A method, system, medium and device for reconstructing a dynamic stress field of a non-harmonic blade

Abstract

The application discloses a non-harmonic blade dynamic stress field reconstruction method, system, medium and equipment, and relates to the technical field of blade vibration monitoring. The application firstly ensures high fidelity of a numerical model by minimizing the deviation of measured vibration displacement and finite element analysis vibration displacement, so as to obtain a multi-order modal dynamic stress field formed by multi-order modal dynamic stresses on each grid node of a non-harmonic blade model through finite element analysis, and determine a mistuning coefficient of vibration frequency of the non-harmonic blade as a non-harmonic quantity according to a reference frequency under a designed blade motion state, and then depth learning is performed on a complex non-linear mapping relationship of full-field multi-order modal dynamic stresses from a local tip position to the full field; when applied, the multi-order modal dynamic stress field of the target non-harmonic blade can be obtained by coupling the multi-order modal, the distortion problem of the linear method in processing the complex vibration mode of the non-harmonic blade is effectively overcome, and the precise prediction of the multi-modal coupling dynamic stress field of the target non-harmonic blade in actual operation is realized.

Description

Technical Field

[0001] This invention relates to the field of blade vibration monitoring technology, and in particular to a method, system, medium, and device for reconstructing the dynamic stress field of an intuned blade. Background Technology

[0002] Ideally, blades in rotating machinery exhibit periodic symmetry relative to the entire rotation, known as harmonic blades. However, in actual engineering practice, various uncertainties, such as the original materials, manufacturing process, installation, and normal wear during operation, can lead to slight differences in the physical parameters (including elastic modulus, density, and geometry) of each blade within the rotation. This disrupts the periodic symmetry of the entire rotation, resulting in blades exhibiting this phenomenon, which are called disharmonic blades. Blade disharmony disrupts the periodic symmetry of the entire blade structure, leading to uneven vibration distribution and localized vibration concentration during blade vibration. This localization further induces high-cycle fatigue in the blade rotor. In high-speed rotating machinery, high-cycle fatigue fracture of blades is one of the contributing factors to catastrophic accidents; therefore, monitoring the vibration stress of blades is essential.

[0003] In the prior art, blade tip timing technology captures the arrival time of rotating blades using non-contact sensors (such as optical fibers or capacitive probes). It can calculate the vibration displacement information of the blades without interfering with the flow field and blade structure, and then use linear fitting to describe the relationship between blade tip vibration displacement and dynamic stress.

[0004] However, tip timing can only measure the vibration displacement at the measured position of the blade, and cannot directly obtain the displacement of other positions on the blade or the blade stress condition, which is of great concern in actual operation. Furthermore, the linear fitting method is difficult to apply to multimodal coupled vibration, especially to non-harmonic blades with severe vibration localization, resulting in low accuracy in dynamic stress monitoring. Summary of the Invention

[0005] Therefore, it is necessary to provide a method, system, medium, and device for reconstructing the dynamic stress field of non-tunable blades to address the above-mentioned technical problems.

[0006] The present invention adopts the following technical solution: This invention provides a method for reconstructing the dynamic stress field of an intuned blade, comprising: Multiple sets of vibration displacement data were obtained from the measurement of the non-harmonious blade; a numerical calculation model of the non-harmonious blade was constructed and finite element analysis was performed on the non-harmonious blade. With the goal of minimizing the deviation between the vibration displacement data of the finite element analysis and the measured vibration displacement data, the amplitude of the airflow excitation force on the non-harmonious blade was adjusted to obtain the multi-mode dynamic stress field formed by the multi-mode dynamic stress on each grid node of the non-harmonious blade model. The vibration frequency of the untuned blade is determined based on the measured vibration displacement data, and the detuning coefficient of the vibration frequency of the untuned blade is determined as the untuning quantity based on the reference frequency of the blade under the designed motion state. Using the vibration displacement data measured by the non-tunable blade as input, the first neural network model is trained to predict the non-tunability of the non-tunable blade. Based on the vibration displacement data, inharmonicity, and grid node coordinates after blade meshing, the second neural network model is trained to predict the multi-mode dynamic stress field of the inharmonic blade. The vibration displacement data of the target non-harmonious blade is obtained, and the multi-mode dynamic stress field of the target non-harmonious blade is predicted by the trained first neural network model and second neural network model, so as to obtain the multi-mode coupled dynamic stress field of the target non-harmonious blade.

