Creep degree prediction method and device, computer equipment, readable storage medium and program product

Abstract

The invention relates to a creep degree prediction method and device, computer equipment, a computer readable storage medium and a computer program product. The method comprises the steps that a target training set is acquired, the target training set comprises a first training set and a second training set, each training set comprises a plurality of samples, and each sample comprises creep degree time sequence data of a first length and material environment data of a second length; training the initial model through creep degree time sequence data and material environment data of each sample in the first training set to obtain a trained initial creep degree prediction model; the parameters of the initial creep degree prediction model are adjusted through the creep degree time sequence data and the material environment data of all the samples in the second training set, the target model parameters are obtained, the target creep degree prediction model is obtained based on the target model parameters, consideration factors are comprehensive, and the accuracy degree of the creep degree prediction value can be improved.

Description

Technical Field

[0001] The present application relates to the field of computer technology, and in particular to a creep prediction method, apparatus, computer equipment, computer-readable storage medium, and computer program product. Background Art

[0002] Concrete creep refers to the property that the strain of concrete continues to increase over time under long-term stress. Under long-term load, the stress on the structure or material remains unchanged, while the strain increases over time. This phenomenon is called creep. In traditional technology, a variety of models have been proposed for the prediction of creep, such as the B3 and B4 models. However, these models only predict concrete creep based on concrete materials and load time, and the factors considered are relatively single, resulting in poor accuracy of the creep prediction results of the models. Summary of the invention

[0003] Based on this, it is necessary to provide a creep prediction method, device, computer equipment, computer-readable storage medium and computer program product that consider multiple factors to improve the accuracy of creep prediction in order to address the above technical problems.

[0004] In a first aspect, the present application provides a creep prediction method, comprising:

[0005] Acquire a target training set, wherein the target training set includes a first training set and a second training set, each of the training sets includes a plurality of samples, and each of the samples includes creep time series data of a first length and material environment data of a second length;

[0006] Training the initial model using the creep time series data of each sample in the first training set and the material environment data to obtain a trained initial creep prediction model;

[0007] The parameters of the initial creep prediction model are adjusted by using the creep time series data of each sample in the second training set and the material environment data to obtain target model parameters, and a target creep prediction model is obtained based on the target model parameters.

[0008] In one embodiment, the creep prediction method further includes:

[0009] Acquire creep time series data of a first target length in a target scene and material environment data corresponding to the creep time series data;

[0010] The creep time series data of the first target length and the material environment data corresponding to the creep time series data are input into the target creep prediction model to obtain the predicted creep at the next moment of the creep time series data.

[0011] In one embodiment, the creep prediction method further includes:

[0012] Collecting initial experimental data of multiple creep experiments, the initial experimental data including experimental numbers, material environment data and original creep time series data, the original creep time series data including time series and creep series, the time series including multiple time points, each of which corresponds to a first creep degree and a second creep degree;

[0013] Performing abnormal supplementation processing on the initial experimental data to obtain first experimental data;

[0014] Based on a preset time interval, sampling the original creep time series data in the first experimental data to obtain target creep time series data with the same time interval;

[0015] Normalization processing is performed based on the material environment data of each creep experiment and the target creep degree time series data in the first experimental data to obtain target experimental data.

[0016] In one embodiment, the creep prediction method further includes:

[0017] Dividing the target creep time series data in the target experimental data to obtain first time series training data, second time series training data and test time series data;

[0018] According to the experimental numbers of the creep experiments, material environment data are added to the first time series training data, the second time series training data and the test time series data to obtain the first training set, the second training set and the test set.

[0019] In one embodiment, the target creep time series data in the target experimental data is divided to obtain first time series training data, second time series training data and test time series data, including:

[0020] Dividing the target creep time series data in the target experimental data according to a preset ratio to obtain a prediction set and a non-prediction set;

[0021] Performing sample division processing on the non-prediction set to obtain first time series training data;

[0022] Performing sample partitioning processing on the prediction set to obtain a target number of samples;

[0023] The target number of samples are extracted respectively based on the first prediction ratio and the second prediction ratio to obtain second time series training data and test time series data, wherein the first prediction ratio is determined based on the second prediction ratio.

[0024] In one embodiment, the performing sample partitioning processing on the non-prediction set to obtain first time series training data includes:

[0025] According to a preset sliding window, the time series in the target creep degree time series data of each creep experiment is slidingly divided to obtain a first number of time series samples; and the creep degree of the time series sample at the next moment is determined as a sample label; based on each of the time series samples and the sample label of each of the time series samples, the first time series training data is obtained;

[0026] The performing sample division processing on the prediction set to obtain a target number of samples includes:

[0027] According to the preset sliding window, the time series in the target creep degree time series data of each creep experiment is slidingly divided to obtain a target number of time series samples; and the creep degree of the time series sample at the next moment is determined as the sample label; based on each of the time series samples and the sample label of each of the time series samples, the target number of samples is obtained.

[0028] In one embodiment, the initial model is trained by using the creep time series data of each sample in the first training set and the material environment data to obtain the trained initial creep prediction model, including:

[0029] Obtaining a first training set, and inputting the creep time series data and the material environment data of each sample in the first training set into an initial model;

[0030] By using the long short-term memory network in the initial model, feature extraction is performed on the time series of each of the samples in the first training set to obtain latent variables, and the latent variables are determined to be creep time series features of the samples;

[0031] Splicing the creep time series characteristics of the sample and the material environment data to obtain a splicing vector, and performing perception processing on the splicing vector through a multi-layer perceptron module in a hidden layer of the initial model to obtain a creep prediction value at the next moment of the time series of the sample;

[0032] Based on the true value of the creep at the next moment and the predicted value of the creep at the next moment, the loss is calculated. When the preset training completion condition is not met, back propagation processing is performed based on the loss to obtain the gradient of the model parameters of the initial model, and based on the gradient of the model parameters, an updated initial model is obtained; based on the updated initial model, the step of obtaining the first training set is re-executed until the preset training completion condition is met to obtain the trained initial creep prediction model.

[0033] In one embodiment, the creep prediction method further includes:

[0034] For each sample at the initial position of the creep experiment in the test set, predict the sample by using the target creep prediction model to obtain a creep prediction value of the sample at the next moment;

[0035] Processing the creep prediction value at the next moment and the sample at the next position of the initial position of the creep experiment by the target creep prediction model to obtain the creep prediction value at the next moment of the next position;

[0036] Based on the creep prediction value of each sample at the next moment, a creep prediction sequence is obtained;

[0037] Calculating deviation data based on a creep degree prediction sequence corresponding to the creep experiment and a creep degree true sequence corresponding to the creep experiment;

[0038] Based on the deviation data of each creep experiment, the accuracy of the target creep degree prediction model is obtained.

