Method for predicting the fatigue life of a water vapor compressor and related device

Abstract

The application discloses a kind of prediction method of fatigue life of water vapor compressor and related device, it is related to mechanical engineering technical field, the method is first in structure parameter source domain sampling obtains multiple current structure parameters, and obtains corresponding initial fatigue life by pre-trained fatigue life prediction model;Then determine target structure parameter and construct its adjacent structure parameter target domain, calculate the sampling probability density of this target domain;The sampling probability density ratio is obtained in combination with the sampling probability density of source domain and target domain;Subsequently, with the sampling probability density ratio as weight, the pre-trained fatigue life prediction model is reconstructed using current structure parameter and its initial fatigue life;Finally, input current structure parameter into reconstructed fatigue life prediction model, output more accurate fatigue life, to effectively avoid the prediction distortion problem caused by source domain and target domain distribution deviation, improve the accuracy of predicted fatigue life of water vapor compressor.

Description

Technical Field

[0001] This application relates to the field of mechanical engineering technology, and in particular to a method and related apparatus for predicting the fatigue life of a steam compressor. Background Technology

[0002] With the continuous pursuit of reliability in industrial equipment, the fatigue life of key components (such as impellers and hubs) of steam compressors, as core rotating machinery in energy, chemical, and refrigeration systems, under complex alternating loads has attracted much attention. To evaluate the reliability of different design schemes, the concepts of structural parameter source domain and structural parameter target domain are often introduced in engineering practice. The structural parameter source domain refers to the distribution area of ​​structural parameters whose fatigue life has been accumulated through high-fidelity simulation or experiments; the structural parameter target domain refers to the distribution area of ​​structural parameters formed by the target structural parameters and their adjacent uncertainties.

[0003] Currently, the target structural parameters of a steam compressor are typically input into a fatigue life prediction model trained based on the source domain of the structural parameters. The generalization ability of this fatigue life prediction model is then used to extrapolate and predict the fatigue life of the steam compressor.

[0004] However, when the target domain of the structural parameters of the steam compressor deviates from the source domain of the structural parameters, the extrapolation prediction using the fatigue life prediction model will produce prediction bias due to the distribution shift between the source domain and the target domain, resulting in a distortion of the predicted fatigue life of the steam compressor. Summary of the Invention

[0005] In view of the above problems, this application provides a method and related apparatus for predicting the fatigue life of a steam compressor, in order to avoid the problem of distortion in the predicted fatigue life of the steam compressor. The specific solution is as follows:

[0006] The first aspect of this application provides a method for predicting the fatigue life of a steam compressor, comprising:

[0007] The structural parameter source domain of the steam compressor is sampled to obtain multiple current structural parameters of the steam compressor, and the multiple current structural parameters of the steam compressor are input into a pre-trained fatigue life prediction model to obtain multiple initial fatigue lives of the steam compressor.

[0008] The target structural parameters of the steam compressor are determined from the source domain of the structural parameters of the steam compressor, and the target domain of the structural parameters of the steam compressor is constructed based on the target structural parameters of the steam compressor.

[0009] Based on multiple current structural parameters of the steam compressor and the target domain of the structural parameters of the steam compressor, the sampling probability density of the target domain of the structural parameters of the steam compressor is calculated.

[0010] Based on the sampling probability density of the source domain of the structural parameters of the steam compressor and the sampling probability density of the target domain of the structural parameters of the steam compressor, the sampling probability density ratio of the steam compressor is calculated.

[0011] Based on multiple current structural parameters of the steam compressor, multiple initial fatigue lives of the steam compressor, and the sampling probability density ratio of the steam compressor, the pre-trained fatigue life prediction model is reconstructed to obtain the reconstructed fatigue life prediction model.

[0012] Multiple current structural parameters of the steam compressor are input into the reconstructed fatigue life prediction model to obtain multiple current fatigue lives of the steam compressor.

[0013] In one possible implementation, after inputting multiple current structural parameters of the steam compressor into the reconstructed fatigue life prediction model to obtain multiple current fatigue lives of the steam compressor, the method further includes:

[0014] Based on the target domain of the structural parameters of the steam compressor, multiple random structural parameters of the steam compressor are generated; and the multiple random structural parameters of the steam compressor are input into the reconstructed fatigue life prediction model to obtain multiple random fatigue lives of the steam compressor.

[0015] Based on multiple random fatigue lives of the steam compressor, the fatigue failure probability of the steam compressor under the target structural parameters is calculated.

[0016] When the fatigue failure probability of the steam compressor under the target structural parameters is greater than a preset fatigue failure probability threshold, the target structural parameters of the steam compressor are updated based on multiple random structural parameters of the steam compressor and the fatigue failure probability of the steam compressor under the target structural parameters. Then, the process returns to the step of constructing the target domain of the structural parameters of the steam compressor based on the target structural parameters of the steam compressor, until the fatigue failure probability of the steam compressor under the target structural parameters is not greater than the fatigue failure probability threshold.

[0017] In one possible implementation, the training process of the fatigue life prediction model includes:

[0018] The structural parameter source domain of the steam compressor is sampled to obtain multiple training structural parameters of the steam compressor;

[0019] Physical simulations were performed on multiple training structural parameters of the steam compressor to obtain multiple training fatigue lives of the steam compressor.