[0007] Optionally, determining the vibration frequency of the untuned blade based on measured vibration displacement data, and determining the detuning coefficient of the vibration frequency of the untuned blade based on the reference frequency under the designed blade motion state, specifically includes: The vibration frequency of the non-coordinated blade is obtained by performing a fast Fourier transform on the measured vibration displacement data. The detuning coefficient of the vibration frequency of the untuned blade is determined by the following formula based on the reference frequency under the designed blade motion state: ; in, To design the reference frequency for the blades under motion conditions, The vibration frequency of the non-harmonic blade.

[0008] Optionally, the first neural network is a one-dimensional convolutional neural network, which includes a first convolutional layer, a first pooling layer, a second convolutional layer, a second pooling layer, and three fully connected layers stacked in sequence, all of which use the ReLU activation function.

[0009] Optionally, the second neural network is a GMM graph convolutional neural network, which includes one input layer, seven intermediate layers and one output layer stacked in sequence. The input layer and the intermediate layers both use Gaussian biased linear units as activation functions, while the output layer does not use an activation function.

[0010] Optionally, the step of training the second neural network model to predict the multi-modal dynamic stress field of the inharmonious blade based on the vibration displacement data, inharmonious amount, and grid node coordinates after blade meshing specifically includes: For each set of measured vibration displacement data, the result after inputting it into a one-dimensional convolutional neural network and passing through the second pooling layer is subjected to inter-channel pooling to obtain the channel pooling result; The channel pooling results, multiple sets of non-harmonic quantities of the non-harmonic blade, and the grid node coordinates after blade meshing are concatenated and used as input to train the second neural network model for predicting the dynamic stress field of the non-harmonic blade.

[0011] Optionally, the prediction of the multi-modal dynamic stress field of the target non-harmonious blade using the trained first neural network model and second neural network model specifically includes: The vibration displacement data is input into the trained first neural network model to determine the detuning coefficient of the vibration frequency of the target untuned blade. The grid node coordinates, vibration displacement data, and detuning coefficient of the vibration frequency after the target non-tuned blade is meshed are input into the trained second neural network model to determine the multi-mode dynamic stress field of the target non-tuned blade surface.

[0012] This invention provides a dynamic stress field reconstruction system for an intunable blade, comprising: The acquisition module is used to acquire multiple sets of vibration displacement data measured on the non-harmonious blade; a numerical calculation model of the non-harmonious blade is constructed to perform finite element analysis on the non-harmonious blade, with the goal of minimizing the deviation between the vibration displacement data of the finite element analysis and the measured vibration displacement data, and the amplitude of the airflow excitation force on the non-harmonious blade is adjusted to obtain the multi-mode dynamic stress field formed by the multi-mode dynamic stress on each grid node of the non-harmonious blade model; The non-harmonicity determination module is used to determine the vibration frequency of the non-harmonic blade based on the measured vibration displacement data, and to determine the detuning coefficient of the vibration frequency of the non-harmonic blade as the non-harmonicity based on the reference frequency of the blade under the designed motion state. The first training module is used to predict and train the first neural network model for the non-tuning amount of the non-tuning blade using the vibration displacement data measured by the non-tuning blade as input. The second training module is used to predict and train the multi-mode dynamic stress field of the non-harmonic blade based on the vibration displacement data, non-harmonicity, and grid node coordinates after blade meshing. The reconstruction module is used to acquire the vibration displacement data of the target aharmonic blade, and predict the multi-mode dynamic stress field of the target aharmonic blade through the trained first neural network model and second neural network model, so as to couple and obtain the multi-mode coupled dynamic stress field of the target aharmonic blade.

[0013] The present invention provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the above-described method for reconstructing the dynamic stress field of anisotropic blades.

[0014] The present invention provides a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the program, it implements the above-mentioned method for reconstructing the dynamic stress field of the non-tunable blade.