[0039] In a second aspect, the present application also provides a creep prediction device, comprising:

[0040] A first acquisition module is used to acquire a target training set, wherein the target training set includes a first training set and a second training set, each of the training sets includes a plurality of samples, and each of the samples includes creep time series data of a first length and material environment data of a second length;

[0041] A first training module, used to train the initial model through the creep time series data of each sample in the first training set and the material environment data to obtain a trained initial creep prediction model;

[0042] The first adjustment module is used to adjust the parameters of the initial creep prediction model through the creep time series data of each sample in the second training set and the material environment data to obtain target model parameters, and obtain a target creep prediction model based on the target model parameters.

[0043] In a third aspect, the present application further provides a computer device, including a memory and a processor, wherein the memory stores a computer program, and the processor implements the steps in the embodiments of the present application when executing the computer program.

[0044] In a fourth aspect, the present application further provides a computer-readable storage medium on which a computer program is stored, and when the computer program is executed by a processor, the steps in the embodiments of the present application are implemented.

[0045] In a fifth aspect, the present application also provides a computer program product, including a computer program, which implements the steps in the embodiments of the present application when executed by a processor.

[0046] The above-mentioned creep prediction method, device, computer equipment, computer-readable storage medium and computer program product, by adopting this method, can predict the creep at subsequent times through a deep learning model, concrete-related material data, environmental data and creep time series data, consider the influencing factors more comprehensively, and ensure the accuracy of the predicted creep. BRIEF DESCRIPTION OF THE DRAWINGS

[0047] In order to more clearly illustrate the technical solutions in the embodiments of the present application or related technologies, the drawings required for use in the embodiments of the present application or related technical descriptions will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other related drawings can be obtained based on these drawings without paying creative work.

[0048] Figure 1 is an application environment diagram of a creep prediction method in an embodiment;

[0049] Figure 2 A schematic diagram of a flow chart of a creep prediction method in one embodiment;

[0050] Figure 3 A schematic diagram of a process for obtaining target experimental data in one embodiment;

[0051] Figure 4 A schematic diagram of a flow chart of a data partitioning step in an embodiment;

[0052] Figure 5 A schematic diagram of a flow chart of steps for dividing time series data in one embodiment;

[0053] Figure 6 A schematic diagram of a process of obtaining first time series training data in one embodiment;

[0054] Figure 7 Schematic diagram of a flow chart of a model training step in one embodiment;

[0055] Figure 8 A schematic diagram of a process of dividing data into training sets and test sets in one embodiment;

[0056] Fig. 9 A schematic diagram of dividing time series data in one embodiment;

[0057] Fig.10 A schematic diagram of a structure in which processing is performed through LSTM and MLP layers in one embodiment;

[0058] Fig.11 A schematic diagram of a model validation process in one embodiment;

[0059] Fig.12 is a structural block diagram of a creep prediction device in one embodiment;

[0060] Fig.13 FIG. 4 is a diagram showing the internal structure of a computer device in one embodiment. DETAILED DESCRIPTION

[0061] In order to make the purpose, technical solution and advantages of the present application more clearly understood, the present application is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.

[0062] There are many creep prediction models in the field of creep prediction in traditional technology, such as B3 and B4 models. The above models are generally based only on material data or creep environment data, and are affected by different experimental test conditions. They cannot guarantee a relatively good prediction accuracy in all environments, resulting in poor accuracy of creep prediction. Based on this, the creep prediction method provided in this embodiment can predict non-stationary time series through deep learning technology to obtain more accurate prediction results. In fact, it comprehensively considers the materials of the creep experiment, the environmental parameters of the creep experiment, and the early data of the creep experiment, and predicts the creep degree of subsequent creep experiments through neural networks such as long short-term memory networks to ensure the stability of prediction performance. Specifically, it can include processes such as building a creep experiment database, preprocessing creep data, building and training a neural network model, creep prediction and model evaluation.

[0063] In one embodiment, Figure 1As shown, a creep prediction method is provided. This embodiment uses the method applied to a terminal as an example for illustration. It is understandable that the method can also be applied to a server, and can also be applied to a system including a terminal and a server, and is implemented through the interaction between the terminal and the server. The above-mentioned terminal can be, but is not limited to, various personal computers, laptops, smart phones, tablet computers, etc. The server can be an independent physical server, or a server cluster or distributed system composed of multiple physical servers, or a cloud server providing cloud computing services. In this embodiment, the creep prediction method includes the following steps:

[0064] Step 102: Obtain a target training set.

[0065] Specifically, the target training set includes a first training set and a second training set, each training set includes a plurality of samples, each sample includes creep time series data of a first length and material environment data of a second length; the format of the data contained in the first training set and the second training set is the same, the first training set is used to train the initial model to obtain the trained initial creep prediction model. The second training set is used to adjust the parameters of the trained initial creep prediction model to obtain the target creep prediction model. The creep time series data of the first length includes a time series of the first length and a creep series, the time series includes a plurality of time points arranged in order of time, the creep series includes the creep corresponding to each time point, that is, the creep of the concrete collected at each time point. The terminal can collect multiple creeps at the same time point, for example, each time point can correspond to the first creep and the second creep. The material environment data of the second length includes multiple experimental material parameters and multiple experimental environment parameters, each experimental material parameter may include one or more of cement, water, gravel, sand, water-cement ratio, water reducer, sand ratio, and aggregate volume ratio; the experimental environment parameters include one or more of temperature, humidity, dry test environment, stress direction, and curing age. The terminal can obtain the first training set and the second training set in the target training set.

[0066] Step 104 , training the initial model using the creep time series data and material environment data of each sample in the first training set to obtain a trained initial creep prediction model.

[0067] Among them, the initial model can be a neural network model to be trained, or a deep learning model to be trained, etc.

[0068] Specifically, the terminal can input the creep time series data of each sample contained in the first training set, and the material environment data corresponding to each sample, into the initial model, train the above samples through the initial model, and obtain the creep prediction value corresponding to each sample. The creep prediction value corresponding to each sample is the creep prediction value of the sample at the next moment. Based on the creep prediction value of each sample at the next moment and the creep true value of each sample at the next moment, the loss of model training is calculated, and based on the loss and the preset training completion conditions, the trained initial creep prediction model is obtained.

[0069] Step 106, adjusting the parameters of the initial creep prediction model through the creep time series data and material environment data of each sample in the second training set to obtain target model parameters, and obtaining a target creep prediction model based on the target model parameters.

[0070] Specifically, the terminal can input the creep time series data of each sample contained in the second training set, and the material environment data corresponding to each sample, into the initial creep prediction model, train the above samples through the initial model, and obtain the creep prediction value corresponding to each sample, the creep prediction value corresponding to each sample is the creep prediction value of the sample at the next moment, based on the creep prediction value of each sample at the next moment and the creep true value of each sample at the next moment, calculate the loss of model training, obtain the target model parameters based on the loss, replace the model parameters in the initial creep prediction model based on the target model parameters, and obtain the target creep prediction model.

