A Multiphysics Field Reconstruction Method for Degraded Flow Passage Components of Gas Turbines

By calibrating a three-dimensional multiphysics field and a degradation primitive disturbance field library for single-unit health benchmarks, the problem of multiphysics field reconstruction of gas turbine flow passage components was solved, enabling rapid and accurate degradation monitoring and management.

CN123069955APending Publication Date: 2026-11-10XIAN THERMAL POWER RES INST CO LTD
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202611217600.2
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2026-08-12
Publication Date
2026-11-10

AI Technical Summary

Technical Problem

Existing digital twin technology for gas turbines cannot effectively resolve the contradictions of no measurement points inside flow passage components, numerous degenerate combinations, high real-time requirements, and strict accuracy requirements, resulting in multi-physics reconstruction distortion and insufficient real-time performance.

Method used

By calibrating a single-unit-specific health benchmark three-dimensional multiphysics field based on actual gas turbine measurement data, degrading it into standardized primitives, constructing a degradation primitive perturbation field library, using finite measurement points to invert and solve the degradation weight coefficients, and generating a real multiphysics field through a nonlinear compensation model.

Benefits of technology

It enables rapid multiphysics reconstruction without exhaustively enumerating degradation scenarios, adapts to real-time health monitoring and full lifecycle management of gas turbines, reduces computational load and model deployment cycle, and improves reconstruction accuracy.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN123069955A_ABST
    Figure CN123069955A_ABST
Patent Text Reader

Abstract

This invention provides a multiphysics field reconstruction method for degradation of gas turbine flow passage components, belonging to the field of gas turbine digital twin technology. It can at least partially solve the problems in existing technologies such as the inability to exhaustively enumerate degradation modes, the lack of measurement points inside the flow passage components leading to imperceptible implicit degradation, and the distortion of multiphysics field reconstruction. This invention includes: calibrating a three-dimensional multiphysics field based on measured data to establish a health baseline; decomposing degradation into standardized degradation primitives and constructing a pure perturbation field library; inverting degradation weight coefficients based on real-time data from finite measurement points; weighting and superimposing the primitive perturbation fields according to the weight coefficients to obtain a coarsely corrected multiphysics field; and superimposing nonlinear correction terms for strongly coupled degradation scenarios to output a true multiphysics field. This invention does not require exhaustively enumerating degradation scenarios; it achieves real-time multiphysics field reconstruction by covering degradation modes through primitive combination.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] This invention relates to the field of digital twin and degradation diagnosis technology for gas turbines, specifically to a multiphysics reconstruction method for degradation of gas turbine flow passage components. Background Technology

[0002] As a core power source in power generation, aerospace, and ship propulsion, gas turbines' flow passage components operate under high-temperature and high-pressure environments for extended periods, enduring multiple effects such as airflow impact, thermal cycling fatigue, and media corrosion. This makes them susceptible to various types of coupled, latent degradation. Typical degradation types of gas turbine flow passage components include fouling, corrosion, increased blade tip clearance, blade damage, and gas seal leakage. These degradations are characterized by their strong concealment and multiple combinations: firstly, there are no direct measurement sensors inside the flow passage components; existing measurement points can only obtain macroscopic parameters at the overall turbine level, making it impossible to pinpoint specific degradation locations, types, and severity; secondly, the degradation combination space grows exponentially. Taking a certain F-class gas turbine as an example, considering only single or multiple stage clearance anomalies, localized or global fouling, randomly located blade tip damage, and gas seal leakage in the 18 compressor stages and 4 turbine stages, the possible combinations are numerous. to kind.

[0003] Existing digital twin technologies for gas turbines generally employ offline, fully three-dimensional computational fluid dynamics (CFD) and computational structural dynamics (CSD) simulations in an exhaustive degradation mode. This involves building a database of thousands to tens of thousands of three-dimensional simulation samples by pre-setting typical degradation scenarios, and then using online interpolation based on measured macroscopic parameters to reconstruct multiphysics fields. However, this approach has three main drawbacks: First, the degradation modes are not fully covered. 3D simulations rely on preset degradation scenarios, but the level, type, and severity of degradation in actual service are random. Unpreset or omitted scenarios will lead to distortion in multiphysics reconstruction.

[0004] Second, real-time performance is severely lacking. A single full 3D CFD / CSD simulation takes 8 to 72 hours. To exhaustively enumerate all combinations of degradation parameters, it is necessary to construct... to The sample database requires several years for initial modeling, and a single update during online interpolation can take several hours, which cannot meet the needs of real-time updates on a minute-by-minute basis.

[0005] Third, baseline deviation and error amplification. Traditional twin models are built based on ideal design geometry, without considering manufacturing errors at the time of the actual machine's delivery and break-in degradation in the early stages of service. This results in an inherent deviation between the health status baseline and the actual condition of the actual machine. After the degradation is superimposed, the deviation is further amplified, making the multiphysics reconstruction accuracy unable to meet the requirements of engineering diagnosis.

[0006] In summary, existing technologies are unable to resolve the core contradictions of the lack of measurement points inside flow passage components, numerous degenerate combinations, high real-time requirements, and stringent accuracy requirements.