[0015] The above-mentioned at least one technical solution adopted in this invention can achieve the following beneficial effects: This invention first ensures high fidelity of the numerical model by minimizing the deviation between the measured vibration displacement and the vibration displacement analyzed by the finite element method (FEM). This allows for the acquisition of the multi-modal dynamic stress field formed by the multi-modal dynamic stress at each grid node of the unharmonized blade model through FEM. The detuning coefficient of the vibration frequency of the unharmonized blade is determined as the unharmonization factor based on the reference frequency under the designed blade motion state. Subsequently, a first neural network is used to learn the mapping relationship between vibration displacement data and the unharmonization factor, while a second neural network is used to learn the mapping relationship between the unharmonization factor at each grid node and the vibration displacement to the multi-modal dynamic stress field. The combination of these two methods enables deep learning of the complex nonlinear mapping relationship from local tip displacement to multi-modal dynamic stress across the entire field. In application, multi-modal coupling can be used to obtain the multi-modal coupled dynamic stress field of the target unharmonized blade, effectively overcoming the distortion problem of linear methods when dealing with complex vibration modes of unharmonized blades. This achieves accurate and rapid prediction of the multi-modal coupled dynamic stress field of the target unharmonized blade in actual operation, significantly improving the accuracy and engineering practicality of dynamic stress monitoring for unharmonized blades. Attached Figure Description

[0016] The accompanying drawings, which are included to provide a further understanding of the invention and form part of this invention, illustrate exemplary embodiments of the invention and are used to explain the invention, but do not constitute an undue limitation of the invention. In the drawings:

[0017] Figure 1 A schematic diagram of the dynamic stress field reconstruction method for an inharmonious blade provided by the present invention; Figure 2 A schematic diagram of an online reconstruction method for multimodal coupled dynamic stress field of an entire circumference non-harmonious blade based on tip timing provided by the present invention; Figure 3 A schematic diagram of a leaf tip timing measurement system provided by the present invention; Figure 4 This is a schematic diagram of a finite element model obtained by mesh generation according to the present invention; Figure 5 A schematic diagram of a first neural network model provided by the present invention; Figure 6 A schematic diagram of a second neural network model provided by the present invention; Figure 7A schematic diagram of an online reconstruction network architecture for a full-circle non-harmonious blade multimodal coupled dynamic stress field based on tip timing is provided for this invention. Figure 8 A schematic diagram of a dynamic stress field reconstruction system for an intunable blade provided by the present invention; Figure 9 A schematic diagram of a computer device for reconstructing the dynamic stress field of an intuned blade, provided by the present invention. Detailed Implementation

[0018] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention will be clearly and completely described below in conjunction with specific embodiments and corresponding drawings. Obviously, the described embodiments are only a part of the embodiments of this invention, and not all of them. Based on the embodiments of this invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this invention.

[0019] Blade vibration exhibits multiple vibration modes. Currently, tip timing technology combined with linear fitting is generally applicable to the first-order mode, neglecting the influence of the remaining modes. This invention proposes a method for reconstructing the dynamic stress field of multi-modal coupled vibration, applicable to multi-modal coupled vibration.

[0020] The technical solutions provided by the various embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

[0021] Figure 1 This is a schematic diagram of a dynamic stress field reconstruction method for an inharmonious blade according to the present invention, which specifically includes the following steps: S101: Obtain multiple sets of vibration displacement data measured on the non-harmonious blade; construct a numerical calculation model of the non-harmonious blade and perform finite element analysis on the non-harmonious blade. With the goal of minimizing the deviation between the vibration displacement data of the finite element analysis and the measured vibration displacement data, adjust the amplitude of the airflow excitation force on the non-harmonious blade to obtain the multi-mode dynamic stress field formed by the multi-mode dynamic stress on each grid node of the non-harmonious blade model.

[0022] S102: Determine the vibration frequency of the untuned blade based on the measured vibration displacement data, and determine the detuning coefficient of the vibration frequency of the untuned blade as the untuning quantity based on the reference frequency of the designed blade motion state.

[0023] S103: Using the vibration displacement data measured by the non-tunable blade as input, the first neural network model is trained to predict the non-tunability of the non-tunable blade.

[0024] S104: Based on the vibration displacement data of the non-harmonious blade, the non-harmonious amount, and the coordinates of the grid nodes after the blade mesh is divided, the second neural network model is trained to predict the multi-mode dynamic stress field of the non-harmonious blade.

[0025] S105: Obtain the vibration displacement data of the target non-harmonious blade, predict the multi-mode dynamic stress field of the target non-harmonious blade through the trained first neural network model and second neural network model, so as to obtain the multi-mode coupled dynamic stress field of the target non-harmonious blade.

[0026] For ease of explanation, the following description focuses solely on the server as the executing entity. The server mentioned in this invention can be a server set up on a business platform, or a device such as a desktop computer or laptop computer capable of executing the solution of this invention.

