A method for predicting the probabilistic low-cycle fatigue life of a turbine blade

Abstract

The application discloses a kind of probability low-cycle fatigue life prediction methods of turbine blade, it is related to blade life prediction technical field.Acquire the multiple groups of sample working condition data of turbine blade to be predicted, and the low-cycle fatigue life of turbine blade to be predicted under each group of sample working condition data is calculated respectively by fatigue life prediction model based on critical plane method;Through multiple groups of sample working condition data and corresponding low-cycle fatigue life, neural network is trained, and low-cycle fatigue life prediction model is obtained;Through Monte Carlo method, the multiple groups of candidate working condition data of turbine blade to be predicted are generated, and the life prediction result corresponding to each group of candidate working condition data is predicted by low-cycle fatigue life prediction model;According to multiple life prediction results, the probability low-cycle fatigue life prediction result of turbine blade to be predicted is determined.The method can improve the accuracy of probability low-cycle fatigue life prediction while improving the efficiency of blade probability low-cycle fatigue life prediction.

Description

Technical Field

[0001] This invention relates to the field of blade life prediction technology, and in particular to a method for predicting the probabilistic low-cycle fatigue life of turbine blades. Background Technology

[0002] As a core component of turbine machinery, blades operate under extreme environments such as high temperature, high pressure, and high speed for extended periods. During frequent start-ups and shutdowns of the unit, blades are subjected to severe alternating loads, leading to significant elasto-plastic cyclic stress and strain in locally weak areas of their structure. This stress and strain accumulate continuously and may eventually cause fatigue failure. Accurate prediction of the probabilistic low-cycle fatigue life of such components helps in developing reasonable operation and maintenance strategies, reducing downtime for maintenance, and improving system reliability. This is of great significance for ensuring the operational safety of turbine machinery and the economic efficiency of unit operation.

[0003] In existing technologies, to ensure the accuracy of probabilistic low-cycle fatigue life prediction, the Smith-Watson-Topper (SWT) model is usually used to predict the life of blades. However, when predicting probabilistic low-cycle fatigue life based on the SWT model, the life prediction results need to be solved through a large number of Monte Carlo simulations or random sampling methods. Each random sampling requires a new finite element simulation to calculate the life through the SWT model. This cyclical process of "sampling-simulation-prediction" results in an excessively long overall prediction process, and the efficiency problem is even more prominent in high-reliability analyses with large sample size requirements. Summary of the Invention

[0004] Therefore, it is necessary to provide a method for predicting the probabilistic low-cycle fatigue life of turbine blades to address the aforementioned technical problems. This method can improve both the accuracy and efficiency of predicting the probabilistic low-cycle fatigue life of turbine blades.

[0005] The present invention adopts the following technical solution:

[0006] This invention provides a method for predicting the probabilistic low-cycle fatigue life of turbine blades, comprising:

[0007] Multiple sets of sample operating condition data of the turbine blade to be predicted were obtained, and the low-cycle fatigue life of the turbine blade to be predicted under each set of sample operating condition data was calculated by the fatigue life prediction model based on the critical plane method. The fatigue life prediction model based on the critical plane method is determined based on the interaction between normal strain and shear strain on the critical plane of the turbine blade and the relationship between mean stress and fatigue life.

[0008] The low-cycle fatigue life prediction model is obtained by training a neural network with multiple sets of sample working condition data and corresponding low-cycle fatigue life. During the training process, the low-cycle fatigue life prediction model adopts the adaptive moment estimation ADAM optimizer and obtains the number of hidden layers and neurons of the neural network through Bayesian optimization.

[0009] Multiple sets of candidate operating condition data for the turbine blade to be predicted are generated using the Monte Carlo method, and the life prediction result corresponding to each set of candidate operating condition data is predicted using a low-cycle fatigue life prediction model.

[0010] Based on multiple lifetime prediction results, the probabilistic low-cycle fatigue lifetime prediction results of the turbine blade to be predicted are determined.

[0011] Optionally, the fatigue life prediction model based on the critical plane method includes the calculation formula for fatigue damage parameters and the life calculation model;

[0012] Fatigue damage parameters The calculation formula is:

[0013] ;

[0014] in, For the normal stress on the candidate plane, The normal strain amplitude on the candidate plane, The shear strain amplitude on the candidate plane;

[0015] The lifetime calculation model is as follows:

[0016] ;

[0017] or ;

[0018] in, To correct the parameters, The fatigue strength coefficient of the material. The elastic modulus of the material. For low-cycle fatigue life, The fatigue strength index of the material. The fatigue ductility coefficient of the material. For average stress, The fatigue ductility index of the material. For the material's yield strength, This refers to the tensile strength of the material.