[0007] Therefore, this field urgently needs a multi-physics reconstruction technology for gas turbine flow passage components that does not require exhaustive enumeration of degradation scenarios, adapts to a limited number of macroscopic measurement points, and balances computational real-time performance with reconstruction accuracy. This technology would overcome the technical bottlenecks of "no internal measurement point sensing blind spots" and "multi-physics distortion caused by the explosion of degradation mode combinations" in flow passage components. Providing reliable full-domain physical field data support for gas turbine degradation location, life assessment, and operation and maintenance decisions is a technical problem that urgently needs to be solved in this field. Summary of the Invention

[0008] The present invention aims to solve at least one of the technical problems existing in the prior art, and provides a multiphysics field reconstruction method for the degradation of gas turbine flow passage components.

[0009] To achieve the above objectives, the present invention provides a multiphysics reconstruction method for the degradation of gas turbine flow passage components, comprising: Step S1: Calibrate the three-dimensional multiphysics field of a single-unit dedicated health benchmark based on the measured data of the gas turbine; Step S2: Decompose the degradation of the gas turbine flow passage component into standardized degradation primitives. For each degradation primitive, obtain the difference between the degradation field and the three-dimensional multiphysics field of the health benchmark through offline simulation as a pure perturbation field, and construct a degradation primitive perturbation field library. Step S3: Based on the real-time operating data of the limited measurement points of the gas turbine, with the goal of minimizing the residual between the calculated and measured values ​​of the three-dimensional multiphysics field of the health benchmark, the degradation weight coefficients corresponding to each degradation primitive are solved by inversion. Step S4: Using the health baseline three-dimensional multiphysics field as a basis, the corresponding pure perturbation fields are weighted and superimposed according to the degradation weight coefficients to obtain the coarsely corrected multiphysics field; and Step S5: When the degradation weight coefficient meets the preset coupling threshold condition, based on the degradation weight coefficient and the operating condition parameters, a nonlinear correction term is generated by a nonlinear compensation model used to compensate for the linear superposition error under the strong coupling degradation scenario. The nonlinear correction term is superimposed on the coarse correction multiphysics field, and the reconstructed real multiphysics field is output.

[0010] Furthermore, in step S1, the measured data includes factory test data, overhaul disassembly measured data, and historical operating data under healthy steady-state conditions; the calibration adopts an inverse parameter calibration method with the goal of minimizing the weighted residual between the measured value and the simulation calculation value, and the parameters to be calibrated include blade assembly clearance, actual blade profile deviation, surface roughness, and sealing leakage coefficient; the healthy benchmark three-dimensional multiphysics field covers pressure field, temperature field, velocity field, and stress field.

[0011] Furthermore, in step S2, the standardized degradation unit is defined by three dimensions: degradation type, degradation region, and degradation gradient. The degradation type includes roughness degradation, tip clearance degradation, blade local defect degradation, and assembly deformation degradation. The degradation region corresponds to each stage of the flow passage components of the compressor and turbine. The degradation gradient is set according to the severity of degradation.

[0012] Furthermore, 3 to 10 gradient samples are constructed for each of the aforementioned degradation types; the calculation formula for the pure perturbation field is as follows: ; in, For the first Class degeneration in the first Pure perturbation field at each gradient To introduce this degraded offline simulation field, The health benchmark is a three-dimensional multiphysics field under the same load and boundary conditions.

[0013] Furthermore, in step S3, the inversion solution employs a weighted least squares objective function with a regularization term and introduces physical boundary constraints. The expression for the objective function is: ; ; in, Let the objective function value be the degradation weight coefficient. For the first Measured values ​​at each measuring point The first three-dimensional multiphysics field corresponding to the health benchmark Calculated values ​​for each measuring point For the first The degenerate primitive perturbation field is mapped to the first... The parameter change at each measuring point For the first The aforementioned degradation weight coefficients For the first The weighting coefficient of each measurement point The regularization coefficient is . For weights and upper limits, The total number of measurement points. This represents the total number of degenerate primitives.

[0014] Furthermore, in step S3, the finite measurement points include temperature, pressure, flow rate, vibration, heat consumption, and output parameters; the inversion solution also includes an abnormal data removal step: when the measured value of a single measurement point deviates from the corresponding value of the three-dimensional multiphysics field of the health benchmark by more than a preset standard deviation multiple, the weight of the measurement point is reset to zero and removed from the inversion calculation.

[0015] Further, in step S5, the preset coupling threshold condition is: any of the degradation weight coefficients is greater than or equal to the first threshold, or there are two or more degradation weight coefficients greater than or equal to the second threshold; the first threshold is greater than the second threshold.

[0016] Furthermore, the inputs to the nonlinear compensation model include the degradation weighting coefficients, unit load, speed, inlet total temperature, and the measurement point residual vector of the coarse-corrected multiphysics field; the output of the nonlinear compensation model is a global nonlinear corrected field; the true multiphysics field satisfies: ; in, For the real multiphysics field, The health benchmark is a three-dimensional multiphysics field. For the first The aforementioned degradation weight coefficients For the first The aforementioned pure perturbation field For the nonlinear correction term, This represents the total number of degenerate primitives.

[0017] Furthermore, the gas turbine flow path component degradation multiphysics reconstruction method also includes a benchmark recalibration step: when it is detected that the gas turbine has completed water washing or overhaul, the steady-state operation data after maintenance is collected, and step S1 is re-executed to update the health benchmark three-dimensional multiphysics field.

[0018] Furthermore, the offline simulation covers at least three typical load conditions, and the pure disturbance field of the non-nominal load conditions is obtained through multi-condition linear interpolation; and in step S4, a uniform weight superposition is used for global uniform degradation, and a partitioned differentiated weight superposition is used for local distributed degradation.

