A method for analyzing the dynamic response of a gas turbine cylinder elastic support structure

Abstract

The application relates to a dynamic response analysis method of a gas turbine cylinder elastic support structure, wherein the method comprises the following steps: generating a finite element simulation result of the gas turbine cylinder elastic support structure based on at least one of an isotropic loss factor, a structure size, an elastic modulus and an acting load of the gas turbine cylinder elastic support structure; constructing a firefly optimization algorithm-back propagation FA-BP neural network model; training the FA-BP neural network model by using a training set; evaluating the trained FA-BP neural network model, and generating a dynamic response result of the gas turbine cylinder elastic support structure after the evaluation result meets a preset condition. The embodiment of the application can realize training and prediction of a dynamic analysis model of an elastic support structure, provide a fast iteration method for design optimization, improve the calculation efficiency of elastic support dynamic simulation in the iteration design process, and realize a dynamic simulation method with high efficiency and precision.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of gas turbines, in particular to a dynamic response analysis method of a gas turbine cylinder elastic support structure. BACKGROUND

[0002] Heavy-duty gas turbines usually operate in extreme harsh conditions such as high temperature, high pressure, high speed, and thermal-fluid-solid multi-field coupling. Among them, the gas cylinder elastic support plays a crucial role in the gas turbine system. The main function of the elastic support structure is to ensure that the gas cylinder is stably fixed on the foundation during operation, while absorbing and isolating the vibration generated by the operation of the gas turbine. It can effectively reduce the vibration generated by the operation of the gas turbine from being transmitted to the foundation, thereby protecting the foundation structure from damage and reducing the vibration impact on the surrounding environment. Secondly, the elastic support helps to maintain the precise position of the gas cylinder, preventing the deviation of the coaxiality caused by vibration, which is directly related to the efficiency and performance of the gas turbine. Furthermore, by absorbing vibration, the elastic support can also reduce the wear and tear of the internal components of the gas cylinder, prolonging the service life of the gas turbine. Finally, a good gas cylinder and foundation support system helps to reduce noise levels, improve the working environment, and meet environmental protection requirements. As can be seen, the elastic support between the gas cylinder and the foundation is a key factor for the stable, efficient and safe operation of the gas turbine, and its design and application must be fully valued. Therefore, the dynamic characteristics analysis of the gas cylinder elastic support during the design process is very important.

[0003] For structural optimization, the prediction of dynamic response is an important factor affecting structural optimization. On the one hand, the accuracy of predicting the dynamic response is the basis of optimization, and overestimation or underestimation of the dynamic response will result in safety or performance deficiencies. At different stages of structural design, experimental methods or numerical calculation methods based on finite elements are usually used. The effectiveness of these methods has been widely verified, ensuring the accuracy of the dynamic response. On the other hand, the efficiency of dynamic response prediction determines the implementation cost of the optimization algorithm. Generally, the estimation of the objective function needs to be achieved by obtaining data through a large number of experiments or calculations, and the more data, the more accurate the estimation of the objective function, which means that all parameter values that meet the constraint conditions need to be traversed. When the number of design parameters is high, the existence of dimension disaster results in the need for tens of thousands of times or more of traversal. Considering the time, this is impossible for experiments. For numerical calculation methods based on finite elements, such a large number of cases is also difficult to achieve. In addition to the huge amount of calculation and the requirement of calculation time, meshing also brings a huge labor cost. This makes the data obtained by these methods insufficient to achieve accurate implementation of the optimization algorithm within a relatively short development cycle. Therefore, an efficient, fast and accurate dynamic response prediction method is of great significance for structural optimization. SUMMARY

[0004] The application provides a dynamic response analysis method of a gas turbine cylinder elastic support structure to solve the problem of low accuracy of dynamic response and low efficiency of dynamic response prediction in the prior art, and the accurate implementation of the optimization algorithm is difficult to complete within the development cycle.

[0005] The first aspect of the application provides a dynamic response analysis method of a gas turbine cylinder elastic support structure, including the following steps: based on at least one of the isotropic loss factor, the structure size, the elastic modulus and the acting load of the gas turbine cylinder elastic support structure, generating a finite element simulation result of the gas turbine cylinder elastic support structure, and extracting dynamic response output data of a target machine learning model from the finite element simulation result; based on the dynamic response output data, constructing a firefly optimization algorithm-back propagation FA-BP neural network model according to a plurality of parameters representing the gas turbine cylinder elastic support structure and at least one external load parameter; training the FA-BP neural network model using a training set to obtain a trained FA-BP neural network model; evaluating the trained FA-BP neural network model based on at least one evaluation index of mean absolute percentage error and coefficient of determination to obtain an evaluation result, and generating a dynamic response result of the gas turbine cylinder elastic support structure after the evaluation result meets a preset condition.

[0006] Optionally, in an embodiment of the application, the extracting of the output result of the preset machine learning model from the finite element simulation result includes: extracting stress values that meet a preset maximum condition in a plurality of three-dimensional intercept points, and predicting a plurality of order natural frequencies of the gas turbine cylinder elastic support structure; determining the output result of the preset machine learning model based on the stress values that meet the preset maximum condition in the plurality of three-dimensional intercept points and the plurality of order natural frequencies of the gas turbine cylinder elastic support structure.