[0020] Based on multiple training structural parameters and multiple training fatigue lives of the steam compressor, a neural network model is trained to obtain a trained fatigue life prediction model. The input of the fatigue life prediction model is the multiple structural parameters of the steam compressor, and the output of the fatigue life prediction model is the multiple fatigue lives of the steam compressor.

[0021] In one possible implementation, calculating the sampling probability density of the structural parameter target domain of the steam compressor based on multiple current structural parameters of the steam compressor and the structural parameter target domain of the steam compressor includes:

[0022] Construct the probability density function corresponding to the target domain of the structural parameters of the steam compressor;

[0023] Input multiple current structural parameters of the steam compressor into the probability density function corresponding to the structural parameter target domain of the steam compressor to obtain the probability density of multiple current structural parameters of the steam compressor under the structural parameter target domain;

[0024] The probability density of multiple current structural parameters of the steam compressor under the structural parameter target domain is determined as the sampling probability density of the structural parameter target domain.

[0025] In one possible implementation, the pre-trained fatigue life prediction model is reconstructed based on multiple current structural parameters of the steam compressor, multiple initial fatigue lives of the steam compressor, and the sampling probability density ratio of the steam compressor to obtain a reconstructed fatigue life prediction model, including:

[0026] For each current structural parameter of the steam compressor, the current structural parameter of the steam compressor and the initial fatigue life of the steam compressor corresponding to the current structural parameter of the steam compressor are determined as a sampling sample of the steam compressor.

[0027] The weights of the multiple samples of the steam compressor are obtained by weighting the multiple samples of the steam compressor based on the sampling probability density ratio of the steam compressor.

[0028] Based on multiple sampled samples from the steam compressor and the weights of these samples, a weighted loss function is used to retrain the pre-trained fatigue life prediction model, resulting in the reconstructed fatigue life prediction model.

[0029] In one possible implementation, calculating the fatigue failure probability of the steam compressor under the target structural parameters based on multiple random fatigue lives of the steam compressor includes:

[0030] For each random fatigue life of the steam compressor, determine whether the random fatigue life of the steam compressor is greater than a preset fatigue life threshold.

[0031] If the random fatigue life of the steam compressor is not greater than the preset fatigue life threshold, then the random fatigue life of the steam compressor is determined as the first target fatigue life of the steam compressor.

[0032] Based on the number of first target fatigue lives of the steam compressor and the number of random fatigue lives of the steam compressor, the fatigue failure probability of the steam compressor under the target structural parameters is calculated.

[0033] In one possible implementation, updating the target structural parameters of the steam compressor based on multiple random structural parameters of the steam compressor and the fatigue failure probability of the steam compressor under the target structural parameters includes:

[0034] For each random fatigue life of the steam compressor, determine whether the random fatigue life of the steam compressor is greater than a preset fatigue life threshold.

[0035] If the random fatigue life of the steam compressor is greater than the preset fatigue life threshold, then the random fatigue life of the steam compressor is determined as the second target fatigue life of the steam compressor.

[0036] Based on the random structural parameters of the steam compressor corresponding to the second target fatigue life of the steam compressor, the mean value of the random structural parameters of the steam compressor is calculated.

[0037] The target structural parameters of the steam compressor are updated based on the mean of the random structural parameters of the steam compressor.

[0038] A second aspect of this application provides a computer program product including computer-readable instructions that, when executed on an electronic device, cause the electronic device to implement a method for predicting the fatigue life of a steam compressor as described in the first aspect or any implementation thereof.

[0039] A third aspect of this application provides an electronic device, comprising at least one processor and a memory connected to the processor, wherein:

[0040] The memory is used to store computer programs;

[0041] The processor is used to execute the computer program so that the electronic device can implement the method for predicting the fatigue life of a steam compressor in the first aspect or any implementation thereof.

[0042] A fourth aspect of this application provides a computer storage medium carrying one or more computer programs that, when executed by an electronic device, enable the electronic device to perform a method for predicting the fatigue life of a steam compressor described in the first aspect or any implementation thereof.

[0043] By employing the above technical solution, this application provides a method and related apparatus for predicting the fatigue life of a steam compressor. The method samples the source domain of structural parameters of the steam compressor to obtain multiple current structural parameters, and inputs these parameters into a pre-trained fatigue life prediction model to obtain multiple initial fatigue lives. It then determines a target structural parameter from the source domain and constructs a target domain based on this target parameter. Based on the multiple current structural parameters and the target domain, it calculates the sampling probability density of the target domain. Combining the obtained sampling probability density of the source domain with the calculated sampling probability density of the target domain, it further calculates the sampling probability density ratio. Using this sampling probability density ratio, multiple current structural parameters, and their corresponding initial fatigue lives, it reconstructs the pre-trained fatigue life prediction model to obtain a reconstructed fatigue life prediction model. Finally, it inputs the multiple current structural parameters into the reconstructed fatigue life prediction model to output more accurate current fatigue lives, thereby effectively avoiding prediction distortion caused by the distribution offset between the source and target domains and improving the accuracy of the predicted fatigue life of the steam compressor. Attached Figure Description

[0044] The above and other features, advantages, and aspects of the embodiments of this disclosure will become more apparent from the accompanying drawings and the following detailed description. Throughout the drawings, the same or similar reference numerals denote the same or similar elements. It should be understood that the drawings are schematic, and the originals and elements are not necessarily drawn to scale.