[0019] The beneficial effects of this invention are as follows: This invention decomposes complex composite degradation into standardized independent primitives. Only a few dozen sets of gradient primitive perturbation fields need to be constructed, which can cover thousands of actual degradation modes through linear combination. This completely replaces the traditional full 3D simulation approach of exhaustively listing each scene. The simulation computation volume in the offline modeling stage can be reduced by more than 70%, and the model deployment cycle is greatly shortened.

[0020] This invention relies on existing conventional measurement points such as temperature, pressure, flow rate and power of the unit, and combines a weighted inversion algorithm constrained by aerodynamic thermodynamics to quantitatively identify implicit degradation without direct measurement points, realizing non-intrusive online monitoring of the health status of flow passage components without the need for additional sensors.

[0021] This invention constructs a two-layer accuracy assurance mechanism of "single-machine benchmark calibration plus nonlinear coupling compensation" to eliminate manufacturing and assembly deviations between the design model and the actual machine. It also corrects vortex interference and flow field coupling errors in strongly degraded scenarios through small-sample machine learning, and the relative error of reconstructed multi-physics fields can meet the accuracy requirements of engineering applications.

[0022] This invention adopts a reduced-order computational architecture of "primitive pre-computation plus online weight inversion plus field superposition", which concentrates the computational load in the offline stage. The online stage only needs to perform weight solving and field superposition operations. The time consumption of a single multiphysics reconstruction can meet the real-time update requirements at the minute level.

[0023] This invention incorporates a dynamic recalibration mechanism for the health reference field, which can automatically trigger reference updates after unit water washing or major overhaul, adapting to the full life cycle health management of gas turbines. Attached Figure Description

[0024] Figure 1 This is a flowchart illustrating the overall steps of the multiphysics field reconstruction method for the degradation of gas turbine flow passage components according to an embodiment of the present invention. Figure 2 This is a schematic diagram illustrating the optimization solution principle of the degradation weight coefficient inversion in an embodiment of the present invention; Figure 3 This is a schematic diagram of the field quantity relationship in the multiphysics superposition reconstruction according to an embodiment of the present invention; Figure 4 This is a schematic diagram of the data flow of each module in the multiphysics reconstruction system according to an embodiment of the present invention. Detailed Implementation

[0025] To make the objectives, technical solutions, and beneficial effects of this application clearer, the following detailed description, in conjunction with the accompanying drawings and specific embodiments, further illustrates this application. It should be understood that the specific embodiments described in this specification are merely for explaining this application and are not intended to limit it.

[0026] The overall technical solution of the multiphysics field reconstruction method for the degradation of gas turbine flow passage components of the present invention will be described below.

[0027] See Figure 1 The multiphysics reconstruction method for the degradation of gas turbine flow passage components of the present invention includes steps S1 to S5 and an optional reference recalibration step. Each step will be described in detail below.

[0028] Step S1 involves calibrating a single-unit-specific health baseline three-dimensional multiphysics field based on measured data from the gas turbine. This health baseline three-dimensional multiphysics field refers to the solution result of a three-dimensional multiphysics calculation model constructed using a reverse parameter calibration method based on the actual geometric state and physical property parameters of a single gas turbine under healthy conditions. It encompasses pressure, temperature, velocity, and stress fields. "Single-unit-specific" means that each gas turbine requires its own independent health baseline field due to differences in manufacturing tolerances, assembly deviations, and initial service run-in degradation, rather than using a uniform design parameter model. "Calibration" refers to the process of reverse-calibrating the geometric and physical property parameters in the model to match the measured values ​​of the actual turbine.

[0029] The measured data includes factory test data, overhaul disassembly measured data, and historical operating data under healthy steady-state conditions. Factory test data provides the performance baseline of the unit in its initial state; overhaul disassembly measured data provides direct measurements of geometric parameters such as blade assembly clearance, actual blade profile deviation, surface roughness, and seal leakage; historical operating data under healthy steady-state conditions provides operating parameters such as pressure ratio, efficiency, exhaust temperature, shaft power, and fuel flow rate of the unit under different loads.

[0030] The calibration employs an inverse parameter calibration method aimed at minimizing the weighted residual between measured and simulated values. The parameter set to be calibrated includes the compressor and turbine tip clearances, equivalent surface roughness of the moving and stationary blades, the sealing leakage coefficient, and inlet boundary conditions. The resulting reference field must meet preset accuracy criteria.

[0031] The output of step S1 is a single-machine dedicated health benchmark three-dimensional multiphysics field with accuracy verification. This output serves as the reference basis for the pure perturbation field calculation in step S2, the calculation basis for weight inversion in step S3, and the basis for field superposition in step S4.

[0032] Step S2 involves decomposing the degradation of the gas turbine flow path components into standardized degradation primitives and constructing a degradation primitive perturbation field library. The standardized degradation primitive refers to the smallest independent perturbation unit defined by three dimensions: degradation type, degradation region, and degradation gradient. Its core characteristics are singularity and superposition—each primitive describes only one type of degradation in a specific region and at a specific severity level, resulting in physical field changes. Different primitives are independent of each other and can be linearly superimposed.

[0033] In different implementations, the degradation element can cover, but is not limited to, the following degradation types: roughness degradation, tip clearance degradation, localized blade defect degradation, and assembly deformation degradation. Roughness degradation covers blade fouling and surface corrosion; tip clearance degradation covers single-stage abnormal wear and multi-stage uniform wear; localized blade defect degradation covers leading edge chipping, trailing edge defects, and localized thinning of the blade body; assembly deformation degradation covers slight bending of the main shaft and abnormal gas seal leakage. The commonality among these degradation types is that their corresponding physical field perturbation quantities can all be obtained through three-dimensional simulation, and these perturbation quantities can be linearly superimposed with the perturbation quantities of other degradation types.