[0071] In the above creep prediction method, the creep at subsequent moments can be predicted through deep learning models, concrete-related material data, environmental data, and creep time series data, taking into account the influencing factors more comprehensively to ensure the accuracy of the predicted creep.

[0072] In one embodiment, Figure 2 As shown, the creep prediction method also includes:

[0073] Step 202: Acquire creep time series data of a first target length in a target scene and material environment data corresponding to the creep time series data.

[0074] Among them, the target scene can be a scene where the creep of the target object needs to be predicted, and the target object can be a building or structure constructed with concrete materials, etc.; the creep time series data of the first target length can include multiple time points and the creep corresponding to each time point, and the creep corresponding to each time point is the creep of the target object detected at each time point; the material environment data corresponding to the creep time series data can be the environment data of the target scene, the material data of the target object in the target scene, and the environment data related to the target object. In other words, the creep time series data can include a time series and a creep sequence, the time series includes each time point arranged in order of early and late, and the creep sequence includes the creep corresponding to each time point. The present disclosure does not limit the first target length, and those skilled in the art can make a specific determination based on the specific circumstances of the actual application scenario.

[0075] Specifically, the terminal needs to predict the creep of the target scene, and can collect creep time series data and material environment data in the target scene to obtain creep time series data of the first target length of the target scene and material environment data in the target scene.

[0076] Step 204 , input the creep time series data of the first target length and the material environment data corresponding to the creep time series data into the target creep prediction model to obtain the predicted creep at the next moment of the creep time series data.

[0077] The next moment of the creep time series data may be the next moment of the time series in the creep time series data, that is, the next time point of the time series in the creep time series data.

[0078] Specifically, the terminal can input the creep time series data and material environment data of the first target length in the target scene into the target creep prediction model. For example, the terminal can extract features of the creep time series data through the long short-term memory network in the target creep prediction model to obtain the creep time series features, splice the creep time series features with the material environment numbers corresponding to the creep time series data in the target scene to obtain the spliced ​​features, and process the spliced ​​features through the target creep prediction model to obtain the predicted creep, that is, obtain the creep prediction value of the next time point of the time series in the creep time series data.

[0079] In this embodiment, by collecting the creep time series data and material environment data of the target object in the target scene, and processing the creep time series data and material environment data through a trained target creep prediction model, an accurate creep prediction value can be obtained while considering multiple influencing factors.

[0080] In an exemplary embodiment, Figure 3 As shown, the creep prediction method also includes:

[0081] Step 302, collecting initial experimental data of multiple creep experiments.

[0082] Among them, the initial experimental data includes the experimental number, material environment data and original creep time series data. The original creep time series data includes a time series and a creep series. The time series includes multiple time points, and each time point corresponds to a first creep degree and a second creep degree.

[0083] Specifically, the terminal can collect the original experimental data corresponding to each creep experiment, that is, the initial experimental data. Each creep experiment is marked with a unique experimental number to distinguish it from other creep experiments. The material environment data can be the material data and experimental environment data of the concrete creep experiment, including one or more of the following parameters: cement, water, gravel, sand, water-reducing agent, water-cement ratio, sand ratio, aggregate volume ratio, loading force, temperature, test environment, stress direction, and curing age. The first creep degree and the second creep degree corresponding to each time point are at that time point. The creep degree collected for the first time is the first creep degree, and the creep degree collected for the second time is the second creep degree; the original creep degree time series data of the creep experiment can be shown in Table 1 below:

[0084] Table 1

[0085]

[0086] Step 304, perform abnormal supplementation processing on the initial experimental data to obtain first experimental data. Based on a preset time interval, sample the original creep time series data in the first experimental data to obtain target creep time series data with the same time interval.

[0087] Among them, the preset time interval can be a pre-configured sampling time interval, and the specific value of the preset time interval can be determined based on the experimental monitoring time range of the creep experiment, and the experimental monitoring time range of the creep experiment can be the load age period; optionally, for the initial experimental data with a load age period of 28 days, the corresponding preset time interval is 8 hours; for the initial experimental data with a load age period of 5-10 years, the corresponding preset time interval can be a value within 20 days-40 days.

[0088] Specifically, the terminal can extract abnormal data from the initial experimental data, process the abnormal data, and obtain the first experimental data; the abnormal data may include jump point data and missing data; the specific processing process may be: the terminal can use the threshold method to filter the initial experimental data to obtain abnormal data, and fill the abnormal data by the moving average method to obtain filled data; the terminal can replace the abnormal data in the initial experimental data by the filled data to obtain the first experimental data. The terminal can sample the original creep time series data in the first experimental data based on a preset time interval to obtain creep time series data with the same time interval, that is, to obtain the target creep time series data.

[0089] Step 306 , performing normalization processing based on the material environment data of each creep experiment and the target creep degree time series data in the first experimental data to obtain target experimental data.

[0090] Specifically, the terminal can normalize the material environment data and target creep degree time series data of each creep experiment in the first experimental data through a preset normalization algorithm to obtain normalized material environment data and normalized target creep degree time series data, that is, to obtain target experimental data.

[0091] Optionally, the preset normalization algorithm is a maximum-minimum normalization algorithm.

[0092] In one example, the terminal can normalize the target creep time series data of each creep experiment in the first experimental data by a maximum and minimum value normalization algorithm to obtain normalized creep time series data. The terminal can obtain the target experimental data based on the normalized creep time series data corresponding to each creep experiment and the material environment data of each creep experiment. For example, the normalized creep time series data can be calculated by the following formula:

[0093] x'=(x-min) / (max-min)

[0094] Among them, x is the target creep time series data, such as creep; x' is the target creep time series data after normalization, such as creep after normalization; min is the minimum value of creep in each target creep time series data, and max is the maximum value of creep in each target creep time series data.

[0095] In this embodiment, by preprocessing and normalizing the original experimental data, the consistency of the data format can be guaranteed, providing a reliable data basis for the model training process and the model application process.

[0096] In an exemplary embodiment, Figure 4 As shown, the creep prediction method also includes:

[0097] Step 402, dividing the target creep time series data in the target experimental data to obtain first time series training data, second time series training data and test time series data.

[0098] Specifically, the terminal can divide the target time delay data to obtain the target creep degree time series data corresponding to each creep experiment, and the material environment data corresponding to each creep experiment. In this way, the terminal can divide the target creep degree time series data to obtain a prediction set and a non-prediction set, and obtain the first time series training data, the second time series training data and the test time series data based on the prediction set and the non-prediction set. The prediction set is the experimental data used to predict the development of creep degree, and the non-prediction set is the experimental data not used to predict the development of creep degree.

