A phase change refrigeration material performance prediction method and related device

Abstract

The application discloses the data phase change material information field, disclose a kind of oriented phase change refrigeration material performance prediction method and related device, construct about phase change refrigeration material component, process condition, temperature induced martensitic phase change enthalpy change data and adiabatic temperature change shape memory alloy database;Utilize database and machine learning model, establish component process-adiabatic temperature change generation model;Utilize component process-adiabatic temperature change generation model to obtain phase change enthalpy change dataset;Extract DSC curve feature in phase change enthalpy change dataset, obtain low-dimensional phase change feature;Based on low-dimensional phase change feature, combine process parameter, utilize artificial neural network model to construct elastic-thermal performance prediction model;Under the condition of given DSC data, utilize elastic-thermal performance prediction model, obtain corresponding adiabatic temperature change under the condition of given DSC data.The application can combine generation model and prediction model under small sample condition, with temperature induced martensitic phase change feature as bridge, and predict shape memory alloy elastic-thermal performance.

Description

TECHNICAL FIELD

[0001] The application belongs to the field of data phase change material information science, and particularly relates to a performance prediction method for phase change refrigeration materials and a related device. BACKGROUND

[0002] Elastocaloric refrigeration is a new refrigeration technology that is most likely to replace traditional gas-liquid compression refrigeration methods, and the size of the elastocaloric effect of the elastocaloric material directly determines the upper limit of the elastocaloric refrigeration effect. However, the equipment for testing the elastocaloric effect by using the direct method is complex and the experimental period is long, so there are few reports on elastocaloric effect at present; and the adiabatic temperature change data obtained by the indirect method are often quite different from the actual values, which limits the development of elastocaloric materials. Therefore, it is necessary to explore a new method for developing high elastocaloric materials.

[0003] In recent years, many methods combining material information science such as machine learning have been used to study the shape memory characteristics, superelasticity and engineering applications of shape memory alloys, but due to the small size of the elastocaloric data, it is difficult to apply machine learning methods to elastocaloric prediction.

[0004] For shape memory alloys, the research on temperature-induced reversible martensitic phase transition is much earlier than the research on elastocaloric effect. The essence of both is based on the entropy change between the parent phase and the martensitic phase of the shape memory alloy, but the causes are different, the former is temperature-induced martensitic phase transition, and the latter is stress-induced martensitic phase transition, and the two are related but difficult to describe using a simple formula. SUMMARY

[0005] To solve the problems in the prior art, the purpose of the present application is to provide a performance prediction method for phase change refrigeration materials and a related device, which can combine a generative model with a prediction model under small sample conditions, use temperature-induced martensitic phase transition characteristics as a bridge, and reasonably predict the elastocaloric performance of shape memory alloys, thereby having guiding significance for accelerating the research and development of new materials.

[0006] To achieve the above purpose, the technical scheme adopted by the present application is as follows:

[0007] A performance prediction method for phase change refrigeration materials, comprising the following processes:

[0008] constructing a shape memory alloy database about the composition, process conditions, temperature-induced martensitic phase transition enthalpy change data and adiabatic temperature change of the phase change refrigeration material;

[0009] establishing a composition-process-adiabatic temperature change generative model by using the shape memory alloy database and a machine learning model;

[0010] obtaining a phase change enthalpy change data set by using the composition-process-adiabatic temperature change generative model;

[0011] extracting DSC curve features in the phase transition enthalpy change dataset to obtain low-dimensional phase transition features;

[0012] Based on the low-dimensional phase transition features, combined with process parameters, an elastic-thermal performance prediction model is constructed by using an artificial neural network model;

[0013] Under the condition of given DSC data, the corresponding adiabatic temperature change under the condition of given DSC data is obtained by using the elastic-thermal performance prediction model, so as to realize the prediction of the elastic-thermal performance of the phase change refrigeration material.

[0014] Preferably, the composition includes elements contained in the phase change refrigeration material;

[0015] The process conditions include part or all of whether hot rolling, cold rolling deformation, whether solid solution, whether quenching, aging treatment temperature and aging treatment time.

[0016] Preferably, the composition process-adiabatic temperature change generation model is established by using the shape memory alloy database and the machine learning model, including:

[0017] Based on the shape memory alloy database, a variational conditional autoencoder generation model is used to establish a mapping mechanism from composition and process conditions to temperature-induced martensitic phase transition enthalpy change data, and a composition process-adiabatic temperature change generation model is obtained, realizing the hidden space generation from composition and process conditions to temperature-induced martensitic phase transition enthalpy change data.

[0018] When the variational conditional autoencoder generation model is used to establish a mapping mechanism from composition and process conditions to temperature-induced martensitic phase transition enthalpy change data, and a composition process-adiabatic temperature change generation model is obtained, the actual numerical value of the composition, cold rolling deformation, aging treatment temperature and aging treatment time is used as the input of the variational conditional autoencoder; Whether hot rolling, whether solid solution, whether quenching is represented by one-hot encoding.