[0034] The pure disturbance field is the difference between the degradation simulation field and the healthy baseline field under the same boundary conditions and load. For each degradation type, 3 to 10 gradient samples are constructed to cover the entire degradation range. Offline simulation covers multiple typical load conditions; the pure disturbance field for non-nominal load conditions is obtained through multi-condition linear interpolation.

[0035] The output of step S2 is the degenerate primitive perturbation field library and the parameter changes of each primitive perturbation field mapped to the measurement point. The former serves as the material for field superposition in step S4, and the latter serves as the calculation parameters for weight inversion in step S3.

[0036] Step S3 involves inverting and solving for the degradation weight coefficients based on real-time operating data from a limited number of gas turbine measuring points. The degradation weight coefficients are normalized parameter vectors characterizing the severity of each degradation element, with values ​​ranging from 0 to 1; a larger value indicates a higher severity of the corresponding degradation. "Inverting and solving" refers to using the calculated values ​​from the health baseline measuring points calibrated in step S1 as a reference, attributing the difference between the real-time operating data and the calculated values ​​to the effects of each degradation element, and quantitatively decoupling the contribution ratio of each degradation element through an optimization algorithm.

[0037] The limited measurement points include operational data collected by the unit's existing monitoring system, such as temperature, pressure, flow rate, vibration, heat consumption, and output parameters. The inversion solution aims to minimize the residuals and can be solved using aerodynamic thermodynamic constraints. In a preferred embodiment, the inversion uses a weighted least squares objective function with a regularization term, while introducing a degenerate prior knowledge base to constrain the solution boundary. The inversion process also includes an outlier removal step to avoid sensor malfunctions and data jumps interfering with the inversion results.

[0038] The solution algorithm can be selected from gradient descent algorithms, evolutionary algorithms, and intelligent search algorithms according to engineering requirements. Gradient descent algorithms include batch gradient descent, stochastic gradient descent, and Adam optimization algorithm; evolutionary algorithms include genetic algorithms and differential evolution algorithms; intelligent search algorithms include particle swarm optimization and simulated annealing algorithms. The commonality of these algorithms is that they all possess residual optimization capabilities, all use the objective function as the optimization object, and are constrained by the laws of aerodynamics and thermodynamics and degenerate prior boundaries.

[0039] The output of step S3 is the degeneracy weight coefficient vector corresponding to each degeneracy primitive. This output serves as the multiplication coefficient for weighted superposition in step S4 and as the input for coupling threshold determination and nonlinear compensation model in step S5.

[0040] Step S4 involves weighted superposition of the primitive perturbation fields using the healthy baseline three-dimensional multiphysics field as a basis to obtain a coarsely corrected multiphysics field. This step utilizes the degradation weight coefficients obtained in step S3 to linearly superimpose the pure perturbation field generated in step S2 onto the healthy baseline field calibrated in step S1, according to the weight magnitude, achieving preliminary correction of global degradation. For uniform global degradation, a uniform weight superposition is used; for locally distributed degradation, a differentiated weight superposition is used.

[0041] The formula for calculating the uniform weight across the entire domain is as follows: ; The formula for calculating the partitioned differential weighting is as follows: ; in, For the first Class degeneration in the first The partition weight of the region This represents the elementary perturbation field in the corresponding region. This represents the total number of regions.

[0042] The output of step S4 is a coarsely corrected multiphysics field, which serves as the basis for the superposition of nonlinear correction terms in step S5.

[0043] Step S5 involves superimposing nonlinear correction terms to output a true multiphysics field for strongly coupled degradation scenarios. When the degradation weight coefficients meet the preset coupling threshold condition, it indicates that linear superposition cannot fully reflect the flow field vortex interference and strong distortion coupling effects under strongly degraded scenarios, and nonlinear compensation correction needs to be introduced.

[0044] The nonlinear compensation model is a machine learning model used to compensate for linear superposition errors under strongly coupled degradation scenarios. This model takes the strongly coupled degradation weights, unit operating parameters, and the residuals of the coarse correction field as inputs, and outputs a global nonlinear compensation disturbance field. In different implementations, the nonlinear compensation model can be implemented using, but is not limited to, the following methods: gradient boosting tree model, support vector regression model, shallow neural network model, and Gaussian process regression model. The commonality of these models is that they can all learn the nonlinear mapping relationship between linear superposition errors and degradation weights and operating parameters under strongly coupled degradation conditions based on small sample training data, and all use the residuals of the coarse correction field as training labels. In a preferred embodiment, a gradient boosting tree model is used as the nonlinear compensation model.

[0045] The preset coupling threshold condition is: any degradation weight coefficient is greater than or equal to a first threshold, or there are two or more degradation weight coefficients greater than or equal to a second threshold, wherein the first threshold is greater than the second threshold. In a preferred embodiment, the first threshold is set to 0.7 and the second threshold is set to 0.4.

[0046] When the degenerate weighting coefficients do not meet the preset coupling threshold condition, the nonlinear correction term is zero, and the true multiphysics field is the coarse-corrected multiphysics field. The final reconstructed true multiphysics field satisfies: ; in, Three-dimensional multiphysics for health benchmarks For the first Each degradation weight coefficient, For the first A pure perturbation field This is a nonlinear correction term.