[0007] Optionally, in an embodiment of the application, the constructing of the firefly optimization algorithm-back propagation FA-BP neural network model according to a plurality of parameters representing the gas turbine cylinder elastic support structure and at least one external load parameter includes: normalizing data in a data set to obtain processed data; based on the processed data, generating input nodes according to the parameters of the gas turbine cylinder elastic support structure and the at least one external load parameter, and generating output nodes according to the stress values that meet the preset maximum condition in the plurality of three-dimensional intercept points and the plurality of order natural frequencies; based on the input nodes and the output nodes, constructing the FA-BP neural network model.

[0008] Optionally, in an embodiment of the present application, the calculation formula of mutual attraction degree in the firefly optimization algorithm-back propagation FA-BP neural network model is as follows:

[0009]

[0010] wherein β0 is the initial attraction degree when r ij =0, γ is the light absorption coefficient, r ij is the Euclidean distance between the firefly individuals;

[0011] The calculation formula of the Euclidean distance between the firefly individuals is as follows:

[0012]

[0013] wherein x id represents the position of firefly i in d-dimensional space, x jd represents the position of firefly j in d-dimensional space, and D represents D-dimensional space.

[0014] The position updating formula of the firefly is as follows:

[0015]

[0016] wherein x j is the position of firefly j with lower brightness, rand() is a random disturbance term, α is a step factor, and t is the iteration number.

[0017] Optionally, in an embodiment of the present application, the calculation formula of the mean absolute percentage error is as follows:

[0018]

[0019] wherein N is the data number, t i is the true value, and y i is the predicted value.

[0020] The calculation formula of the determination coefficient is as follows:

[0021]

[0022] wherein N is the data number, y i is the true value, y i is the predicted value, and is the mean value of all true values.

[0023] The second aspect embodiment of the application provides a device for analyzing the dynamic response of a gas turbine cylinder elastic support structure, comprising: an extraction module configured to generate a finite element simulation result of the gas turbine cylinder elastic support structure based on at least one of an isotropic loss factor, a structure size, an elastic modulus, and an acting load of the gas turbine cylinder elastic support structure, and extract dynamic response output data of a target machine learning model from the finite element simulation result; a construction module configured to construct a firefly optimization algorithm-back propagation (FA-BP) neural network model based on the dynamic response output data and according to a plurality of parameters representing the gas turbine cylinder elastic support structure and at least one external load parameter; a training module configured to train the FA-BP neural network model using a training set to obtain a trained FA-BP neural network model; and an evaluation module configured to evaluate the trained FA-BP neural network model based on at least one of a mean absolute percentage error and a coefficient of determination to obtain an evaluation result, and generate a dynamic response result of the gas turbine cylinder elastic support structure after the evaluation result meets a preset condition.

[0024] Optionally, in an embodiment of the application, the extraction module comprises an extraction unit configured to extract stress values of a plurality of three-dimensional intercept points that meet a preset maximum condition, and predict a plurality of natural frequencies of the gas turbine cylinder elastic support structure; and a determination unit configured to determine an output result of the preset machine learning model based on the stress values of the plurality of three-dimensional intercept points that meet the preset maximum condition and the plurality of natural frequencies of the gas turbine cylinder elastic support structure.

[0025] Optionally, in an embodiment of the application, the construction module comprises a processing unit configured to perform normalization processing on data in a data set to obtain processed data; a generation unit configured to generate input nodes based on the parameters of the gas turbine cylinder elastic support structure and the at least one external load parameter, and generate output nodes based on the stress values of the plurality of three-dimensional intercept points that meet the preset maximum condition and the plurality of natural frequencies; and a construction unit configured to construct the FA-BP neural network model based on the input nodes and the output nodes.

[0026] Optionally, in an embodiment of the application, a calculation formula of mutual attraction degree in the firefly optimization algorithm-back propagation (FA-BP) neural network model is as follows:

[0027]

[0028] wherein β0 is an initial attraction degree when t ij = 0, γ is a light absorption coefficient, r ij is the Euclidean distance between firefly individuals.

[0029] The Euclidean distance between the firefly individuals is calculated by the following formula:

[0030]

[0031] where x id represents the position of firefly i in d-dimensional space, x jd represents the position of firefly j in d-dimensional space, and D represents the d-dimensional space.

[0032] The position update formula of the firefly is:

[0033]

[0034] where x j is the position of firefly j with lower brightness, rand() is a random disturbance term, a is a step factor, and t is the iteration number.

[0035] Optionally, in an embodiment of the present application, the calculation formula of the mean absolute percentage error is:

[0036]

[0037] where N is the number of data, t i is the true value, and y i is the predicted value.

[0038] The calculation formula of the determination coefficient is:

[0039]

[0040] where N is the number of data, t i is the true value, y i is the predicted value, is the mean value of all true values.

[0041] The third aspect embodiment of the present application provides an electronic device, comprising 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 dynamic response analysis method of the gas turbine cylinder elastic support structure as described in the above embodiments.

[0042] The fourth aspect embodiment of the present application provides a computer readable storage medium, which stores a computer program, and the program is executed by a processor to implement the dynamic response analysis method of the gas turbine cylinder elastic support structure as described above.

[0043] The fifth aspect of the embodiment of the present application provides a computer program product, the computer program product stores a computer program, and the program is executed by a processor to realize the method for analyzing the dynamic response of the elastic support structure of the gas turbine cylinder.

[0044] The embodiment of the present application can use artificial intelligence technology to realize training and prediction of the dynamic analysis model of the elastic support structure, provide a fast iteration method for design optimization, realize improvement of the calculation efficiency of the elastic support dynamic simulation in the iteration design process, and realize the dynamic simulation method with consideration of efficiency and accuracy. Thus, the problems in the related art, such as low accuracy of the dynamic response and low efficiency of the dynamic response prediction, and difficulty in accurate implementation of the optimization algorithm within the development cycle, are solved.