[0019] Optionally, the fatigue life prediction model based on the critical plane method includes the calculation formula for fatigue damage parameters and the life calculation model;

[0020] The low-cycle fatigue life of the turbine blade to be predicted was calculated under each set of sample operating conditions using a fatigue life prediction model based on the critical plane method, including:

[0021] For any set of sample operating condition data, multi-physics coupled numerical solution is performed based on the sample operating condition data to obtain the stress and strain distribution data of the turbine blade to be predicted under the sample operating condition data.

[0022] The detection location of the turbine blade to be predicted is determined, and the stress tensor and strain tensor of the detection location are extracted from the stress-strain distribution data.

[0023] The stress tensor and strain tensor are transformed to multiple candidate planes through coordinate transformation, and the damage parameters under each candidate plane are calculated using the fatigue damage parameter calculation formula based on the stress and strain under each candidate plane.

[0024] The candidate plane corresponding to the largest damage parameter is determined as the critical plane, and the stress and strain data on the critical plane are substituted into the life calculation model to obtain the low-cycle fatigue life under the sample working condition data.

[0025] Optionally, the neural network is a multi-layer fully connected deep neural network; the multi-layer fully connected deep neural network includes an input layer, multiple hidden layers, and an output layer; the neural network is trained using multiple sets of sample working condition data and corresponding low-cycle fatigue life data to obtain a low-cycle fatigue life prediction model, including:

[0026] Bayesian optimization techniques were used to determine the number of hidden layers and neurons in a multi-layer fully connected deep neural network.

[0027] After determining the number of hidden layers and neurons in the multilayer fully connected deep neural network, the multilayer fully connected deep neural network is trained with multiple sets of sample working condition data as input and low-cycle fatigue life as output to obtain a low-cycle fatigue life prediction model; during the training process of the multilayer fully connected deep neural network, the adaptive moment estimation ADAM optimizer is used to update the network parameters.

[0028] Optionally, before training a multi-layer fully connected deep neural network, the method further includes:

[0029] The data from multiple sets of sample operating conditions and their corresponding low-cycle fatigue lives were normalized.

[0030] Optionally, based on multiple lifetime prediction results, the probabilistic low-cycle fatigue lifetime prediction results for the turbine blade to be predicted are determined, including:

[0031] Based on multiple lifetime prediction results, the probability density function and cumulative distribution function of the lifetime of the turbine blade to be predicted are statistically analyzed.

[0032] The probability density function and cumulative distribution function are used as the probabilistic low-cycle fatigue life prediction results for the turbine blade to be predicted.

[0033] This invention provides a device for predicting the probabilistic low-cycle fatigue life of a turbine blade, comprising:

[0034] The acquisition module is used to acquire multiple sets of sample operating condition data of the turbine blade to be predicted, and calculate the low-cycle fatigue life of the turbine blade to be predicted under each set of sample operating condition data using the fatigue life prediction model based on the critical plane method. The fatigue life prediction model based on the critical plane method is determined based on the interaction between normal strain and shear strain on the critical plane of the turbine blade and the relationship between mean stress and fatigue life.

[0035] The module is used to train a neural network using multiple sets of sample working condition data and corresponding low-cycle fatigue life data to obtain a low-cycle fatigue life prediction model. During the training process, the low-cycle fatigue life prediction model adopts the adaptive moment estimation ADAM optimizer and obtains the number of hidden layers and neurons of the neural network through Bayesian optimization.

[0036] The prediction module is used to generate multiple sets of candidate operating condition data for the turbine blade to be predicted using the Monte Carlo method, and to predict the life prediction result corresponding to each set of candidate operating condition data using a low-cycle fatigue life prediction model.

[0037] The determination module is used to determine the probabilistic low-cycle fatigue life prediction result of the turbine blade to be predicted based on multiple life prediction results.

[0038] The present invention provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the above-described method for predicting the probabilistic low-cycle fatigue life of turbine blades.