[0099] Step 404 , according to the experimental number of each creep experiment, material environment data is added to the first time series training data, the second time series training data and the test time series data respectively to obtain a first training set, a second training set and a test set.

[0100] Specifically, the terminal can obtain the material environment data corresponding to each experimental number based on the experimental number of each creep experiment contained in the first time series training data, and obtain the first training set based on the material environment data of each creep experiment and the first time series training data; correspondingly, the terminal can obtain the material environment data corresponding to each experimental number based on the experimental number of each creep experiment contained in the second time series training data, and obtain the second training set based on the material environment data of each creep experiment and the second time series training data; similarly, the terminal can obtain the material environment data corresponding to each experimental number based on the experimental number of each creep experiment contained in the test time series data, and obtain the test training set based on the material environment data of each creep experiment and the test time series data.

[0101] In this embodiment, the target experimental data is randomly divided into multiple data sets according to the experimental labels to obtain multiple data sets for different purposes, thereby ensuring the randomness of the data and improving the quality of the data used for model training and model testing.

[0102] In an exemplary embodiment, Figure 5 As shown, the specific implementation process of the step of "dividing the target creep time series data in the target experimental data to obtain the first time series training data, the second time series training data and the test time series data" may include:

[0103] Step 502 , dividing the target creep time series data in the target experimental data according to a preset ratio to obtain a prediction set and a non-prediction set.

[0104] The preset ratio may be pre-configured, for example, may be determined based on an actual application scenario. Optionally, the preset ratio may be 9:1.

[0105] Specifically, the terminal can divide the target creep time series data in the target experimental data according to the experimental numbers and preset ratios of each creep experiment, and obtain creep time series data of a first ratio and creep time series data of a second ratio. The terminal can determine the creep time series data of the first ratio as a prediction set, and determine the creep time series data of the second ratio as a non-prediction set. The first ratio can be smaller than the second ratio. For example, when the preset ratio is 9:1, the first ratio can be 10%, and the second ratio can be 90%.

[0106] Step 504: perform sample partitioning processing on the non-prediction set to obtain first time series training data.

[0107] Specifically, the terminal may perform sliding division on the creep time series data in the non-prediction set through a preset sliding window to obtain a plurality of samples, and obtain the first time series training data based on each sample.

[0108] Step 506: perform sample partitioning processing on the prediction set to obtain a target number of samples.

[0109] Specifically, the terminal can perform sliding division on the creep time series data in the prediction set through a preset sliding window to obtain a target number of samples. Optionally, since each creep time series data corresponds to a first creep and a second creep at each time point, the terminal can perform sliding division on the time series in the creep time series data in the prediction set and the first creep sequence according to the preset sliding window to obtain a target number of samples.

[0110] Step 508 , extracting the target number of samples based on the first prediction ratio and the second prediction ratio respectively to obtain second time series training data and test time series data.

[0111] The first prediction ratio is determined based on the second prediction ratio. The second prediction ratio may be β, and the corresponding first prediction ratio may be 1-β.

[0112] Specifically, the terminal may determine the target number of samples of the first prediction ratio as the second time series training data, and determine the target number of samples of the second prediction ratio as the test time series data. Optionally, the terminal may also determine the time series in the creep time series data in the prediction set and the creep corresponding to each time series as the control data set.

[0113] In this embodiment, the target creep time series data is divided according to the first prediction ratio and the second ratio, so as to improve the quality of the training data.

[0114] In an exemplary embodiment, Figure 6 As shown, the specific implementation process of the step of "performing sample division processing on the non-prediction set to obtain the first time series training data" may include:

[0115] Step 602: Slidingly divide the time series in the target creep degree time series data of each creep experiment according to the preset sliding window to obtain a first number of time series samples. And determine the creep degree of the time series sample at the next moment as the sample label. Based on each time series sample and the sample label of each time series sample, obtain the first time series training data.

[0116] The length of the preset sliding window may be s, and the specific value of s may be determined based on an actual application scenario; the length of the time series in the target creep degree time series of each creep experiment may be denoted as L.

[0117] Specifically, for the target creep time series data of each creep experiment, the terminal can perform sliding sampling processing on the target creep time series data through a preset sliding window to obtain a first number of time series samples, and the length of each time series sample is the length of the preset sliding window. The terminal can also determine the creep degree of the time series of each time series sample at the next moment after the target experimental data as the sample label of the time series sample, obtain a sample based on each time series sample and the sample label, and form the first time series training data based on multiple samples. Optionally, the first number n can be calculated by the following formula, n=L-s+1. The first number can be obtained based on the first target value, the difference between the length of the target creep time series sequence and the length of the preset sliding window, and the first target value can be 1.

[0118] Optionally, the specific implementation process of the step of "performing sample division processing on the prediction set to obtain a target number of samples" may include:

[0119] According to the preset sliding window, the time series in the target creep degree time series data of each creep experiment is slidingly divided to obtain the target number of time series samples. And the creep degree of the next moment of the time series sample is determined as the sample label. Based on each time series sample and the sample label of each time series sample, the target number of samples is obtained.

[0120] Specifically, for the target creep time series data of each creep experiment in the prediction set, the terminal can perform sliding sampling processing on the target creep time series data through a preset sliding window to obtain a first number of time series samples, and the length of each time series sample is the length of the preset sliding window. The terminal can also determine the creep degree of the next moment of the time series of each time series sample after the target experimental data as a sample label of the time series sample, and obtain multiple samples based on each time series sample and the sample label.

[0121] In this embodiment, by dividing the target creep time series data by a preset sliding window, the quality of the obtained time series data can be improved, providing a reliable data basis for the subsequent model training process, and further improving the model prediction effect of the creep.

[0122] In one embodiment, Figure 7 As shown, the specific processing process of the step "training the initial model through the creep time series data and material environment data of each sample in the first training set to obtain the trained initial creep prediction model" includes:

[0123] Step 702, obtaining a first training set, and inputting creep time series data and material environment data of each sample in the first training set into an initial model.

[0124] Specifically, the terminal may input the creep time series data of each sample included in the first training set, and the material environment data corresponding to each sample, into the initial model.

[0125] Step 704, extracting features from the time series of each sample in the first training set through the long short-term memory network in the initial model to obtain latent variables, and determining the latent variables as the creep time series features of the samples.