[0045] Figure 1 A flowchart illustrating a method for predicting the fatigue life of a steam compressor provided in this application embodiment;

[0046] Figure 2 A mesh diagram illustrating the physical simulation of a steam compressor hub-blade configuration provided in this application embodiment;

[0047] Figure 3 This is a schematic diagram of the hardware structure of an electronic device provided in an embodiment of this application. Detailed Implementation

[0048] The embodiments of this application are described below with reference to the accompanying drawings. The terminology used in the implementation section of this application is for explaining specific embodiments only and is not intended to limit the scope of this application.

[0049] The embodiments of this application will now be described with reference to the accompanying drawings. Those skilled in the art will recognize that, with technological advancements and the emergence of new scenarios, the technical solutions provided in the embodiments of this application are equally applicable to similar technical problems.

[0050] The terms "first," "second," etc., used in the specification, claims, and accompanying drawings of this application are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such terms are interchangeable where appropriate; this is merely a way of distinguishing objects with the same attributes in the embodiments of this application. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion, so that a process, method, system, product, or apparatus that comprises a series of elements is not necessarily limited to those elements, but may include other elements not explicitly listed or inherent to those processes, methods, products, or apparatuses.

[0051] To avoid the problem of distorted prediction of the fatigue life of a steam compressor, this application provides a method for predicting the fatigue life of a steam compressor. The following description, in conjunction with the accompanying drawings and specific embodiments, provides a more detailed explanation of the method for predicting the fatigue life of a steam compressor provided in this application.

[0052] Please see the appendix Figure 1 , Figure 1 This is a flowchart illustrating a method for predicting the fatigue life of a steam compressor, provided in an embodiment of this application. The method may include the following steps:

[0053] Step S101: Sample the source domain of the structural parameters of the steam compressor to obtain multiple current structural parameters of the steam compressor, and input the multiple current structural parameters of the steam compressor into the pre-trained fatigue life prediction model to obtain multiple initial fatigue lives of the steam compressor.

[0054] It should be noted that steam compressors are turbine machines that use high-temperature, high-pressure steam as the working medium. They are widely used in industrial fields such as chemical engineering, seawater desalination, and waste heat recovery, and their operating environment is extremely harsh. They not only face the risk of water erosion from high temperatures, wet steam, and even phase changes, but also must withstand alternating thermo-mechanical coupling loads, which can easily lead to fatigue damage to critical rotating components such as impellers and hubs. Compared to ordinary steam compressors, steam compressors have significantly higher requirements for structural reliability and fatigue life prediction accuracy.

[0055] In this application, the training process of the fatigue life prediction model includes: sampling the source domain of structural parameters of the steam compressor to obtain multiple training structural parameters of the steam compressor; performing physical simulation on the multiple training structural parameters of the steam compressor to obtain multiple training fatigue lives of the steam compressor; and training a neural network model based on the multiple training structural parameters and multiple training fatigue lives of the steam compressor to obtain a trained fatigue life prediction model. The input of the fatigue life prediction model is the multiple structural parameters of the steam compressor, and the output of the fatigue life prediction model is the multiple fatigue lives of the steam compressor.

[0056] Source domain of structural parameters of steam compressor This is an engineering-level design space database that covers the key structural parameters of the impeller-hub of a steam compressor, primarily including the impeller root fillet radius R1, the hub-shaft contact length L1, and the radial thickness L2 of the hub bottom support structure. It encompasses all possible combinations within its reasonable engineering range. If represented in a three-dimensional coordinate system, with the horizontal, vertical, and axial axes corresponding to the impeller root fillet radius R1, hub-shaft contact length L1, and hub bottom support structure L2 respectively, then the entire cubic region constitutes the source domain of the steam compressor's structural parameters. Geometric representation of the steam compressor. Source domain of each structural parameter of the steam compressor. All of these can have their upper limits pre-set based on engineering experience and manufacturing constraints. and lower limit These reflect manufacturing processes, assembly constraints, and strength requirements. For example, R1 is between 5.0mm and 7.0mm, L1 is between 10.0mm and 15.0mm, and L2 is between 45.0mm and 55.0mm.

[0057] Specifically, regarding the structural parameter source domain of a continuous and high-dimensional steam compressor Latin hypercube sampling (LHS) can be used. While ensuring a limited number of samples, it distributes them evenly across each parameter dimension to maximize coverage of the entire design space and avoid sample clustering or omissions. Specifically, for each structural parameter dimension, its structural parameter source domain is... Divide into n equal parts s 80 (e.g.) subintervals (n) s (where n is the total number of samples), then randomly select a training structure parameter within each sub-interval, and then form n through random permutations and combinations. s A set of three-dimensional parameter points, i.e., multiple structural parameters used for training. This ensures that each parameter is explored evenly across its entire range, while also spreading the sample points as widely as possible in three-dimensional space, thereby improving the generalization ability of subsequent models.

[0058] Next, we can analyze these n... s Training structure parameters obtained from group sampling The entire physical simulation process is executed sequentially. Specifically, firstly, the structural parameters (R1, L1, L2) for each training set are input into professional Computer-Aided Engineering (CAE) software to construct the corresponding three-dimensional geometric model. Then, Computational Fluid Dynamics (CFD) is used to simulate the flow of high-temperature and high-pressure steam inside the steam compressor, and the resulting unsteady pressure load is mapped onto the structural finite element model for fluid-structure interaction (FSI) analysis to obtain the dynamic stress and strain response of the hub-blade system under real operating conditions. Finally, based on the stress / strain cycle history obtained from the FSI simulation, a low-cycle fatigue model (Smith-Watson-Topper, SWT) can be applied to calculate the expected fatigue life under this specific geometric configuration. The low-cycle fatigue model considers the effects of maximum stress and plastic strain, making it suitable for low-cycle fatigue analysis. Through this process, a realistic, physically verified fatigue life of the steam compressor is obtained for each training structural parameter. For easier understanding, please refer to... Figure 2 , Figure 2 A mesh diagram illustrating the physical simulation of a steam compressor hub-blade configuration provided in this application embodiment.

