A method for monitoring the service process of a steam turbine rotor based on digital twinning

Abstract

The application provides a steam turbine rotor service process monitoring method based on digital twinning, and relates to the technical field of steam turbine monitoring. The steam turbine rotor service process monitoring method based on digital twinning can monitor the steam turbine rotor service process in real time from different angles by fusing multiple digital twinning sub-models to construct a digital twinning model, and can better reflect the complex and changeable operating conditions in the steam turbine rotor service process by monitoring the steam turbine rotor service process according to real-time vibration signals. The method makes up for the shortcomings of traditional monitoring technologies, such as poor real-time performance, slow data processing and analysis, excessive consumption of human and material resources, and is suitable for complex and changeable conditions of the steam turbine rotor service process, and provides a new idea for optimizing the steam turbine rotor service monitoring technology.

Description

Technical Field

[0001] This invention relates to the field of steam turbine monitoring technology, and more specifically, to a method for monitoring the service process of a steam turbine rotor based on digital twins. Background Technology

[0002] Steam turbines play a crucial role in energy conversion and are the most important rotating equipment in power plants. However, the turbine rotor is also the component that operates under the most severe conditions during unit operation and is the weakest link in the unit's lifespan. Therefore, it is essential to strengthen real-time monitoring of the unit's rotor performance.

[0003] Steam turbine rotors are high-speed rotating components, and precise control and adjustment of the rotor's service process is a technical challenge in current research. Traditional information acquisition technology consumes a lot of manpower and time. At the same time, the speed of manual data processing is far from keeping up with the speed of data generation and acquisition, making it impossible to achieve real-time monitoring of the steam turbine rotor's service process. Summary of the Invention

[0004] The following provides a brief overview of one or more aspects to offer a basic understanding of them. This overview is not an exhaustive summary of all conceived aspects, nor is it intended to identify key or decisive elements of all aspects, nor to define the scope of any or all aspects. Its sole purpose is to present some concepts of one or more aspects in a simplified form to prepare for the more detailed descriptions that follow.

[0005] The present invention aims to provide a method for monitoring the service process of a steam turbine rotor based on digital twins, which can realize real-time monitoring of the service process of the steam turbine rotor.

[0006] Embodiments of the present invention can be implemented in the following ways:

[0007] A method for monitoring the service process of a steam turbine rotor based on digital twins, comprising:

[0008] Acquire historical failure data of the turbine rotor, initial characteristic information of the turbine rotor service process system, and initial service condition parameters;

[0009] An executable digital twin model of the turbine rotor service process is constructed based on the historical failure data of the turbine rotor, the initial characteristic information, and the initial service condition parameters; wherein, the digital twin model includes a binary classification prediction model, a temperature field model, a constitutive model, a continuous damage model, and a degradation model;

[0010] The digital twin sub-models are then merged to form a digital twin model;

[0011] The digital twin model is optimized;

[0012] Real-time vibration signals of the turbine rotor during its service life are acquired and noise reduction is performed.

[0013] The digital twin model determines whether the turbine rotor has a fault based on the noise-reduced vibration signal and analyzes the cause of the fault; if a fault occurs, it indicates that the turbine rotor has begun to degrade, and the parameters of the degradation model are updated.

[0014] Optionally, the step of optimizing the digital twin model includes:

[0015] To acquire vibration signals and real-time service condition parameters of the turbine rotor during its service life;

[0016] The vibration signal is subjected to noise reduction processing to obtain a noise-reduced signal;

[0017] The digital twin model obtains simulated vibration signals based on the real-time service condition parameters;

[0018] The noise-reduced signal is compared with the simulated vibration signal to obtain the deviation between the noise-reduced signal and the simulated vibration signal;

[0019] If the deviation exceeds the preset range, the extended Kalman filter algorithm is used and the digital twin model is optimized according to the deviation.

[0020] Repeat the above steps until the obtained deviation is within the preset range.

[0021] Optionally, the initial feature information includes the geometric structure information used, material parameters, number of rotor layers, rotor inner and outer diameters, and historical vibration signal data.

[0022] Optionally, the steps for constructing the binary classification prediction model include:

[0023] The historical vibration signal data is used to train a binary classification prediction model using a logistic regression algorithm to obtain the binary classification prediction model.