[0126] Among them, the initial model may include at least a long short-term memory network and a hidden layer. The long short-term memory network may include multiple LSTM computing units, each LSTM computing unit corresponds to each time point of the creep time series data of the sample one by one, for example, each LSTM computing unit may include a first LSTM computing layer, a second LSTM computing layer and a third LSTM computing layer. The first LSTM computing layer in the LSTM computing unit is connected to the second LSTM computing layer, and the second LSTM computing layer is connected to the third LSTM computing layer, that is, the output result of the first LSTM computing layer is input to the second LSTM computing layer, and the output result of the second LSTM computing layer is input to the third LSTM computing layer; the computing layer in each LSTM computing unit is connected head to tail with the LSTM computing layer of the next LSTM computing unit of the LSTM computing unit, that is, the first LSTM computing layer in the current LSTM computing unit is connected to the first LSTM computing layer of the next LSTM computing unit of the current LSTM computing unit, and the output result of the first LSTM computing layer in the current LSTM computing unit is input to the first LSTM computing layer of the next LSTM computing unit, and other structures are similar and will not be repeated here.

[0127] Specifically, for each sample, the terminal can input the creep time series data of the sample into the LSTM computing unit in the long short-term memory network. The terminal can input the creep time series data corresponding to each time point in the creep time series data of the sample into the first LSTM computing layer in each LSTM computing unit, perform feature extraction on the creep time series data of the sample through each LSTM computing layer in each LSTM computing unit, obtain the hidden variable at the last time series, and determine the hidden variable as the creep time series feature of the sample.

[0128] Step 706, splicing the creep time series characteristics and material environment data of the sample to obtain a splicing vector, and performing perception processing on the splicing vector through the multi-layer perceptron module in the hidden layer of the initial model to obtain the creep prediction value of the next moment of the sample time series.

[0129] Specifically, the terminal can splice the creep time series features corresponding to each sample with the material environment data corresponding to each sample to obtain the splicing vectors corresponding to each sample. In this way, the terminal can process the splicing vectors through the MLP layer in the target creep prediction model to obtain the creep prediction value, which is the creep prediction value at the next moment of the time series in the time series data of the sample. For example, the time series of the sample can be t1-t6, and the corresponding creep prediction value predicted by the sample is the creep at time t7.

[0130] Step 708, based on the true value of creep at the next moment and the predicted value of creep at the next moment, the loss is calculated, and when the preset training completion condition is not met, back propagation processing is performed based on the loss to obtain the gradient of the model parameters of the initial model, and based on the gradient of the model parameters, an updated initial model is obtained. Based on the updated initial model, the step of obtaining the first training set is re-executed until the preset training completion condition is met, and a trained initial creep prediction model is obtained.

[0131] Specifically, the terminal can collect the true value of the creep of the sample in the target experimental data, that is, obtain the sample label of the sample, and the sample label of the sample is the true value of the creep at the next moment. The terminal can calculate the loss of the predicted creep value and the true value of the creep of the sample through a preset loss function to obtain the loss corresponding to the sample. If the preset training completion condition is not met at present, that is, the preset number of iterations has not been reached at present, or the calculated loss does not converge, it is determined that the preset training completion condition is not met at present. In this way, the terminal can perform back propagation processing based on the loss to obtain the gradient of the model parameters of the initial model, and calculate the updated model parameters based on the gradient, and replace the model parameters of the initial model based on the updated model parameters to obtain the updated initial model.

[0132] Based on this, the terminal can re-execute the steps of obtaining the first training set in the above embodiment based on the updated initial model, and input the creep time series data and material environment data of each sample in the first training set into the initial model until it is determined that the preset training completion conditions are met, and the current updated initial model is determined as the trained initial creep prediction model.

[0133] Optionally, the terminal may re-execute the steps of the method of the above embodiment until the calculated loss meets the preset convergence condition, or the number of iterations has reached the target number, and it is determined that the preset training completion condition is currently met.

[0134] In this embodiment, by considering the creep time series data and material environmental parameters, the model can be trained by integrating various data to improve the prediction performance of the model. The role of the creep time series characteristics in model training is also enhanced, which further improves the accuracy of model prediction and achieves stable and reliable long-term prediction of creep.

[0135] In an exemplary embodiment, the creep prediction method further includes:

[0136] For each sample at the initial position of the creep experiment in the test set, the target creep prediction model is used to predict the sample to obtain the creep prediction value of the sample at the next moment. The creep prediction value at the next moment and the sample at the next position of the initial position of the creep experiment are processed by the target creep prediction model to obtain the creep prediction value at the next moment of the next position. Based on the creep prediction value of each sample at the next moment, a creep prediction sequence is obtained. Based on the creep prediction sequence corresponding to the creep experiment and the creep true sequence corresponding to the creep experiment, the deviation data is calculated. Based on the deviation data of each creep experiment, the accuracy of the target creep prediction model is obtained.

[0137] Specifically, for each creep experiment in the test set, the creep time series data corresponding to each creep experiment may have a length of L; the terminal may obtain the sample at the initial position in the creep time series data. For example, when the sample length is s, the creep time series data of the sample at the initial position is x1-x s .x s is the true value of creep at the sth time point and the sth time point; the sample at the initial position is processed by the target creep prediction model to obtain the creep prediction value corresponding to the sample, that is, the creep prediction value of the next moment of the time series of the sample is obtained, that is, x s+1 Predicted creep value at time.

[0138] The terminal can calculate the sample of the next position based on the sample of the initial position, and x s+1 The creep prediction value at the moment is used to obtain the updated sample at the next position. For example, the sample at the next position based on the sample at the initial position is x2-x s+1 , correspondingly, the terminal will x2-x s and x s+1 Time point and x s+1 The creep prediction value at time is input into the target creep prediction model to obtain x s+3 The creep prediction value at time x is obtained by repeating the above process. L The creep prediction value at each moment is calculated based on the creep prediction corresponding to each moment and the creep true value corresponding to each moment, and the deviation data is calculated. Based on the comparison between the deviation data and the loss corresponding to the sample, it is determined whether the creep experiment has passed the accuracy check. Based on the number of creep experiments that have passed the accuracy check, the accuracy of the target creep prediction model is obtained. Optionally, if the deviation data is greater than the loss, it is determined that the creep experiment has passed the accuracy check.

[0139] In this embodiment, the accuracy of the model is quantified to improve the accuracy of model evaluation.

[0140] In the following, in conjunction with a specific embodiment, a specific implementation process of the above creep prediction method is described in detail, which may include:

[0141] like Figure 8 As shown, it can be a flow chart of data set division. Specifically, the target experimental data is divided to obtain time series data and material environment data; each time series data and material environment data are associated with each other through the experimental number. The time series data is the target creep time series data.

[0142] The time series data is divided according to a preset ratio, for example, divided according to a ratio of 9:1, to obtain a non-prediction set and a prediction set. The non-prediction set is subjected to sliding sampling processing through a preset sliding window to obtain a time series training set a, that is, to obtain the first time series training data; the prediction set includes the first creep degree (creep function 1) and the second creep degree (creep function 2) corresponding to each time point respectively; the creep degree time series data corresponding to the creep function 1 of the prediction set is subjected to sliding sampling processing through a preset sliding window to obtain the time series training data, and the time series training data is divided according to the first prediction ratio 1-β and the second prediction ratio β to obtain the second time series training data (time series training set b) and the test time series data (that is, the time series test set); the terminal can also determine the creep degree time series data corresponding to the creep function 2 in the prediction set as the control set.

