Global time-varying load prediction method, system and equipment for high-temperature rotating part

Abstract

The invention discloses a global time-varying load prediction method, system and device for a high-temperature rotating component, and relates to the technical field of mechanical engineering and computer scientificity, and the method comprises the steps: obtaining a target working condition parameter of a to-be-measured rotating component, calculating a local outlier factor of the target working condition parameter, and calculating the local outlier factor of the to-be-measured rotating component; judging whether the local outlier factor meets a first preset condition or not to obtain a first judgment result; when the first judgment result is yes, a plurality of reference working condition parameters matched with the target working condition parameters are searched in an agent model library, and the target working condition parameters are predicted according to the global time-varying load prediction model of the corresponding high-temperature rotating part and the time-varying load prediction model of the dangerous position; and determining a global time-varying load prediction value with a dangerous position of the to-be-detected rotating component corresponding to the target working condition parameter. According to the method, the efficiency and the accuracy of predicting the time-varying load of the rotating part are improved.

Description

Technical Field

[0001] The present application relates to the fields of mechanical engineering and computer science and technology, and in particular to a method, system and device for predicting global time-varying loads of high-temperature rotating components. Background Art

[0002] In the fields of aerospace, energy and power, mechanical engineering, etc., there are generally rotating parts that are subjected to high temperatures. Under the interaction of heat and force, rotating parts are often the key to equipment system failure and the focus of life assessment. With the increasingly stringent operating conditions and the continuous improvement of energy efficiency and reliability requirements, it is particularly important to accurately predict the stress and temperature field of rotating parts. These two parameters are not only related to the life assessment and failure prediction of components, but also directly related to changes in material properties and potential mechanical failure risks.

[0003] With the development of computer technology, domestic and foreign scholars have applied some order reduction methods to the construction of time-varying load fields, such as Proper Orthogonal Decomposition (POD) and Galerkin projection method. These technologies still have certain limitations. They are mainly manifested in the following aspects: 1) The prediction accuracy of complex areas such as stress mutation is insufficient. The constructed order reduction model does not consider the correlation of data in the time dimension; 2) The global model has a large demand for computing resources and cannot be flexibly applied under multiple working conditions. Ensuring the applicability of the global model under multiple working conditions requires complex models and a lot of training time. Summary of the invention

[0004] The purpose of the present application is to provide a global time-varying load prediction method, system and device for high-temperature rotating parts, which improves the efficiency and accuracy of the time-varying load prediction of rotating parts.

[0005] To achieve the above objectives, this application provides the following solutions:

[0006] In a first aspect, the present application provides a global time-varying load prediction method for a high-temperature rotating component, the global time-varying load prediction method for a high-temperature rotating component comprising:

[0007] Obtaining target operating parameters of the rotating component to be tested;

[0008] Calculating a local outlier factor of a target operating condition parameter, and determining whether the local outlier factor satisfies a first preset condition, to obtain a first determination result;

[0009] When the first judgment result is yes, searching for multiple reference operating condition parameters matching the target operating condition parameters in the proxy model library, and using the global time-varying load prediction model of the high-temperature rotating component corresponding to the multiple reference operating condition parameters and the time-varying load prediction model of the dangerous position of the rotating component, determine the global time-varying load prediction value with the dangerous position of the rotating component to be tested corresponding to the target operating condition parameter;

[0010] The proxy model library includes a global time-varying load prediction model of a high-temperature rotating component and a time-varying load prediction model of a dangerous position of a rotating component corresponding to a plurality of different operating parameters;

[0011] The global time-varying load prediction model of the high-temperature rotating component is determined based on a first training set, a multivariate regression model, reduced-order singular values, and reduced-order right singular vectors; the first training set includes sample operating parameters and corresponding reduced-order left singular vector sample values; the reduced-order left singular vector sample values, reduced-order singular values, and reduced-order right singular vectors are obtained by sequentially performing finite element simulation, singular value decomposition, and reduced-order on multiple sample operating parameters of the rotating component;

[0012] The time-varying load prediction model for the dangerous position of the rotating component is obtained by training the time convolution network model using the second training set; the second training set includes sample operating condition parameters and corresponding time-varying load sample values ​​of the dangerous position; wherein the time-varying load sample values ​​of the dangerous position are obtained by sequentially performing finite element simulation on multiple sample operating condition parameters of the rotating component; the dangerous position is the position where the time-varying load exceeds a preset threshold in the finite element simulation results.

[0013] In a second aspect, the present application provides a global time-varying load prediction system for a high-temperature rotating component, the global time-varying load prediction system for a high-temperature rotating component is used to implement the global time-varying load prediction method for a high-temperature rotating component according to any one of claims 1 to 6, and the global time-varying load prediction system for a high-temperature rotating component comprises:

[0014] A data acquisition unit, used to acquire target operating parameters of the rotating component to be tested;

[0015] A first judgment result determination unit, used for calculating a local outlier factor of a target operating condition parameter, and judging whether the local outlier factor satisfies a first preset condition, to obtain a first judgment result;

[0016] a time-varying load prediction value determination unit, configured to, when the first judgment result is yes, search for a plurality of reference operating condition parameters matching the target operating condition parameters in the proxy model library, and determine a global time-varying load prediction value with a dangerous position of the rotating component to be tested corresponding to the target operating condition parameters by using a global time-varying load prediction model of a high-temperature rotating component and a time-varying load prediction model of a dangerous position of the rotating component corresponding to the plurality of reference operating condition parameters;

[0017] The proxy model library includes a global time-varying load prediction model of a high-temperature rotating component and a time-varying load prediction model of a dangerous position of a rotating component corresponding to a plurality of different operating parameters;