[0024] The historical vibration signal data includes vibration signal characteristic parameters under six states of turbine rotor: no fault, imbalance, misalignment, rubbing, loose support, and shaft crack. The vibration signal characteristic parameters include waveform index, peak value index, pulse index, margin index, and kurtosis index.

[0025] Optionally, the steps of constructing the degradation model include:

[0026] The degradation model is established based on the historical failure data of the turbine rotor and using a stochastic process model; the expression of the degradation model is:

[0027]

[0028] Where f(m) is the degradation of the turbine rotor at time m; f(0) is the initial degradation of the turbine rotor; α m β is a time-varying parameter representing the degradation rate of the turbine rotor at time m; m σ represents the nonlinearity of the turbine rotor at time m; b is the diffusion coefficient of the degradation process; B(m) is the standard Brownian motion.

[0029] Optionally, after updating the parameters of the degradation model, the method further includes: predicting the remaining life of the turbine rotor;

[0030] The steps for predicting the remaining life of the turbine rotor include:

[0031] The parameters of the degradation model are updated to time m, and the state parameter update equation in the degradation model is:

[0032]

[0033] Where N(0,σ) represents random error;

[0034] The state at time m+k is calculated using the aforementioned degradation model, and the state at time m+k is:

[0035]

[0036] Find the minimum value of k that makes the inequality hold, where the minimum value of k is the remaining life of the turbine rotor predicted at time m; the inequality is:

[0037] f(m+k)≥threshold

[0038] Where threshold is the failure threshold.

[0039] Optionally, the steps for constructing the temperature field model include:

[0040] The turbine rotor is considered as an infinitely long cylinder with a central hole and a uniform initial temperature;

[0041] The temperature field of a steam turbine rotor is calculated using the finite element method; the mathematical model and boundary conditions are as follows:

[0042]

[0043] Where t is the rotor temperature in °C; τ is time in seconds; a is the thermal conductivity coefficient; λ is the thermal conductivity of the rotor material in W / (m·K); R is the radius of any point on the rotor in meters; t0 is the initial rotor temperature in °C; R1 is the inner diameter of the rotor in meters; R2 is the outer diameter of the rotor in meters; t f α is the steam temperature, in °C; α is the heat transfer coefficient, in W / (m³). 2 ·k);

[0044] The obtained partial differential equations are transformed into algebraic equations to obtain the temperature field model; and / or,

[0045] The construction of the constitutive model includes: simulating the elastoplastic strain-stress relationship of the material based on the Ramberg-Osgood model in the finite element model calculation; obtaining the creep behavior of the turbine rotor at high temperature using the time-hardened Norton-Bailey constitutive equation; and / or

[0046] The construction of the continuous damage model includes:

[0047] The Lemaitre continuous damage mechanics model was adopted, and the formula for calculating creep fatigue was used to analyze the creep fatigue damage of the turbine rotor.

[0048] Optionally, the step of constructing an executable digital twin model of the turbine rotor service process based on the initial feature information and initial service condition parameters includes:

[0049] The initial feature information and the initial service condition parameters are processed to obtain effective and relevant initial data;

[0050] The digital twin model is constructed based on the effective relevant initial data.

[0051] Optionally, the real-time vibration signal is obtained using an absolute vibration measurement method; wherein, the absolute vibration measurement method is to measure the absolute vibration of the turbine rotor relative to the sensor.

[0052] Optionally, the step of denoising the real-time vibration signal includes: using Kalman filtering and particle filtering to denoise the real-time vibration signal.

[0053] Optionally, the step of updating the parameters of the degradation model includes:

[0054] The parameters of the degradation model are updated by using an unscented particle filter algorithm combined with the real-time vibration signal.