[0059] The neural network model, acting as a surrogate model, does not replace physical simulation but rather reproduces simulation results at extremely low cost within the already validated simulation region. Once trained, it only requires new R1, L1, and L2 values ​​to output predicted lifespan within milliseconds, significantly accelerating design iterations. This neural network model is a feedforward structure, containing an input layer, several hidden layers, and an output layer. The input layer receives three structural parameters, and the output layer provides a single fatigue life value. The hidden layers extract features layer by layer using nonlinear activation functions (such as the hyperbolic tangent function), ultimately fitting a complex mapping relationship between structural parameters and fatigue life. The training process is essentially supervised learning: multiple training structural parameters of the steam compressor obtained in the first two stages are used... Multiple training fatigue life tests for steam compressors As a training set, the neural network continuously adjusts its internal weights and biases to make its predictions approximate the actual simulation values. A loss function (such as mean squared error) measures the prediction deviation, and an optimization algorithm (such as gradient descent) drives the model to converge to the optimal solution. The model's predicted values ​​are compared with the actual simulation values, and the error signal is backpropagated to update the model parameters until the error converges to an acceptable level (e.g., ≤5%), resulting in a well-trained fatigue life prediction model. In mechanical design, a 5% life prediction error is typically within an acceptable safety margin; and it indicates that the neural network has successfully captured the dominant physical mechanism between structural parameters and fatigue life, rather than simply fitting noise. (Well-trained fatigue life prediction model) , ,in, As the activation function, it can be set directly. Sure; The weight matrix can be obtained through training. The bias vector can be learned through training. These are the multiple structural parameters of the steam compressor; This study describes the fatigue life prediction model for a steam compressor. The trained model's inference speed far exceeds that of physical simulation, achieving predictions that are both efficient and accurate.

[0060] At this point, instead of using the initial 80 training samples, intelligent sampling is performed again from the source domain design space to obtain a new parameter set on a larger scale. The method used remains Latin hypercube sampling, ensuring that sampling points are uniformly distributed throughout the design space and avoiding local clustering. For example, 2000 samples are taken, approximately 25 times the original source domain sample size (80 samples). The sampling range still covers the entire design space, i.e., R1 is between 5.0 mm and 7.0 mm, L1 is between 10.0 mm and 15.0 mm, and L2 is between 45.0 mm and 55.0 mm. The resulting multiple current structural parameters of the steam compressor are represented as a 2000-row, 3-column matrix. The current structural parameters of the steam compressor are... The input is fed into a pre-trained fatigue life prediction model, which automatically performs forward propagation calculations and outputs the corresponding fatigue life prediction values. For example, consider the 2000 sets of current structural parameters of a steam compressor. The data is input into a pre-trained fatigue life prediction model to obtain multiple initial fatigue lives of the steam compressor. Each of them This represents the number of cycle lives predicted by the i-th set of structural parameters. Without needing to call time-consuming simulation programs such as CFD / FSI / SWT, multiple initial fatigue lives of the steam compressor can be obtained solely through a pre-trained fatigue life prediction model.

[0061] Step S102: Determine the target structural parameters of the steam compressor from the source domain of the structural parameters of the steam compressor, and construct the target domain of the structural parameters of the steam compressor based on the target structural parameters of the steam compressor.

[0062] In this application, the target structural parameters It is the source domain of structural parameters of the steam compressor. The optimization algorithm focuses on candidate design schemes in the k-th iteration. For example, in a certain round of optimization, the current optimal solution is: This means R1 = 6.1 mm, L1 = 12.8 mm, and L2 = 49.5 mm. These target structural parameters are not fixed but dynamically updated during the optimization process, representing a dynamically changing local design region. The target domain for the structural parameters of the steam compressor. It is not a discrete set, but a set surrounding the current design point. The continuous distribution region has a mathematical form that is a normal distribution. The center point is The standard deviation is It is usually set based on experience (e.g., 0.1).

[0063] It should be noted that the normal distribution has good mathematical properties, naturally expressing a search preference that is dense at the center and sparse at the edges; and the standard deviation can be flexibly adjusted. Controlling the search range facilitates subsequent comparative analysis with the uniform distribution of the source domain.

[0064] Step S103: Based on multiple current structural parameters of the steam compressor and the target domain of the structural parameters of the steam compressor, calculate the sampling probability density of the target domain of the structural parameters of the steam compressor.

[0065] In this application, firstly, a probability density function corresponding to the target domain of the structural parameters of the steam compressor can be constructed. Then, multiple current structural parameters of the steam compressor can be input into the probability density function corresponding to the target domain of the structural parameters to obtain the probability density of the multiple current structural parameters of the steam compressor under the target domain of the structural parameters. Finally, the probability density of the multiple current structural parameters of the steam compressor under the target domain of the structural parameters can be determined as the sampling probability density of the target domain of the structural parameters.