[0018] The global time-varying load prediction model of the high-temperature rotating component is determined based on a first training set, a multivariate regression model, reduced-order singular values, and reduced-order right singular vectors; the first training set includes sample operating parameters and corresponding reduced-order left singular vector sample values; the reduced-order left singular vector sample values, reduced-order singular values, and reduced-order right singular vectors are obtained by sequentially performing finite element simulation, singular value decomposition, and reduced-order on multiple sample operating parameters of the rotating component;

[0019] The time-varying load prediction model for the dangerous position of the rotating component is obtained by training the time convolution network model using the second training set; the second training set includes sample operating condition parameters and corresponding time-varying load sample values ​​of the dangerous position; wherein the time-varying load sample values ​​of the dangerous position are obtained by sequentially performing finite element simulation on multiple sample operating condition parameters of the rotating component; the dangerous position is the position where the time-varying load exceeds a preset threshold in the finite element simulation results.

[0020] In a third aspect, the present application provides a computer device comprising: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement any of the above-described methods for predicting global time-varying loads of high-temperature rotating components.

[0021] According to the specific embodiments provided in this application, this application has the following technical effects:

[0022] The present application discloses a global time-varying load prediction method, system and device for high-temperature rotating parts. By constructing a proxy model library, the global time-varying load prediction model of high-temperature rotating parts under corresponding working condition parameters and the time-varying load prediction model of the dangerous position of the rotating parts are combined to predict the global time-varying load with dangerous positions of the rotating parts to be tested corresponding to the target working condition parameters. Among them, the constructed proxy model library contains proxy models under multiple working condition parameters, and the existing proxy models in the proxy model library can be used to quickly monitor the time-varying load of the target parameter working condition online. At the same time, due to the addition of the time-varying load prediction model of the dangerous position of the rotating parts, the time-varying load of the dangerous position can be accurately predicted, which further improves the accuracy of the global time-varying load prediction of high-temperature rotating parts. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0024] Figure 1 A schematic flow chart of a method for predicting global time-varying loads of high-temperature rotating components provided in one embodiment of the present application;

[0025] Figure 2 A schematic diagram of operating parameters provided for another embodiment of the application;

[0026] Figure 3 A schematic diagram of the construction and prediction process of a global time-varying load prediction model for a high-temperature rotating component provided by another embodiment of the present application;

[0027] Figure 4 A schematic diagram of a dangerous location determined based on finite element simulation provided by another embodiment of the present application;

[0028] Figure 5 A schematic diagram of constructing a temporal convolutional network provided in another embodiment of the present application;

[0029] Figure 6 A visualization diagram of a local outlier factor of a target operating condition in a proxy model library provided by another embodiment of the present application;

[0030] Figure 7 A schematic diagram of the structure of a computer device provided in one embodiment of the present application. DETAILED DESCRIPTION

[0031] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0032] In recent years, Finite Element Analysis (FEA) has become a common tool for evaluating the dynamic behavior and thermodynamic response of rotating parts. However, due to its high dimensionality and computationally intensive characteristics, traditional finite element models are unable to cope with complex working conditions and a large number of design iterations, especially under the requirements of real-time monitoring and rapid response. The present application provides a global time-varying load prediction method, system and device for high-temperature rotating parts, which can efficiently and accurately realize the time-varying load prediction of rotating parts.

[0033] In order to make the above-mentioned objects, features and advantages of the present application more obvious and easy to understand, the present application is further described in detail below in conjunction with the accompanying drawings and specific implementation methods.

[0034] In an exemplary embodiment, Figure 1 As shown, a global time-varying load prediction method for a high-temperature rotating component is provided, comprising the following steps S1 to S3. Among them:

[0035] Step S1, obtaining target operating parameters of the rotating component to be tested. Specifically, the present application mainly performs global time-varying load prediction on the rotating component in a high temperature and harsh environment.

[0036] Step S2, calculate the local outlier factor of the target operating condition parameter, and determine whether the local outlier factor satisfies a first preset condition, to obtain a first determination result; the first preset condition is that the local outlier factor is close to 1. If the local outlier factor is close to 1, it means that the existing time-varying load prediction model of the rotating component can be used for interpolation to perform time-varying load prediction for the new operating condition (target operating condition).

[0037] Step S3, when the first judgment result is yes, search for multiple reference operating condition parameters that match the target operating condition parameters in the proxy model library, and use the global time-varying load prediction model of the high-temperature rotating component corresponding to the multiple reference operating condition parameters and the time-varying load prediction model of the dangerous position of the rotating component to be tested to determine the global time-varying load prediction value with the dangerous position of the rotating component to be tested corresponding to the target operating condition parameters.

[0038] The proxy model library includes global time-varying load prediction models of high-temperature rotating parts and time-varying load prediction models of dangerous positions of rotating parts corresponding to multiple different operating parameters.

[0039] When constructing the proxy model library, according to the sampling of the working condition parameter combination, the global time-varying load prediction model of the high-temperature rotating parts corresponding to multiple working conditions and the time-varying load prediction model of the dangerous position of the rotating parts (i.e., the proxy model) are trained, and the working condition parameters are used as the key and the constructed proxy model is used as the value, and saved to the proxy model library. The corresponding proxy model can be found through the working condition parameters. The proxy model in the proxy model library can be used to quickly predict the time-varying load data of all nodes of the finite element model, and the time-varying load data of the dangerous position can be accurately predicted.