[0027] This invention employs a neural network algorithm to identify the blade's inharmonicity based on the vibration displacement measured online at the blade tip. Furthermore, it reconstructs the multi-mode dynamic stress field of the blade using the vibration displacement and the identified inharmonicity, and calculates the mode shape conversion coefficient to couple the multi-mode dynamic stress field.

[0028] Figure 2 This diagram illustrates an online reconstruction method for the multimodal coupled dynamic stress field of a fully untuned blade based on tip timing, as described in this invention. The main idea of ​​this method is as follows: First, based on a tip-timing measurement system suitable for fully untuned blades, the vibration displacement data of the untuned blade is acquired online. Then, an online identification method for blade untuning quantization is established based on a convolutional neural network to obtain the degree of blade detuning. Finally, a method for reconstructing the blade dynamic stress field under multimodal coupling is established based on a graph convolutional neural network. By inputting the characteristics of the real-time measured vibration displacement data from tip timing and the identified degree of detuning, the overall dynamic stress of the blade participating in each vibration mode is obtained. The multimodal coupled dynamic stress field of the blade can then be calculated based on the mode shape transformation coefficient.

[0029] Specifically, the server can first obtain the vibration displacement data of the non-harmonious blades online based on the pre-established blade tip timing measurement system.

[0030] The blade tip timing measurement system includes a blade tip timing probe, a signal acquisition instrument, and a host computer. The probe is arranged along the circumference of the entire non-harmonious blade. The vibration displacement of the non-harmonious blade is calculated by monitoring the difference between the actual arrival time and the ideal arrival time of the blade tip position. Figure 3 This is a schematic diagram of a leaf tip timing measurement system according to the present invention.

[0031] Based on this, the server can construct a numerical calculation model of the non-harmonious blade and analyze the multi-mode dynamic stress field of the blade and the vibration displacement data under finite element analysis.

[0032] The server can establish a three-dimensional geometric model based on the actual dimensions of the entire blade, mesh the three-dimensional geometric model, set actual material geometric parameters and multiple sets of different intunability parameters, and conduct multimodal finite element analysis of the entire intuned blade. The intunability parameter is the detuning coefficient of the vibration frequency of the intuned blade, determined based on the reference frequency under the designed blade motion state.

[0033] During the analysis, the amplitude of the airflow excitation force on the blade can be adjusted based on the measured vibration displacement data, and the boundary conditions of the numerical calculation model can be continuously optimized to ensure that the error between the numerical results and the experimental results meets the pre-set accuracy requirements. Finally, the multi-mode dynamic stress and multi-mode vibration displacement data under finite element analysis on each mesh node of the blade model are obtained. Figure 4 This is a schematic diagram of a finite element model obtained by mesh generation in this invention.

[0034] Then, the server can establish an online identification method for blade anharmonic quantization based on convolutional neural networks. Before model training, sample data must first be constructed.

[0035] In theoretical design and finite element analysis, it is usually assumed that all blades on an impeller are identical, possess the same natural frequency, and exhibit consistent vibration behavior. However, factors such as manufacturing errors can cause the blades to exhibit dissonant characteristics, resulting in a deviation between the vibration frequency of the dissonant blades and the design value. In this case, the entire ring of blades no longer vibrates at a single frequency; instead, each blade has its own unique natural frequency, forming a frequency-dispersed band.

[0036] Therefore, for the reference frequency under the design blade motion state, in one or more embodiments of the present invention, the design value can be directly used, or the frequency closest to the design value among the entire ring of non-harmonious blades can be selected as the reference frequency. Subsequently, the relative deviation between the obtained vibration frequency and the reference frequency is used as the non-harmonious quantity. For the latter, although each blade has a deviation, the deviation of most blades usually follows a normal distribution. The blade with the frequency closest to the design value often statistically represents the ideal state of the entire ring of blades. Using it as a reference often best reflects the design intention and dynamic characteristics of the overall blade.

[0037] Specifically, the server can first conduct non-harmonious blade rotational vibration tests to obtain the frequency that is closest to the design value under motion conditions as the reference frequency. The aforementioned steps allow for the measurement of the non-harmonious blades at any given time. By performing a Fast Fourier Transform on the vibration displacement data measured at the blade tip at regular intervals, the vibration frequency of the non-harmonic blade can be obtained. Then, the detuning coefficient of the vibration frequency of the untuned blade can be determined using the following formula based on the reference frequency under the designed blade motion state: This completes the identification of the detuned state of the non-tunable blades.

