates maintenance recommendations based on risk assessment results and provides optimal maintenance decisions through multi-objective optimization algorithms.

[0032] Preferably, the data acquisition unit includes a vibration sensor, a temperature sensor, a pressure sensor and a rotation speed sensor, and the sensors transmit data to the data processing unit via a wireless sensor network.

[0033] Preferably, the preprocessing module of the data processing unit uses Kalman filtering or wavelet transform method to perform denoising and time synchronization processing on the sensor signal.

[0034] Preferably, the decision support module adopts a particle swarm optimization (PSO) algorithm to provide optimal maintenance decisions to reduce maintenance costs and downtime.

[0035] The present invention provides a method and system for evaluating the risk of vibration failure of a steam turbine, which has the following beneficial effects:

[0036] 1. The present invention adopts a technical solution based on multi-dimensional data collection and fusion, deploys vibration, temperature, pressure and speed sensors at key parts of the steam turbine, obtains the equipment operation status in real time, and performs multi-dimensional fusion and feature extraction on the data, thereby achieving comprehensive, real-time monitoring and early prediction of steam turbine vibration faults. Compared with the existing technology that only relies on a single sensor or periodic inspection, the present invention can continuously monitor and timely predict potential faults during the operation of the equipment, thereby avoiding the problem that traditional methods cannot respond in real time, and significantly improving the accuracy and timeliness of fault warning.

[0037] 2. The present invention uses a dynamic risk assessment model and an intelligent decision support system, combined with Bayesian networks, fuzzy logic and multi-objective optimization algorithms, to comprehensively assess the operating risks of steam turbines and put forward scientific and reasonable maintenance suggestions based on real-time acquisition of equipment status information. Compared with the static risk assessment method commonly used in the prior art, the dynamic risk assessment adopted by the present invention can continuously update the assessment results based on the real-time data of the equipment, ensure the real-time and accuracy of maintenance decisions, and effectively reduce the problem of traditional solutions being unable to cope with sudden failures or overly conservative maintenance decisions.

[0038] 3. The present invention is based on the technical solution of machine learning including LSTM and multi-objective optimization algorithm including particle swarm optimization PSO, which can accurately predict the remaining service life RUL of the equipment and optimize the maintenance strategy on this basis, avoiding unnecessary downtime and excessive maintenance. Compared with the solution relying on experience and regular maintenance in the prior art, the present invention can provide a more accurate maintenance plan based on real-time monitoring data and accurate fault prediction results, effectively reducing the unplanned downtime and maintenance costs of the equipment, and improving the operating reliability and economic benefits of the steam turbine. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] Figure 1 is a flow chart of the method of the present invention;

[0040] Figure 2 It is a schematic diagram of the system framework of the present invention. DETAILED DESCRIPTION

[0041] The following will be combined with the drawings in the specification of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0042] Please see attached Figure 1 - Attachment Figure 2 , an embodiment of the present invention provides a method for evaluating the risk of vibration failure of a steam turbine, which collects real-time data such as vibration, temperature, pressure and rotation speed of the steam turbine through a variety of sensors, wherein the sensors include an accelerometer, a displacement sensor, a temperature sensor, a pressure sensor and a rotation speed sensor;

[0043] Preprocess the collected multi-source data, including denoising, time synchronization and data standardization, to eliminate environmental noise and time differences between data;

[0044] Perform multi-dimensional fusion on the pre-processed data, and use weighted average method or other data fusion algorithms to combine the data from different sensors into a unified comprehensive health assessment value;

[0045] Based on multi-source data and fusion results, dynamic risk assessment is performed using Bayesian networks or fuzzy logic models to assess the current failure risk of the turbine and adjust the assessment results based on real-time data.

[0046] Specifically, the key operating data of the turbine, such as vibration, temperature, pressure and speed, are collected in real time through a variety of sensors, including accelerometers, displacement sensors, temperature sensors, pressure sensors and speed sensors. These sensors transmit the collected data to the data processing platform through the network system. After being transmitted to the data processing platform, the collected multi-source data is firstly subjected to denoising processing, and the environmental noise and interference in the signal are removed by using methods such as Kalman filtering or wavelet transform to ensure the high quality of the data; then the data from different sources are synchronized and standardized to solve the inconsistency and scale difference of the data in time, and ensure the accuracy of subsequent analysis. Then, the results of different sensors are fused using the weighted average method or other data fusion algorithms, and the data information from each sensor is integrated into a unified comprehensive health assessment value. Finally, based on the multi-source data and the fused assessment results, a Bayesian network or fuzzy logic model is used to perform dynamic risk assessment to assess the current fault risk of the turbine, and the assessment results can be dynamically adjusted according to the changes in real-time data, thereby providing more accurate and timely risk warnings, helping to achieve early diagnosis of faults and reasonable maintenance decisions.