[0055] The beneficial effects of the digital twin-based turbine rotor service process monitoring method provided by the embodiments of the present invention include:

[0056] This invention provides a method for monitoring the service process of a steam turbine rotor based on digital twins. The method includes acquiring historical failure data of the steam turbine rotor, initial characteristic information of the steam turbine rotor service process system, and initial service condition parameters; constructing an executable digital twin sub-model of the steam turbine rotor service process based on the historical failure data, initial characteristic information, and initial service condition parameters. This digital twin sub-model includes a binary prediction model, a temperature field model, a constitutive model, a continuous damage model, and a degradation model; fusing the digital twin sub-models to form a digital twin model; optimizing the digital twin model; acquiring real-time vibration signals during the steam turbine rotor service process and performing noise reduction processing; and using the noise-reduced vibration signals to determine whether a fault has occurred in the steam turbine rotor and analyze the cause of the fault. If a fault occurs, it indicates that the steam turbine rotor has begun to degrade, and the parameters of the degradation model are updated. The digital twin model, constructed from multiple sub-models, enables real-time monitoring of the turbine rotor's service life from various perspectives. Simultaneously, it monitors the turbine rotor's service life based on real-time vibration signals, providing a better reflection of the complex and ever-changing operating conditions during turbine rotor service. This method overcomes the shortcomings of traditional monitoring techniques, such as poor real-time performance, slow data processing and analysis, and excessive consumption of human and material resources. It is suitable for the complex and variable conditions of turbine rotor service life and provides a new approach to optimizing turbine rotor service monitoring technology. Attached Figure Description

[0057] The above-described features and advantages of the present invention will be better understood after reading the following detailed description of embodiments of the present disclosure in conjunction with the accompanying drawings. In the drawings, components are not necessarily drawn to scale, and components having similar related characteristics or features may have the same or similar reference numerals.

[0058] Figure 1 A flowchart of a method for monitoring the service process of a steam turbine rotor based on digital twins according to one aspect of the present invention is shown;

[0059] Figure 2 A flowchart illustrating the steps of a digital twin-based method for monitoring the service life of a steam turbine rotor according to one aspect of the present invention is shown.

[0060] Figure 3 An experimental test set classification confusion matrix diagram is shown according to one aspect of the present invention;

[0061] Figure 4 A schematic diagram of the vibration detection system according to one aspect of the present invention is shown from a first-view perspective.

[0062] Figure 5 A schematic diagram of the vibration detection system according to one aspect of the present invention is shown from a second perspective.

[0063] Figure 6 A flowchart for predicting the remaining life of a rotor according to one aspect of the present invention is shown.

[0064] Figure label:

[0065] 10-Steam turbine rotor; 11-Eddy current sensor; 12-Shaft vibration detector. Detailed Implementation

[0066] The present invention will now be described in detail with reference to the accompanying drawings and specific embodiments. It should be noted that the aspects described below with reference to the accompanying drawings and specific embodiments are merely exemplary and should not be construed as limiting the scope of protection of the present invention in any way.

[0067] In the description of this invention, it should be noted that if terms such as "upper," "lower," "inner," "outer," or "vertical" appear, the orientation or positional relationship indicated is based on the orientation or positional relationship shown in the accompanying drawings, or the orientation or positional relationship in which the product of this invention is usually placed when in use, and does not indicate or imply that the device or element referred to must have a specific orientation or be constructed and operated in a specific orientation, and therefore should not be construed as a limitation of this invention.

[0068] At the same time, it should be noted that the terms "first" and "second" are used only for distinguishing descriptions and should not be interpreted as indicating or implying relative importance.

[0069] In the description of this invention, it should also be noted that, unless otherwise explicitly specified or limited, the terms "installation," "connection," and "linking" should be interpreted broadly. For example, they can refer to a fixed connection, an integral connection, or a detachable connection; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium, or a connection within two components, etc. Those skilled in the art can understand the specific meaning of the above terms in this invention according to the specific circumstances.

[0070] The operating conditions of steam turbine rotors are complex, and the speed of manual data processing cannot keep up with the speed of data generation and acquisition, making real-time monitoring of the steam turbine rotor's service life impossible. Digital twin technology, however, establishes qualitative and quantitative monitoring of the physical entity, enabling dynamic synchronization between the physical entity and a virtual model. It features real-time performance, high fidelity, and high integration. Therefore, by establishing a digital twin model, the inventors can achieve accurate and rapid fault early warning and life prediction in complex environments, representing a crucial means to improve the system safety of steam turbine rotors during service.