[0045] Additional aspects and advantages of the present application will be in part apparent and in part pointed out hereinafter. BRIEF DESCRIPTION OF DRAWINGS

[0046] The above and / or additional aspects and advantages of the present application will become apparent and be readily appreciated from the following description, including the accompanying drawings, in which:

[0047] Figure 1 A flowchart of a method for analyzing the dynamic response of the elastic support structure of the gas turbine cylinder according to an embodiment of the present application is provided.

[0048] Figure 2 A variable size diagram of the elastic support of the cylinder according to an embodiment of the present application is provided.

[0049] Figure 3 A three-dimensional intersection point diagram in the elastic support of the cylinder according to an embodiment of the present application is provided.

[0050] FIG. 4(a) is a diagram of the maximum stress at the three-dimensional intersection point 1 according to an embodiment of the present application.

[0051] FIG. 4(b) is a diagram of the maximum stress at the three-dimensional intersection point 2 according to an embodiment of the present application.

[0052] FIG. 4(c) is a diagram of the maximum stress at the three-dimensional intersection point 2 according to an embodiment of the present application.

[0053] FIG. 4(d) is a diagram of the first order natural frequency according to an embodiment of the present application.

[0054] FIG. 4(e) is a diagram of the second order natural frequency according to an embodiment of the present application.

[0055] FIG. 4(f) is a diagram of the third order natural frequency according to an embodiment of the present application.

[0056] Fig. 4(g) is a schematic diagram of the 4th order natural frequency according to an embodiment of the present application;

[0057] Fig. 4(h) is a schematic diagram of the 5th order natural frequency according to an embodiment of the present application;

[0058] Figure 5 Fig. 4(h) is a schematic diagram of the 5th order natural frequency according to an embodiment of the present application;

[0059] Figure 6 Fig. 4(h) is a schematic diagram of the 5th order natural frequency according to an embodiment of the present application;

[0060] Figure 7 Fig. 4(h) is a schematic diagram of the 5th order natural frequency according to an embodiment of the present application; DETAILED DESCRIPTION

[0061] Embodiments of the present application are described in detail below with reference to the accompanying drawings, in which the same or similar components have the same or similar designations and functions throughout the various figures and embodiments. The embodiments described below are examples in which similar or identical components have the same or similar designations and functions throughout the various figures and embodiments. The embodiments described below are intended to explain the present application, and are not intended to limit the present application.

[0062] A method for analyzing the dynamic response of a gas turbine cylinder elastic support structure is described below with reference to the accompanying drawings. In view of the low accuracy of the dynamic response and the low efficiency of the dynamic response prediction in the related art mentioned in the background, it is difficult to accurately implement the optimization algorithm within the development cycle. The present application provides a method for analyzing the dynamic response of a gas turbine cylinder elastic support structure. In this method, artificial intelligence technology can be used to train and predict the dynamic analysis model of the elastic support structure, providing a fast iterative method for design optimization, improving the calculation efficiency of the elastic support dynamics simulation in the iterative design process, and achieving a dynamic simulation method that balances efficiency and accuracy. Thus, the problems of low accuracy of the dynamic response and low efficiency of the dynamic response prediction in the related art are solved.

[0063] Specifically, Figure 1 Fig. 4(h) is a schematic diagram of the 5th order natural frequency according to an embodiment of the present application;

[0064] As Figure 1 shown, the method for analyzing the dynamic response of a gas turbine cylinder elastic support structure includes the following steps:

[0065] In step S101, based on at least one of the isotropic loss factor, the structural size, the elastic modulus and the acting load of the gas turbine cylinder elastic support structure, a finite element simulation result of the gas turbine cylinder elastic support structure is generated, and a dynamic response output data of a target machine learning model is extracted from the finite element simulation result.

[0066] It can be understood that machine learning is increasingly applied in the field of structure optimization. After a certain amount of data is obtained through finite element calculation or experiment, a black box mathematical model can be used to approximate the real physical model according to the data to describe the relationship between the parameters and the dynamic response. In the subsequent parameter traversal process, only the calculation of the function is needed to obtain the dynamic response value, thereby greatly reducing the traversal cost.

[0067] In actual execution, for the structure type of the gas turbine cylinder elastic support model, the embodiment of the application can obtain the finite element simulation result of the gas turbine cylinder elastic support structure by changing the isotropic loss factor of the structure, changing the structural size, changing the elastic modulus, changing the acting load, etc. As for the machine learning model of the rotor, a simulation analysis model of the machine learning model is obtained through pre-training to realize the prediction of the simulation result of the cylinder elastic support structure under different parameters.

[0068] From the perspective of method verification, in order to generate a large amount of training data, a static load loading method is used to obtain the load response of the cylinder elastic support structure. The variable structure parameter data of the cylinder elastic support includes the thickness of plate 1 and the thickness of plate 2 Figure 2 ), the isotropic loss factor, the elastic modulus and the applied load. The input parameter space has 5 dimensions, different numbers of parameter values are taken at equal intervals on each dimension within a certain parameter floating range, and a parameter grid of 5x5x10 3 is generated in the parameter space. Table 1 is a list of cylinder elastic support structure parameters, wherein, as shown in Table 1:

[0069] Table 1

[0070]

[0071]

[0072] Optionally, in an embodiment of the application, the output result of the preset machine learning model is extracted from the finite element simulation result, including: extracting stress values that meet a preset maximum condition in a plurality of three-dimensional intercept points, and predicting a plurality of orders of natural frequencies of the gas turbine cylinder elastic support structure; based on the stress values that meet the preset maximum condition in the plurality of three-dimensional intercept points and the plurality of orders of natural frequencies of the gas turbine cylinder elastic support structure, determining the output result of the preset machine learning model.