[0047] The invention also includes a benchmark recalibration step. When it is detected that the gas turbine has completed water washing, overhaul, or major structural repair, steady-state operation data after maintenance is collected, and step S1 is re-executed to update the three-dimensional multiphysics field of the health benchmark, so as to avoid evaluation deviation caused by benchmark drift after performance recovery.

[0048] The above is a description of the overall technical solution of the present invention. The following detailed explanation of each step is provided through specific embodiments.

[0049] Example 1 This embodiment takes the compressor section of a 390MW F-class gas turbine as the application object, and verifies the basic reconstruction capability of the method for a combined degradation scenario of compressor fouling and increased clearance.

[0050] Step S1: Calibrate the three-dimensional multiphysics field of the single-unit exclusive health benchmark based on the measured data of the gas turbine.

[0051] This embodiment collects steady-state operating data and disassembly / maintenance records after the unit's first major overhaul as calibration inputs. The steady-state operating data covers the entire range above 50% load, and the operating parameters include pressure ratio, overall unit efficiency, exhaust temperature, shaft power, and fuel flow rate. The disassembly / maintenance records provide measured values ​​of tip clearance for each stage of the blades, equivalent sand grain roughness on the blade surface, and measured values ​​of sealing clearance.

[0052] The set of parameters to be calibrated includes: tip clearances of the compressor stages 1 through 18, equivalent surface roughness of the moving and stationary blades at each stage, sealing leakage coefficients, and inlet boundary conditions. Calibration employs an inverse parameter calibration method, aiming to minimize the weighted residual between measured and simulated values. The objective function for calibration is: ; in, This is the set of geometric and physical property parameters to be calibrated. For the first The measured values ​​of each measuring point or performance parameter. The first obtained by mapping the health reference field Calculated values ​​for each parameter. For the first The weighting coefficients of each parameter.

[0053] After calibration, the compressor's overall efficiency baseline was 89.0%, the pressure ratio calculation error was 0.72%, the efficiency calculation error was 1.12%, and the exhaust temperature calculation error was 5.3℃, all meeting the calibration accuracy requirements. The acceptance criteria for calibration accuracy are: overall efficiency calculation error not exceeding 1.5%, pressure ratio calculation error not exceeding 1.0%, and exhaust temperature calculation error not exceeding 8℃.

[0054] Step S2: Decompose the degradation of the flow passage component into standardized degradation primitives and construct a library of degradation primitive perturbation fields.

[0055] In this embodiment, the compressor is divided into high-pressure, medium-pressure, and low-pressure stages. The high-pressure stages are stages 1 to 7, the medium-pressure stages are stages 8 to 14, and the low-pressure stages are stages 15 to 18. Two types of primitives are constructed: roughness degradation and blade tip clearance degradation. Each type has 3 degradation gradients, resulting in a total of 18 primitive samples.

[0056] The degradation primitives adopt a three-level coding rule: "Type T - Region Z - Gradient G". T is the degradation type code, where R represents roughness degradation, S represents tip clearance degradation, D represents blade defect degradation, and A represents assembly deformation degradation. Z is the region code, with C1 to C18 corresponding to compressor stages 1 to 18, and T1 to T4 corresponding to turbine stages 1 to 4. G is the gradient code, with G1 to G5 corresponding to five degradation gradients from light to heavy.

[0057] Gradient quantization definition of roughness degradation primitive: using equivalent sand grain roughness As a quantitative indicator, mild dirt buildup corresponds to... Moderate buildup corresponds to Severe fouling and corrosion correspond to The corresponding decreases in single-level isentropic efficiency are approximately 0.5%, 1.5%, and 3.5%, respectively.

[0058] Gradient quantization definition of the blade tip clearance degradation primitive: using the operating clearance calibrated with a healthy reference field as the reference value. Among them, the reference clearance of each stage of the compressor The gap increase gradient is set to +10%, +20%, and +30%, corresponding to the absolute gap increments of [missing values]. , and .

[0059] After completing the full 3D CFD calculations in offline simulation, the pure perturbation field is extracted by subtracting the degenerate field from the reference field, forming a standardized degenerate primitive perturbation field library. The formula for calculating the pure perturbation field is: ; in, For the first Class degeneration in the first Pure perturbation field at each gradient To introduce this degraded offline simulation field, A three-dimensional multiphysics field serves as the health benchmark under the same load and boundary conditions.

[0060] Step S3: Solve for the degradation weight coefficients based on the real-time operating data of a limited number of gas turbine measuring points.

[0061] In this embodiment, the unit is selected to operate under stable conditions at 75% load. The actual measured operating data are input as follows: compressor efficiency 88.2%, exhaust temperature 8°C higher than baseline, and pressure ratio 0.8% lower than baseline.

[0062] Measurement point selection and data preprocessing: Six core operating measurement points were selected as inputs for the inversion. The compressor inlet total temperature was sampled at 1Hz, and outliers were removed using a 30-second moving average. The compressor outlet total pressure was sampled at 1Hz, and processed with temperature compensation and a 30-second moving average. The turbine exhaust average temperature was sampled at 1Hz, corrected for thermocouple array mean. The fuel mass flow rate was sampled at 1Hz, converted to standard operating conditions after temperature and pressure compensation. The unit output shaft power was sampled at 1Hz, smoothed using a 1-minute moving average. The overall heat rate was sampled once per minute, corrected to ISO standard operating conditions.