[0040] The global time-varying load prediction model of high-temperature rotating components is determined based on a first training set, a multivariate regression model, reduced-order singular values ​​and reduced-order right singular vectors; the first training set includes sample operating condition parameters and corresponding reduced-order left singular vector sample values; the reduced-order left singular vector sample values, reduced-order singular values ​​and reduced-order right singular vectors are obtained by performing finite element simulation, singular value decomposition and reduction in sequence on multiple sample operating condition parameters of the rotating components.

[0041] The time-varying load prediction model for the dangerous position of the rotating component is obtained by training the time convolution network model using the second training set; the second training set includes sample operating condition parameters and corresponding time-varying load sample values ​​of the dangerous position; wherein the time-varying load sample values ​​of the dangerous position are obtained by sequentially performing finite element simulation on multiple sample operating condition parameters of the rotating component; the dangerous position is the position where the time-varying load exceeds a preset threshold in the finite element simulation results.

[0042] As an optional implementation, in step S2, calculating the local outlier factor of the target operating condition parameter specifically includes:

[0043] Step S21, calculate the reachable distance of the local sample point (ie, the target operating condition parameter), first calculate the distance between the nearest neighbor points of the local sample point (ie, the sample operating condition parameter combination), the formula is as follows:

[0044] dist k* (o,P)=max{d k* (o),d(o,P)}(1)

[0045] Among them, o is a local sample point, P is a point in the k*th nearest neighbor set of point o, and dist k* (o,P) is the reachable distance of the local sample point o; d k* (o) is the k*th distance of the local sample point o; d(o,P) is the distance between the k*th point P and the local sample point o.

[0046] Step S22, calculate the local reachability density, the formula is as follows:

[0047]

[0048] Among them, ld k* (o) is the local reachable density of the local sample point o; N k* (o) represents the k*th distance neighborhood of the local sample point o, which represents the set of all points within the k*th distance of the local sample point o.

[0049] Step S23, calculate the local outlier factor, the formula is as follows:

[0050]

[0051] Among them, lof k* (o) is the local outlier factor of the local sample point o. If the ratio is greater than 1, it means that the density of the sample point o is smaller than the density of its surrounding points, and it is necessary to rebuild the time-varying load prediction model of rotating components to fill the sample space; if the ratio is less than 1, it means that the density of the local sample point o is greater than the density of its surrounding points, and interpolation calculation can be performed directly in the sample space.

[0052] As an optional implementation, the global time-varying load prediction method for high-temperature rotating components further includes:

[0053] Step S4: when the first judgment result is no, judging whether the local outlier factor meets a second preset condition, and obtaining a second judgment result. The second preset condition is that the local outlier factor is much greater than 1.

[0054] Step S5, when the second judgment result is yes, finite element simulation is performed based on the target operating condition parameters to obtain the time-varying load values ​​of all position points on the rotating component to be measured. When the local outlier factor does not meet the first preset condition and the second preset condition, finite element simulation is performed again on the target operating condition parameters, and the corresponding time-varying load values ​​of all position points on the rotating component to be measured are obtained using the finite element simulation results.

[0055] As an optional implementation, in step S3, multiple reference operating condition parameters matching the target operating condition parameters are searched in the proxy model library, and the global time-varying load prediction model of the high-temperature rotating component corresponding to the multiple reference operating condition parameters and the time-varying load prediction model of the dangerous position of the rotating component are used to determine the global time-varying load prediction value with the dangerous position of the rotating component to be tested corresponding to the target operating condition parameter, specifically including:

[0056] Step S31, using the k-nearest neighbor algorithm, searching the proxy model library for operating parameters that are close to the target operating parameters, and obtaining multiple reference operating parameters that match the target operating parameters. The calculation formula is as follows:

[0057]

[0058] Where X represents the target operating condition parameter; X j represents the jth reference condition parameter; s represents the dimension of the data, x δ The value of the δ-dimensional data representing the target operating condition parameters, represents the value of the δth dimension data of the jth sample point (i.e., reference operating condition parameter) in the sample space, a δ Represents the weight of the δth dimension data.

[0059] Step S32, determining the distance weight of the corresponding reference operating condition parameter according to the distance between the target operating condition parameter and each reference operating condition parameter; that is, assigning different distance weights to the reference operating condition parameters according to the distance between each target operating condition parameter and each reference operating condition parameter.

[0060] Step S33, determining the global time-varying load prediction value of the high-temperature rotating component corresponding to each reference operating condition parameter according to the global time-varying load prediction model of the high-temperature rotating component corresponding to each reference operating condition parameter.

[0061] Step S34, determining the predicted time-varying load value of the dangerous position of the rotating component corresponding to each reference operating condition parameter according to the predicted time-varying load model of the dangerous position of the rotating component corresponding to each reference operating condition parameter.

[0062] Step S35, based on the global time-varying load prediction value of the high-temperature rotating component corresponding to each reference operating condition parameter and the time-varying load prediction value of the dangerous position of the rotating component corresponding to each reference operating condition parameter, obtain the global time-varying load prediction value with the dangerous position of the rotating component corresponding to each reference operating condition parameter; that is, combine the global time-varying load prediction value of the high-temperature rotating component corresponding to each reference operating condition parameter with the time-varying load prediction value of the dangerous position, replace the time-varying load of the corresponding dangerous position output by the global time-varying load prediction model with the time-varying load prediction value of the dangerous position, and realize accurate prediction of the time-varying load data of the dangerous position.

[0063] Step S36, based on the global time-varying load prediction value with dangerous position of the rotating component corresponding to each reference operating condition parameter and the corresponding distance weight, determine the global time-varying load prediction value with dangerous position of the rotating component to be tested corresponding to the target operating condition parameter.