[0066] For any structural parameter of a steam compressor Its target domain of structural parameters The probability density is as follows This function measures With target structural parameters The closer: the closer The higher the probability density, the farther away; The lower the probability density, the better. Multiple current structural parameters of the steam compressor are input into the probability density function corresponding to the structural parameter target domain of the steam compressor to obtain the probability density of multiple current structural parameters of the steam compressor under the structural parameter target domain. For example, the 2000 sets of current structural parameters generated in step S101... Substituting each parameter into the above formula, we obtain the probability density of multiple current structural parameters of the steam compressor under the target domain of structural parameters. The probability density of multiple current structural parameters of the steam compressor under the structural parameter target domain is the sampling probability density of the structural parameter target domain.

[0067] Step S104: Based on the sampling probability density of the source domain of the structural parameters of the steam compressor and the sampling probability density of the target domain of the structural parameters of the steam compressor, calculate the sampling probability density ratio of the steam compressor.

[0068] Obtain multiple current structural parameters of the steam compressor in the structural parameter source domain. The probability density of the steam compressor in the structural parameter source domain is given by the following parameters. The probability density below is the sampling probability density of the source domain of the structural parameters of the steam compressor. Sampling probability density of the source domain based on the structural parameters of a steam compressor. Sampling probability density of the target domain of structural parameters of steam compressor The sampling probability density ratio of the steam compressor was calculated. ,in, This indicates that the probability density of each structural parameter sampled within the source domain is the same. Therefore, the sampling probability density of this steam compressor is higher than that of other structural parameters. Essentially, it is the magnification factor of the target domain probability of the structural parameters relative to the source domain probability of the structural parameters.

[0069] Step S105: Based on multiple current structural parameters of the steam compressor, multiple initial fatigue lives of the steam compressor, and the sampling probability density ratio of the steam compressor, the pre-trained fatigue life prediction model is reconstructed to obtain the reconstructed fatigue life prediction model.

[0070] In this application, firstly, for each current structural parameter of the steam compressor, the current structural parameter of the steam compressor and the initial fatigue life of the steam compressor corresponding to the current structural parameter can be determined as a sampling sample of the steam compressor. Then, based on the sampling probability density ratio of the steam compressor, weights can be assigned to multiple sampling samples of the steam compressor to obtain the weights of multiple sampling samples of the steam compressor. Finally, based on the multiple sampling samples of the steam compressor and their weights, a weighted loss function can be used to retrain the pre-trained fatigue life prediction model to obtain the reconstructed fatigue life prediction model.

[0071] Specifically, for each current structural parameter of the steam compressor, the current structural parameters of the steam compressor are... And the initial fatigue life of the steam compressor corresponding to the current structural parameters of the steam compressor. A sample identified as a steam compressor The sampling probability density ratio calculated using step S104 Multiple sampled samples from the steam compressor are assigned corresponding weights. These weights reflect the relative importance of each sample within the current optimization objective (i.e., the region near the target structural parameters); samples closer to the current design point have higher weights. The larger the value, the stronger its influence in model training; samples far from the target region are automatically weakened. This non-uniform weighting mechanism effectively avoids the local accuracy loss caused by global average fitting, making the model learning process more targeted and efficient. Finally, based on the weighted sample set, a weighted loss function is constructed, and the pre-trained fatigue life prediction model is retrained using this function as the target. Specifically, by minimizing the weighted mean square error, i.e., minimizing the prediction bias of high-weight samples, the neural network parameters are dynamically optimized to obtain a reconstructed fatigue life prediction model specifically adapted to the current design region. The reconstructed fatigue life prediction model inherits the generalization ability of the source domain model, while exhibiting higher prediction accuracy in the target domain, realizing an intelligent evolution from a general agent to a local agent.

[0072] The core of the weight reconstruction process is to introduce the sampling probability density ratio. As a weighting factor, the transfer samples are differentiated, thereby guiding the neural network model to learn the performance patterns of the current optimization target region more efficiently and accurately. The key to this mechanism is transforming the originally uniformly distributed training data into priority-based supervision signals, enabling the model to focus on learning important samples and ignore irrelevant samples during retraining. In the weighted loss function... middle, The number of transfer samples, and the predicted response value for each sample. (i.e., initial fatigue life) is assigned a weight. This weight determines its influence during the training process. When When this value is 0, it indicates that the sample is located in the core region of the target domain near the current design point, possessing high engineering value and information density. The predicted response corresponding to this sample at this time... These are considered high-priority supervisory signals, and the model will focus on learning their input-output relationships to improve prediction accuracy in key regions. For example, if a set of structural parameters is close to the optimal solution, its fatigue life prediction results should be given high priority to ensure that the model can accurately capture the nonlinear behavior in that region. Conversely, when... When this occurs, it indicates that the sample is far from the current optimization focus and belongs to the marginal or low-relevance region. At this point, the predicted response for the sample... The model is assigned low weights, or even ignored in its learning. This prevents the model from deviating from the main optimization direction due to overfitting non-critical regions, effectively preventing resource waste and overfitting risks. It achieves an intelligent transition from global generalization to local focus, so that the reconstructed fatigue life prediction model not only inherits the breadth of knowledge from the source domain, but also has the depth capability of being highly sensitive to specific design points, truly realizing dynamic, adaptive, and efficient model transfer and optimization.

[0073] Step S106: Input multiple current structural parameters of the steam compressor into the reconstructed fatigue life prediction model to obtain multiple current fatigue lives of the steam compressor.