[0039] The present invention provides a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the above-mentioned method for predicting the probabilistic low-cycle fatigue life of turbine blades.

[0040] The above-mentioned at least one technical solution adopted in this invention can achieve the following beneficial effects:

[0041] In this invention, the fatigue life prediction model based on the critical plane method is determined based on the interaction between normal strain and shear strain on the critical plane of the turbine blade and the relationship between mean stress and fatigue life. Therefore, the fatigue life prediction model based on the critical plane method has high accuracy in predicting low-cycle fatigue life. On this basis, the fatigue life prediction model based on the critical plane method is used to predict the low-cycle fatigue life of a limited sample operating data. This data is then used as training data. A neural network is used to learn the mapping relationship between the sample operating data and the low-cycle fatigue life to construct the low-cycle fatigue life prediction model. Then, based on the trained low-cycle fatigue life prediction model, the life prediction results of a large number of candidate operating data are predicted. In this way, while ensuring prediction accuracy, the repetitive consumption of high-cost calculation steps can be greatly reduced, thereby improving the efficiency of blade probabilistic low-cycle fatigue life prediction. Attached Figure Description

[0042] The accompanying drawings, which are included to provide a further understanding of the invention and form part of this invention, illustrate exemplary embodiments of the invention and are used to explain the invention, but do not constitute an undue limitation of the invention. In the drawings:

[0043] Figure 1 A schematic flowchart of a method for predicting the probabilistic low-cycle fatigue life of a turbine blade provided by the present invention.

[0044] Figure 2 A schematic diagram of a gas turbine blade model provided by the present invention;

[0045] Figure 3 This invention provides a schematic diagram of arbitrary planar coordinate transformation;

[0046] Figure 4 This is a schematic diagram of the structure of a low-cycle fatigue life prediction model provided by the present invention;

[0047] Figure 5 This is a schematic diagram of a computer device for predicting the probabilistic low-cycle fatigue life of turbine blades, as provided by the present invention. Detailed Implementation

[0048] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention will be clearly and completely described below in conjunction with specific embodiments and corresponding drawings. Obviously, the described embodiments are only a part of the embodiments of this invention, and not all of them. Based on the embodiments of this invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this invention.

[0049] Accurate probabilistic low-cycle fatigue life prediction helps in formulating reasonable operation and maintenance strategies, reducing downtime for maintenance, and improving system reliability, which is of great significance for ensuring the safe operation of turbine machinery and the economic efficiency of unit operation. However, current probabilistic low-cycle fatigue life prediction models for turbine blades still have certain limitations in terms of prediction accuracy and efficiency, and cannot fully meet the needs of actual engineering.

[0050] Based on this, the present invention provides a method for predicting the probabilistic low-cycle fatigue life of turbine blades, which improves the efficiency and accuracy of predicting the probabilistic low-cycle fatigue life of turbine blades.

[0051] The technical solutions provided by the various embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

[0052] Figure 1 This is a schematic diagram of a probabilistic low-cycle fatigue life prediction method for turbine blades according to the present invention, which specifically includes the following steps:

[0053] S101, acquire multiple sets of sample operating condition data of the turbine blade to be predicted, and calculate the low-cycle fatigue life of the turbine blade to be predicted under each set of sample operating condition data using the fatigue life prediction model based on the critical plane method. The fatigue life prediction model based on the critical plane method is determined based on the interaction between normal strain and shear strain on the critical plane of the turbine blade and the relationship between mean stress and fatigue life.

[0054] Among them, the fatigue life prediction model based on the critical plane method includes the calculation formula for fatigue damage parameters and the life calculation model.

[0055] Fatigue damage parameters The calculation formula is:

[0056] (1);

[0057] in, For the normal stress on the candidate plane, The normal strain amplitude on the candidate plane. The shear strain amplitude on the candidate plane.

[0058] The lifetime calculation model is as follows:

[0059] (2);

[0060] or (3);

[0061] in, To correct the parameters, The fatigue strength coefficient of the material. The elastic modulus of the material. For low-cycle fatigue life, The fatigue strength index of the material. The fatigue ductility coefficient of the material. For average stress, The fatigue ductility index of the material. For the material's yield strength, This represents the tensile strength of the material. It should be noted that the correction parameters can be determined based on the material of the turbine blade. The specific calculation formula is as follows, namely, formula (3) or .