[0073] It can be understood that the stress value satisfying the preset maximum condition in the embodiment of the application can be a maximum stress value.

[0074] In the embodiment of the application, in order to obtain the main stress conditions at different positions and the simulation analysis model, three three-dimensional intercepts for prediction are selected at different positions, as shown in Figure 3 The positions of the three-dimensional intercepts are as shown in Figure 3 After simulation, the maximum stresses at the three-dimensional intercepts in the simulation time are extracted, and these variables will be used as the output values of the machine learning model. In addition, the first five natural frequencies of the elastic support structure of the gas cylinder are predicted using the machine learning model.

[0075] In summary, the output values of the machine learning model proposed in the application are a total of 8 categories, i.e., the maximum stress at the three-dimensional intercept 1, the maximum stress at the three-dimensional intercept 2, the maximum stress at the three-dimensional intercept 3, the first-order natural frequency, the second-order natural frequency, the third-order natural frequency, the fourth-order natural frequency, and the fifth-order natural frequency, as shown in Figure 4(a) to Figure 4(h) .

[0076] It should be noted that the preset maximum condition can be set by those skilled in the art according to actual conditions, and is not specifically limited herein.

[0077] In step S102, based on the dynamic response output data, a firefly optimization algorithm-back propagation FA-BP neural network model is constructed according to a plurality of parameters representing the elastic support structure of the gas cylinder of the gas turbine and at least one external load parameter.

[0078] In actual execution, the embodiment of the application can construct a firefly optimization algorithm-back propagation FA-BP neural network model based on the dynamic response output data according to a plurality of parameters representing the elastic support structure of the gas cylinder of the gas turbine and at least one external load parameter. In the embodiment of the application, five parameters representing the elastic support structure of the gas cylinder and one external load parameter are used, i.e., the input of the FA-BP neural network model is six parameters, so six input nodes are set. The hidden layer nodes are set according to the empirical formula and the trial-and-error method, where m and n are the numbers of neurons in the input layer and the output layer, respectively. a is a positive integer between 1 and 10. The application sets the number of hidden layer nodes to 5. The output is one node.

[0079] Optionally, in an embodiment of the present application, a firefly optimization algorithm-back propagation FA-BP neural network model is constructed according to a plurality of parameters characterizing the elastic support structure of the gas turbine cylinder and at least one external load parameter, including: performing normalization processing on the data in the data set to obtain processed data; based on the processed data, generating input nodes according to the parameters of the elastic support structure of the gas turbine cylinder and at least one external load parameter, and generating output nodes according to the stress values in the plurality of three-dimensional intersection points that meet the preset maximum condition and the plurality of natural frequencies; and based on the input nodes and the output nodes, constructing the FA-BP neural network model.

[0080] As a possible implementation manner, the network in the embodiment of the present application adopts a Sigmoid function as an activation function, and in order to make the data set within the value range of the activation function, the data is normalized to obtain processed data. During the running of the FA-BP network, the size of the learning rate can affect the stability of the network. After step-by-step testing, the learning rate is set to 0.001. The number of training times is set to 1000. The target error is set to 0.00001. The learning function is Trainlm.

[0081] The firefly algorithm (FA) is adopted in the present application to optimize the network weights and thresholds, enhance the robustness of the network, and improve the prediction performance. FA is one of the popular algorithms in the field of swarm intelligence. The inspiration comes from how fireflies signal each other by flashing their lights to attract mates or warn potential predators. In order to construct the search model of FA, there are three idealized rules. First, fireflies do not distinguish between genders, and they are attracted to each other only considering the brightness of the individual light emission. Second, the attraction is proportional to the brightness of the light emission and inversely proportional to the distance between individuals. Third, the brightness of the firefly is determined by the value of the objective function to be optimized. For this algorithm, each firefly represents a solution to the problem. The brightness of the firefly represents the fitness of its position. The higher the brightness, the better the position of the firefly in the solution space. The position and data are adjusted and updated by comparing the brightness, and then the optimal solution of the problem is obtained.

[0082] Suppose the position of firefly i in D-dimensional space is x i =(x i1 ,x i2 ,…,x iD ). Then, the brightness of the firefly is equal to the function value at this position, denoted as F i =f(x i ). If the brightness of firefly i is higher than that of firefly j, firefly j will be attracted to firefly i.

[0083] Optionally, in an embodiment of the present application, the mutual attraction degree in the firefly optimization algorithm-back propagation FA-BP neural network model is calculated according to the following formula:

[0084]

[0085] wherein β0 is the initial attraction degree when r ij = 0, γ is the light absorption coefficient, r ij is the Euclidean distance between firefly individuals;

[0086] The calculation formula of the Euclidean distance between firefly individuals is as follows:

[0087]

[0088] wherein x id represents the position of firefly i in a d-dimensional space, x jd represents the position of firefly j in a d-dimensional space, and D represents a D-dimensional space.