[0074] In this application, multiple current structural parameters of the steam compressor are input into the reconstructed fatigue life prediction model to obtain multiple corresponding current fatigue lives. These current structural parameters refer to the latest generated or updated set of structural design schemes during the optimization iteration process (e.g., a combination of R1, L1, and L2 parameters sampled around the current design point). The reconstructed fatigue life prediction model refers to the target domain-specific neural network model obtained after weighted retraining based on the sampling probability density ratio in step S105. This model has undergone local accuracy enhancement for the current design region, enabling it to more accurately reflect the fatigue response characteristics of the steam compressor near the target structure. By batch inputting these current structural parameters into this high-precision surrogate model, the corresponding current fatigue life prediction values ​​can be quickly output.

[0075] Furthermore, the method may further include the following steps: generating multiple random structural parameters of the steam compressor based on the target domain of the steam compressor's structural parameters; inputting these multiple random structural parameters into the reconstructed fatigue life prediction model to obtain multiple random fatigue lives of the steam compressor; calculating the fatigue failure probability of the steam compressor under the target structural parameters based on these multiple random fatigue lives; when the fatigue failure probability of the steam compressor under the target structural parameters is greater than a preset fatigue failure probability threshold, updating the target structural parameters of the steam compressor based on the multiple random structural parameters and the fatigue failure probability of the steam compressor under the target structural parameters, and returning to the step of constructing the target domain of the steam compressor's structural parameters based on the target structural parameters, until the fatigue failure probability of the steam compressor under the target structural parameters is not greater than the fatigue failure probability threshold.

[0076] Specifically, to further evaluate the reliability of the current design scheme, the target domain of the already constructed steam compressor structural parameters can be used as a basis. Generate a large number of random structure parameters i = 1, 2, ..., N mcs , where N mcs Typically, these samples range from hundreds to thousands (e.g., 2000). They are obtained from the target domain using the Monte Carlo method to simulate parameter fluctuations and uncertainties that may exist during actual manufacturing and operation. Subsequently, these random structure parameters... Inputting the data into the reconstructed fatigue life prediction model quickly yields multiple corresponding stochastic fatigue life prediction values. ,in The unit is the number of cycles, representing the expected fatigue life of the structure under typical operating conditions of a steam compressor.

[0077] To determine whether each sample meets safety requirements, a preset fatigue life threshold needs to be set. This fatigue life threshold It can be determined by engineering specifications or design briefs (e.g.) Each cycle corresponds to the equipment's design life. This applies to each random fatigue life of the steam compressor. Determining the random fatigue life of a steam compressor Is it greater than the preset fatigue life threshold? If the random fatigue life of the steam compressor... Not greater than the preset fatigue life threshold This indicates that the combination of structural parameters does not have sufficient fatigue life under the current operating conditions, and it belongs to the failure sample. Therefore, the random fatigue life of the steam compressor will be... The first target fatigue life of the steam compressor was determined. The number of first target fatigue lives of the steam compressor obtained was [number missing]. , hour ; hour And the number of random fatigue lives of steam compressors. The fatigue failure probability of the steam compressor under the target structural parameters was calculated. .

[0078] For each random fatigue life of a steam compressor Determine whether the random fatigue life of the steam compressor exceeds a preset fatigue life threshold. If the random fatigue life of the steam compressor... Greater than the preset fatigue life threshold This indicates that the combination of structural parameters has sufficient fatigue life under the current operating conditions and belongs to the safe sample. Therefore, the random fatigue life of the steam compressor will be... The second target fatigue life was determined for the steam compressor. Based on the stochastic structural parameters of the steam compressor corresponding to the second target fatigue life, the mean value of the stochastic structural parameters of the steam compressor is calculated. Based on the mean value of the stochastic structural parameters of the steam compressor, the target structural parameters of the steam compressor are updated to achieve migration to the safe region.

[0079] Furthermore, the gradient of the failure probability with respect to each structural parameter can be calculated using the first-order scoring function method, the expression of which is: ,in, Represents the failure probability gradient; This represents the joint probability density, defined by the probability model. This gradient guides the optimization algorithm to search in the direction of reducing failure risk until the objective function changes by less than 0.5% in three consecutive iterations and all constraints are satisfied, for example... The final output is the optimal design, with an isentropic efficiency of ≥90% or a shaft power of ≥500kW.

[0080] After the update is complete, the process returns to the step of "constructing the target domain of structural parameters based on the new target structural parameters," regenerating the target domain, sampling, predicting, and evaluating, forming a dynamic closed-loop optimization iterative process. This process is repeated until the probability of fatigue failure drops below the threshold, while simultaneously satisfying other engineering constraints such as efficiency and power.

[0081] In summary, this application provides a method for predicting the fatigue life of a steam compressor. This method samples the source domain of structural parameters of the steam compressor to obtain multiple current structural parameters, which are then input into a pre-trained fatigue life prediction model to obtain multiple initial fatigue lives. A target structural parameter is determined from the source domain, and a target domain is constructed based on this target parameter. The sampling probability density of the target domain is calculated based on the multiple current structural parameters and the target domain. The sampling probability density ratio is further calculated by combining the obtained sampling probability density of the source domain and the calculated sampling probability density of the target domain. Using this sampling probability density ratio, multiple current structural parameters, and their corresponding initial fatigue lives, the pre-trained fatigue life prediction model is reconstructed to obtain a reconstructed fatigue life prediction model. Finally, the multiple current structural parameters are input into the reconstructed fatigue life prediction model to output more accurate current fatigue lives, thereby effectively avoiding prediction distortion caused by the distribution offset between the source and target domains and improving the accuracy of the predicted fatigue life of the steam compressor.

