A multi-source random response collaborative gas turbine performance robust design method
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- SHANGHAI JIAOTONG UNIV
- Filing Date
- 2026-03-23
- Publication Date
- 2026-06-26
Smart Images
Figure CN122286979A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of gas turbine performance design, and more specifically to a robust gas turbine performance design method based on multi-source stochastic response coordination. Background Technology
[0002] As a core power component of equipment, the stability and reliability of gas turbine performance are crucial for operational safety. However, during the design, manufacturing, and use of gas turbines, they face various random challenges that can lead to a decline in actual performance or even safety accidents.
[0003] Traditional gas turbine overall performance design methods have significant drawbacks: In the design phase, simulation models simplify physical processes and employ empirical functions, introducing random biases that lead to a disconnect between designed and actual performance. In the manufacturing phase, factors such as component manufacturing processes and precision introduce random fluctuations in performance parameters, which are addressed solely with safety margins without actively suppressing performance propagation. In the operational phase, environmental disturbances such as intake parameters and combustion heat release are not jointly optimized with static design parameters, easily causing sudden performance changes. Existing methods often treat the randomness of a single aspect in isolation, lacking collaborative optimization mechanisms, resulting in conservative design results, poor adaptability to operating conditions, and high safety risks. Summary of the Invention
[0004] The purpose of this invention is to provide a robust design method for gas turbine performance based on multi-source stochastic response coordination, in order to solve the problems of poor adaptability to operating conditions and high safety risks in existing technologies.
[0005] The technical solution of the present invention to solve the above-mentioned technical problems is as follows: A robust design method for gas turbine performance based on multi-source stochastic response coordination includes the following steps: S1: Analysis of the impact of stochasticity in the design model: A simulation model of the overall performance of the gas turbine is established based on the aerodynamic thermodynamic mechanism equation. Random sampling is performed on the simplified modeling parameters of the simulation model to generate multiple sets of parameter samples. Under multi-condition control input, the core performance parameter response data of the gas turbine components and the whole machine are obtained through simulation calculation using the aerodynamic thermodynamic model. Statistical analysis is performed on the distribution characteristics of the core performance parameters to construct the probability density function of the core performance parameters. S2: Analysis of the impact of manufacturing randomness: Based on the performance test data of the core components of the gas turbine, a random distribution model of the component performance parameters is established; the fluctuation range and deviation boundary of the component performance parameters are determined, and the influence law of the randomness of the component performance parameters on the core performance of the gas turbine is analyzed; based on the aerodynamic thermodynamic model, different combinations of component performance parameters are generated through an active learning sampling strategy to obtain the distribution characteristics of the core performance parameters of the whole machine, and the probability density function of the core performance parameters is constructed. S3: Analysis of the impact of stochasticity during use: Establish a high-dimensional nonlinear stochastic dynamic model of the gas path system that includes random disturbances in the operating environment parameters and random disturbances in the combustion chamber energy release; solve the generalized probability density evolution equation corresponding to the model to analyze the impact of stochasticity during use on the gas path state parameters and the core performance of the whole machine, and obtain the response distribution of the gas path state parameters and the distribution of the core performance of the whole machine; combine the parameter distributions obtained in steps S1 and S2 to establish the constraints on the gas path state parameters and performance parameters, construct a robust optimization model for the overall performance of the gas turbine, solve for the robust distribution of the core performance parameters of the whole machine, and construct the probability density function of the core performance parameters; Step S4: Robust Optimization of Overall Performance Based on Multi-Source Stochasticity: Based on a nonlinear aero-thermodynamic model, the risk levels of the impact of design model randomness, manufacturing randomness, and usage randomness on the overall performance of the gas turbine are determined at different task stages, and corresponding weight coefficients are assigned according to the risk levels; the probability density functions of the core performance parameters obtained in steps S1, S2, and S3 are fused to generate a comprehensive probability density function characterizing multi-source randomness; with the distribution characteristics of component thermodynamic state parameters and core performance parameters as constraints, and with the goal of minimizing the dispersion and optimizing the expected value of the comprehensive probability density function, an optimization objective function is constructed; the optimization objective function is solved to obtain the optimal combination of overall performance parameters, thus completing the robust design of the overall performance of the gas turbine.