[0143] In this way, the terminal can add material environment parameters to each time series sample in the time series training set a, time series training set b and time series test set through the experiment number to obtain training set a, training set b and test set, that is, the first training set, the second training set and the test set.

[0144] Specifically, all data sets are first divided into time series data and material environment data. The two types of data are associated with experimental numbers. In the subsequent processing of time series data, the data always has the corresponding experimental number. Then, the time series data set is randomly divided into a prediction set (experimental data used to predict creep development) and a non-prediction set (experimental data not used to predict creep development) according to the experimental number in a certain proportion. For subsequent model training and testing, the creep time series data is further processed using the sliding window method. First, the window size s is selected, and the recommended value is 6. Assuming that the length of a time series is L, after processing by this method, n=L-s+1 fixed time series samples of length s can be obtained, and the next value of each time series sample is used as the label to be predicted for the sample.

[0145] like Fig. 9As shown, it can be a schematic diagram divided according to the preset sliding window. The target creep time series data can be the original data in the figure, for example, it can include x1, x2, x3, x4, x5, x6, x7, x8, ..., xL; the length of the preset sliding window can be 6, then the sampled time series samples can be seq1: x1-x6, seq2: x2-x7, seq3: x3-x8, .... Specifically, each creep data in the non-prediction set experiment is processed as above to obtain the time series training set a. Considering that each experiment contains two time series data of "creep degree 1" and "creep degree 2", each experiment in the time series training set a should include 2n time series of fixed length s and 2n prediction labels. Similarly, the "creep 1" data of each experiment in the prediction set is also processed as above to obtain n sequence samples, and the prediction ratio is set at the same time; the first (1-β)*n samples of these n sequence samples and the labels to be predicted are used as each experimental data in the training set b, so as to construct the time series training set b; the last β*n samples of these n sequence samples and the labels to be predicted are used as each experimental data in the test set, so as to construct the time series test set. The "creep 2" data in each prediction set experiment is used as a control set for subsequent model verification. The experiments in the time series training set b, the time series test set and the control set are one-to-one corresponding, and the creep development prediction diagram and control diagram can be drawn based on this.

[0146] like Fig.10 As shown, it can be a schematic diagram of the process of extracting creep time series features and determining the creep prediction value at the next moment. The length of each sample can be s, and the data included can be X1, X2, ..., Xs, X1 represents the creep time series data at the first time point, and the creep time series data at the first time point can be the creep at the first time point and the creep corresponding to the first time point, Xs represents the creep at the sth time point and the creep corresponding to the sth time point, the creep time series data at the first time point in the sample is input to the first LSTM calculation layer of the first LSTM calculation unit, and correspondingly, the creep time series data at other time points are respectively input to the first LSTM calculation layer of each corresponding LSTM calculation unit, and the sample is processed by the first LSTM calculation unit, the second LSTM calculation unit and the third LSTM calculation unit to obtain a hidden vector, which can be recorded as hs1, hs2, hs3, hs4, .... The corresponding experimental materials and environmental parameters of the sample can be recorded as m1, m2, m3, m4, .... The latent vector and the experimental material and environmental parameters of the sample are spliced ​​to obtain a spliced ​​vector, and the spliced ​​vector is input into the MLP layer in the target creep prediction model to obtain the output result of the target creep prediction model, that is, the creep prediction value X's+1 at the next moment.

[0147] like Fig.11As shown, it can be a schematic diagram of a prediction method and the effectiveness verification of the model: Since this embodiment is a prediction of the future development of creep, that is, a dynamic long-term prediction, during the test process, the predicted values need to be used to continue predicting subsequent predicted values (rolling prediction). The specific prediction method can be as follows: The test process is carried out for each experiment number in the test set. First, for the first sample of a certain experiment in the test set, its time series of length s is used as the known initial sequence. The time series at 1 to s is combined with material and environmental parameters to predict the value at s + 1. This predicted value is regarded as a known value and added to the sequence. Then, the values at 2 to s + 1 are combined with material and environmental parameters to predict the value at s + 2, and so on, until all samples in this experiment number are predicted. Then, it is compared with the true value of this experiment number in the test set, the RMSE (Root Mean Square Error) is calculated, the image of the predicted value and the true value is drawn, and finally, the predicted values of the subsequent creep development of each creep experiment in the test set and their deviation b1 (quantified by RMSE) from the true value can be obtained. Among them, s can be 6.

[0148] Based on the control set obtained in the above embodiment, the control data "creep degree 2" and the test set data "creep degree 1" are two sets of data measured under the same experimental conditions. If the deviation between the model predicted value and the true value is less than the deviation between the control experiment and the true value, it is ensured that the model error can be accepted, that is, the model accuracy check is passed; that is, the prediction of the creep development in the subsequent period of the test set by the model is reasonable. The RMSE is calculated respectively for the true value of each experiment in the test set and the values of the corresponding experiment and time period in the control set as the deviation b2 between the two. The sizes of b1 and b2 for each experiment are compared, and the ratio of the number of experiments with b1 < b2 to the total number of experiments in the test set is used as the model accuracy acc. The larger acc is, the more accurate the prediction of the subsequent development of the creep experiment by the model is.

[0149] In the creep degree prediction method provided in this embodiment, by processing the training set a, the training set b, and the test set, a target creep degree prediction model that meets the preset training completion conditions can be obtained, and reliable prediction of the future development of concrete creep can be realized.

[0150] It should be understood that, although the various steps in the flowcharts involved in the above-mentioned embodiments are displayed in sequence according to the indication of the arrows, these steps are not necessarily executed in sequence according to the order indicated by the arrows. Unless there is a clear explanation in this article, the execution of these steps does not have a strict order restriction, and these steps can be executed in other orders. Moreover, at least a part of the steps in the flowcharts involved in the above-mentioned embodiments can include multiple steps or multiple stages, and these steps or stages are not necessarily executed at the same time, but can be executed at different times, and the execution order of these steps or stages is not necessarily to be carried out in sequence, but can be executed in turn or alternately with other steps or at least a part of the steps or stages in other steps.

[0151] Based on the same inventive concept, the embodiment of the present application also provides a creep prediction device for implementing the creep prediction method involved above. The implementation scheme for solving the problem provided by the device is similar to the implementation scheme recorded in the above method, so the specific limitations in one or more creep prediction device embodiments provided below can refer to the limitations of the creep prediction method above, and will not be repeated here.