[0071] Figure 1This is a flowchart of the turbine rotor service process monitoring method based on digital twin provided in this embodiment. Figure 2 This is a flowchart illustrating the steps of the digital twin-based turbine rotor service process monitoring method provided in this embodiment. Please refer to the attached diagram. Figure 1 and Figure 2 This embodiment provides a method for monitoring the service process of a steam turbine rotor based on digital twins. The method includes:

[0072] S01: Obtain historical failure data of the turbine rotor, initial characteristic information of the turbine rotor service process system, and initial service condition parameters.

[0073] The historical failure data of the turbine rotor refers to currently known historical failure data. Initial characteristic information includes, but is not limited to, the geometry, material parameters, number of rotor layers, rotor inner and outer diameters, and other component dimensional information, as well as historical vibration signal data, all within the turbine rotor's service life system. Initial service condition parameters include, but are not limited to, temperature, pressure, and rotor speed. Specifically, temperature can be measured using a temperature sensor, pressure can be measured using a pressure sensor, and rotor speed can be measured using a photoelectric sensor.

[0074] S02: Construct an executable digital twin model of the turbine rotor service process based on historical failure data, initial characteristic information, and initial service condition parameters of the turbine rotor.

[0075] Digital twin models include, but are not limited to, binary classification prediction models, temperature field models, constitutive models, continuous damage models, and degradation models. The degradation model is established based on historical failure data of the turbine rotor and utilizes a stochastic process model; its expression is as follows:

[0076]

[0077] In the formula, f(m) represents the degradation of the product at time m; f(0) represents the initial degradation of the product; α m β is a time-varying parameter representing the degradation rate of the turbine rotor at time m; m σ represents the nonlinearity of the turbine rotor at time m; b β is the diffusion coefficient of the degradation process; β(m) is the standard Brownian motion.

[0078] Meanwhile, other models in the digital twin sub-model are constructed based on the initial characteristic information and initial service condition parameters of the turbine rotor service process system. Specifically, the steps for constructing the digital twin model based on the initial characteristic information and initial service condition parameters of the turbine rotor service process system include:

[0079] S21: Process the initial characteristic information and initial service condition parameters to obtain effective and relevant initial data.

[0080] The initial feature information and initial service condition parameters are analyzed and processed using Hadoop, Spark, or Flink software to obtain effective relevant initial data. In other words, the effective relevant initial data is the data output after processing the initial feature information and initial service condition parameters using Hadoop, Spark, or Flink software.

[0081] S22: Construct a digital twin model based on the effective relevant initial data.

[0082] The digital twin models in step S22 include, but are not limited to, binary classification prediction models, temperature field models, constitutive models, and progressive damage models. By using multiple digital twin models to monitor the turbine rotor's service process from different perspectives, the monitoring becomes more accurate and comprehensive.

[0083] The steps to build a binary classification prediction model include:

[0084] Using the historical vibration signal data obtained in step S01, a binary classification prediction model is trained using a logistic regression algorithm, thereby constructing a binary classification prediction model as a digital twin model. Specifically, the historical vibration signal data includes vibration signal characteristic parameters under six states of the turbine rotor: fault-free, unbalanced, misaligned, rubbing, loose supports, and shaft cracks. The vibration signal characteristic parameters mainly refer to time-domain statistical indicators, including waveform indicators, peak values, impulse indicators, margin indicators, and kurtosis indicators. Data such as rotational speed, sampling frequency, and time-domain statistical indicators are used as training set samples, and a logistic regression algorithm is selected for training.

[0085] Logistic regression, also known as regression analysis, is a multivariate statistical method primarily used to describe the optimal mapping relationship between a set of independent variables and a response variable, where the response variable exhibits a dichotomous property. The independent variables can be binary, continuous, discrete, or a mixture of all three, while the response variable is a binary variable. The relationship between the probabilities of the independent and response variables follows an S-shaped curve. Using logistic regression requires establishing a mapping to transform the original real values ​​into 0 / 1 values, achieved using the sigmoid function. Figure 3 The diagram shows the classification confusion matrix of the experimental test set of this binary classification prediction model. The horizontal axis, predictedclass, represents the model's prediction result, and the vertical axis, trueclass, represents the true state label. The values ​​on the diagonal of the matrix represent the number of times the prediction model correctly classifies. As shown in the figure, the model's accuracy in distinguishing the rotor service state type on the test set is above 95% (i.e., the prediction is correct when the predicted structure matches the true state).