[0047] The evaluation method further comprises:

[0048] Use machine learning algorithms or deep learning algorithms (including support vector machines (SVMs) and long short-term memory networks (LSTMs)) to extract features and predict faults from vibration signals, and predict the remaining service life (RUL) of the equipment;

[0049] Based on the risk assessment results, maintenance recommendations are generated and multi-objective optimization algorithms (including particle swarm optimization PSO) are used to provide optimal maintenance decisions to reduce downtime and maintenance costs.

[0050] Specifically, the system first analyzes the characteristics of the vibration signal in the time domain and frequency domain, extracts the key features related to the fault mode, and uses the support vector machine (SVM) for classification to determine whether there is a fault and issue an early warning; at the same time, the long short-term memory network (LSTM) is used to model the time series data of the vibration signal to accurately predict the remaining useful life (RUL) of the equipment. Based on the dynamic risk assessment results, the system will generate maintenance recommendations for different fault modes and working conditions, and optimize maintenance decisions through multi-objective optimization algorithms (such as particle swarm optimization PSO), comprehensively considering factors such as fault repair costs, equipment downtime, and maintenance personnel configuration, and giving the optimal maintenance decision, thereby minimizing downtime, reducing maintenance costs, and improving equipment reliability and operating efficiency.

[0051] The data preprocessing also includes using Kalman filtering or wavelet transform method to remove noise in sensor signals to ensure data quality;

[0052] Multi-dimensional data fusion uses a weighted average method, which calculates the comprehensive health assessment value through the following formula:

[0053]

[0054] Among them, w i is the weight coefficient of the i-th sensor, X i is the data of the i-th sensor.

[0055] Specifically, Kalman filtering or wavelet transform methods are used to remove noise from sensor signals. Kalman filtering can effectively estimate the optimal state value from noisy signals, thereby improving the quality and accuracy of signals; wavelet transform analyzes signals in different frequency bands through multi-scale decomposition, removes high-frequency noise, and retains effective low-frequency information, thereby improving data reliability. The data fusion part uses the weighted average method to combine data from different sensors into a comprehensive health assessment value. In this process, different weight coefficients are assigned according to the reliability, accuracy and historical performance of each sensor, and a more accurate health assessment result is obtained through weighted averaging, thereby providing more stable and accurate data support for subsequent risk assessment and fault prediction.

[0056] Dynamic risk assessment is modeled using a Bayesian network, and the joint probability distribution of the Bayesian network is:

[0057]

[0058] Among them, X1, X2, …, X n is the relevant variable, Pa(X i ) is the variable X i The parent node of .

[0059] Specifically, dynamic risk assessment is modeled using Bayesian networks. By describing the conditional dependencies between various variables, Bayesian networks can effectively capture the interactions and impacts between various operating data of the steam turbine and potential faults. Based on the multi-dimensional data collected by sensors, the network establishes the probabilistic relationship between various fault factors (such as vibration, temperature, pressure, etc.) and fault risks, and updates the risk assessment results through Bayesian reasoning. Each node represents a variable, and the edges between nodes represent conditional dependencies. By calculating the joint probability distribution, the failure risk of the current equipment can be dynamically evaluated, and the evaluation results can be adjusted according to changes in real-time data, thereby providing accurate fault prediction and real-time risk assessment. This method can effectively deal with the complex working conditions and potential failure modes of equipment, and help achieve more accurate risk management.

[0060] Fault prediction uses a long short-term memory network (LSTM) to model the time series data of the vibration signal and predict the remaining service life of the turbine based on the model.

[0061] Specifically, fault prediction uses a long short-term memory network (LSTM) to model the time series data of vibration signals. LSTM can effectively capture long-term dependencies in signals and is particularly suitable for processing vibration data with time series characteristics. By training historical vibration data, the LSTM model can learn the equipment's operating mode, vibration change trends, and potential signs of failure, thereby predicting the remaining useful life (RUL) of the turbine. This method can continuously monitor the health status of the equipment based on real-time data and provide early warnings before a failure occurs, helping maintenance personnel to develop reasonable maintenance plans, avoiding equipment downtime and production interruptions caused by sudden failures, and maximizing the service life of the equipment.