[0062] Specifically, taking a gas turbine blade as an example, the blade model can be found in [reference needed]. Figure 2 .

[0063] The input samples covering the operating condition space are generated by the Latin Hypercube Sampling (LHS) method. These are multiple sets of sample operating condition data of the turbine blade to be predicted. Each set of sample operating condition data includes, but is not limited to, the thermal fatigue characteristics of the blade material, operating temperature, pressure, speed, and cooling airflow parameters.

[0064] The low-cycle fatigue life of the turbine blade to be predicted is calculated under each set of sample operating conditions using a fatigue life prediction model based on the critical plane method, including the following steps:

[0065] S201. For any set of sample operating condition data, perform multi-physics coupled numerical solution based on the sample operating condition data to obtain the stress and strain distribution data of the turbine blade to be predicted under the sample operating condition data.

[0066] For any set of sample operating conditions data, multi-physics coupled numerical solutions are performed through thermo-fluid-structure interaction numerical analysis based on the sample operating conditions data to obtain the detailed stress and strain distribution of the turbine blade to be predicted under the sample operating conditions data.

[0067] Specifically, in this embodiment, a three-dimensional high-fidelity geometric model of the turbine blade to be predicted and its surrounding fluid domain is constructed based on sample operating condition data; then, a fine mesh is generated for the geometric model; subsequently, multi-physics coupled numerical solution is performed to obtain stress and strain distribution data of the turbine blade to be predicted under extreme operating conditions such as high temperature and high pressure.

[0068] S202, determine the detection location of the turbine blade to be predicted, and extract the stress tensor σ and strain tensor ε of the detection location from the stress-strain distribution data.

[0069] Specifically, determining the detection location includes: based on the results of the thermo-fluid-structure interaction numerical analysis in step S201, selecting the area with the largest stress or strain value as the detection location for focused investigation; at the same time, based on actual analysis needs, specific areas of interest such as the blade root, leading edge, and trailing edge can also be selected as detection locations.

[0070] Extract the complete stress tensor at the selected detection location. With strain tensor The extracted stress tensor With strain tensor The mathematical expression is:

[0071] (4);

[0072] Among them, stress tensor The strain tensor is used to describe the distribution intensity of internal forces at the detection location in various directions. Used to describe the degree of deformation of the detection location in various directions. The components include ; The components include .

[0073] S203 transforms the stress tensor and strain tensor to multiple candidate planes through coordinate transformation, and calculates the damage parameters of each candidate plane based on the stress and strain under each candidate plane using the fatigue damage parameter calculation formula; and determines the candidate plane corresponding to the largest damage parameter as the critical plane.

[0074] Specifically, see the appendix. Figure 3 The provided diagram illustrates arbitrary planar coordinate transformation. At the detection position determined in step S202, the coordinates are transformed by a rotation angle. θ (Plane normal and) x (angle between axes) and ϕ (Plane normal and) z The candidate plane is generated by the angle between the axes, and the stress tensor is transformed by coordinate transformation. σ and strain tensor ε Transform to any candidate plane to obtain the stress and strain in that coordinate system. Substitute the stress and strain obtained after transformation to the candidate plane into the fatigue damage parameter calculation formula to calculate the damage parameters for that candidate plane. d .

[0075] Wherein, coordinate transformation matrix as follows:

[0076] (5).

[0077] The stress and strain after coordinate transformation are as follows:

[0078] (6).

[0079] It should be noted that in the formula for calculating fatigue damage parameters, , , In this context, the matrix index numbers indicate the position within the matrix, for example, express The component in the 3rd row and 3rd column.

[0080] By rotation angle θ and ϕ Using a small angle as the step size, the algorithm iterates through all candidate planes, repeating the above operation to calculate the fatigue damage parameters for each candidate plane. d Select fatigue damage parameters d The largest candidate plane is taken as the critical plane. σ n,max Let Δ be the normal stress on the critical plane. ε n,max Let Δ be the normal strain amplitude on the critical plane. γ max It is denoted as the shear strain amplitude on the critical plane.

[0081] S204 substitutes the stress-strain data on the critical plane into the life calculation model to obtain the low-cycle fatigue life under the sample working condition data.

[0082] The fatigue life prediction model based on the critical plane method obtains the low-cycle fatigue life; this fatigue life prediction model considers the influence of the interaction between mean stress and normal strain and shear strain on fatigue life, thus improving the prediction accuracy.