[0019] By sampling in the hidden space, the data after sampling is decoded by the decoder of the composition process-adiabatic temperature change generation model to generate temperature-induced martensitic phase transition data under specific composition and process conditions, and is used as the phase transition enthalpy change dataset.

[0020] Preferably, the DSC curve features in the phase transition enthalpy change dataset are extracted to obtain low-dimensional phase transition features, including:

[0021] Based on the phase transition enthalpy change dataset, the DSC curve features are automatically extracted by using synchronous extrusion wavelet transform combined with variational autoencoder to obtain low-dimensional phase transition features.

[0022] Preferably, based on the phase transition enthalpy change dataset, the DSC curve features are automatically extracted by using synchronous extrusion wavelet transform combined with variational autoencoder, including:

[0023] The DSC curve is processed by using a synchronous extrusion wavelet transform to generate a time-frequency graph, and then the time-frequency graph is input into a variational autoencoder, and the hidden layer variable in the variational autoencoder is taken as a low-dimensional phase change feature.

[0024] Preferably, based on the low-dimensional phase change feature, a thermomechanical property prediction model is constructed by using an artificial neural network model in combination with process parameters, including:

[0025] The low-dimensional phase change feature and the process parameter are taken as inputs of the artificial neural network model, and the corresponding adiabatic temperature change is taken as an output of the artificial neural network model, the artificial neural network model is trained, and a mapping relationship between the DSC data and the process condition and the adiabatic temperature change is constructed, and the trained artificial neural network model is taken as the thermomechanical property prediction model.

[0026] Preferably, the artificial neural network model adopts a fully connected neural network and comprises an input layer, a hidden layer and an output layer; wherein a Dropout layer is added after the hidden layer, and the Dropout layer is selected as a trick for training the fully connected neural network; three hidden layers are included, and the node numbers of the three hidden layers are 16, 32 and 16 respectively; and the output layer is an adiabatic temperature change value.

[0027] In each training batch, half of the hidden layer node values are set to 0 to ignore half of the feature detectors.

[0028] The application also provides a phase change refrigeration material performance prediction system, including:

[0029] A database construction module is configured to construct a shape memory alloy database about components, process conditions, temperature-induced martensitic phase change enthalpy change data and adiabatic temperature changes of the phase change refrigeration material.

[0030] A first model establishment module is configured to establish a component process-adiabatic temperature change generation model by using the shape memory alloy database and a machine learning model.

[0031] A data set acquisition module is configured to acquire a phase change enthalpy change data set by using the component process-adiabatic temperature change generation model.

[0032] A feature extraction module is configured to extract DSC curve features in the phase change enthalpy change data set to obtain low-dimensional phase change features.

[0033] A second model establishment module is configured to construct a thermomechanical property prediction model by using an artificial neural network model in combination with process parameters based on the low-dimensional phase change features.

[0034] The prediction module is used for obtaining corresponding adiabatic temperature change under the condition of the given DSC data by using the elastic-thermal performance prediction model under the condition of the given DSC data, so as to realize the prediction of the elastic-thermal performance of the phase change refrigeration material.

[0035] The application further provides an electronic device, comprising:

[0036] one or more processors;

[0037] a storage device having one or more programs stored thereon;

[0038] When the one or more programs are executed by the one or more processors, the one or more processors realize the phase change refrigeration material performance prediction method as described above.

[0039] The application further provides a storage medium having a computer program stored thereon, wherein the computer program is executed by a processor to realize the phase change refrigeration material performance prediction method as described above.

[0040] Compared with the prior art, the application has the following beneficial effects:

[0041] The phase change refrigeration material performance prediction method mainly includes data generation and performance prediction, and can realize the rapid and accurate prediction of the elastic-thermal performance of the phase change refrigeration material under the condition of a small sample, which has important guiding significance for guiding the development of new materials. In order to generate the temperature-induced martensitic phase change enthalpy data under the process parameters of a specific composition, the application takes the composition process parameters as the constraint condition of the generation model, establishes the composition process-adiabatic temperature change generation model of the temperature-induced martensitic phase change enthalpy under the process parameters of a specific composition, and realizes the hidden space generation from the composition process to the temperature-induced martensitic phase change data. The application proposes to use the synchronous extrusion wavelet transform combined with the variational autoencoder to realize the phase change feature extraction of the high-dimensional DSC data, realizes the dimension reduction of the high-dimensional data, and greatly reduces the possibility of overfitting of the model. BRIEF DESCRIPTION OF DRAWINGS

[0042] Figure 1 The figure is the implementation block diagram of the phase change refrigeration material performance prediction method in the embodiment of the application.

[0043] Figure 2 The figure is the DSC data feature extraction diagram of the synchronous extrusion wavelet transform combined with the variational autoencoder in the method of the application.

[0044] FIG. 3(a) is a fitting result diagram of the extracted features in the embodiment of the present application on an extreme gradient boosting decision tree; FIG. 3(b) is a fitting result diagram of the extracted features in the embodiment of the present application on a random forest model; FIG. 3(c) is a fitting result diagram of the extracted features in the embodiment of the present application on a gradient boosting regression model; FIG. 3(d) is a fitting result diagram of the extracted features in the embodiment of the present application on an AdaBoost model; FIG. 3(e) is a fitting result diagram of the extracted features in the embodiment of the present application on a CatBoost model; and FIG. 3(f) is a fitting result diagram of the extracted features in the embodiment of the present application on a fully connected artificial neural network.