[0064] Specifically, the time-varying load field data under the target working condition is obtained by distance weighted interpolation, and the formula is as follows:

[0065]

[0066] Where m represents the number of reference conditions, x i is the predicted value of the global time-varying load with the ith dangerous location; ai represents the distance weight of the i-th reference working condition parameter; x′ is the global time-varying load prediction result with dangerous positions obtained by weighted interpolation of the global time-varying load prediction model with dangerous positions of multiple rotating parts.

[0067] As an optional implementation, in step S3, the training process of the global time-varying load prediction model of the high-temperature rotating component specifically includes:

[0068] Step S311, obtaining a first training sample.

[0069] Step S312: input the first training sample into a multiple regression model, and output the predicted value of the reduced left singular vector. The multiple regression model is trained according to the input operating condition parameters and the truncated left singular vector as input and output respectively, and the mapping of the operating condition parameters to the left singular vector is realized through the multiple regression model.

[0070] Step S313, constructing a first loss function according to the reduced left singular vector prediction value and the reduced left singular vector sample value, and adjusting the parameters of the multivariate regression model according to the first loss function to obtain a trained multivariate regression model.

[0071] Step S314, performing matrix operations on the reduced left singular vector prediction values, the reduced singular values, and the reduced right singular vectors to obtain a global time-varying load prediction model for high-temperature rotating components, that is, reconstructing the reduced-order physical field matrix through the retained singular values ​​and their corresponding left and right singular vector matrices to obtain the time-varying load field of the finite element model of the rotating component under the corresponding working conditions for cloud map visualization.

[0072] As an optional implementation, in step S3, finite element simulation, singular value decomposition and order reduction are sequentially performed on multiple sample operating condition parameters of the rotating component, specifically including:

[0073] Step S321 , performing finite element simulation based on a plurality of sample operating condition parameters of the rotating component, and obtaining finite element simulation results of time-varying loads of the rotating component under different sample operating condition parameters.

[0074] Specifically, according to the physical data of the rotating component to be tested, a geometric model of the finite element rotating component is established; according to the range of working condition parameters input by the finite element simulation, a sampling method is used to generate a working condition data group, and a finite element simulation of the corresponding time-varying load is performed. Among them, the working condition data group is used for model simulation under different working conditions. The physical data includes load data, geometric dimensions and material data of the rotating component to be tested; the load data includes stress strain data and temperature data.

[0075] Specifically, finite element modeling is performed in Abaqus finite element analysis software to complete steps such as geometric modeling, material property assignment, meshing, load application and boundary conditions, so as to obtain a finite element simulation model of the rotating component.

[0076] Step S322, according to the finite element simulation results of the time-varying load of the rotating component under different sample operating parameters, the time-varying load value corresponding to each grid unit in all analysis steps of the rotating component under the corresponding sample operating parameters is obtained, and the physical field matrix of the rotating component under the corresponding sample operating parameters is constructed based on the time-varying load value corresponding to each grid unit in all analysis steps.

[0077] Specifically, the constructed physical field matrix is ​​as follows:

[0078]

[0079] Among them, L is the selected analysis step; N is the number of grid nodes, and A is the constructed physical field matrix.

[0080] Step S323, performing singular value decomposition on the physical field matrix of the rotating component under each sample operating parameter to obtain a decomposed singular value vector matrix of the rotating component under each sample operating parameter; the decomposed singular value vector matrix includes: singular values, left singular vectors and right singular vectors.

[0081] The physical field matrix is ​​subjected to singular value decomposition according to the singular value decomposition formula, and the formula is as follows:

[0082] A=UΣV T (7)

[0083] Among them, its shape is m×n; U is a square matrix of shape m×m, each of which is called a left singular vector; Σ is an m×n matrix, except for the non-zero elements on the diagonal, the elements in other positions are zero; V T is the transposed matrix of V, which has a shape of n×n, and each orthogonal vector in it is called a right singular vector.

[0084] Step S324, reducing the order of the decomposed singular value vector matrix of the rotating component under each sample operating parameter to obtain the reduced order singular values, reduced order left singular vector sample values ​​and reduced order right singular vectors of the rotating component under each sample operating parameter.

[0085] Among them, according to the energy proportion of the singular value elements after decomposition, the first k1 elements whose energy proportion is greater than 99.99% of the total energy are selected, and the singular value vector matrix after the original physical field decomposition is truncated (i.e., reduced in order) according to the k1 value. At the same time, the product of the truncated singular value and the corresponding left and right singular value vectors is used to approximate the original physical field matrix. The formula is as follows:

[0086]

[0087] Among them, k1 is the order taken; is the matrix composed of the reduced left singular vectors; is the reduced singular value matrix, which is a diagonal matrix; is the matrix of the right singular vectors after the reduction and then the transposed matrix. Then, the left singular vector matrix Each row of Right Singular Vector Matrix By performing matrix operations, we can obtain the time-varying load value of each grid node in the finite element model at the corresponding time.

[0088] As an optional implementation, in step S3, the training process of the time-varying load prediction model of the dangerous position of the rotating component specifically includes:

[0089] Step S331, obtaining a second training sample. According to the finite element simulation results of the time-varying load under the corresponding working condition, the position points where the time-varying load exceeds the preset threshold are selected as dangerous positions, and the time-varying load data of these dangerous positions are extracted separately as the time-varying load sample values ​​of the dangerous positions. According to the working condition input corresponding to each finite element analysis step, the input is packaged in a time window, that is, the working condition parameter input of the current analysis step and the working condition parameter input corresponding to several previous analysis steps constitute the input of a time convolution network to consider the mutual correlation of data in the time dimension, and the extracted time-varying load data of the dangerous position is used as the output in combination with the sliding window method to constitute the training data of the time convolution network (i.e., the second training sample).