[0006] According to the method described in claim 1, the simplified modeling parameters in step S1 are the intake duct total pressure recovery coefficient and the combustion efficiency empirical formula correction coefficient, and the multi-condition control input parameters include fuel quantity, adjustable guide vane angle, and tail nozzle throat area.
[0007] Further solution: The random sampling in step S1 adopts the Monte Carlo sampling method, and the total number of parameter samples meets the confidence requirements for the statistical characteristic analysis of random response.
[0008] Further solution: The core components mentioned in step S2 include a fan, a compressor, and a turbine. The random distribution model for the performance parameters of these components adopts a truncated normal distribution, the expression of which is: In the formula, θ represents the component performance parameter, and μ represents the mean value of the component performance parameter. The variance of component performance parameters, , Let Φ(⋅) be the upper and lower limits of the cutoff interval, and let Φ(⋅) be the cumulative distribution function of the standard normal distribution.
[0009] Further solution: the upper and lower limits of the cutoff interval , The mean μ and variance are determined by the allowable deviation range of the component performance parameters. It was obtained from the statistical analysis of component performance test data.
[0010] Further solution: The expression for the generalized probability density evolution equation in step S3 is: In the formula, p(x,t) is the probability density function of the gas path state parameter x at time t, and v(x,t) is the drift term of the state parameter; the equation is solved using the latent representation method, which models the randomness of the process as a stochastic process, and its expression is: In the formula, θ(t) is the random perturbation parameter at time t. Let ε be the mean of the random perturbation parameters, ε be the random intensity coefficient, and ξ(t) be the standard Brownian motion.
[0011] Further solution: The weighting coefficients mentioned in step S4 satisfy... in To design the randomness weighting coefficients of the model, To create random weighting coefficients, The randomness weighting coefficient is used during the process; the takeoff phase satisfies... > = The cruise phase meets > = .
[0012] Further options: The core performance parameters mentioned in steps S1, S2, and S3 include surge margin, turbine inlet temperature, engine speed, thrust, and fuel consumption rate.
[0013] Further solution: The expression for the comprehensive probability density function mentioned in step S4 is: ; The expression for the optimization objective function is: In the formula, For the comprehensive probability density function, , , Let J be the probability density function under three types of randomness, and J be the objective function. The variance of the overall probability density function. This represents the expected value of the overall probability density function. , The target weight coefficients are satisfied. + =1.
[0014] Further solution: The objective function described in step S4 is solved using a multi-objective Bayesian optimization algorithm. The optimization results must meet the verification criteria of reducing the standard deviation of thrust fluctuation by 30% to 40% and reducing the probability of turbine inlet temperature exceeding the limit to below 0.1%.
[0015] The present invention has the following beneficial effects: This method, by co-quantifying and jointly representing three types of stochasticity—design model, manufacturing, and usage process—changes the traditional approach of treating single stochasticities in isolation, effectively overcoming the problems of conservative design results and significant performance losses. Based on a mechanism of dynamically allocating weight coefficients according to task stages, the optimization objective can adaptively focus on key sources of stochasticity in different task stages, achieving a synergistic optimal balance between overall gas turbine performance and robustness.
[0016] Meanwhile, this method, based on a general probability distribution representation and collaborative optimization framework, possesses good engineering flexibility and model scalability, and can adapt to the design requirements of different types of gas turbines. After applying this method, the sensitivity of gas turbines to random factors in actual use is significantly reduced, performance fluctuations are effectively suppressed, the stable operating boundary is expanded, and their operational safety, reliability, and adaptability to operating conditions are systematically improved. Attached Figure Description
[0017] Figure 1 Flowchart of a robust design method for gas turbine performance based on multi-source stochastic response coordination.
[0018] Figure 2 Schematic diagram of the overall performance robust design effect of gas turbine. Detailed Implementation
[0019] To make the objectives, technical solutions, and advantages of this invention clearer, the following detailed description of the multi-source stochastic response collaborative robust design method for gas turbine performance, in conjunction with the accompanying drawings and specific embodiments, is provided. This embodiment uses a high-bypass ratio aero-gas turbine high-pressure compressor as an application object, fully covering the quantitative analysis and collaborative optimization of three types of stochasticity in design, manufacturing, and use, verifying the effectiveness and feasibility of this method.