[0062] The multi-objective optimization algorithm is a particle swarm optimization PSO algorithm, which is calculated by the following formula:

[0063]

[0064]

[0065] in, is the position of the particle, is the velocity of the particle, ω is the inertia weight, c1, c2 are acceleration constants, and r1, r2 are random factors.

[0066] Specifically, the multi-objective optimization algorithm adopts the particle swarm optimization (PSO) algorithm. PSO is an optimization method that simulates the foraging behavior of bird flocks. It solves the problem by searching for the optimal solution in the solution space. In this method, each particle represents a possible solution. The particle adjusts its position based on its own experience and the experience of the group, and gradually approaches the optimal solution. The update process of the particle's speed and position takes into account the current particle's position, the historical optimal position, and the global optimal position in the group, thereby guiding the search process to approach the optimal solution. Through particle swarm optimization, multiple optimization objectives can be considered at the same time, such as fault repair cost, downtime and maintenance cost, thereby providing a comprehensive optimal maintenance decision to ensure the efficient operation and economic benefits of the equipment. The advantage of the PSO algorithm is that it can handle complex multi-objective optimization problems and has strong global search capabilities, which is suitable for dynamically changing complex systems.

[0067] A steam turbine vibration failure risk assessment system, comprising:

[0068] Data acquisition unit: multiple sensor modules are used to collect data such as vibration, temperature, pressure and speed of the steam turbine, and transmit the data to the data processing unit;

[0069] Data processing unit: including preprocessing module, data fusion module, risk assessment module, fault prediction module and decision support module, used to process, analyze, evaluate and predict the collected data;

[0070] Risk assessment module: used to assess the failure risk of steam turbines through Bayesian network or fuzzy logic model based on preprocessed and fused data;

[0071] Fault prediction module: used to predict the fault risk and remaining service life UL of the steam turbine through support vector machine or LSTM algorithm;

[0072] Decision support module: Generates maintenance recommendations based on risk assessment results and provides optimal maintenance decisions through multi-objective optimization algorithms.

[0073] Specifically, the data acquisition unit collects key operating data such as vibration, temperature, pressure and speed of the steam turbine in real time through multiple sensor modules. The sensor modules include accelerometers, temperature sensors, pressure sensors and speed sensors. These sensors transmit data to the data processing unit through a wireless sensor network. The data processing unit contains multiple functional modules, among which the preprocessing module performs denoising, time synchronization and standardization on the collected data to ensure data quality; the data fusion module combines the data from different sensors into a unified health assessment value through weighted average method or other fusion algorithms; the risk assessment module evaluates the failure risk of the steam turbine through Bayesian network or fuzzy logic model based on the fused data, and updates the risk level in real time; the fault prediction module uses support vector machine (SVM) or long short-term memory network (LSTM) algorithm to extract key features from the vibration signal and predict the remaining service life (RUL) of the steam turbine to help warn of potential faults in advance; the decision support module generates personalized maintenance suggestions based on the risk assessment results, and comprehensively considers factors such as maintenance cost and downtime through multi-objective optimization algorithms (such as particle swarm optimization PSO) to provide the best maintenance decision, thereby achieving accurate and efficient equipment maintenance and fault management. The system can monitor the operating status of the steam turbine in real time, make accurate fault predictions, reduce unplanned downtime and maintenance costs, and improve equipment reliability and operating efficiency.

[0074] The data acquisition unit includes a vibration sensor, a temperature sensor, a pressure sensor and a rotation speed sensor, and the sensors transmit data to the data processing unit through a wireless sensor network.

[0075] The preprocessing module of the data processing unit uses Kalman filtering or wavelet transform method to perform denoising and time synchronization processing on the sensor signal.

[0076] The decision support module uses the particle swarm optimization (PSO) algorithm to provide optimal maintenance decisions to reduce maintenance costs and downtime.

[0077] Specifically, the preprocessing module of the data processing unit uses Kalman filtering or wavelet transform methods to denoise and time synchronize the sensor signals to eliminate environmental noise and time differences between different sensors, ensure high quality and consistency of data, and improve the accuracy of subsequent analysis. The decision support module uses a particle swarm optimization (PSO) algorithm to comprehensively consider multiple factors such as maintenance costs, equipment downtime, maintenance cycles, etc. based on the collected real-time data and evaluation results. Through multi-objective optimization, it provides the best maintenance decision, thereby effectively reducing maintenance costs, reducing unplanned downtime, ensuring efficient and safe operation of the turbine, and improving the economy and reliability of the overall equipment.