[0090] Step S332: input the second training sample into the time convolutional network model, and output the time-varying load prediction value of the dangerous position.

[0091] Step S333, construct a second loss function according to the time-varying load prediction value of the dangerous position and the time-varying load sample value of the dangerous position, and use the second loss function to adjust the parameters of the time convolution network model to obtain a time-varying load prediction model for the dangerous position of the rotating component. Wherein, by adjusting the corresponding hyperparameters in real time, when the second loss function reaches the set value or the training round reaches the maximum value, the training is stopped, and the corresponding model and parameters are saved.

[0092] Beneficial effects of this application:

[0093] When the present application is applied, the local outlier factor corresponding to the new working condition is calculated. If it is close to 1, the k-nearest neighbor algorithm is used to find several groups of proxy models with similar working condition parameters in the proxy model database, and weighted interpolation is performed by distance to obtain the corresponding time-varying load prediction value under the new working condition parameters. The present application establishes a proxy model database by combining singular value decomposition, time convolution network, k-nearest neighbor algorithm, etc., to achieve efficient and accurate prediction of the time-varying load field data of the rotating parts, which can greatly shorten the calculation time and realize the online monitoring of the time-varying load data of the rotating parts.

[0094] A specific embodiment is listed below to further illustrate the global time-varying load prediction method for high-temperature rotating components of the present application.

[0095] Step 100: Establish a finite element component model based on the physical data of the rotating component to be tested; the physical data includes load data, geometric dimensions and rotating component material data; the load data includes stress strain data and temperature data. Among them, the rotating component to be tested is a steam turbine rotor, and the rotor is used in a steam turbine. For the high and medium pressure rotors of the steam turbine, its structure is axisymmetric, so it is only necessary to establish finite element modeling of half of the rotor in the axial section, CAX8RT grid unit, a total of 20972 units; the material is 30Cr1Mo1V. The operating parameters are: the initial temperature of the rotor is between 20 and 215°C, the initial temperature of the steam is between 320°C and 400°C, the steam heating rate is between 0.7 and 2°C / min, and the speed increase rate is between 20 and 100r / min. 2 , the load increase rate is 0.2~0.8% / min.

[0096] Step 200: Based on the variation range of the input working condition parameters of the rotating component to be tested, Latin hypercube sampling is used to generate multiple sets of input working condition parameters as inputs for finite element simulation. Figure 2 Specifically, the rotor initial temperature, the rotor speed increase rate, the rotor load increase rate, the steam initial temperature, and the steam heating rate are sampled to generate 50 groups of operating condition parameter combinations.

[0097] Step 300: According to the working condition parameter combination generated in step 200, set the boundary conditions of finite element simulation, perform finite element simulation, obtain the corresponding stress temperature field data (i.e., instantaneous variable load field data), and export it. At the same time, determine the dangerous position of the rotating part to be tested. The schematic diagram of the dangerous position is as follows: Figure 4 The exported data is shown in Table 1.

[0098] Table 1 Simulation result data for a specified working condition

[0099]

[0100] Step 400, combining the derived component stress and temperature field data in the order of the analysis steps to obtain a physical field matrix.

[0101] Step 500, perform singular value decomposition and order reduction on the physical field matrix, and finally truncate the singular values ​​according to the principle of singular value energy ratio, further reduce the order, and obtain singular values ​​and left singular value vector matrix, right singular value vector matrix, wherein the left singular value vector matrix includes left singular vector sample values.

[0102] Step 600: Use the sample operating condition parameters and the left singular vector sample values ​​as the first training samples to train a multivariate regression model.

[0103] Step 700, after obtaining the trained multivariate regression model, use the trained multivariate regression model to output the left singular vector prediction value, and then perform matrix operations on the reduced left singular vector prediction value, the reduced singular value, and the reduced right singular vector to obtain a reduced-order model (i.e., a global time-varying load prediction model for high-temperature rotating components), and obtain the time-varying load field of the finite element model of the rotating component under the corresponding working condition for cloud map visualization. Specifically, the construction and prediction process of the global time-varying load prediction model for high-temperature rotating components is as follows: Figure 3 shown.

[0104] Step 800, based on the derived time-varying load data of the dangerous position, combined with the input working condition data and the sliding window method, the data is packaged into a time frame to obtain a second training sample. The second training sample is used to train the temporal convolutional network, and the hyperparameters are adjusted to obtain the optimal model. Figure 4 As shown in the figure, the schematic diagram of the sliding window method to construct a temporal convolutional network is as follows Figure 5 As shown. The temporal convolutional network is a network model based on time series. The size of the time window will have a certain impact on the prediction results. In order to quantify the impact of different time windows on the load prediction results, the root mean square error (RMSE), mean absolute error (MAE), Pearson correlation coefficient (Pearson Coff), and determination coefficient (R-squared, R 2 ) is used as the evaluation index to further analyze the prediction results. The specific calculation formula is as follows:

[0105]

[0106]

[0107] Where n is the total number of sample conditions; i is the serial number of the sample condition; yi is the true value of the time-varying load; is the predicted value of the time-varying load, is the average value of the true value of the time-varying load; is the Pearson correlation coefficient value between the true value of the time-varying load and the predicted value of the time-varying load; is the standard deviation of the true value of the time-varying load, is the standard deviation of the predicted value of the time-varying load; cov is the covariance, E is the expectation, is the mean of the true values ​​of the time-varying load; is the mean of the predicted values ​​of the time-varying load.

[0108] The stress prediction results of different time window sizes are shown in Table 2.