[0089] In the operation process, the position updating formula of the firefly is as follows:

[0090]

[0091] wherein x j is the position of the firefly j with lower brightness, rand() is a random disturbance term, i.e., a Gaussian random number; α is a step factor, which can be adjusted according to the pseudo-time step number in each iteration, so that the algorithm can balance between global search and local search in the search process; and t is the iteration number. Table 2 is a parameter setting table of the FA algorithm, wherein the related parameters of the firefly algorithm are shown in Table 2:

[0092] Table 2

[0093] Parameters Value Firefly number 20 Pseudo time step 50 α 0.25 β 0.2 Gamma 1

[0094] In step S103, the FA-BP neural network model is trained by using the training set, and a trained FA-BP neural network model is obtained.

[0095] In the actual execution process, the data set can be divided into a training set, a validation set and a test set by a non-replacement random sampling generation manner from the total data set, and the division ratio is 0.6, 0.2 and 0.2. The training set is used for training the model, the FA-BP neural network model is trained by using the training set, and a trained FA-BP neural network model is obtained, which contains 15000 samples. The validation set is used for selecting model hyperparameters, and contains 5000 samples. The test set is used for finally evaluating the model effect, and contains 5000 samples.

[0096] The embodiment of the application utilizes artificial intelligence technology to realize training and prediction of a dynamic analysis model of an elastic support structure, provides a rapid iteration method for design optimization, realizes improvement of the calculation efficiency of the elastic support dynamics simulation in the iteration design process, and realizes a dynamics simulation method with both efficiency and accuracy.

[0097] In step S104, the trained FA-BP neural network model is evaluated based on at least one of the evaluation indexes of the mean absolute percentage error and the coefficient of determination, an evaluation result is obtained, and the dynamics response result of the elastic support structure of the gas turbine cylinder is generated after the evaluation result meets a preset condition.

[0098] Specifically, 25,000 different cylinder elastic support configurations are designed by different combinations of parameters and changes in size, among which 20,000 groups are used as the training set and the verification set of the network, and the remaining 5,000 groups are used as the test set of the network. In order to verify the effectiveness of the prediction of the network, the trained FA-BP neural network model is evaluated based on at least one of the evaluation indexes of the mean absolute percentage error and the coefficient of determination, and an evaluation result is obtained. Since the magnitude of the predicted parameters is different, the mean absolute percentage error (MAPE) and the coefficient of determination (R 2 ) are mainly used for evaluation.

[0099] Among them, in an embodiment of the application, the MAPE is generally expressed in the form of percentage. It can not only measure the error between the predicted value and the true value, but also consider the proportion of the error and the true value. The calculation formula of the mean absolute percentage error is:

[0100]

[0101] Among them, N is the number of data, t i is the true value, y i is the predicted value;

[0102] R 2 can evaluate the fitting degree of the network. The larger R 2 , the higher the degree of explanation of the independent variable to the dependent variable, and the higher the percentage of the change caused by the independent variable in the total change. It is explained that the observation points are more concentrated near the regression straight line. The calculation formula of the coefficient of determination is:

[0103]

[0104] Among them, N is the number of data, t i is the true value, y i is the predicted value, is the mean of all true values.

[0105] The embodiment of the application relates to a simplified and efficient gas turbine dynamics response parameter prediction method, and solves the problems of large calculation scale, long iteration process duration and large calculation workload in the dynamics analysis process of the elastic support of a gas turbine cylinder.

[0106] Specifically, the working principle of the dynamics response analysis method of the gas turbine cylinder elastic support structure in the embodiment of the application is described in detail in one specific embodiment.

[0107] Embodiment:

[0108] In Figure 2 The application is verified in the gas turbine cylinder elastic support structure shown in the figure. The prediction result of the proposed FA-BP model is shown in the figure. Figure 4(a) to Figure 4(h) The prediction result of 5000 test sets is shown in the figure, wherein, since the natural frequency is only affected by the size and the elastic modulus in the simulation process, the training set and the verification set are 250 in total, and the test set is 50. By comparing the prediction result with the expected result, it can be seen that the overall prediction result is relatively close to the real data, the fitting degree of the proposed FA-BP network is relatively high, the error is relatively small, and the FA-BP network can be used as an effective tool for predicting the dynamics response of the cylinder elastic support.

[0109] In addition, the MAPE and R 2 Two types of parameters quantitatively evaluate the prediction performance of the FA-BP network. The specific results are shown in Table 3. It can be seen that the R 2 are higher than 99%, indicating that the FA-BP model has a good fitting degree and a high quality. The MAPE has a low value, which means that the prediction result of the FA-BP model deviates from the real value to a small extent, and the prediction performance is high.

[0110] Table 3

[0111] MAPE [R 2 ]]> Three-dimensional intercept 1 0.056252951 0.997324685 Three-dimensional intercept 2 0.045319823 0.999242074 Three-dimensional intercept 3 0.001092034 0.999998059 First order natural frequency 0.005064654 0.999954379 Second order natural frequency 0.006589843 0.999941106 Third order natural frequency 0.002106843 0.999999053 Fourth order natural frequency 0.002071067 0.999783146 Fifth order natural frequency 0.003655218 0.99964841

[0112] As shown in the figure, the embodiment of the application can include the following steps: Figure 5

[0113] Step S501: constructing a gas turbine cylinder elastic support model of different configurations.

[0114] Step S502: isotropic loss factor, structure size, elastic modulus and acting load.

[0115] Step S503: finite element simulation.

[0116] Step S504: dynamics response output of the cylinder elastic support of different configurations.

[0117] Step S505: FA-BP neural network model.​

[0118] Step S506: new configuration parameters.

[0119] Step S507: dynamic response prediction.