[0020] Analysis of the impact of randomness on the design model Based on aerodynamic thermodynamic equations, a performance simulation model of a gas turbine was built. This model includes thermodynamic calculation modules for core components such as the intake, combustion chamber, compressor, turbine, and exhaust nozzle, and strictly adheres to the laws of conservation of mass, energy, and momentum.
[0021] For the key parameters in the simulation model that need to be simplified, the intake total pressure recovery coefficient (simplification deviation range ±1.2%~±1.5%) and the combustion chamber efficiency empirical formula correction coefficient (fluctuation range ±2.5%~±3%) are selected as random sampling objects. The Monte Carlo random sampling method is used to generate no less than 500 sets of thermodynamic parameter samples. The sample size is determined through statistical testing to meet the confidence requirements of the statistical characteristic analysis of random response.
[0022] A multi-condition control input matrix is set up, with input parameters including fuel supply, adjustable guide vane installation angle, and nozzle throat area, covering typical gas turbine operating conditions such as takeoff, cruise, climb, and landing. The sampled parameters are input one by one into the overall engine performance simulation model, and response data for core performance parameters such as fan surge margin, compressor outlet temperature and pressure, turbine inlet temperature, overall engine speed, thrust, and fuel consumption rate are obtained through iterative calculations.
[0023] Normality tests and distribution characteristic analyses were performed on the response data. The probability density functions of key overall performance parameters such as thrust and fuel consumption rate were constructed using the variational autoencoder density estimation method. A quantitative mapping relationship between the simplified parameters of the design model and the overall performance indicators was established, and the impact of the randomness of the design model on performance was quantitatively characterized.
[0024] Analysis of the impact of manufacturing randomness Test data on the flow and efficiency characteristics of core components such as fans, compressors, and turbines were collected. Based on manufacturing process specifications and measured tolerance data, the fluctuation boundaries of component performance parameters were determined. Among them, the standard deviation of the compressor blade mounting angle tolerance was 0.05°, and the standard deviation of the turbine clearance machining error was 0.1mm.
[0025] A stochastic distribution model for component performance parameters is established using a truncated normal distribution. The mean of the distribution is determined by the statistical average of the component performance test data, the variance is calculated from the measured tolerance data, and the upper and lower limits of the truncation interval are defined based on the allowable deviation range of the component design. This model quantitatively describes the stochastic fluctuations in component efficiency and flow rate caused by manufacturing randomness.
[0026] Based on the gas turbine aero-thermodynamic model, an active learning sampling strategy is introduced. Using the uncertainty of performance response as the sampling criterion, priority is given to selecting component parameter combinations that significantly affect the overall engine performance for simulation, generating 5000 sets of random component performance parameter combinations. These parameter combinations are then input into the simulation model to obtain the response dataset of the core performance parameters of the entire engine, and the distribution characteristics of thrust and fuel consumption rate are statistically analyzed.
[0027] A thrust reduction exceeding 5% was set as the performance degradation threshold. High-risk tolerance combinations leading to this degradation were identified, and key influencing factors of manufacturing randomness were determined. Based on statistically obtained performance distribution data, a joint probability density function of thrust and fuel consumption rate was constructed to quantitatively analyze the impact of manufacturing randomness on overall engine performance.
[0028] Analysis of the randomness of the process A high-dimensional nonlinear stochastic dynamic model of the gas path system was established, incorporating environmental disturbances and combustion chamber energy release disturbances. The model's state variables include key gas path parameters such as intake temperature, intake pressure, combustion chamber heat release, turbine inlet temperature, and compressor outlet pressure. The random disturbance range of intake temperature is ±15K, and the noise intensity coefficient of combustion chamber heat release fluctuation is set to 2%–5%.
[0029] For the generalized probability density evolution equation corresponding to this model, the latent representation method is used for solution. First, the high-dimensional gas path state equation is reduced to a low-dimensional latent space. Then, probability density evolution calculations are performed in the latent space, and finally, the results are mapped back to the original state space. During the solution process, the randomness of intake air temperature, pressure, and combustion chamber heat release is modeled as a stochastic process, with the following expression: ; The random intensity coefficient ε is selected based on the percentage of inlet temperature and pressure and combustion heat release under given conditions, and ξ(t) is the standard Brownian motion.