[0086] The steps for constructing a temperature field model include:

[0087] The finite element method is used to calculate the temperature field of the steam turbine. During the calculation, the turbine rotor is considered as an infinitely long cylinder with a central opening and uniform initial temperature. The mathematical model and boundary conditions are as follows:

[0088]

[0089] Where t is the rotor temperature in °C; τ is time in seconds; a is the thermal conductivity coefficient; λ is the thermal conductivity of the rotor material in W / (m·K); R is the radius of any point on the rotor in meters; t0 is the initial rotor temperature in °C; R1 is the inner diameter of the rotor in meters; R2 is the outer diameter of the rotor in meters; t f α is the steam temperature, in °C; α is the heat transfer coefficient, in W / (m³). 2 ·k).

[0090] Furthermore, the obtained partial differential equations can be transformed into algebraic equations, specifically by taking any cross-section of the turbine rotor, dividing this cross-section into n layers of equal radial thickness, and using the temperature of a point on a certain layer to represent the temperature of that layer. Therefore, when the number of layers is sufficiently large, i.e., n is sufficiently large, a sufficiently accurate temperature field can be obtained, allowing for the establishment of a more precise temperature field model.

[0091] The construction of the constitutive model includes:

[0092] In finite element model calculations, the Ramberg-Osgood model is used to simulate the elastoplastic strain-stress relationship of materials; the elastoplastic strain-stress relationship can be expressed as:

[0093]

[0094] Where ε is the total strain; ε el For elastic strain; ε pl ε represents plastic strain; σ is Von Mises stress (MPa); E is the elastic modulus (MPa); K and n′ are temperature-dependent material parameters. The specific values ​​of K and n′ are provided by the turbine manufacturer. Thus, the stress σ can be calculated using the above formula and the obtained total strain ε.

[0095] The creep behavior of the turbine rotor at high temperature is obtained using the time-hardened Norton-Bailey constitutive equation, and the creep strain can be expressed as:

[0096] ε c ×Aσ n τ m

[0097] Where, ε cLet σ be the creep strain; σ be the Von Mises stress (MPa); τ be the time (s); and A, n, and m be material parameters provided by the turbine manufacturer. The creep strain can be calculated using the stress σ obtained from the above formula expressing the elastic-plastic strain-stress relationship.

[0098] When constructing a constitutive model, the above-mentioned elastic-plastic strain-stress relationship expression and creep strain expression must be used. In this way, the constitutive model established in the above manner can obtain the creep behavior of the turbine rotor under high temperature conditions, that is, the constitutive model can output creep strain.

[0099] The construction of the continuous damage model includes:

[0100] The progressive damage model adopts the Lemaitre progressive damage mechanics model and uses the formula for calculating creep fatigue to analyze the creep fatigue damage of the rotor under high temperature and high pressure. The expression for creep fatigue can be:

[0101]

[0102]

[0103] Where D represents the total damage from creep fatigue; N represents the number of alternating strain cycles; D c For creep damage, dD c For each time increment step (d) τ The cumulative creep damage increment; D f For fatigue damage; dD f R represents the cumulative fatigue damage increment per alternating strain (dN); v The multiaxiality factor reflects the multiaxial effects; Δε is the total strain; σ H υ is the hydrostatic pressure, in MPa; σ is Poisson's ratio; eq The stress is the equivalent stress, in MPa; α1, α2, λ, γ, and Ω are material-related constants. Rainflow counting can be used to extract alternating strain.

[0104] S03: Merge the digital twin sub-models to form a digital twin model.

[0105] The Simulink visualization platform in Matlab is used to merge the runnable digital twin sub-models to form a digital twin model. However, due to the differences in the software and methods used to build the digital twin sub-models, as well as the variations in data types, it is necessary to debug the coordination and compatibility between the interfaces of each digital twin sub-model during the fusion process. This ensures that the data between the various digital twin sub-models can be mutually converted and matched, thereby constructing a multiphysics integrated simulation platform and forming a complete digital twin model.

[0106] S04: Optimize the digital twin model.