[0120] Step S508: high-precision and high-efficiency gas turbine cylinder elastic support dynamic response analysis method.

[0121] The dynamic response analysis method of the gas turbine cylinder elastic support structure according to the embodiments of the present application can utilize artificial intelligence technology to realize training and prediction of a dynamic analysis model of the elastic support structure, provide a rapid iteration method for design optimization, improve the calculation efficiency of the elastic support dynamic simulation in the iteration design process, and realize a dynamic simulation method with both efficiency and precision. Thus, the problem of low accuracy of dynamic response and low efficiency of dynamic response prediction in the related art, which makes it difficult to complete the accurate implementation of the optimization algorithm within the development cycle, is solved.

[0122] Next, a dynamic response analysis device of a gas turbine cylinder elastic support structure according to the embodiments of the present application is described with reference to the accompanying drawings.

[0123] Figure 6 is a structural schematic diagram of a dynamic response analysis device of a gas turbine cylinder elastic support structure according to the embodiments of the present application.

[0124] As shown in Figure 6 , the dynamic response analysis device 10 of the gas turbine cylinder elastic support structure includes an extraction module 100, a construction module 200, a training module 300, and an evaluation module 400.

[0125] Specifically, the extraction module 100 is configured to generate a finite element simulation result of the gas turbine cylinder elastic support structure based on at least one of the isotropic loss factor, the structure size, the elastic modulus, and the acting load of the gas turbine cylinder elastic support structure, and extract dynamic response output data of a target machine learning model from the finite element simulation result.

[0126] The construction module 200 is configured to construct a firefly optimization algorithm-back propagation FA-BP neural network model according to a plurality of parameters representing the gas turbine cylinder elastic support structure and at least one external load parameter based on the dynamic response output data.

[0127] The training module 300 is configured to train the FA-BP neural network model using a training set to obtain a trained FA-BP neural network model.

[0128] The evaluation module 400 is configured to evaluate the trained FA-BP neural network model based on at least one of the evaluation indexes of the mean absolute percentage error and the coefficient of determination, to obtain an evaluation result, and to generate the dynamic response result of the elastic support structure of the gas turbine cylinder when the evaluation result meets a preset condition.

[0129] Optionally, in an embodiment of the present application, the extraction module 100 comprises an extraction unit and a determination unit.

[0130] The extraction unit is configured to extract the stress values that meet the preset maximum condition from the plurality of three-dimensional intersection points, and to predict the multi-order natural frequencies of the elastic support structure of the gas turbine cylinder.

[0131] The determination unit is configured to determine the output result of the preset machine learning model based on the stress values that meet the preset maximum condition from the plurality of three-dimensional intersection points and the multi-order natural frequencies of the elastic support structure of the gas turbine cylinder.

[0132] Optionally, in an embodiment of the present application, the construction module 200 comprises a processing unit, a generation unit and a construction unit.

[0133] The processing unit is configured to perform normalization processing on the data in the data set to obtain processed data.

[0134] The generation unit is configured to generate input nodes based on the processed data according to the parameters of the elastic support structure of the gas turbine cylinder and at least one external load parameter, and to generate output nodes based on the stress values that meet the preset maximum condition from the plurality of three-dimensional intersection points and the multi-order natural frequencies.

[0135] The construction unit is configured to construct the FA-BP neural network model based on the input nodes and the output nodes.

[0136] Optionally, in an embodiment of the present application, the calculation formula of the mutual attraction degree in the firefly optimization algorithm-back propagation FA-BP neural network model is as follows:

[0137]

[0138] wherein, β0 is the initial attraction degree when r ij = 0, γ is the light absorption coefficient, and r ij is the Euclidean distance between the firefly individuals.

[0139] The calculation formula of the Euclidean distance between the firefly individuals is as follows:

[0140]

[0141] wherein, x id represents the position of the firefly i in a d-dimensional space, and x jdrepresents the position of the firefly j in a d-dimensional space, D represents a D-dimensional space;

[0142] The position updating formula of the firefly is:

[0143]

[0144] wherein x j represents the position of the firefly j with lower brightness, rand() represents a random disturbance term, a represents a step factor, and t represents the number of iterations.

[0145] Optionally, in an embodiment of the present application, the calculation formula of the mean absolute percentage error is:

[0146]

[0147] wherein N represents the number of data, t i represents a true value, and y i represents a predicted value.

[0148] The calculation formula of the coefficient of determination is:

[0149]

[0150] wherein N represents the number of data, t i represents a true value, and y i represents a predicted value, is the mean value of all true values.

[0151] It should be noted that the aforementioned explanation and description of the embodiment of the method for analyzing the dynamic response of the elastic support structure of the gas turbine cylinder also applies to the embodiment of the device for analyzing the dynamic response of the elastic support structure of the gas turbine cylinder, which will not be described here.

[0152] The device for analyzing the dynamic response of the elastic support structure of the gas turbine cylinder according to the embodiment of the present application can use artificial intelligence technology to train and predict the dynamic analysis model of the elastic support structure, provide a fast iterative method for design optimization, improve the calculation efficiency of the dynamic simulation of the elastic support in the iterative design process, and realize a dynamic simulation method with both efficiency and accuracy. Therefore, the problem of low accuracy of dynamic response and low efficiency of dynamic response prediction in the related art is solved, and the accurate implementation of the optimization algorithm in the development cycle is difficult to achieve.

[0153] Figure 7 The structure schematic diagram of the electronic device provided in the embodiment of the present application is shown. The electronic device can include:

[0154] The memory 701, the processor 702, and the computer program stored in the memory 701 and executable on the processor 702.