[0082] This application also provides an electronic device in its embodiments. (See reference...) Figure 3 The diagram illustrates a structural schematic suitable for implementing the electronic device in the embodiments of this application. The electronic device in the embodiments of this application may include, but is not limited to, fixed terminals such as mobile phones, laptops, PDAs (personal digital assistants), PADs (tablet computers), desktop computers, etc. Figure 3 The electronic device shown is merely an example and should not impose any limitation on the functionality and scope of use of the embodiments of this application.

[0083] like Figure 3As shown, the electronic device may include a processing unit (e.g., a central processing unit, a graphics processing unit, etc.) 301, which can perform various appropriate actions and processes according to a program stored in a read-only memory (ROM) 302 or a program loaded from a storage device 308 into a random access memory (RAM) 303. When the electronic device is powered on, the RAM 303 also stores various programs and data required for the operation of the electronic device. The processing unit 301, ROM 302, and RAM 303 are interconnected via a bus 304. An input / output (I / O) interface 305 is also connected to the bus 304.

[0084] Typically, the following devices can be connected to I / O interface 305: input devices 306 including, for example, touchscreens, touchpads, keyboards, mice, cameras, microphones, accelerometers, gyroscopes, etc.; output devices 307 including, for example, liquid crystal displays (LCDs), speakers, vibrators, etc.; storage devices 308 including, for example, memory cards, hard drives, etc.; and communication devices 309. Communication device 309 allows electronic devices to communicate wirelessly or wiredly with other devices to exchange data. Although Figure 3 Electronic devices with various devices are shown, but it should be understood that it is not required to implement or have all of the devices shown. More or fewer devices may be implemented or have alternatively.

[0085] This application also provides a computer program product including computer-readable instructions, which, when executed on an electronic device, cause the electronic device to implement any of the methods for predicting the fatigue life of a steam compressor provided in this application.

[0086] This application also provides a computer-readable storage medium carrying one or more computer programs. When the one or more computer programs are executed by an electronic device, the electronic device can implement any of the methods for predicting the fatigue life of a steam compressor provided in this application.

[0087] It should also be noted that the device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. In addition, in the device embodiment drawings provided in this application, the connection relationship between modules indicates that they have a communication connection, which can be implemented as one or more communication buses or signal lines.

[0088] Through the above description of the embodiments, those skilled in the art can clearly understand that this application can be implemented by means of software plus necessary general-purpose hardware, or it can be implemented by special-purpose hardware including application-specific integrated circuits, special-purpose CPUs, special-purpose memory, special-purpose components, etc. Generally, any function performed by a computer program can be easily implemented by corresponding hardware, and the specific hardware structure used to implement the same function can also be diverse, such as analog circuits, digital circuits, or special-purpose circuits. However, for this application, software program implementation is more often the preferred implementation method. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product is stored in a readable storage medium, such as a computer floppy disk, USB flash drive, mobile hard disk, ROM, RAM, magnetic disk, or optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, training equipment, or network device, etc.) to execute the methods described in the various embodiments of this application.

[0089] In the above embodiments, the implementation can be achieved, in whole or in part, through software, hardware, firmware, or any combination thereof. When implemented in software, it can be implemented, in whole or in part, in the form of a computer program product.

[0090] The computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, all or part of the processes or functions described in the embodiments of this application are generated. The computer may be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions may be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, the computer instructions may be transmitted from one website, computer, training device, or data center to another website, computer, training device, or data center via wired (e.g., coaxial cable, fiber optic, digital subscriber line (DSL)) or wireless (e.g., infrared, wireless, microwave, etc.) means. The computer-readable storage medium may be any available medium that a computer can store or a data storage device such as a training device or data center that integrates one or more available media. The available media may be magnetic media (e.g., floppy disks, hard disks, magnetic tapes), optical media (e.g., DVDs), or semiconductor media (e.g., solid-state drives (SSDs)).

Claims

1. A method for predicting the fatigue life of a steam compressor, characterized in that, include: The structural parameter source domain of the steam compressor is sampled to obtain multiple current structural parameters of the steam compressor, and the multiple current structural parameters of the steam compressor are input into a pre-trained fatigue life prediction model to obtain multiple initial fatigue lives of the steam compressor. The target structural parameters of the steam compressor are determined from the source domain of the structural parameters of the steam compressor, and the target domain of the structural parameters of the steam compressor is constructed based on the target structural parameters of the steam compressor. Based on multiple current structural parameters of the steam compressor and the target domain of the structural parameters of the steam compressor, the sampling probability density of the target domain of the structural parameters of the steam compressor is calculated. Based on the sampling probability density of the source domain of the structural parameters of the steam compressor and the sampling probability density of the target domain of the structural parameters of the steam compressor, the sampling probability density ratio of the steam compressor is calculated. Based on multiple current structural parameters of the steam compressor, multiple initial fatigue lives of the steam compressor, and the sampling probability density ratio of the steam compressor, the pre-trained fatigue life prediction model is reconstructed to obtain the reconstructed fatigue life prediction model. Multiple current structural parameters of the steam compressor are input into the reconstructed fatigue life prediction model to obtain multiple current fatigue lives of the steam compressor.