[0063] The inversion objective function is in the form of weighted least squares with L2 regularization: ; ; in, The first corresponding to the health baseline field Calculated values ​​for each measuring point For the first The elementary perturbation field is mapped to the first... The parameter change at each measuring point The L2 regularization coefficient is... For weights and upper limits.

[0064] In this embodiment, the key parameter values ​​are: the total number of measuring points and performance parameters. Regularization coefficient Weight and upper limit Iteration stopping threshold .

[0065] Abnormal data removal rules: When the measured value of a single measurement point deviates from the corresponding value of the healthy baseline. When the standard deviation is exceeded, the point is identified as an abnormal measurement point, its weight is automatically reset to zero, and it is removed from the inversion calculation.

[0066] The initial residuals for the inversion are: exhaust temperature 8.0℃, pressure ratio 0.8%, and efficiency 0.8%. After iterative convergence optimization using the gradient descent algorithm, the degradation weight coefficients are obtained: roughness element weights. Corresponding to moderate fouling; blade tip clearance element weight This corresponds to a slight increase in clearance. After inversion, the residuals converge to: exhaust temperature 0.6℃, pressure ratio 0.08%, and efficiency 0.07%.

[0067] Step S4: Using the healthy baseline three-dimensional multiphysics field as a basis, the elementary perturbation fields are weighted and superimposed to obtain the coarsely corrected multiphysics field.

[0068] This embodiment uses a global linear superposition method to complete the coarse correction: ; in, , .

[0069] Step S5: Determine whether the preset coupling threshold condition is met.

[0070] In this embodiment, the two degradation weight coefficients are 0.6 and 0.3, respectively. The preset coupling threshold condition is: a single degradation weight is greater than or equal to 0.7, or two or more degradation weights are greater than or equal to 0.4. Only one degradation weight is greater than 0.4, which does not meet the condition that two or more degradation weights are greater than or equal to 0.4. Therefore, the degradation degree has not reached the strong coupling threshold, and the nonlinear correction term... The true multiphysics field is the coarse-corrected multiphysics field.

[0071] The final reconstructed three-dimensional pressure, temperature, and velocity fields, including those resulting from fouling and increased gaps, were obtained. Compared with the measured results from subsequent overhaul disassembly, the relative error of the field reconstruction was 1.2%. The formula for calculating the relative error of the field reconstruction is as follows: ; in, The real multiphysics field reconstructed by this method, This serves as a benchmark verification field obtained from overhaul disassembly measurements or borehole inspections. The degradation type, location, and degree are consistent with the overhaul results, verifying the accuracy of the method under typical degradation scenarios.

[0072] Example 2 This embodiment uses a 300MW gas turbine section as the application object. It verifies the effect of the nonlinear compensation mechanism on improving reconstruction accuracy in a strongly coupled degradation scenario where severe fouling, increased clearance, and blade defects coexist in the turbine section. The operating conditions are 100% load, ambient temperature 25℃, and inlet pressure 101.3kPa.

[0073] Step S1: The calibration process for the health baseline three-dimensional multiphysics field in this embodiment is consistent with the calibration process in Embodiment 1. Steady-state operating data and disassembly and maintenance records after the unit's major overhaul are collected as calibration inputs. The individual unit's dedicated health baseline three-dimensional multiphysics field is obtained through reverse parameter calibration.

[0074] Step S2: The degradation primitive perturbation field library in this embodiment is expanded based on Embodiment 1, and includes three types of primitives: roughness degradation, blade tip clearance degradation, and blade local defect degradation.

[0075] The localized defect degradation unit covers three typical types of blade damage: leading edge chipping, trailing edge chipping, and localized thinning of the blade body. Geometric quantification of defects is defined as follows: leading edge chipping area accounts for 2% to 5% of the blade's projected area, with a chordal length of 3mm to 8mm; trailing edge chipping length is 2mm to 6mm, and the chipping depth does not exceed 15% of the maximum blade thickness; localized thinning of the blade body reduces thickness by 5% to 20%, with the thinned area accounting for 3% to 10% of the blade body area. Local aerodynamic distortion field, vortex shedding field, and temperature hotspot field are obtained through offline 3D simulation. The pure perturbation field of each unit is extracted according to the pure perturbation field calculation formula.

[0076] Step S3: Input measurement points include: compressor outlet pressure 1.58MPa, turbine exhaust average temperature 586℃, fuel flow rate 12.3kg / s, and total output power 302MW.

[0077] The structure and parameter settings of the inversion objective function are the same as in Example 1. Through residual optimization inversion under mechanistic constraints, the degradation weight coefficients of each primitive are obtained: roughness primitive weights. This corresponds to severe fouling; blade tip clearance element weights This corresponds to a moderate increase in gap size; the weight of the leaf defect element. This corresponds to a slight tail edge defect.

[0078] Step S4: Linearly superimpose the pure perturbation fields of the three types of primitives onto the healthy reference field according to the degradation weight coefficient to obtain the coarsely corrected multiphysics field.

[0079] Step S5: Determine whether the preset coupling threshold condition is met. In this embodiment... This satisfies the condition that the single degeneracy weight is greater than or equal to the first threshold (0.7). Meanwhile... and The condition that two or more degradation weights are greater than or equal to the second threshold (0.4) is met. Therefore, nonlinear compensation correction is automatically triggered.

[0080] In this embodiment, the nonlinear compensation model adopts a gradient boosting tree model. The model input includes a degenerate weight coefficient vector. Unit load Rotation speed , total inlet temperature and coarse correction field measurement point residual vector The model output is a global nonlinear corrected field. The mapping relationship is as follows: ; in The mapping function represents the nonlinear compensation model.