[0155] The processor 702 implements the method for analyzing the dynamic response of the elastic support structure of the gas turbine cylinder according to the above embodiments when executing the program.

[0156] Further, the electronic device further comprises:

[0157] The communication interface 703 is configured to communicate between the memory 701 and the processor 702.

[0158] The memory 701 is configured to store a computer program executable on the processor 702.

[0159] The memory 701 can include a high-speed RAM memory, and can also include a non-volatile memory, for example, at least one disk memory.

[0160] If the memory 701, the processor 702 and the communication interface 703 are implemented independently, the communication interface 703, the memory 701 and the processor 702 can be connected to each other through a bus and complete the communication between each other. The bus can be an Industry Standard Architecture (ISA) bus, a Peripheral Component (PCI) bus or an Extended Industry Standard Architecture (EISA) bus, etc. The bus can be divided into an address bus, a data bus, a control bus, etc. For the convenience of representation, Figure 7 Only one thick line is used in the figure, but it does not mean that there is only one bus or only one type of bus.

[0161] Optionally, in a specific implementation, if the memory 701, the processor 702 and the communication interface 703 are integrated on a chip, the memory 701, the processor 702 and the communication interface 703 can complete the communication between each other through an internal interface.

[0162] The processor 702 can be a central processing unit (CPU), or an application specific integrated circuit (ASIC), or one or more integrated circuits configured to implement the embodiments of the present application.

[0163] The embodiments also provide a computer readable storage medium, which stores a computer program, and the program is executed by the processor to implement the method for analyzing the dynamic response of the elastic support structure of the gas turbine cylinder.

[0164] The embodiment of the present application further provides a computer program product, which stores a computer program, and the computer program is executed by a processor to realize the method for analyzing the dynamic response of the elastic supporting structure of the gas turbine cylinder.

[0165] In the description of the specification, the description of the terms "one embodiment", "some embodiments", "an example", "a specific example", or "some examples" means that the specific features, structures, materials or characteristics described in connection with the embodiment or example are included in at least one embodiment or example of the present application. In the specification, the illustrative description of the above terms does not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described can be combined in any appropriate manner in any one or N embodiments or examples. In addition, the person skilled in the art can combine and combine the different embodiments or examples described in the specification and the features of the different embodiments or examples without contradiction.

[0166] In addition, the terms "first", "second" are only used for descriptive purposes, and cannot be understood as indicating or implying relative importance or implicitly indicating the number of the indicated technical features. Therefore, the features defined with "first", "second" can explicitly or implicitly include at least one of the features. In the description of the present application, the meaning of "N" is at least two, for example, two, three, etc., unless otherwise specifically limited.

[0167] Any process or method descriptions in flow charts or otherwise described herein can be understood as representing code modules, segments, or portions of code that include one or more executable instructions for performing a specified logic function or process, and the various embodiments of the application contemplate that the modules, segments, or portions of code may be implemented in hardware, software, or a combination of both. The preferred embodiments of the application contemplate that the various processes or methods described in the specification can be implemented in a computer program product tangibly embodied in a machine-readable storage medium, for execution by a processor. The machine-readable storage medium can include, but is not limited to, one or more types of storage devices that are external to the processor, such as a disk drive, a solid state drive, a flash drive, a USB drive, a DVD, a CD, a ROM, a RAM, a cache, a semiconductor memory, etc. The computer program product can also include, but is not limited to, one or more types of computer program products that are internal to the processor, such as a register, a cache, a semiconductor memory, etc.

[0168] The logic and / or steps represented in the flowcharts and / or described herein, for example, can be considered as a sequence of executable instructions stored in a computer readable medium, which can be executed by an instruction execution system, apparatus or device, such as a computer-based system, a processor-based system, or other system that can fetch the instructions from the instruction execution system, apparatus or device and execute the instructions, or a combination of them. For the purposes of this specification, a "computer readable medium" can be any apparatus that can contain, store, communicate, propagate, or transport the program for use by or in connection with the instruction execution system, apparatus or device. The computer readable medium can be a computer readable storage medium or a computer readable signal medium. The computer readable storage medium can include, but is not limited to, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or a propagation medium. The computer readable signal medium can include, but is not limited to, a computer readable medium that facilitates transfer of the program from one place to another. A specific example of a computer readable medium is a non-transitory computer-readable storage medium. A specific example of a computer readable signal medium is a source or destination of the computer readable medium. Another specific example of a computer readable signal medium is a computer readable signal travelling through space. Thus, a computer readable medium can take many forms of hardware to carry out the program for use by or in connection with the instruction execution system, apparatus or device.

[0169] It should be understood that aspects of the application can be implemented in hardware, software, firmware or combinations thereof. In the above embodiments, the N steps or methods can be implemented in software or firmware stored in a memory and executed by a suitable instruction execution system. As such, if implemented in hardware, and in another embodiment, any of the following technologies, known in the art, or their combinations can be used: discrete logic circuitry having logic gates for implementing logic functions on data signals, application specific integrated circuits having appropriate combinational logic gates, programmable gate arrays (PGA), field programmable gate arrays (FPGA), and the like.

[0170] Those skilled in the art can understand that all or part of the steps carried out by the above-mentioned embodiment methods can be completed by programs instructing related hardware, and the programs can be stored in a computer readable storage medium. When the programs are executed, one or a combination of the steps of the method embodiments is included.