[0030] By solving the equations, the probability density distribution of key parameters such as turbine inlet temperature and compressor outlet pressure under extreme operating conditions is predicted. Combining the component performance parameter boundaries obtained from the stochastic analysis of the design model and the surge margin threshold obtained from the stochastic analysis of manufacturing, the constraints of the robust optimization model are constructed. It is determined that a turbine inlet temperature exceeding 1730K is a dangerous operating condition, and a surge margin not less than 85% of the design value is a safe boundary. This completes the quantification of the stochasticity of the usage process and the setting of the constraints.
[0031] Robust optimization of overall performance of multi-source stochastic collaboration Based on a nonlinear aero-thermodynamic model, Failure Mode and Effects Analysis (FMEA) was used to assess the risk level of three types of randomness on the overall aircraft performance during different mission phases. During takeoff, environmental disturbances are prone to causing sudden changes in intake parameters, resulting in the highest risk level. During cruise, the long-term cumulative effect of component manufacturing tolerances becomes prominent, increasing the risk level.
[0032] Weighting coefficients are dynamically allocated based on risk level: the weighting coefficient for the takeoff phase is set to... =0.2、 =0.2、 =0.6; the weighting coefficient for the cruise phase is adjusted to =0.5、 =0.2、 =0.3, and satisfies .
[0033] The probability density functions of thrust and fuel consumption rate under three types of stochasticity—design model, manufacturing, and usage—are weighted and superimposed to generate a comprehensive probability density function representing multi-source stochasticity. Its expression is as follows: A multi-objective optimization function is constructed with the objectives of minimizing the dispersion of the comprehensive probability density function (to improve performance stability) and optimizing the expected value (to ensure core performance of thrust and fuel consumption rate). The multi-objective Bayesian optimization algorithm is used to solve for the optimal combination of overall performance parameters, and the number of optimization iterations is set to 200. The surrogate model is a Gaussian process regression model.
[0034] The optimized performance parameter combination was substituted into the whole machine simulation model for verification. The results showed that the thrust fluctuation standard deviation was reduced by 30% to 40%, the fuel consumption rate distribution kurtosis was reduced by 5%, the fan surge margin safety boundary was expanded by 10% under extreme conditions, and the turbine inlet temperature over-limit probability was reduced from 0.5% to 0.1%, thus completing the robust design of the overall performance of the gas turbine.
[0035] This embodiment demonstrates through a complete process verification that the method can effectively synergistically quantify the effects of three types of randomness, achieve the optimal balance between overall performance and robustness, and significantly improve the operational safety and reliability of gas turbines.
[0036] Summary of working principles: The core working principle of this method is to achieve a synergistic improvement in the overall performance and disturbance resistance of gas turbines through independent quantification of multi-source randomness, dynamic weight allocation, probabilistic feature fusion, and dual-objective robust optimization. First, to address model simplification biases in the design phase, random sampling combined with aerodynamic-thermodynamic simulation is used to quantify their impact on performance parameters and construct a probability density function. For component tolerances in the manufacturing phase, a stochastic model is established based on a truncated normal distribution, and active learning sampling focuses on combinations of high-impact parameters to clarify the performance propagation path of manufacturing randomness. For environmental disturbances in the usage phase, a high-dimensional nonlinear stochastic dynamics model and latent representation method are used to analyze the probabilistic evolution of gas path state parameters and delineate performance safety boundaries. Second, based on failure mode impact analysis, the risk levels of three types of randomness in different mission phases are determined, and weight coefficients are dynamically allocated to adaptively focus the optimization objective on key disturbance sources. Finally, by weighted fusion of the performance probability density functions of the three types of randomness, an optimization model is constructed with the goal of minimizing the dispersion of the comprehensive probability density function and optimizing the expected value. The optimal combination of performance parameters is solved by using a multi-objective Bayesian optimization algorithm to offset the superposition effect of multi-source randomness at the system level, and finally achieves performance stability and robustness improvement of the gas turbine throughout the entire mission cycle.
[0037] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.