[0109] Table 2 Time-varying load prediction results of different time window sizes (mean values ​​of indicators at different dangerous locations)

[0110] Time window size RMSE MAE Pearson Coff <![CDATA[R 2 ]]> 4 4.740466 3.144075 0.998462 0.996068 6 3.482561 2.300573 0.999188 0.997779 8 3.672514 2.921857 0.999430 0.997602 10 2.523119 1.813433 0.999613 0.998848 12 3.388574 2.589993 0.998605 0.997966

[0111] According to Table 2, the best effect is achieved when the time window size is 10. Therefore, the time window size of 10 is selected as the size of the second sample data package.

[0112] Step 900 , combining the global time-varying load prediction model of the high-temperature rotating component and the time-varying load prediction model of the dangerous position of the rotating component for predicting the time-varying load of the rotating component under different working conditions.

[0113] Step 1000, different operating condition parameters and corresponding proxy models are saved in a proxy model library, and the corresponding proxy model can be found through the operating condition parameters for subsequent calling.

[0114] Step 1100, when encountering new operating parameters that require time-varying load prediction, first calculate the local outlier factor of the operating parameters in the proxy model to determine whether to perform k-nearest neighbor interpolation or whether to add a corresponding proxy model. Specifically, ten groups of sample operating parameters are used to calculate the local outlier factor in the proxy model sample space, and the data of three dimensions of steam temperature rise rate, speed rise rate and load rise rate are used to distinguish different operating conditions. As shown in Table 3 below. The visualization results of the local outlier factor of the target operating condition in the proxy model library are shown in Figure 6 shown.

[0115] Table 3 Local outlier factors of test condition sample points

[0116]

[0117] Step 1200, according to the new operating condition parameters, the k-nearest neighbor algorithm is used to select the closest groups of operating conditions in the proxy model library as reference operating conditions, and the time-varying load prediction value of the rotating component to be tested under the new operating condition is obtained according to the distance weighted interpolation. Specifically, taking stress interpolation as an example, taking the 9th group of sample operating condition parameters as an example, the steam temperature rise rate of the startup condition is 1.395903℃ / min, and the speed rise rate is 39.07527r / min 2 , the load increase rate is 0.3525% / min. The reference operating condition information of the five groups of proxy models closest to the target operating condition parameters found by the k-nearest neighbor algorithm is shown in Table 4, including the steam temperature increase rate, speed increase rate, load increase rate and the Euclidean distance between them and the target operating condition parameters.

[0118] Table 4 Parameter information of the five groups of proxy models closest to the target working conditions

[0119]

[0120]

[0121] Step 1300, among multiple proxy models selected from the proxy model library, distance weighted interpolation is used to predict the time-varying load value under the target working condition parameters. Specifically, different numbers of groups k involved in interpolation also affect the interpolation results. By comparing the RMSE, MAE, Pearson Coff, R 2 , the results are shown in Table 5.

[0122] Table 5 Comparison of time-varying load prediction results at different k

[0123] k RMSE MAE Pearson Coff <![CDATA[R 2 ]]> 1 6.320601 5.586039 0.999032 0.994596 2 4.003973 3.215975 0.999520 0.997983 3 8.076193 7.234912 0.997815 0.991207 4 8.688017 7.735146 0.997704 0.989933 5 7.714432 6.883803 0.992140 0.998109

[0124] It can be seen that when k=2, the best interpolation effect is achieved, so in this embodiment, k=2 is used as the number of interpolation groups.

[0125] In summary, this embodiment uses the finite element simulation method instead of actually accumulating a certain amount of data, and establishes a proxy model database by generating a large number of operating condition parameter combinations, combining singular value decomposition reduction technology, time convolution network, etc., and interpolating the time-varying load field data of the components under the new operating conditions through the k-nearest neighbor algorithm, thereby realizing fast and accurate online monitoring of the time-varying loads of rotating components.

[0126] Based on the same inventive concept, the embodiment of the present application also provides a global time-varying load prediction system for high-temperature rotating parts for implementing the global time-varying load prediction method for high-temperature rotating parts involved above. The implementation scheme for solving the problem provided by the system is similar to the implementation scheme recorded in the above method, so the specific limitations in the embodiments of the global time-varying load prediction system for one or more high-temperature rotating parts provided below can be referred to the limitations of the global time-varying load prediction method for high-temperature rotating parts above, and will not be repeated here.

[0127] In an exemplary embodiment, a global time-varying load prediction system for a high-temperature rotating component is provided, comprising:

[0128] The data acquisition unit is used to acquire the target operating parameters of the rotating component to be tested.

[0129] The first judgment result determination unit is used to calculate the local outlier factor of the target operating condition parameter, and determine whether the local outlier factor meets a first preset condition to obtain a first judgment result.

[0130] A time-varying load prediction value determination unit is used to search for multiple reference operating condition parameters that match the target operating condition parameters in the proxy model library when the first judgment result is yes, and use the global time-varying load prediction model of the high-temperature rotating component corresponding to the multiple reference operating condition parameters and the time-varying load prediction model of the dangerous position of the rotating component to be tested to determine the global time-varying load prediction value with the dangerous position of the rotating component to be tested corresponding to the target operating condition parameters.

[0131] The proxy model library includes global time-varying load prediction models of high-temperature rotating parts and time-varying load prediction models of dangerous positions of rotating parts corresponding to multiple different operating parameters.

[0132] The global time-varying load prediction model of high-temperature rotating components is determined based on a first training set, a multivariate regression model, reduced-order singular values ​​and reduced-order right singular vectors; the first training set includes sample operating condition parameters and corresponding reduced-order left singular vector sample values; the reduced-order left singular vector sample values, reduced-order singular values ​​and reduced-order right singular vectors are obtained by performing finite element simulation, singular value decomposition and reduction in sequence on multiple sample operating condition parameters of the rotating components.