[0171] In addition, each of the functional units in the various embodiments of the present application can be integrated in one processing module, or each of the units can be physically present separately, or two or more units can be integrated in one module. The integrated module can be realized in the form of hardware or in the form of a software functional module. When the integrated module is realized in the form of a software functional module and sold or used as an independent product, it can also be stored in a computer readable storage medium.

[0172] The storage medium mentioned above can be a read-only memory, a magnetic disk or an optical disk, etc. Although the embodiments of the present application have been shown and described above, it should be understood that the above embodiments are exemplary and should not be construed as limiting the present application, and those skilled in the art can make changes, modifications, replacements and variations to the above embodiments within the scope of the present application.

Claims

1. A method for analyzing the dynamic response of an elastic support structure for a gas turbine cylinder, characterized in that, Includes the following steps: Based on at least one of the isotropic loss factor, structural dimensions, elastic modulus, and applied load of the gas turbine cylinder elastic support structure, generate the finite element simulation results of the gas turbine cylinder elastic support structure, and extract the dynamic response output data of the target machine learning model from the finite element simulation results; Based on the dynamic response output data, and according to multiple parameters characterizing the elastic support structure of the gas turbine cylinder and at least one external load parameter, a firefly optimization algorithm-backpropagation FA-BP neural network model is constructed. The FA-BP neural network model is trained using the training set to obtain the trained FA-BP neural network model; The trained FA-BP neural network model is evaluated based on at least one of the evaluation metrics, namely, mean absolute percentage error and coefficient of determination, to obtain the evaluation result. After the evaluation result meets the preset conditions, the dynamic response result of the gas turbine cylinder elastic support structure is generated. The step of extracting the output of the preset machine learning model from the finite element simulation results includes: Extract the stress values ​​that satisfy the preset maximum conditions from multiple three-dimensional intercept points, and predict the multi-order natural frequencies of the elastic support structure of the gas turbine cylinder; Based on the stress values ​​that satisfy the preset maximum conditions at the multiple three-dimensional intercept points and the multi-order natural frequencies of the elastic support structure of the gas turbine cylinder, the dynamic response output data of the target machine learning model is determined.

2. The method according to claim 1, characterized in that, The firefly optimization algorithm-backpropagation (FA-BP) neural network model is constructed based on multiple parameters characterizing the elastic support structure of the gas turbine cylinder and at least one external load parameter, including: The data in the dataset is normalized to obtain the processed data; Based on the processed data, an input node is generated according to the parameters of the gas turbine cylinder elastic support structure and the at least one external load parameter, and an output node is generated according to the stress value that satisfies the preset maximum condition and the multi-order natural frequency among the multiple three-dimensional intercept points. The FA-BP neural network model is constructed based on the input node and the output node.

3. The method according to claim 1, characterized in that, The formula for calculating the mutual attraction in the firefly optimization algorithm-backpropagation FA-BP neural network model is as follows: ， in, for Initial attraction at that time The light absorption coefficient is... The Euclidean distance between individual fireflies; The formula for calculating the Euclidean distance between individual fireflies is: ， in, express Fireflies in 3D Location, express Fireflies in 3D Location, express 3D space; The formula for updating the firefly's position is: ， in, For fireflies with even lower brightness j The position, rand() is a random perturbation term, Step size factor t This represents the number of iterations.

4. The method according to claim 1, characterized in that, The formula for calculating the mean absolute percentage error is: ， in, For the number of data points, For the true value, This is a predicted value; The formula for calculating the coefficient of determination is as follows: ， in, The number of data items. For the true value, The predicted value, This is the mean of all true values.

5. A dynamic response analysis device for an elastic support structure of a gas turbine cylinder, characterized in that, include: The extraction module is used to generate finite element simulation results of the gas turbine cylinder elastic support structure based on at least one of the isotropic loss factor, structural dimensions, elastic modulus and applied load, and extract the dynamic response output data of the target machine learning model from the finite element simulation results. The construction module is used to construct a firefly optimization algorithm-backpropagation FA-BP neural network model based on the dynamic response output data, according to multiple parameters characterizing the elastic support structure of the gas turbine cylinder and at least one external load parameter; The training module is used to train the FA-BP neural network model using the training set to obtain the trained FA-BP neural network model. An evaluation module is used to evaluate the trained FA-BP neural network model based on at least one of the evaluation indicators, namely, mean absolute percentage error and coefficient of determination, to obtain evaluation results, and to generate the dynamic response results of the gas turbine cylinder elastic support structure after the evaluation results meet preset conditions. The extraction module includes: An extraction unit is used to extract stress values ​​that meet preset maximum conditions from multiple three-dimensional intercept points and predict the multi-order natural frequencies of the gas turbine cylinder elastic support structure. The determining unit is used to determine the dynamic response output data of the target machine learning model based on the stress values ​​that satisfy the preset maximum conditions among the multiple three-dimensional intercept points and the multi-order natural frequencies of the elastic support structure of the gas turbine cylinder.

6. An electronic device, characterized in that, include: The system includes 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 a dynamic response analysis method for an elastic support structure of a gas turbine cylinder as described in any one of claims 1-4.

7. A computer-readable storage medium having a computer program stored thereon, characterized in that, The program is executed by the processor to implement the dynamic response analysis method for an elastic support structure of a gas turbine cylinder as described in any one of claims 1-4.

8. A computer program product, comprising a computer program, characterized in that, The computer program is executed to implement a dynamic response analysis method for a gas turbine cylinder elastic support structure as described in any one of claims 1-4.

Images