[0081] The final true multiphysics field is the superposition of the coarse correction field and the nonlinear correction field: ; Implementation effect comparison: When only linear superposition coarse correction is used, the relative error of field reconstruction is 5.8%, which cannot meet the engineering accuracy requirements; after introducing nonlinear compensation to correct the interference and coupling distortion of the flow field vortex system, the relative error of the field is reduced to 1.5%. Subsequent borehole inspection confirmed that there was surface fouling and slight trailing edge defects on the turbine's second-stage blade, and the measured tip clearance was about 22% larger than the baseline, which matched the degradation type, location and degree obtained by inversion.

[0082] Example 3 This embodiment verifies the baseline self-updating mechanism after unit operation and maintenance, ensuring the baseline consistency of degradation assessment throughout the entire life cycle. The implementation scenario is the baseline recalibration of a certain F-class gas turbine after the compressor online water washing is completed.

[0083] The execution process of steps S1 to S5 is the same as that of Embodiment 1 and Embodiment 2. The following focuses on the execution process of the benchmark recalibration step.

[0084] Triggering Decision: After the water washing is completed, the system automatically collects steady-state operating data for 24 consecutive hours, covering the entire load range from 50% to 100%, with a sampling period of 1 second. Calculations show that the overall efficiency of the machine increases by 2.1% after water washing compared to before washing, and the average exhaust temperature decreases by 11℃. The performance recovery exceeds the preset benchmark update threshold, meeting the recalibration triggering conditions.

[0085] Recalibration execution: Using the measured data after water washing as input, re-execute step S1 to calibrate the geometric and physical property parameters and generate an updated health baseline three-dimensional multiphysics field. After recalibration, the pressure ratio calculation error of the new reference field was 0.65%, the efficiency calculation error was 1.08%, and the exhaust temperature calculation error was 4.9℃, all of which were below the calibration qualification threshold.

[0086] Effect Verification: The updated reference field, used as the new reference for subsequent degradation inversion, effectively eliminated the reference drift caused by the performance recovery after water washing. Hole inspection confirmed that the blade surface was clean and free of obvious scale after water washing, and the performance parameters were restored to near-new machine levels, consistent with the reference characteristics after recalibration, thus verifying the rationality and accuracy of the dynamic reference update mechanism.

[0087] The present invention also discloses a multi-physics field reconstruction system for the degradation of gas turbine flow passage components, which is used to realize the engineering implementation of the above method and includes five functional modules.

[0088] Health reference field calibration module: The inputs are factory test data, maintenance measurement data and health steady-state operation data, and the output is a single-machine dedicated health reference three-dimensional multiphysics field. The data is output to the field superposition and reconstruction module.

[0089] Typical primitive library construction module: The input is geometric parameters, physical property parameters and working boundary conditions, and the output is the pure disturbance field and the measurement point parameter mapping value of various primitives. The data is output to the weight coefficient inversion module and the field superposition reconstruction module respectively.

[0090] The weight coefficient inversion module takes real-time measurement data, healthy baseline field measurement values, and primitive measurement point mapping values ​​as inputs, and outputs the weight coefficients of each degenerate primitive. The data is then output to the field overlay and reconstruction module. This module can employ gradient descent algorithms, evolutionary algorithms, or intelligent search algorithms.

[0091] Field overlay and reconstruction module: The inputs are the healthy reference field, the elementary disturbance field, the weighting coefficients and the operating condition parameters. It performs linear overlay and nonlinear compensation calculations and outputs the reconstructed real three-dimensional multiphysics field. The data is output to the visualization output module.

[0092] Visualization output module: The input is reconstructed multiphysics data, and the output is visualization results such as cloud maps of temperature and pressure fields, performance degradation curves, and degradation assessment reports.

[0093] In summary, the embodiments of the present invention have at least the following technical effects: This invention decomposes complex composite degradation into standardized independent primitives, requiring only the construction of dozens of gradient primitive perturbation fields to cover thousands of actual degradation modes, reducing the simulation computation amount in the offline modeling stage by more than 70%.

[0094] This invention relies on existing measuring points in the unit and combines a weighted inversion algorithm with mechanistic constraints to achieve quantitative sensing of implicit degradation such as blade fouling, increased blade tip clearance, and localized blade defects, without the need for additional sensors.

[0095] The dual-layer correction mechanism constructed in this invention enables the relative errors of the pressure, temperature, and efficiency indices of the reconstructed multiphysics field to meet the accuracy requirements of engineering applications.

[0096] This invention adopts a reduced-order computing architecture, and the time required for a single full three-dimensional multiphysics reconstruction can meet the real-time update requirements at the minute level, making it suitable for gas turbine digital twin scenarios.

[0097] This invention incorporates a dynamic recalibration mechanism for the health reference field, adapting to the health management needs of gas turbines throughout their entire life cycle.

[0098] It is understood that the above embodiments are merely exemplary implementations used to illustrate the principles of the present invention, and the present invention is not limited thereto. For those skilled in the art, various modifications and improvements can be made without departing from the spirit and essence of the present invention, and these modifications and improvements are also considered to be within the scope of protection of the present invention.