Claims
1. A robust design method for gas turbine performance based on multi-source stochastic response coordination, characterized in that, Includes the following steps: S1: Analysis of the impact of stochasticity in the design model: A simulation model of the overall performance of the gas turbine is established based on the aerodynamic thermodynamic mechanism equation. Random sampling is performed on the simplified modeling parameters of the simulation model to generate multiple sets of parameter samples. Under multiple operating condition control inputs, the core performance parameter response data of the gas turbine components and the whole machine are obtained through simulation calculation using the aerodynamic thermodynamic model. Statistical analysis is performed on the distribution characteristics of the core performance parameters to construct the probability density function of the core performance parameters. S2: Analysis of the impact of manufacturing randomness: Based on the performance test data of the core components of the gas turbine, a random distribution model of the component performance parameters is established; S3: Analysis of the impact of randomness in the usage process: Establish a high-dimensional nonlinear stochastic dynamic model of the gas path system that includes random disturbances in usage environment parameters and random disturbances in combustion chamber energy release; Step S4: Robust Optimization of Overall Performance Based on Multi-Source Stochasticity: Based on a nonlinear aero-thermodynamic model, the risk levels of the impact of design model randomness, manufacturing randomness, and usage randomness on the overall performance of the gas turbine are determined at different task stages, and corresponding weight coefficients are assigned according to the risk levels; the probability density functions of the core performance parameters obtained in steps S1, S2, and S3 are fused to generate a comprehensive probability density function characterizing multi-source randomness; with the distribution characteristics of component thermodynamic state parameters and core performance parameters as constraints, and with the goal of minimizing the dispersion and optimizing the expected value of the comprehensive probability density function, an optimization objective function is constructed; the optimization objective function is solved to obtain the optimal combination of overall performance parameters, thus completing the robust design of the overall performance of the gas turbine.
2. The robust design method for gas turbine performance based on multi-source stochastic response coordination according to claim 1, characterized in that, The simplified modeling parameters mentioned in step S1 are the intake duct total pressure recovery coefficient and the combustion efficiency empirical formula correction coefficient. The multi-condition control input parameters include fuel quantity, adjustable guide vane angle, and tail nozzle throat area.
3. The robust design method for gas turbine performance based on multi-source stochastic response coordination according to claim 1, characterized in that, The random sampling in step S1 adopts the Monte Carlo sampling method, and the total number of parameter samples meets the confidence requirements for the statistical characteristic analysis of random response.
4. The robust design method for gas turbine performance based on multi-source stochastic response coordination according to claim 1, characterized in that, Step S2 also includes determining the fluctuation range and deviation boundary of the component performance parameters, analyzing the influence of the randomness of the component performance parameters on the core performance of the gas turbine; based on the aerodynamic thermodynamic model, generating different combinations of component performance parameters through an active learning sampling strategy, obtaining the distribution characteristics of the core performance parameters of the whole machine, and constructing the probability density function of the core performance parameters.
5. The robust design method for gas turbine performance based on multi-source stochastic response coordination according to claim 4, characterized in that, Step S3 also includes solving the generalized probability density evolution equation corresponding to the model, analyzing the influence of the randomness of the usage process on the gas path state parameters and the core performance of the whole machine, obtaining the response distribution of the gas path state parameters and the distribution of the core performance of the whole machine; combining the parameter distributions obtained in steps S1 and S2, establishing the constraints on the gas path state parameters and performance parameters, constructing a robust optimization model for the overall performance of the gas turbine, solving for the robust distribution of the core performance parameters of the whole machine, and constructing the probability density function of the core performance parameters.
6. The robust design method for gas turbine performance based on multi-source stochastic response coordination according to claim 5, characterized in that, The weighting coefficients described in step S4 satisfy in To design the randomness weighting coefficients of the model, To create random weighting coefficients, The randomness weighting coefficient is used during the process; the takeoff phase satisfies... > = The cruise phase meets > = .
7. The robust design method for gas turbine performance based on multi-source stochastic response coordination according to claim 5, characterized in that, The core performance parameters mentioned in steps S1, S2, and S3 include surge margin, turbine inlet temperature, engine speed, thrust, and fuel consumption rate.
8. The robust design method for gas turbine performance based on multi-source stochastic response coordination according to claim 1, characterized in that, The expression for the comprehensive probability density function mentioned in step S4 is: ； The expression for the optimization objective function is: In the formula, For the comprehensive probability density function, , , Let J be the probability density function under three types of randomness, and J be the objective function. The variance of the overall probability density function. This represents the expected value of the overall probability density function. , The target weight coefficients are satisfied. + =1.