[0107] By optimizing the digital twin model, the constructed digital twin model can accurately predict the turbine rotor. Optimization of the digital twin model requires real-time acquisition of the turbine rotor's state during service, using this real-time state as the basis for optimization. However, in the process of monitoring the turbine rotor's state, it is difficult to accurately measure its operating state. In this embodiment, vibration signals are used, which can effectively reflect the operating state of the turbine rotor under complex operating conditions. Specifically, the steps for optimizing the digital twin model include:

[0108] S41: Acquire vibration signals and real-time service condition parameters of the turbine rotor during its service life.

[0109] When acquiring vibration signals during the service life of a steam turbine rotor, an absolute vibration measurement method can be used, which measures the absolute vibration of the steam turbine rotor relative to the sensor probe. Figure 4 The structure of the vibration detection system is shown from a first-view perspective. Figure 5 The structure of the vibration detection system is shown from a second-view perspective. (Refer to...) Figure 4 and Figure 5 In this embodiment, vibration signals are obtained through a vibration detection system. The vibration detection system includes a turbine rotor 10, a shaft vibration detector 12, and two eddy current sensors 11. Both eddy current sensors 11 are connected to the shaft vibration detector 12, and the two eddy current sensors 11 are distributed radially outside the turbine rotor 10 at 90° intervals along its circumference. When the turbine rotor 10 vibrates, the gap distance between the outer circumferential surface of the turbine rotor 10 and the eddy current sensors 11 changes. The shaft vibration detector 12 converts this distance change into a voltage change, thereby obtaining the vibration signal of the turbine rotor 10 during its service life.

[0110] S42: Perform noise reduction processing on the vibration signal to obtain a noise-reduced signal.

[0111] Kalman filtering and particle filtering are used to denoise the acquired vibration signals in order to reduce the noise data in the acquired vibration signals.

[0112] S43: The digital twin model obtains simulated vibration signals based on real-time service condition parameters.

[0113] The real-time service condition parameters obtained in step S41 are input into the digital twin model, and the digital twin model simulation is used to obtain simulated vibration signals.

[0114] S44: Compare the noise-reduced signal with the simulated vibration signal to obtain the deviation between the noise-reduced signal and the simulated vibration signal.

[0115] The denoised signal and the simulated vibration signal are compared, and their deviations are calculated. The comparison parameters of the vibration signal mainly refer to time-domain statistical indicators, including waveform indicators, peak value indicators, impulse indicators, margin indicators, and kurtosis indicators.

[0116] S45: If the deviation exceeds the preset range, the extended Kalman filter algorithm is used and the digital twin model is optimized according to the deviation.

[0117] In this embodiment, the preset range is ±0.5. If the deviation exceeds ±0.5, it indicates a significant error between the digital twin model and the turbine rotor service process system. This error will affect the accuracy of the prediction results, therefore, the digital twin model needs to be optimized based on the deviation value. Specifically, the extended Kalman filter algorithm is used to optimize the digital twin model based on the deviation. If the deviation is ±0.5, it indicates that the deviation is within the preset range.

[0118] Further, repeat steps S41-S45 until the deviation is within the preset range, indicating that the simulation structure output by the digital twin model is consistent with the actual detection data.

[0119] S05: Acquire real-time vibration signals of the turbine rotor during its service life and perform noise reduction processing.

[0120] It should be noted that the vibration signal obtained in step S41 is also the real-time vibration signal during the service of the turbine rotor. The real-time vibration signal can be noise-reduced in the same way as in step S05 and step S42. Therefore, step S05 can be set to be executed during the execution of step S04. At this time, steps S41 and S42 in step S04 do not need to be repeated.

[0121] S06: The digital twin model determines whether the turbine rotor has a fault based on the noise-reduced vibration signal and analyzes the cause of the fault.

[0122] like Figure 6As shown, the denoised vibration signal obtained in step S05 is input into the optimized digital twin model. The binary classification prediction model in the digital twin model judges the working state of the turbine rotor based on the denoised vibration signal and outputs a judgment result on whether the turbine rotor has a fault. That is, the result output by the binary classification prediction model is: the turbine rotor has a fault, or the turbine rotor has not a fault. If the turbine rotor has a fault, it means that the turbine rotor has begun to degrade. If the turbine rotor has not experienced a fault, it means that the turbine rotor has not yet begun to degrade. Therefore, the result output by the binary classification prediction model can also be regarded as a judgment result on whether the turbine rotor has begun to degrade.