[0038] Furthermore, based on the numerical calculation model and finite element analysis, small perturbations can be added to the vibration frequency and rotational speed of the intuned blades to generate simulated vibration displacement data and corresponding intunability quantities, forming a sufficient dataset for training and validating neural networks. Specifically, the density detuning coefficient of the blades can be set during the finite element analysis. With elastic modulus detuning coefficient The range is set to random variation within ±6%, allowing the analyzed blade vibration frequency to fluctuate within a certain range. Simultaneously, the blade rotation speed is set to fluctuate within ±10 rpm.

[0039] The dataset is divided into training and validation sets proportionally. Vibration displacement data of the non-harmonious blades are used as input, and the non-harmonious amount of the non-harmonious blades is used as output. A first neural network model is trained using data from the training set, and its performance is evaluated using data from the validation set. Training stops when the prediction accuracy of the first neural network model for the blade non-harmonious amount in the validation set data meets the requirements. After training, the non-harmonious amount of the blades can be identified online using the vibration displacement data measured in step S101.

[0040] Subsequently, a method for reconstructing the blade dynamic stress field under multimodal coupling can be established based on the second neural network model. In the preceding steps, the vibration displacement data of the non-harmonious blade, the corresponding non-harmonious amount of the non-harmonious blade, and the corresponding multimodal dynamic stress field of the non-harmonious blade can be obtained online, forming training and testing data samples for the reconstruction model. The second neural network can be a GMM (Graph Convolutional Neural Network).

[0041] Similarly, the dataset is divided into training and validation sets according to the proportion. The graph convolutional neural network is trained using the data in the training set. Due to the characteristics of the graph convolutional neural network, the network input is the coordinates of the finite element mesh nodes of the blade, the real-time displacement signal measured at the blade tip, and the non-harmonic quantity identified by the measured displacement signal. The output is the multi-mode dynamic stress of all nodes, that is, the multi-mode dynamic stress field of the blade.

[0042] Furthermore, since an excessively large input dimension can cause an exponential increase in network parameters and increase the difficulty of network training, it is necessary to extract features from the original measured displacement signal to reduce the input dimension. In one or more embodiments of the present invention, the vibration displacement data processed by the second pooling layer of the second neural network can be subjected to inter-channel pooling operation and concatenated with the identified blade anharmonicity as the node input features of the online dynamic stress field reconstruction network.

[0043] The dynamic stress field reconstruction effect of the network was tested using validation set data, and training was stopped when the requirements were met. After training, the network can reconstruct the multi-mode dynamic stress field of the corresponding blade using the vibration displacement obtained by online measurement and the inharmonicity of the inharmonic blade identified online.

[0044] The dynamic stress field of multimodal coupling of non-harmonious blades cannot be directly obtained by adding the dynamic stress fields of each order. It is necessary to obtain the mode transformation coefficient based on the measured vibration displacement data and the multimodal vibration displacement data under finite element analysis, and then couple the multimodal dynamic stress fields.

[0045] Specifically, the server can perform a fast Fourier transform on the measured vibration displacement data to decouple the amplitudes of each order. and phase angle At the same time, the corresponding position on the blade is obtained based on the position of the sensor. each order of modal displacement The modal transformation coefficients of each order mode can be obtained from the formula. The calculation shows that the dynamic stress field of multimodal coupling can be obtained by superimposing the dynamic stress fields of each order mode according to the corresponding mode shape transformation coefficient.

[0046] Finally, in practical applications, the server can acquire the vibration displacement data of the target untuned blade, input the vibration displacement data into the trained first neural network model, determine the detuning coefficient of the vibration frequency of the target untuned blade, and input the grid node coordinates, vibration displacement data and vibration frequency detuning coefficient of the target untuned blade into the trained second neural network model to predict the multi-mode dynamic stress field of the target untuned blade, so as to couple and obtain the multi-mode coupled dynamic stress field of the target untuned blade.