2. The method for predicting the fatigue life of a steam compressor according to claim 1, characterized in that, After inputting multiple current structural parameters of the steam compressor into the reconstructed fatigue life prediction model to obtain multiple current fatigue lives of the steam compressor, the method further includes: Based on the target domain of the structural parameters of the steam compressor, multiple random structural parameters of the steam compressor are generated; and the multiple random structural parameters of the steam compressor are input into the reconstructed fatigue life prediction model to obtain multiple random fatigue lives of the steam compressor. Based on multiple random fatigue lives of the steam compressor, the fatigue failure probability of the steam compressor under the target structural parameters is calculated. When the fatigue failure probability of the steam compressor under the target structural parameters is greater than a preset fatigue failure probability threshold, the target structural parameters of the steam compressor are updated based on multiple random structural parameters of the steam compressor and the fatigue failure probability of the steam compressor under the target structural parameters. Then, the process returns to the step of constructing the target domain of the structural parameters of the steam compressor based on the target structural parameters of the steam compressor, until the fatigue failure probability of the steam compressor under the target structural parameters is not greater than the fatigue failure probability threshold.

3. The method for predicting the fatigue life of a steam compressor according to claim 1, characterized in that, The training process of the fatigue life prediction model includes: The structural parameter source domain of the steam compressor is sampled to obtain multiple training structural parameters of the steam compressor; Physical simulations were performed on multiple training structural parameters of the steam compressor to obtain multiple training fatigue lives of the steam compressor. Based on multiple training structural parameters and multiple training fatigue lives of the steam compressor, a neural network model is trained to obtain a trained fatigue life prediction model. The input of the fatigue life prediction model is the multiple structural parameters of the steam compressor, and the output of the fatigue life prediction model is the multiple fatigue lives of the steam compressor.

4. The method for predicting the fatigue life of a steam compressor according to claim 1, characterized in that, The step of calculating the sampling probability density of the structural parameter target domain of the steam compressor based on multiple current structural parameters of the steam compressor and the structural parameter target domain of the steam compressor includes: Construct the probability density function corresponding to the target domain of the structural parameters of the steam compressor; Input multiple current structural parameters of the steam compressor into the probability density function corresponding to the structural parameter target domain of the steam compressor to obtain the probability density of multiple current structural parameters of the steam compressor under the structural parameter target domain; The probability density of multiple current structural parameters of the steam compressor under the structural parameter target domain is determined as the sampling probability density of the structural parameter target domain.

5. The method for predicting the fatigue life of a steam compressor according to claim 1, characterized in that, The pre-trained fatigue life prediction model is reconstructed based on multiple current structural parameters of the steam compressor, multiple initial fatigue lives of the steam compressor, and the sampling probability density ratio of the steam compressor, to obtain a reconstructed fatigue life prediction model, including: For each current structural parameter of the steam compressor, the current structural parameter of the steam compressor and the initial fatigue life of the steam compressor corresponding to the current structural parameter of the steam compressor are determined as a sampling sample of the steam compressor. The weights of the multiple samples of the steam compressor are obtained by weighting the multiple samples of the steam compressor based on the sampling probability density ratio of the steam compressor. Based on multiple sampled samples from the steam compressor and the weights of these samples, a weighted loss function is used to retrain the pre-trained fatigue life prediction model, resulting in the reconstructed fatigue life prediction model.

6. The method for predicting the fatigue life of a steam compressor according to claim 2, characterized in that, The calculation of the fatigue failure probability of the steam compressor under the target structural parameters based on multiple random fatigue lives of the steam compressor includes: For each random fatigue life of the steam compressor, determine whether the random fatigue life of the steam compressor is greater than a preset fatigue life threshold. If the random fatigue life of the steam compressor is not greater than the preset fatigue life threshold, then the random fatigue life of the steam compressor is determined as the first target fatigue life of the steam compressor. Based on the number of first target fatigue lives of the steam compressor and the number of random fatigue lives of the steam compressor, the fatigue failure probability of the steam compressor under the target structural parameters is calculated.

7. The method for predicting the fatigue life of a steam compressor according to claim 2, characterized in that, The updating of the target structural parameters of the steam compressor based on multiple random structural parameters of the steam compressor and the fatigue failure probability of the steam compressor under the target structural parameters includes: For each random fatigue life of the steam compressor, determine whether the random fatigue life of the steam compressor is greater than a preset fatigue life threshold. If the random fatigue life of the steam compressor is greater than the preset fatigue life threshold, then the random fatigue life of the steam compressor is determined as the second target fatigue life of the steam compressor. Based on the random structural parameters of the steam compressor corresponding to the second target fatigue life of the steam compressor, the mean value of the random structural parameters of the steam compressor is calculated. The target structural parameters of the steam compressor are updated based on the mean of the random structural parameters of the steam compressor.

8. A computer program product, characterized in that, Includes computer-readable instructions that, when executed on an electronic device, cause the electronic device to implement the method for predicting the fatigue life of a steam compressor as described in any one of claims 1 to 7.

9. An electronic device, characterized in that, It includes at least one processor and a memory connected to the processor, wherein: The memory is used to store computer programs; The processor is used to execute the computer program to enable the electronic device to implement the method for predicting the fatigue life of a steam compressor as described in any one of claims 1 to 7.

10. A computer storage medium, characterized in that, The storage medium carries one or more computer programs that, when executed by an electronic device, enable the electronic device to implement the method for predicting the fatigue life of a steam compressor as described in any one of claims 1 to 7.

Images