[0078] Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A method for evaluating the risk of vibration failure of a steam turbine, characterized in that: The following steps are involved: Collecting real-time data such as vibration, temperature, pressure and speed of the steam turbine through a variety of sensors, including accelerometers, displacement sensors, temperature sensors, pressure sensors and speed sensors; Preprocess the collected multi-source data, including denoising, time synchronization and data standardization, to eliminate environmental noise and time differences between data; Perform multi-dimensional fusion on the pre-processed data, and use weighted average method or other data fusion algorithms to combine the data from different sensors into a unified comprehensive health assessment value; Based on multi-source data and fusion results, dynamic risk assessment is performed using Bayesian networks or fuzzy logic models to assess the current failure risk of the turbine and adjust the assessment results based on real-time data.

2. A method for evaluating the risk of vibration failure of a steam turbine according to claim 1, characterized in that: The evaluation method further comprises: Use machine learning algorithms or deep learning algorithms (including support vector machines (SVMs) and long short-term memory networks (LSTMs)) to extract features and predict faults from vibration signals, and predict the remaining service life (RUL) of the equipment; Based on the risk assessment results, maintenance recommendations are generated and multi-objective optimization algorithms (including particle swarm optimization PSO) are used to provide optimal maintenance decisions to reduce downtime and maintenance costs.

3. The method for evaluating the risk of vibration failure of a steam turbine according to claim 1, characterized in that: The data preprocessing also includes using Kalman filtering or wavelet transform method to remove noise in sensor signals to ensure data quality; Multi-dimensional data fusion uses a weighted average method, which calculates the comprehensive health assessment value through the following formula: Among them, w i is the weight coefficient of the i-th sensor, X i is the data of the ith sensor.

4. The method for evaluating the risk of vibration failure of a steam turbine according to claim 1, characterized in that: Dynamic risk assessment is modeled using a Bayesian network, and the joint probability distribution of the Bayesian network is: Among them, X1, X2, …, X n is the relevant variable, Pa ( X i) For variable X i The parent node of .

5. The method for evaluating the risk of vibration failure of a steam turbine according to claim 1, characterized in that: Fault prediction uses a long short-term memory network (LSTM) to model the time series data of the vibration signal and predict the remaining service life of the turbine based on the model.

6. A method for evaluating the risk of vibration failure of a steam turbine according to claim 1, characterized in that: The multi-objective optimization algorithm is a particle swarm optimization PSO algorithm, which is calculated by the following formula: in, is the position of the particle, is the velocity of the particle, ω is the inertia weight, c1, c2 are acceleration constants, and r1, r2 are random factors.

7. A steam turbine vibration failure risk assessment system, used to implement a steam turbine vibration failure risk assessment method as described in claims 1-6, characterized in that: include: Data acquisition unit: multiple sensor modules are used to collect data such as vibration, temperature, pressure and speed of the steam turbine, and transmit the data to the data processing unit; Data processing unit: including preprocessing module, data fusion module, risk assessment module, fault prediction module and decision support module, used to process, analyze, evaluate and predict the collected data; Risk assessment module: used to assess the failure risk of steam turbines through Bayesian network or fuzzy logic model based on preprocessed and fused data; Fault prediction module: used to predict the fault risk and remaining service life UL of the steam turbine through support vector machine or LSTM algorithm; Decision support module: Generates maintenance recommendations based on risk assessment results and provides optimal maintenance decisions through multi-objective optimization algorithms.

8. A steam turbine vibration failure risk assessment system according to claim 7, characterized in that: The data acquisition unit includes a vibration sensor, a temperature sensor, a pressure sensor and a rotation speed sensor, and the sensors transmit data to the data processing unit through a wireless sensor network.

9. The steam turbine vibration failure risk assessment system according to claim 7, characterized in that: The preprocessing module of the data processing unit uses Kalman filtering or wavelet transform method to perform denoising and time synchronization processing on the sensor signal.

10. The steam turbine vibration failure risk assessment system according to claim 7, characterized in that: The decision support module uses the particle swarm optimization (PSO) algorithm to provide optimal maintenance decisions to reduce maintenance costs and downtime.

Images