[0083] S102, a low-cycle fatigue life prediction model is obtained by training a neural network with multiple sets of sample working condition data and corresponding low-cycle fatigue life data. During the training process, the low-cycle fatigue life prediction model adopts the adaptive moment estimation ADAM optimizer and obtains the number of hidden layers and neurons of the neural network through Bayesian optimization.

[0084] Optionally, before training the multi-layer fully connected deep neural network, the multiple sets of sample operating condition data and the corresponding low-cycle fatigue life are normalized. It should be noted that, as mentioned in this invention, training the neural network using multiple sets of sample operating condition data and the corresponding low-cycle fatigue life refers to training the neural network using normalized multiple sets of sample operating condition data and the corresponding low-cycle fatigue life.

[0085] Normalizing multiple sets of sample operating condition data and their corresponding low-cycle fatigue life can ensure data quality and model training stability.

[0086] In one embodiment, the neural network is a multi-layer fully connected deep neural network; the multi-layer fully connected deep neural network includes an input layer, multiple hidden layers, and an output layer; the low-cycle fatigue life prediction model is obtained by training the neural network with multiple sets of sample working condition data and corresponding low-cycle fatigue life, including: using Bayesian optimization techniques to determine the number of hidden layers and the number of neurons in the multi-layer fully connected deep neural network; after determining the number of hidden layers and the number of neurons in the multi-layer fully connected deep neural network, the multi-layer fully connected deep neural network is trained with multiple sets of sample working condition data as input and low-cycle fatigue life as output to obtain the low-cycle fatigue life prediction model; during the training process of the multi-layer fully connected deep neural network, the network parameters are updated using an Adaptive Moment Estimation (ADAM) optimizer.

[0087] The low-cycle fatigue life prediction model is a Bayesian Optimization-Deep Neural Network (BO-DNN), and its structure is as follows: Figure 4 As shown, it consists of an input layer, several hidden layers, and an output layer.

[0088] The input is the working condition sample data obtained in step S101, and the output parameter is the predicted low-cycle fatigue life. The specific mapping relationship is as follows:

[0089] (7);

[0090] in, x These are the input parameters for the low-cycle fatigue life prediction model. These are the predicted values ​​from the low-cycle fatigue life prediction model. These are the learnable parameters for the low-cycle fatigue life prediction model. Hidden layers employ activation functions such as ReLU, Tanh, or Swish to enhance nonlinear fitting capabilities. The model structure can be dynamically adjusted based on data complexity; the number of hidden layers and neurons per layer are determined using Bayesian optimization (BO). During training, ADAM or RMSprop optimizers are used, combined with techniques such as Dropout to prevent overfitting and improve the model's generalization ability.

[0091] After obtaining the low-cycle fatigue life prediction model, the predicted life value of the turbine blade can be output under given operating conditions. To verify its accuracy, a dual evaluation strategy of cross-validation and independent test sets can be adopted to determine the accuracy of the low-cycle fatigue life prediction model, ensuring that the prediction error is controlled and meets the reliability requirements of practical engineering applications.

[0092] S103 generates multiple sets of candidate operating condition data for the turbine blade to be predicted using the Monte Carlo method, and predicts the life prediction result corresponding to each set of candidate operating condition data using a low-cycle fatigue life prediction model.

[0093] The number of candidate operating condition data groups is much larger than the number of sample operating condition data groups.

[0094] The candidate operating condition data for each group includes, but is not limited to: the thermal fatigue characteristics of the blade material, operating temperature, pressure, rotational speed, and cooling airflow parameters.

[0095] Based on the low-cycle fatigue life prediction model obtained through training, this invention introduces a probabilistic life assessment method, which uses Monte Carlo Simulation (MCS) to extend the trained model to obtain the probability distribution of life prediction.

[0096] Specifically, the operating condition data of the turbine blade to be predicted is first assigned an appropriate statistical distribution (such as normal, log-normal or Weibull distribution), and then multiple sets of candidate operating condition data of the turbine blade to be predicted are generated by the Monte Carlo method, that is, the multiple sets of candidate operating condition data conform to a certain statistical distribution.

[0097] Each set of candidate operating condition data is input into the low-cycle fatigue life prediction model one by one to obtain a series of life prediction results, namely low-cycle fatigue life.