[0123] Meanwhile, the digital twin model analyzes the causes of turbine rotor failures based on real-time monitoring information of the turbine rotor. For example, it judges the temperature difference between the upper and lower parts of the turbine rotor based on the color depth of the turbine rotor temperature field cloud map. When the upper part of the turbine rotor is hotter than the lower part due to uneven heat transfer, the existence of the temperature difference may cause the turbine rotor to bend, becoming one of the causes of failures during service, and timely warnings are issued.

[0124] If the turbine rotor begins to degrade, the parameters of the degradation model are updated. Specifically, starting from the early degradation point detected, the Unscented Particle Filter (UPF) algorithm, combined with real-time measured vibration signals, is used to update the degradation model parameters in the second step. The UPF algorithm incorporates the latest observation information into the particle weight update process, thus enabling parameter updates and state estimation of the degradation model based on rotor operation data at different times. Specifically, the update process is as follows: The UPF algorithm samples the initial distribution to generate an original particle set and calculates the weight coefficient corresponding to each particle. Then, the UKF algorithm (Unscented Kalman Filter) is used to calculate the state of each particle, thereby constructing the importance sampling distribution of the particle filter, and updating the weights of each particle based on real-time measured vibration signals. Next, the particles are rearranged according to their weights, removing particles with low weights and retaining particles with high weights, and resampling is performed to increase the number of effective particles. Finally, state estimation is performed using the updated particles and their weights, and iterative updates are performed.

[0125] S07: Predict the remaining life of the turbine rotor.

[0126] If the turbine rotor begins to degrade, its remaining life is predicted. The steps for predicting the remaining life of the turbine rotor include updating the parameters of the degradation model in the digital twin model to time m, estimating the state at time m+k using the degradation model, and obtaining the minimum value of k that makes the inequality hold. The minimum value of k is the predicted remaining life of the turbine rotor at time m. Specifically, the state parameter update equation of the degradation model is:

[0127]

[0128] Here, N(0,σ) represents the random error. Specifically, the random error is Gaussian white noise in the state equation.

[0129] The state at time m+k estimated by the degradation model is:

[0130]

[0131] The inequality is:

[0132] f(m+k)≥threshold

[0133] Where threshold is the failure threshold.

[0134] like Figure 6 As shown, after the degradation model predicts the remaining life of the turbine rotor at time m, if the turbine rotor has not failed, then let m = m+1, update the parameters of the degradation model to time m+1, and predict the remaining life of the turbine rotor at time m+1, until the turbine rotor fails.

[0135] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in the present invention should be included within the scope of protection of the present invention.

Claims

1. A method for monitoring the service process of a steam turbine rotor based on digital twins, characterized in that, include: Acquire historical failure data of the turbine rotor, initial characteristic information of the turbine rotor service process system, and initial service condition parameters; An executable digital twin model of the turbine rotor service process is constructed based on the historical failure data of the turbine rotor, the initial characteristic information, and the initial service condition parameters; wherein, the digital twin model includes a binary classification prediction model, a temperature field model, a constitutive model, a continuous damage model, and a degradation model; The digital twin sub-models are then merged to form a digital twin model; The digital twin model is optimized; Real-time vibration signals of the turbine rotor during its service life are acquired and noise reduction is performed. The digital twin model determines whether the turbine rotor has a fault based on the noise-reduced vibration signal and analyzes the cause of the fault; if a fault occurs, it indicates that the turbine rotor has begun to degrade, and the parameters of the degradation model are updated. The steps for constructing the degradation model include: The degradation model is established based on the historical failure data of the turbine rotor and using a stochastic process model; the expression of the degradation model is: in, For the turbine rotor in m The amount of degradation at any given time; This represents the initial degradation of the turbine rotor. These are time-varying parameters, representing the turbine rotor's state during operation. m The rate of degradation at any given moment; For the turbine rotor in m Nonlinearity at time; The diffusion coefficient is the coefficient for the degradation process. This is standard Brownian motion; After updating the parameters of the degradation model, the method further includes: predicting the remaining life of the turbine rotor; The steps for predicting the remaining life of the turbine rotor include: The parameters of the degradation model are updated to m At time t, the state parameter update equation in the degradation model is: in, This is due to random error; Calculate using the degradation model m+k The state at that moment, the state described m+k The state at that moment is: Obtain the inequality that makes the inequality true. The minimum value, the The minimum value is m The remaining lifespan of the turbine rotor as predicted at any given time; the inequality is: threshold Where threshold is the failure threshold.