[0047] based on Figure 1The proposed method for reconstructing the dynamic stress field of an out-of-tuned blade first ensures high fidelity of the numerical model by minimizing the deviation between the measured vibration displacement and the vibration displacement analyzed by the finite element method. This allows for the acquisition of the multi-modal dynamic stress field formed by the multi-modal dynamic stress at each grid node of the out-of-tuned blade model through finite element analysis. The detuning coefficient of the vibration frequency of the out-of-tuned blade is determined as the out-of-tuning quantity based on the reference frequency under the designed blade motion state. Subsequently, a first neural network is used to learn the mapping relationship between vibration displacement data and the out-of-tuning quantity, while a second neural network is used to learn the mapping relationship between the out-of-tuning quantity at each grid node and the vibration displacement to the multi-modal dynamic stress field. The combination of these two methods enables deep learning of the complex nonlinear mapping relationship from local displacement at the blade tip to multi-modal dynamic stress across the entire field. In application, multi-modal coupling can be used to obtain the multi-modal coupled dynamic stress field of the target out-of-tuned blade, effectively overcoming the distortion problem of linear methods when dealing with complex vibration modes of out-of-tuned blades. This achieves accurate and rapid prediction of the multi-modal coupled dynamic stress field of the target out-of-tuned blade in actual operation, significantly improving the accuracy and engineering practicality of dynamic stress monitoring for out-of-tuned blades.

[0048] This invention provides a method and system for online reconstruction of multimodal coupled dynamic stress field of a full-circle unharmonic blade based on tip timing. It can identify the blade unharmonicity online based on the vibration displacement measured by tip timing, monitor the blade status in real time, and avoid safety problems caused by vibration localization.

[0049] This invention provides a method and system for online reconstruction of the multimodal coupled dynamic stress field of a blade with full-circle non-harmonicity based on tip timing. After obtaining the vibration displacement and non-harmonicity of the blade, the whole-field dynamic stress reconstruction of the blade participating in each vibration mode can be completed online. The stress status of key positions of the blade can be obtained quickly and accurately based on the mode conversion coefficient.

[0050] When applying the dynamic stress field reconstruction method for non-harmonious blades provided by this invention, it is not necessary to consider... Figure 1 The steps shown are executed in sequence. The specific execution order of each step can be determined as needed, and this invention does not impose any restrictions on it.

[0051] Furthermore, the present invention also provides an embodiment of applying the method of the present invention, in which the method is utilized. Figure 2 The blade tip timing measurement system shown measures the tip vibration displacement of the entire circumference of anharmonic blade online. Then, a three-dimensional model is established based on the actual blade geometry and mesh generation is completed. The resulting finite element model is used to set the anharmonicity of the entire circumference blade finite element model, and finite element analysis is carried out to obtain 3000 sets of data samples of multimodal dynamic stress field and vibration displacement of the anharmonic blade. These data samples, together with the actual measured vibration displacement and actual blade anharmonicity data, form a dataset. The dataset is then divided into a training set and a validation set in a 7:3 ratio.

[0052] Figure 5 This is a schematic diagram of a first neural network model in this invention. In this embodiment, a one-dimensional convolutional neural network is used for online identification of blade anharmonicity. It includes two convolutional layers, two pooling layers, and three fully connected layers, with ReLU as the activation function. The input is the vibration displacement measured by the blade tip timing system, with a data length of 200. The output is the blade anharmonicity. The loss function of the training and validation sets gradually decreases and converges to 10% of the initial loss function value during the training process. -3 Training stops when the value is less than a certain multiple. Based on this model and the leaf tip timing measurement system, online identification of leaf dissonance can be achieved.

[0053] Figure 6 This is a schematic diagram of a second neural network model in this invention. In this embodiment, a GMM (Graph Convolutional Neural Network) is used for online reconstruction of the multimodal coupled dynamic stress field of the blade. It includes one input layer, one output layer, and seven intermediate layers. Both the input and intermediate layers use Gaussian-biased linear units as activation functions, while the output layer does not use an activation function. The input consists of the coordinates of the mesh nodes of the finite element model of the blade body, the blade tip vibration displacement, and the blade incoordination. The output is the multimodal dynamic stress field. The loss function of the training and validation sets gradually decreases and converges to 10% of the initial loss function value during the training process. -3 Training stops when the value drops below a certain value. Based on the two models mentioned above and the tip timing measurement system, online reconstruction of the multimodal dynamic stress field of the non-harmonic blade can be achieved.

[0054] Figure 7 This is a schematic diagram of an online reconstruction network architecture for the multimodal coupled dynamic stress field of a full-circle unharmonic blade based on tip timing, as described in this invention. The input dimension of the online unharmonicity identification network is too long. To simplify the input as node input features for the online dynamic stress field reconstruction network, inter-channel pooling is performed on the vibration displacement data processed by the second pooling layer, and this data is concatenated with the identified blade unharmonicity as the node input features of the online dynamic stress field reconstruction network. Based on the measured vibration displacement at tip timing and the multimodal vibration displacement field of the blade, mode transformation coefficients of different orders can be analyzed, and then the multimodal coupled dynamic stress field can be calculated.