[0152] In an exemplary embodiment, Fig.12 As shown, a creep prediction device 1200 is provided, comprising:

[0153] A first acquisition module 1202 is used to acquire a target training set, wherein the target training set includes a first training set and a second training set, each of the training sets includes a plurality of samples, each of the samples includes creep time series data of a first length and material environment data of a second length;

[0154] A first training module 1204 is used to train the initial model through the creep time series data of each sample in the first training set and the material environment data to obtain a trained initial creep prediction model;

[0155] The first adjustment module 1206 is used to adjust the parameters of the initial creep prediction model through the creep time series data of each sample in the second training set and the material environment data to obtain target model parameters, and obtain a target creep prediction model based on the target model parameters.

[0156] In one embodiment, the creep prediction device further comprises:

[0157] A second acquisition module is used to acquire creep time series data of a first target length in a target scene and material environment data corresponding to the creep time series data;

[0158] An input module is used to input the creep time series data of the first target length and the material environment data corresponding to the creep time series data into the target creep prediction model to obtain the predicted creep at the next moment of the creep time series data.

[0159] In one embodiment, the creep prediction device further comprises:

[0160] A collection module, used for collecting initial experimental data of multiple creep experiments, wherein the initial experimental data includes an experimental number, material environment data, and original creep time series data, wherein the original creep time series data includes a time series and a creep series, wherein the time series includes multiple time points, and each of the time points corresponds to a first creep degree and a second creep degree;

[0161] A first processing module, used for performing abnormal supplementation processing on the initial experimental data to obtain first experimental data;

[0162] A sampling module, used for sampling the original creep time series data in the first experimental data based on a preset time interval to obtain target creep time series data with the same time interval;

[0163] The normalization module is used to perform normalization processing based on the material environment data of each creep experiment and the target creep degree time series data in the first experimental data to obtain target experimental data.

[0164] In one embodiment, the creep prediction device further comprises:

[0165] A division module, used for dividing the target creep time series data in the target experimental data to obtain first time series training data, second time series training data and test time series data;

[0166] An adding module is used to add material environment data to the first time series training data, the second time series training data and the test time series data according to the experimental number of each creep experiment to obtain the first training set, the second training set and the test set.

[0167] In one embodiment, the partitioning module is specifically used for:

[0168] Dividing the target creep time series data in the target experimental data according to a preset ratio to obtain a prediction set and a non-prediction set;

[0169] Performing sample division processing on the non-prediction set to obtain first time series training data;

[0170] Performing sample partitioning processing on the prediction set to obtain a target number of samples;

[0171] The target number of samples are extracted respectively based on the first prediction ratio and the second prediction ratio to obtain second time series training data and test time series data, wherein the first prediction ratio is determined based on the second prediction ratio.

[0172] In one embodiment, the partitioning module is further specifically used for:

[0173] According to a preset sliding window, the time series in the target creep degree time series data of each creep experiment is slidingly divided to obtain a first number of time series samples; and the creep degree of the time series sample at the next moment is determined as a sample label; based on each of the time series samples and the sample label of each of the time series samples, the first time series training data is obtained;

[0174] The division module is also specifically used to: perform sliding division on the time series in the target creep degree time series data of each creep experiment according to a preset sliding window to obtain a target number of time series samples; and determine the creep degree of the time series sample at the next moment as a sample label; and obtain a target number of samples based on each time series sample and the sample label of each time series sample.

[0175] In one embodiment, the first module is specifically used for:

[0176] Obtaining a first training set, and inputting the creep time series data and the material environment data of each sample in the first training set into an initial model;

[0177] By using the long short-term memory network in the initial model, feature extraction is performed on the time series of each of the samples in the first training set to obtain latent variables, and the latent variables are determined to be creep time series features of the samples;

[0178] Splicing the creep time series characteristics of the sample and the material environment data to obtain a splicing vector, and performing perception processing on the splicing vector through a multi-layer perceptron module in a hidden layer of the initial model to obtain a creep prediction value at the next moment of the time series of the sample;

[0179] Based on the true value of the creep at the next moment and the predicted value of the creep at the next moment, the loss is calculated. When the preset training completion condition is not met, back propagation processing is performed based on the loss to obtain the gradient of the model parameters of the initial model, and based on the gradient of the model parameters, an updated initial model is obtained; based on the updated initial model, the step of obtaining the first training set is re-executed until the preset training completion condition is met to obtain the trained initial creep prediction model.

[0180] In one embodiment, the creep prediction device further comprises:

[0181] A prediction module, for predicting the sample at the initial position of each creep experiment in the test set by using the target creep prediction model to obtain a creep prediction value of the sample at the next moment;

[0182] A second processing module is used to process the creep prediction value at the next moment and the sample at the next position of the initial position of the creep experiment through a target creep prediction model to obtain the creep prediction value at the next moment of the next position;

[0183] A sequence determination module, used to obtain a creep prediction sequence based on the creep prediction value of each sample at the next moment;

[0184] A first calculation module, used for calculating deviation data based on a creep degree prediction sequence corresponding to the creep experiment and a creep degree true sequence corresponding to the creep experiment;

[0185] The accuracy determination module is used to obtain the accuracy of the target creep prediction model based on the deviation data of each creep experiment.

[0186] Each module in the above creep prediction device can be implemented in whole or in part by software, hardware, or a combination thereof. Each module can be embedded in or independent of a processor in a computer device in the form of hardware, or can be stored in a memory in a computer device in the form of software, so that the processor can call and execute operations corresponding to each module.

[0187] In an exemplary embodiment, a computer device is provided. The computer device may be a server, and its internal structure diagram may be as follows: Fig.13 As shown. The computer device includes a processor, a memory, an input / output interface (Input / Output, referred to as I / O) and a communication interface. The processor, the memory and the input / output interface are connected through a system bus, and the communication interface is connected to the system bus through the input / output interface. The processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program and a database. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The database of the computer device is used to store creep sequence data. The input / output interface of the computer device is used to exchange information between the processor and an external device. The communication interface of the computer device is used to communicate with an external terminal through a network connection. When the computer program is executed by the processor, a creep prediction method is implemented.

[0188] Those skilled in the art will understand that Fig.13 The structure shown in the figure is only a block diagram of a part of the structure related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied. The specific computer device may include more or fewer components than those shown in the figure, or combine certain components, or have a different arrangement of components.

[0189] In an exemplary embodiment, a computer device is provided, including a memory and a processor, wherein a computer program is stored in the memory, and the processor implements the steps in the embodiments of the present application when executing the computer program.

[0190] In one embodiment, a computer-readable storage medium is provided, on which a computer program is stored. When the computer program is executed by a processor, the steps in the embodiments of the present application are implemented.

[0191] In one embodiment, a computer program product is provided, including a computer program, which implements the steps in the embodiments of the present application when executed by a processor.