[0098] S104, based on multiple life prediction results, determine the probabilistic low-cycle fatigue life prediction result of the turbine blade to be predicted.

[0099] In one embodiment, determining the probabilistic low-cycle fatigue life prediction result of the turbine blade to be predicted based on multiple life prediction results includes: statistically analyzing the probability density function and cumulative distribution function of the life of the turbine blade to be predicted based on the multiple life prediction results; and determining the probability density function, cumulative distribution function, and confidence interval as the probabilistic low-cycle fatigue life prediction result of the turbine blade to be predicted.

[0100] The probability low-cycle fatigue life prediction results of the turbine blade to be predicted represent the probability low-cycle fatigue life prediction results of the turbine blade to be predicted under random factors.

[0101] This invention proposes a probabilistic low-cycle fatigue life prediction method for turbine blades, improving the accuracy of such predictions. Specifically, this invention first constructs an improved SWT fatigue life prediction model. This model comprehensively considers the interaction between normal strain and shear strain on the critical plane, as well as the influence of mean stress and other factors on fatigue life, thereby improving the model's accuracy in predicting short low-cycle fatigue lives.

[0102] Furthermore, this method constructs a low-cycle fatigue life prediction model and automatically searches for the optimal hyperparameters in the model through Bayesian optimization. Compared with traditional methods, while ensuring accuracy, Bayesian optimization is used to flexibly determine the structure of the neural network, giving it the characteristics of adaptability, high flexibility, and high efficiency.

[0103] When applying the probabilistic low-cycle fatigue life prediction method for turbine blades provided by this invention, it is not necessary to rely on... Figure 1 The steps shown are executed in sequence. The specific execution order of each step can be determined as needed, and this invention does not impose any restrictions on it.

[0104] The above describes a method for predicting the probabilistic low-cycle fatigue life of turbine blades according to one or more embodiments of the present invention. Based on the same idea, the present invention also provides a corresponding device for predicting the probabilistic low-cycle fatigue life of turbine blades, the device comprising:

[0105] The acquisition module is used to acquire multiple sets of sample operating condition data of the turbine blade to be predicted, and calculate the low-cycle fatigue life of the turbine blade to be predicted under each set of sample operating condition data using the fatigue life prediction model based on the critical plane method. The fatigue life prediction model based on the critical plane method is determined based on the interaction between normal strain and shear strain on the critical plane of the turbine blade and the relationship between mean stress and fatigue life.

[0106] The module is used to train a neural network using multiple sets of sample working condition data and corresponding low-cycle fatigue life data to obtain a low-cycle fatigue life prediction model. During the training process, the low-cycle fatigue life prediction model adopts the adaptive moment estimation ADAM optimizer and obtains the number of hidden layers and neurons of the neural network through Bayesian optimization.

[0107] The prediction module is used to generate multiple sets of candidate operating condition data for the turbine blade to be predicted using the Monte Carlo method, and to predict the life prediction result corresponding to each set of candidate operating condition data using a low-cycle fatigue life prediction model.

[0108] The determination module is used to determine the probabilistic low-cycle fatigue life prediction result of the turbine blade to be predicted based on multiple life prediction results.

[0109] Specific limitations regarding the probabilistic low-cycle fatigue life prediction device for turbine blades can be found in the limitations of the probabilistic low-cycle fatigue life prediction method for turbine blades mentioned above, and will not be repeated here. Each module in the aforementioned probabilistic low-cycle fatigue life prediction device for turbine blades can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in or independent of the processor in a computer device in hardware form, or stored in the memory of a computer device in software form, so that the processor can call and execute the corresponding operations of each module.

[0110] The present invention also provides a computer-readable storage medium storing a computer program that can be used to execute the above-described... Figure 1 The method provided is a probabilistic low-cycle fatigue life prediction method for turbine blades.

[0111] The present invention also provides Figure 5 The schematic diagram of the computer device shown is as follows: Figure 5 As shown, at the hardware level, this computer device includes a processor, internal bus, network interface, memory, and non-volatile memory, and may also include other hardware required for business operations. The processor reads the corresponding computer program from the non-volatile memory into memory and then executes it to achieve the above. Figure 1 The method provided is a probabilistic low-cycle fatigue life prediction method for turbine blades.