[0055] The above describes a method for reconstructing the dynamic stress field of an intuned blade according to one or more embodiments of the present invention. Based on the same idea, the present invention also provides a corresponding dynamic stress field reconstruction system for an intuned blade, such as... Figure 8 As shown.

[0056] Figure 8 A schematic diagram of a dynamic stress field reconstruction system for an intunable blade provided by the present invention includes: The acquisition module 201 is used to acquire multiple sets of vibration displacement data obtained from the measurement of the non-harmonious blade; to construct a numerical calculation model of the non-harmonious blade and perform finite element analysis on the non-harmonious blade, with the goal of minimizing the deviation between the vibration displacement data of the finite element analysis and the measured vibration displacement data, the amplitude of the airflow excitation force on the non-harmonious blade is adjusted to obtain the multi-mode dynamic stress field formed by the multi-mode dynamic stress on each grid node of the non-harmonious blade model; The non-harmonicity determination module 202 is used to determine the vibration frequency of the non-harmonic blade based on the measured vibration displacement data, and to determine the detuning coefficient of the vibration frequency of the non-harmonic blade as the non-harmonicity based on the reference frequency of the designed blade motion state. The first training module 203 is used to train the first neural network model to predict the non-tuning amount of the non-tuning blade using the vibration displacement data measured by the non-tuning blade as input. The second training module 204 is used to predict and train the multi-mode dynamic stress field of the non-harmonic blade based on the vibration displacement data, non-harmonicity, and grid node coordinates after the blade mesh is divided. The reconstruction module 205 is used to acquire the vibration displacement data of the target aharmonic blade, and predict the multi-mode dynamic stress field of the target aharmonic blade through the trained first neural network model and second neural network model, so as to couple and obtain the multi-mode coupled dynamic stress field of the target aharmonic blade.

[0057] Specific limitations regarding the dynamic stress field reconstruction system for non-tunable blades can be found in the limitations of the dynamic stress field reconstruction method for non-tunable blades mentioned above, and will not be repeated here. Each module in the aforementioned dynamic stress field reconstruction system for non-tunable blades can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in or independent of the processor in a computer device, or stored in the memory of a computer device as software, so that the processor can call and execute the corresponding operations of each module.

[0058] The present invention also provides a computer-readable storage medium storing a computer program that can be used to execute the above-described... Figure 1 A method for reconstructing the dynamic stress field of non-harmonious blades is provided.

[0059] The present invention also provides Figure 9 The schematic diagram of the computer device shown is as follows: Figure 9 As shown, at the hardware level, this computer device includes a processor, internal bus, network interface, memory, and non-volatile memory, and may also include other hardware required for business operations. The processor reads the corresponding computer program from the non-volatile memory into memory and then executes it to achieve the above. Figure 1A method for reconstructing the dynamic stress field of non-harmonious blades is provided.

[0060] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium, and when executed, it can include the processes of the embodiments of the methods described above. Any references to memory, storage, databases, or other media used in the embodiments provided by this invention can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, or optical storage, etc. Volatile memory can include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM can be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM), etc.

[0061] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this invention.

Claims

1. A method for reconstructing the dynamic stress field of an inharmonious blade, characterized in that, include: Acquire multiple sets of vibration displacement data obtained from measurements of non-harmonious blades; A numerical calculation model of the non-harmonious blade was constructed, and finite element analysis was performed on the non-harmonious blade. With the goal of minimizing the deviation between the vibration displacement data of the finite element analysis and the measured vibration displacement data, the amplitude of the airflow excitation force on the non-harmonious blade was adjusted to obtain the multi-mode dynamic stress field formed by the multi-mode dynamic stress on each grid node of the non-harmonious blade model. The vibration frequency of the untuned blade is determined based on the measured vibration displacement data, and the detuning coefficient of the vibration frequency of the untuned blade is determined as the untuning quantity based on the reference frequency of the blade under the designed motion state. Using the vibration displacement data measured by the non-tunable blade as input, the first neural network model is trained to predict the non-tunability of the non-tunable blade. Based on the vibration displacement data, inharmonicity, and grid node coordinates after blade meshing, the second neural network model is trained to predict the multi-mode dynamic stress field of the inharmonic blade. The vibration displacement data of the target non-harmonious blade is obtained, and the multi-mode dynamic stress field of the target non-harmonious blade is predicted by the trained first neural network model and second neural network model, so as to obtain the multi-mode coupled dynamic stress field of the target non-harmonious blade.