[0133] The time-varying load prediction model for the dangerous position of the rotating component is obtained by training the time convolution network model using the second training set; the second training set includes sample operating condition parameters and corresponding time-varying load sample values ​​of the dangerous position; wherein the time-varying load sample values ​​of the dangerous position are obtained by sequentially performing finite element simulation on multiple sample operating condition parameters of the rotating component; the dangerous position is the position where the time-varying load exceeds a preset threshold in the finite element simulation results.

[0134] In an exemplary embodiment, a computer device is provided, including: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement a global time-varying load prediction method for a high-temperature rotating component.

[0135] In an exemplary embodiment, a computer device is provided. The computer device may be a server or a terminal. The internal structure diagram thereof may be as follows: Figure 7 As shown. The computer device includes a processor, a memory, an input / output interface (Input / Output, referred to as I / O) and a communication interface. Among them, the processor, the memory and the input / output interface are connected through a system bus, and the communication interface is connected to the system bus through the input / output interface. Among them, the processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program and a database. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The input / output interface of the computer device is used to exchange information between the processor and an external device. The communication interface of the computer device is used to communicate with an external terminal through a network connection. When the computer program is executed by the processor, a global time-varying load prediction method for a high-temperature rotating component is implemented.

[0136] Those skilled in the art will understand that Figure 7 The structure shown in the figure is only a block diagram of a part of the structure related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied. The specific computer device may include more or fewer components than those shown in the figure, or combine certain components, or have a different arrangement of components.

[0137] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, stored data, displayed data, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of relevant data must comply with relevant regulations.

[0138] Those of ordinary skill in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to the memory, database or other medium used in the embodiments provided in the present application 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, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM may be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM).

[0139] The database involved in each embodiment provided in this application may include at least one of a relational database and a non-relational database. The non-relational database may include a distributed database based on blockchain, etc., but is not limited thereto. The processor involved in each embodiment provided in this application may be a general-purpose processor, a central processing unit, a graphics processor, a digital signal processor, a programmable logic device, a data processing logic device based on quantum computing, etc., but is not limited thereto.

[0140] The technical features of the above embodiments may be combined arbitrarily. To make the description concise, 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 specification.

[0141] This article uses specific examples to illustrate the principles and implementation methods of this application. The description of the above embodiments is only used to help understand the method and core ideas of this application. At the same time, for those skilled in the art, according to the ideas of this application, there will be changes in the specific implementation methods and application scope. In summary, the content of this specification should not be understood as limiting this application.

Claims

1. A global time-varying load prediction method for high-temperature rotating parts, characterized in that: The global time-varying load prediction method for high-temperature rotating components includes: Obtaining target operating parameters of the rotating component to be tested; Calculating a local outlier factor of a target operating condition parameter, and determining whether the local outlier factor satisfies a first preset condition, to obtain a first determination result; When the first judgment result is yes, searching for multiple reference operating condition parameters matching the target operating condition parameters in the proxy model library, and using the global time-varying load prediction model of the high-temperature rotating component corresponding to the multiple reference operating condition parameters and the time-varying load prediction model of the dangerous position of the rotating component, determine the global time-varying load prediction value with the dangerous position of the rotating component to be tested corresponding to the target operating condition parameter; The proxy model library includes a global time-varying load prediction model of a high-temperature rotating component and a time-varying load prediction model of a dangerous position of a rotating component corresponding to a plurality of different operating parameters; The global time-varying load prediction model of the high-temperature rotating component is determined based on a first training set, a multivariate regression model, reduced-order singular values, and reduced-order right singular vectors; the first training set includes sample operating parameters and corresponding reduced-order left singular vector sample values; the reduced-order left singular vector sample values, reduced-order singular values, and reduced-order right singular vectors are obtained by sequentially performing finite element simulation, singular value decomposition, and reduced-order on multiple sample operating parameters of the rotating component; The time-varying load prediction model for the dangerous position of the rotating component is obtained by training the time convolution network model using the second training set; the second training set includes sample operating condition parameters and corresponding time-varying load sample values ​​of the dangerous position; wherein the time-varying load sample values ​​of the dangerous position are obtained by sequentially performing finite element simulation on multiple sample operating condition parameters of the rotating component; the dangerous position is the position where the time-varying load exceeds a preset threshold in the finite element simulation results.

2. The global time-varying load prediction method for high-temperature rotating parts according to claim 1, characterized in that: The global time-varying load prediction method for high-temperature rotating components also includes: When the first judgment result is no, judging whether the local outlier factor satisfies a second preset condition, and obtaining a second judgment result; When the second judgment result is yes, a finite element simulation is performed based on the target operating condition parameters to obtain time-varying load values ​​of all position points on the rotating component to be measured.