[0112] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium, and when executed, it can include the processes of the embodiments of the methods described above. Any references to memory, storage, databases, or other media used in the embodiments provided by this invention can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, or optical storage, etc. Volatile memory can include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM can be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM), etc.

[0113] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this invention.

Claims

1. A method for predicting the probabilistic low-cycle fatigue life of a turbine blade, characterized in that, include: Multiple sets of sample operating condition data of the turbine blade to be predicted were obtained, and the low-cycle fatigue life of the turbine blade to be predicted under each set of sample operating condition data was calculated by the fatigue life prediction model based on the critical plane method. The fatigue life prediction model based on the critical plane method is determined based on the interaction between normal strain and shear strain on the critical plane of the turbine blade and the relationship between mean stress and fatigue life. By training a neural network with multiple sets of sample working condition data and corresponding low-cycle fatigue life data, a low-cycle fatigue life prediction model is obtained. The low-cycle fatigue life prediction model uses the adaptive moment estimation ADAM optimizer during training and obtains the number of hidden layers and neurons of the neural network through Bayesian optimization. Multiple sets of candidate operating condition data for the turbine blade to be predicted are generated using the Monte Carlo method, and the life prediction result corresponding to each set of candidate operating condition data is predicted using a low-cycle fatigue life prediction model. Based on multiple lifetime prediction results, the probabilistic low-cycle fatigue lifetime prediction results of the turbine blade to be predicted are determined. The fatigue life prediction model based on the critical plane method includes the calculation formula for fatigue damage parameters and the life calculation model. Fatigue damage parameters The calculation formula is: ； in, For the normal stress on the candidate plane, The normal strain amplitude on the candidate plane. The shear strain amplitude on the candidate plane; The lifetime calculation model is as follows: ； or ; in, To correct the parameters, The fatigue strength coefficient of the material. The elastic modulus of the material. For low-cycle fatigue life, The fatigue strength index of the material. The fatigue ductility coefficient of the material. For average stress, The fatigue ductility index of the material. For the material's yield strength, The tensile strength of the material; The fatigue life prediction model based on the critical plane method is used to calculate the low-cycle fatigue life of the turbine blade under each set of sample operating conditions, including: For any set of sample operating condition data, multi-physics coupled numerical solution is performed based on the sample operating condition data to obtain the stress and strain distribution data of the turbine blade to be predicted under the sample operating condition data. The detection location of the turbine blade to be predicted is determined, and the stress tensor and strain tensor of the detection location are extracted from the stress-strain distribution data. The stress tensor and strain tensor are transformed to multiple candidate planes through coordinate transformation, and the damage parameters under each candidate plane are calculated using the fatigue damage parameter calculation formula based on the stress and strain under each candidate plane. The candidate plane corresponding to the largest damage parameter is determined as the critical plane, and the stress and strain data on the critical plane are substituted into the life calculation model to obtain the low-cycle fatigue life under the sample working condition data.

2. The method according to claim 1, characterized in that, The neural network is a multi-layer fully connected deep neural network; the multi-layer fully connected deep neural network includes an input layer, multiple hidden layers, and an output layer; by training the neural network with multiple sets of sample working condition data and corresponding low-cycle fatigue life data, a low-cycle fatigue life prediction model is obtained, including: Bayesian optimization techniques were used to determine the number of hidden layers and neurons in a multi-layer fully connected deep neural network. After determining the number of hidden layers and neurons in the multilayer fully connected deep neural network, the multilayer fully connected deep neural network is trained with multiple sets of sample working condition data as input and low-cycle fatigue life as output to obtain a low-cycle fatigue life prediction model; during the training process of the multilayer fully connected deep neural network, the adaptive moment estimation ADAM optimizer is used to update the network parameters.

3. The method according to claim 1, characterized in that, Before training a multi-layer fully connected deep neural network, the method further includes: The data from multiple sets of sample operating conditions and their corresponding low-cycle fatigue lives were normalized.

4. The method according to claim 1, characterized in that, Based on multiple lifetime prediction results, the probabilistic low-cycle fatigue lifetime prediction results for the turbine blade to be predicted are determined, including: Based on multiple lifetime prediction results, the probability density function and cumulative distribution function of the lifetime of the turbine blade to be predicted are statistically analyzed. The probability density function and cumulative distribution function are used as the probabilistic low-cycle fatigue life prediction results for the turbine blade to be predicted.

Images