2. The method for reconstructing the dynamic stress field of an intuned blade as described in claim 1, characterized in that, The process of determining the vibration frequency of the untuned blade based on measured vibration displacement data, and determining the detuning coefficient of the untuned blade's vibration frequency based on the reference frequency under the designed blade motion state, specifically includes: The vibration frequency of the non-coordinated blade is obtained by performing a fast Fourier transform on the measured vibration displacement data. The detuning coefficient of the vibration frequency of the untuned blade is determined by the following formula based on the reference frequency under the designed blade motion state: ； in, To design the reference frequency for the blades under motion conditions, The vibration frequency of the non-harmonic blade.

3. The method for reconstructing the dynamic stress field of an intuned blade as described in claim 1, characterized in that, The first neural network is a one-dimensional convolutional neural network, which includes a first convolutional layer, a first pooling layer, a second convolutional layer, a second pooling layer, and three fully connected layers stacked in sequence, all of which use the ReLU activation function.

4. The method for reconstructing the dynamic stress field of an intuned blade as described in claim 1, characterized in that, The second neural network is a GMM graph convolutional neural network, which includes one input layer, seven intermediate layers and one output layer stacked in sequence. Both the input layer and the intermediate layers use Gaussian biased linear units as activation functions.

5. The method for reconstructing the dynamic stress field of an inharmonious blade as described in claim 3, characterized in that, The process of training the second neural network model to predict the multi-modal dynamic stress field of the inharmonious blade based on the vibration displacement data, inharmonious amount, and grid node coordinates after blade meshing includes: For each set of measured vibration displacement data, the result after inputting it into a one-dimensional convolutional neural network and passing through the second pooling layer is subjected to inter-channel pooling to obtain the channel pooling result; The channel pooling results, multiple sets of non-harmonic quantities of the non-harmonic blade, and the grid node coordinates after blade meshing are concatenated and used as input to train the second neural network model for predicting the dynamic stress field of the non-harmonic blade.

6. The method for reconstructing the dynamic stress field of an intuned blade as described in claim 1, characterized in that, The prediction of the multi-modal dynamic stress field of the target intuned blade using the trained first and second neural network models specifically includes: The vibration displacement data is input into the trained first neural network model to determine the detuning coefficient of the vibration frequency of the target untuned blade. The grid node coordinates, vibration displacement data, and detuning coefficient of the vibration frequency after the target non-tunable blade is meshed are input into the trained second neural network model to determine the multi-mode dynamic stress field of the target non-tunable blade surface.

7. A dynamic stress field reconstruction system for an inharmonious blade, characterized in that, include: The acquisition module is used to acquire multiple sets of vibration displacement data obtained from measurements of non-harmonious blades; A numerical calculation model of the non-harmonious blade was constructed, and finite element analysis was performed on the non-harmonious blade. With the goal of minimizing the deviation between the vibration displacement data of the finite element analysis and the measured vibration displacement data, the amplitude of the airflow excitation force on the non-harmonious blade was adjusted to obtain the multi-mode dynamic stress field formed by the multi-mode dynamic stress on each grid node of the non-harmonious blade model. The non-harmonicity determination module is used to determine the vibration frequency of the non-harmonic blade based on the measured vibration displacement data, and to determine the detuning coefficient of the vibration frequency of the non-harmonic blade as the non-harmonicity based on the reference frequency of the blade under the designed motion state. The first training module is used to train the first neural network model to predict the dissonance of the dissonance of the dissonance of the dissonance blade using the vibration displacement data measured by the dissonance blade as input. The second training module is used to predict and train the second neural network model for multi-mode dynamic stress field of the non-harmonic blade based on the vibration displacement data, non-harmonicity, and grid node coordinates after blade meshing. The reconstruction module is used to acquire the vibration displacement data of the target aharmonic blade, and predict the multi-mode dynamic stress field of the target aharmonic blade through the trained first neural network model and second neural network model, so as to obtain the multi-mode coupled dynamic stress field of the target aharmonic blade.

8. A computer-readable storage medium, characterized in that, The storage medium stores a computer program, which, when executed by a processor, implements the method as described in any one of claims 1 to 6.

9. A computer device, characterized in that, The method includes 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 method as described in any one of claims 1 to 6.

Images