[0192] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, stored data, displayed data, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of relevant data must comply with relevant regulations.

[0193] A person of ordinary skill in the art can understand that all or part of the processes in the above-mentioned embodiment method can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to the memory, database or other medium used in the embodiments provided in the present application can include at least one of non-volatile memory and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. As an illustration and not limitation, RAM can be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM). The database involved in each embodiment provided in this application may include at least one of a relational database and a non-relational database. Non-relational databases may include distributed databases based on blockchains, etc., but are not limited to this. The processor involved in each embodiment provided in this application may be a general-purpose processor, a central processing unit, a graphics processor, a digital signal processor, a programmable logic device, a data processing logic device based on quantum computing, an artificial intelligence (AI) processor, etc., but are not limited to this.

[0194] The technical features of the above embodiments may be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this application.

[0195] The above-described embodiments only express several implementation methods of the present application, and the descriptions thereof are relatively specific and detailed, but they cannot be understood as limiting the scope of the present application. It should be pointed out that, for a person of ordinary skill in the art, several variations and improvements can be made without departing from the concept of the present application, and these all belong to the protection scope of the present application. Therefore, the protection scope of the present application shall be subject to the attached claims.

Claims

1. A creep prediction method, characterized in that: The method comprises: Acquire a target training set, wherein the target training set includes a first training set and a second training set, each of the training sets includes a plurality of samples, and each of the samples includes creep time series data of a first length and material environment data of a second length; Training the initial model using the creep time series data of each sample in the first training set and the material environment data to obtain a trained initial creep prediction model; The parameters of the initial creep prediction model are adjusted by using the creep time series data of each sample in the second training set and the material environment data to obtain target model parameters, and a target creep prediction model is obtained based on the target model parameters.

2. The method according to claim 1, characterized in that The method further comprises: Acquire creep time series data of a first target length in a target scene and material environment data corresponding to the creep time series data; The creep time series data of the first target length and the material environment data corresponding to the creep time series data are input into the target creep prediction model to obtain the predicted creep at the next moment of the creep time series data.

3. The method according to claim 1, characterized in that The method further comprises: Collecting initial experimental data of multiple creep experiments, the initial experimental data including experimental numbers, material environment data and original creep time series data, the original creep time series data including time series and creep series, the time series including multiple time points, each of which corresponds to a first creep degree and a second creep degree; Performing abnormal supplementation processing on the initial experimental data to obtain first experimental data; Based on a preset time interval, sampling the original creep time series data in the first experimental data to obtain target creep time series data with the same time interval; Normalization processing is performed based on the material environment data of each creep experiment and the target creep degree time series data in the first experimental data to obtain target experimental data.

4. The method according to claim 3, characterized in that The method further comprises: Dividing the target creep time series data in the target experimental data to obtain first time series training data, second time series training data and test time series data; According to the experimental numbers of the creep experiments, material environment data are added to the first time series training data, the second time series training data and the test time series data to obtain the first training set, the second training set and the test set.

5. The method according to claim 4, characterized in that The step of dividing the target creep time series data in the target experimental data to obtain first time series training data, second time series training data and test time series data includes: Dividing the target creep time series data in the target experimental data according to a preset ratio to obtain a prediction set and a non-prediction set; Performing sample division processing on the non-prediction set to obtain first time series training data; Performing sample partitioning processing on the prediction set to obtain a target number of samples; The target number of samples are extracted respectively based on the first prediction ratio and the second prediction ratio to obtain second time series training data and test time series data, wherein the first prediction ratio is determined based on the second prediction ratio.

6. The method according to claim 5, characterized in that The performing sample division processing on the non-prediction set to obtain first time series training data includes: According to a preset sliding window, the time series in the target creep degree time series data of each creep experiment is slidingly divided to obtain a first number of time series samples; and the creep degree of the time series sample at the next moment is determined as a sample label; based on each of the time series samples and the sample label of each of the time series samples, the first time series training data is obtained; The performing sample division processing on the prediction set to obtain a target number of samples includes: According to the preset sliding window, the time series in the target creep degree time series data of each creep experiment is slidingly divided to obtain a target number of time series samples; and the creep degree of the time series sample at the next moment is determined as the sample label; based on each of the time series samples and the sample label of each of the time series samples, the target number of samples is obtained.

7. The method according to claim 1, characterized in that The initial model is trained by using the creep time series data of each sample in the first training set and the material environment data to obtain a trained initial creep prediction model, including: Obtaining a first training set, and inputting the creep time series data and the material environment data of each sample in the first training set into an initial model; By using the long short-term memory network in the initial model, feature extraction is performed on the time series of each of the samples in the first training set to obtain latent variables, and the latent variables are determined to be creep time series features of the samples; Splicing the creep time series characteristics of the sample and the material environment data to obtain a splicing vector, and performing perception processing on the splicing vector through a multi-layer perceptron module in a hidden layer of the initial model to obtain a creep prediction value at the next moment of the time series of the sample; Based on the true value of the creep at the next moment and the predicted value of the creep at the next moment, the loss is calculated. When the preset training completion condition is not met, back propagation processing is performed based on the loss to obtain the gradient of the model parameters of the initial model, and based on the gradient of the model parameters, an updated initial model is obtained; based on the updated initial model, the step of obtaining the first training set is re-executed until the preset training completion condition is met to obtain the trained initial creep prediction model.

8. The method according to claim 4, characterized in that The method further comprises: For each sample at the initial position of the creep experiment in the test set, predict the sample by using the target creep prediction model to obtain a creep prediction value of the sample at the next moment; Processing the creep prediction value at the next moment and the sample at the next position of the initial position of the creep experiment by the target creep prediction model to obtain the creep prediction value at the next moment of the next position; Based on the creep prediction value of each sample at the next moment, a creep prediction sequence is obtained; Calculating deviation data based on a creep degree prediction sequence corresponding to the creep experiment and a creep degree true sequence corresponding to the creep experiment; Based on the deviation data of each creep experiment, the accuracy of the target creep degree prediction model is obtained.

9. A creep prediction device, characterized in that: The device comprises: A first acquisition module is used to acquire a target training set, wherein the target training set includes a first training set and a second training set, each of the training sets includes a plurality of samples, and each of the samples includes creep time series data of a first length and material environment data of a second length; A first training module, used to train the initial model through the creep time series data of each sample in the first training set and the material environment data to obtain a trained initial creep prediction model; The first adjustment module is used to adjust the parameters of the initial creep prediction model through the creep time series data of each sample in the second training set and the material environment data to obtain target model parameters, and obtain a target creep prediction model based on the target model parameters.

10. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 8 are implemented.

11. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 8 are implemented.

12. A computer program product, comprising a computer program, characterized in that When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 8 are implemented.

Images