Claims

1. A method for reconstructing the multiphysics field of degraded flow passage components in a gas turbine, characterized in that, include: Step S1: Calibrate the three-dimensional multiphysics field of a single-unit dedicated health benchmark based on the measured data of the gas turbine; Step S2: Decompose the degradation of the gas turbine flow passage component into standardized degradation primitives. For each degradation primitive, obtain the difference between the degradation field and the three-dimensional multiphysics field of the health benchmark through offline simulation as a pure perturbation field, and construct a degradation primitive perturbation field library. Step S3: Based on the real-time operating data of the limited measurement points of the gas turbine, with the goal of minimizing the residual between the calculated and measured values ​​of the three-dimensional multiphysics field of the health benchmark, the degradation weight coefficients corresponding to each degradation primitive are solved by inversion. Step S4: Using the health benchmark three-dimensional multiphysics field as a basis, the corresponding pure perturbation fields are weighted and superimposed according to the degradation weight coefficient to obtain the coarsely corrected multiphysics field; as well as Step S5: When the degradation weight coefficient meets the preset coupling threshold condition, based on the degradation weight coefficient and the operating condition parameters, a nonlinear correction term is generated by a nonlinear compensation model used to compensate for the linear superposition error under the strong coupling degradation scenario. The nonlinear correction term is superimposed on the coarse correction multiphysics field, and the reconstructed real multiphysics field is output.

2. The multiphysics reconstruction method for degraded gas turbine flow passage components according to claim 1, characterized in that, In step S1, the measured data includes factory test data, overhaul disassembly measured data, and historical operating data under healthy steady-state conditions; the calibration adopts an inverse parameter calibration method with the goal of minimizing the weighted residual between the measured value and the simulation calculation value. The parameters to be calibrated include blade assembly clearance, actual blade profile deviation, surface roughness, and sealing leakage coefficient; the healthy benchmark three-dimensional multiphysics field covers pressure field, temperature field, velocity field, and stress field.

3. The multiphysics reconstruction method for degraded gas turbine flow passage components according to claim 1, characterized in that, In step S2, the standardized degradation unit is defined by three dimensions: degradation type, degradation region, and degradation gradient. The degradation type includes roughness degradation, tip clearance degradation, blade local defect degradation, and assembly deformation degradation. The degradation region corresponds to each stage of the flow passage components of the compressor and turbine. The degradation gradient is set according to the severity of degradation.

4. The multiphysics reconstruction method for degraded gas turbine flow passage components according to claim 3, characterized in that, For each of the aforementioned degradation types, 3 to 10 gradient samples are constructed; the formula for calculating the pure perturbation field is as follows: ; in, For the first Class degeneration in the first Pure perturbation field at each gradient To introduce this degraded offline simulation field, The health benchmark is a three-dimensional multiphysics field under the same load and boundary conditions.

5. The multiphysics reconstruction method for degraded gas turbine flow passage components according to claim 1, characterized in that, In step S3, the inversion solution employs a weighted least squares objective function with a regularization term and introduces physical boundary constraints. The expression for the objective function is: ; ; in, Let the objective function value be the degradation weight coefficient. For the first Measured values ​​at each measuring point The first three-dimensional multiphysics field corresponding to the health benchmark Calculated values ​​for each measuring point For the first The degenerate primitive perturbation field is mapped to the first... The parameter change at each measuring point For the first The aforementioned degradation weight coefficients For the first The weighting coefficient of each measurement point The regularization coefficient is . For weights and upper limits, The total number of measurement points. This represents the total number of degenerate primitives.

6. The multiphysics reconstruction method for degraded gas turbine flow passage components according to claim 5, characterized in that, In step S3, the finite measurement points include temperature, pressure, flow rate, vibration, heat consumption, and output parameters; the inversion solution also includes an abnormal data removal step: when the measured value of a single measurement point deviates from the corresponding value of the three-dimensional multiphysics field of the health benchmark by more than a preset standard deviation multiple, the weight of the measurement point is reset to zero and removed from the inversion calculation.

7. The multiphysics reconstruction method for degraded flow passage components of a gas turbine according to claim 1, characterized in that, In step S5, the preset coupling threshold condition is: any of the degradation weight coefficients is greater than or equal to the first threshold, or there are two or more degradation weight coefficients greater than or equal to the second threshold; the first threshold is greater than the second threshold.

8. The multiphysics reconstruction method for degraded gas turbine flow passage components according to claim 7, characterized in that, In step S5, the inputs to the nonlinear compensation model include the degradation weight coefficient, unit load, speed, inlet total temperature, and the measurement point residual vector of the coarse-corrected multiphysics field. The output of the nonlinear compensation model is a global nonlinear corrected field. The real multiphysics field satisfies: ; in, For the real multiphysics field, The health benchmark is a three-dimensional multiphysics field. For the first The aforementioned degradation weight coefficients For the first The aforementioned pure perturbation field For the nonlinear correction term, This represents the total number of degenerate primitives.

9. The multiphysics reconstruction method for degraded gas turbine flow passage components according to any one of claims 1 to 8, characterized in that, It also includes a benchmark recalibration step: when it is detected that the gas turbine has completed water washing or overhaul, the steady-state operation data after maintenance is collected, and step S1 is re-executed to update the health benchmark three-dimensional multiphysics field.

10. The multiphysics reconstruction method for degraded gas turbine flow passage components according to any one of claims 1 to 8, characterized in that, In step S2, the offline simulation covers at least three typical load conditions, and the pure disturbance field of the non-nominal load conditions is obtained through multi-condition linear interpolation. Furthermore, in step S4, uniform weight superposition is used for uniform degradation across the entire domain, while differentiated weight superposition is used for localized degradation.