2. The method for monitoring the service process of a steam turbine rotor based on digital twins according to claim 1, characterized in that, The steps for optimizing the digital twin model include: To acquire vibration signals and real-time service condition parameters of the turbine rotor during its service life; The vibration signal is subjected to noise reduction processing to obtain a noise-reduced signal; The digital twin model obtains simulated vibration signals based on the real-time service condition parameters; The noise-reduced signal is compared with the simulated vibration signal to obtain the deviation between the noise-reduced signal and the simulated vibration signal; If the deviation exceeds the preset range, the extended Kalman filter algorithm is used and the digital twin model is optimized according to the deviation. Repeat the above steps until the obtained deviation is within the preset range.

3. The method for monitoring the service process of a steam turbine rotor based on digital twins according to claim 1, characterized in that, The initial feature information includes the geometric structure information used, material parameters, number of rotor layers, rotor inner and outer diameters, and historical vibration signal data.

4. The method for monitoring the service process of a steam turbine rotor based on digital twins according to claim 3, characterized in that, The steps for constructing the binary classification prediction model include: The historical vibration signal data is used to train a binary classification prediction model using a logistic regression algorithm to obtain the binary classification prediction model. The historical vibration signal data includes vibration signal characteristic parameters under six states of turbine rotor: no fault, imbalance, misalignment, rubbing, loose support, and shaft crack. The vibration signal characteristic parameters include waveform index, peak value index, pulse index, margin index, and kurtosis index.

5. The method for monitoring the service process of a steam turbine rotor based on digital twins according to claim 1, characterized in that, The steps for constructing the temperature field model include: The turbine rotor is considered as an infinitely long cylinder with a central hole and a uniform initial temperature; The temperature field of a steam turbine rotor is calculated using the finite element method; the mathematical model and boundary conditions are as follows: in, Rotor temperature, in °C; Time, in seconds; Thermal conductivity; is the thermal conductivity of the rotor material, in W / (m·K); Let be the radius of any point on the rotor, in meters. The initial temperature of the rotor is expressed in °C. The rotor's inner diameter is in meters (m). The rotor's outer diameter is in meters (m). Steam temperature, in °C; The heat transfer coefficient is expressed in W / (m³). 2 ·k); The obtained partial differential equations are transformed into algebraic equations to obtain the temperature field model; and / or, The construction of the constitutive model includes: simulating the elastic-plastic strain-stress relationship of the material based on the Ramberg-Osgood model in the finite element model calculation; The creep behavior of the turbine rotor at high temperatures was obtained using the time-hardened Norton-Bailey constitutive equations; and / or The construction of the continuous damage model includes: The Lemaitre continuous damage mechanics model was adopted, and the formula for calculating creep fatigue was used to analyze the creep fatigue damage of the turbine rotor.

6. The method for monitoring the service process of a steam turbine rotor based on digital twins according to claim 1, characterized in that, The steps for constructing an executable digital twin model of the turbine rotor service process based on the initial feature information and initial service condition parameters include: The initial feature information and the initial service condition parameters are processed to obtain effective and relevant initial data; The digital twin model is constructed based on the effective relevant initial data.

7. The method for monitoring the service process of a steam turbine rotor based on digital twins according to claim 1, characterized in that, The real-time vibration signal is obtained using an absolute vibration measurement method; wherein, the absolute vibration measurement method is to measure the absolute vibration of the turbine rotor relative to the sensor.

8. The method for monitoring the service process of a steam turbine rotor based on digital twins according to claim 1, characterized in that, The steps for denoising the real-time vibration signal include: using Kalman filtering and particle filtering to denoise the real-time vibration signal.

9. The method for monitoring the service process of a steam turbine rotor based on digital twins according to claim 1, characterized in that, The steps for updating the parameters of the degradation model include: The parameters of the degradation model are updated by using an unscented particle filter algorithm combined with the real-time vibration signal.

Images