3. The global time-varying load prediction method for high-temperature rotating parts according to claim 1, characterized in that: Searching for multiple reference operating condition parameters matching the target operating condition parameters in the proxy model library, and using the global time-varying load prediction model of the high-temperature rotating component corresponding to the multiple reference operating condition parameters and the time-varying load prediction model of the dangerous position of the rotating component, determine the global time-varying load prediction value with the dangerous position of the rotating component to be tested corresponding to the target operating condition parameters, specifically including: Using the k-nearest neighbor algorithm, the operating condition parameters close to the target operating condition parameters are searched in the proxy model library to obtain multiple reference operating condition parameters matching the target operating condition parameters; According to the distance between the target operating condition parameter and each reference operating condition parameter, a distance weight of the corresponding reference operating condition parameter is determined; Determine the global time-varying load prediction value of the high-temperature rotating component corresponding to each reference operating condition parameter according to the global time-varying load prediction model of the high-temperature rotating component corresponding to each reference operating condition parameter; Determine the predicted time-varying load value of the dangerous position of the rotating component corresponding to each reference operating condition parameter according to the predicted time-varying load model of the dangerous position of the rotating component corresponding to each reference operating condition parameter; Based on the global time-varying load prediction value of the high-temperature rotating component corresponding to each reference operating condition parameter and the time-varying load prediction value of the dangerous position of the rotating component corresponding to each reference operating condition parameter, a global time-varying load prediction value with a dangerous position of the rotating component corresponding to each reference operating condition parameter is obtained; Based on the global time-varying load prediction value with a dangerous position of the rotating component corresponding to each reference operating condition parameter and the corresponding distance weight, the global time-varying load prediction value with a dangerous position of the rotating component to be tested corresponding to the target operating condition parameter is determined.

4. The global time-varying load prediction method for high-temperature rotating parts according to claim 1, characterized in that: The training process of the global time-varying load prediction model for high-temperature rotating components includes: Obtain a first training sample; Inputting the first training sample into a multivariate regression model, and outputting a reduced-order left singular vector prediction value; Constructing a first loss function according to the predicted value of the reduced left singular vector and the sample value of the reduced left singular vector, and adjusting the parameters of the multivariate regression model according to the first loss function to obtain a trained multivariate regression model; Matrix operations are performed on the predicted values ​​of the reduced left singular vectors, the reduced singular values ​​and the reduced right singular vectors to obtain a global time-varying load prediction model for high-temperature rotating components.

5. The global time-varying load prediction method for high-temperature rotating parts according to claim 4, characterized in that: Finite element simulation, singular value decomposition and order reduction are performed on multiple sample working condition parameters of rotating parts in turn, including: Finite element simulation is performed based on multiple sample working condition parameters of the rotating component to obtain finite element simulation results of time-varying loads of the rotating component under different sample working condition parameters; According to the finite element simulation results of the time-varying load of the rotating component under different sample working conditions, the time-varying load value corresponding to each grid unit of the rotating component under all analysis steps under the corresponding sample working conditions is obtained, and the physical field matrix of the rotating component under the corresponding sample working conditions is constructed based on the time-varying load value corresponding to each grid unit under all analysis steps; Performing singular value decomposition on the physical field matrix of the rotating component under each sample operating condition parameter to obtain a decomposed singular value vector matrix of the rotating component under each sample operating condition parameter; the decomposed singular value vector matrix includes: singular values, left singular vectors and right singular vectors; The decomposed singular value vector matrix of the rotating component under each sample operating condition parameter is reduced in order to obtain the reduced singular values ​​of the rotating component under each sample operating condition parameter, the reduced left singular vector sample values ​​and the reduced right singular vectors.

6. The global time-varying load prediction method for high-temperature rotating parts according to claim 1, characterized in that: The training process of the time-varying load prediction model for hazardous locations of rotating components includes: Obtain a second training sample; Inputting the second training sample into the time convolutional network model, and outputting the time-varying load prediction value of the dangerous position; A second loss function is constructed according to the time-varying load prediction value of the dangerous position and the time-varying load sample value of the dangerous position, and the parameters of the time convolutional network model are adjusted using the second loss function to obtain a time-varying load prediction model for the dangerous position of the rotating component.

7. A global time-varying load prediction system for high-temperature rotating parts, characterized in that: The global time-varying load prediction system for high-temperature rotating parts is used to implement the global time-varying load prediction method for high-temperature rotating parts according to any one of claims 1 to 6, and the global time-varying load prediction system for high-temperature rotating parts includes: A data acquisition unit, used to acquire target operating parameters of the rotating component to be tested; A first judgment result determination unit, used for calculating a local outlier factor of a target operating condition parameter, and judging whether the local outlier factor satisfies a first preset condition, to obtain a first judgment result; a time-varying load prediction value determination unit, configured to, when the first judgment result is yes, search for a plurality of reference operating condition parameters matching the target operating condition parameters in the proxy model library, and determine a global time-varying load prediction value with a dangerous position of the rotating component to be tested corresponding to the target operating condition parameters by using a global time-varying load prediction model of a high-temperature rotating component and a time-varying load prediction model of a dangerous position of the rotating component corresponding to the plurality of reference operating condition parameters; The proxy model library includes a global time-varying load prediction model of a high-temperature rotating component and a time-varying load prediction model of a dangerous position of a rotating component corresponding to a plurality of different operating parameters; The global time-varying load prediction model of the high-temperature rotating component is determined based on a first training set, a multivariate regression model, reduced-order singular values, and reduced-order right singular vectors; the first training set includes sample operating parameters and corresponding reduced-order left singular vector sample values; the reduced-order left singular vector sample values, reduced-order singular values, and reduced-order right singular vectors are obtained by sequentially performing finite element simulation, singular value decomposition, and reduced-order on multiple sample operating parameters of the rotating component; The time-varying load prediction model for the dangerous position of the rotating component is obtained by training the time convolution network model using the second training set; the second training set includes sample operating condition parameters and corresponding time-varying load sample values ​​of the dangerous position; wherein the time-varying load sample values ​​of the dangerous position are obtained by sequentially performing finite element simulation on multiple sample operating condition parameters of the rotating component; the dangerous position is the position where the time-varying load exceeds a preset threshold in the finite element simulation results.

8. A computer device comprising: A memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the global time-varying load prediction method for high-temperature rotating components according to any one of claims 1 to 6.

Images