[0045] FIG. 4(a) is a three-dimensional prediction surface diagram of the Hf2A400T2 alloy in the embodiment of the present application; FIG. 4(b) is a projection of the three-dimensional prediction surface in FIG. 4(a) on the temperature axis; and FIG. 4(c) is a projection of the three-dimensional prediction surface in FIG. 4(a) on the strain axis. DETAILED DESCRIPTION

[0046] The embodiments of the present application will be described in detail below with reference to the accompanying drawings. The embodiments are implemented on the premise of the technical solutions of the present application, and detailed implementation manners and specific operation processes are given, but the protection scope of the present application is not limited to the following embodiments.

[0047] Reference Figure 1 The performance prediction method for the phase change refrigeration material according to the present application includes the following processes:

[0048] A shape memory alloy database of composition-technique-phase change enthalpy-temperature change is constructed, a machine learning model with good performance is developed based on the above database to establish a mapping mechanism from the composition and technique data to the temperature-induced martensitic phase change data (DSC curve), a composition-technique-adiabatic temperature change generation model of low-dimension-high-dimension mapping is established, and a large amount of phase change enthalpy data (DSC curve) is obtained.

[0049] Based on the above easily obtained large amount of inexpensive DSC data, the elastic-thermal performance of the shape memory alloy is predicted, the DSC curve features are extracted, the relationship between the temperature-induced martensitic phase change and the stress-induced martensitic phase change is established, and a latent heat-adiabatic temperature change prediction model is developed.

[0050] The elastic-thermal performance of the phase change refrigeration material is predicted according to the phase change enthalpy-adiabatic temperature change prediction model.

[0051] In the present application, in order to generate temperature-induced martensitic phase transition enthalpy data under specific composition process parameters, the composition process parameters are used as constraint conditions for generating a model, a temperature-induced martensitic phase transition enthalpy data generation model under specific composition process parameters is established, and hidden space generation from composition process to temperature-induced martensitic phase transition data is realized. In addition, a large amount of temperature-induced martensitic phase transition enthalpy data is generated by sampling in the hidden space based on the generation model. The present application mainly uses conditional variational autoencoder (CVAE) for phase transition enthalpy data generation.

[0052] Conditional Variational Autoencoder (CVAE) is a generation model based on Variational Autoencoder (VAE), which can generate data with specified features according to given conditions. Compared with traditional VAE model, CVAE adds conditional input in both encoder and decoder, which makes the generated samples more controllable and can be used for many unsupervised learning tasks.

[0053] The main idea of CVAE is to embed conditional information into the prior distribution of a set of latent variables (z), so that the model can be influenced by the conditional information when generating data. Specifically, assuming that the goal is to generate data y from given conditions x and latent variables z, the optimization objective of CVAE can be represented as:

[0054]

[0055] where, represents the probability of decoder generating y, represents the distribution of encoder mapping x to latent variable z. Since this integral is difficult to solve, the present application uses a reparameterization trick to approximate the gradient. Specifically, the present application reparameterizes the latent variable z by sample noise, which is represented as:

[0056]

[0057] where μ and σ are the mean and variance of the encoder network output, is the noise sampled from the standard normal distribution N(0,1). This reparameterization trick can make the gradient smoothly backpropagate to the network, and can freely vary within the limits of mean and variance during training.

[0058] In summary, CVAE is a generative model based on conditional information, which can generate data with specified characteristics by embedding conditional information in latent variables. Although its objective function is complex, it can be efficiently trained and generate high-quality samples through reparameterization techniques and other optimization methods.

[0059] In the present application, the composition process parameters are used as the constraint conditions of the CVAE generative model, and the temperature-induced martensitic transformation data are used as the generative model data. The generative model is trained to output the spectrum data under the condition of specific composition process parameters.

[0060] Based on the large amount of DSC data generated by the above process, combined with the process condition data, an artificial neural network model is developed to establish the mapping relationship between DSC data and adiabatic temperature change, thereby realizing the rapid prediction of adiabatic temperature change in the service space.

[0061] Embodiment

[0062] In this embodiment, TiNi-based shape memory alloy is taken as the research object, and a composition, process condition and corresponding temperature-induced martensitic transformation data (DSC data) library is established.

[0063] As shown in Figure 1 The performance prediction method for phase change refrigeration materials in this embodiment includes the following steps:

[0064] Step one, build the DSC curve and adiabatic temperature change database of shape memory alloy. The shape memory alloy database used to build the composition process-DSC curve generative model mainly comes from the accumulation of the experimental accumulation of the research group. The database includes 89 types of Ti-Ni-based shape memory alloys with different compositions and processes and the corresponding DSC curves. The composition elements include Ti, Ni, Cu, Hf, Co, Zr, Fe, Pd, Ta, and the process conditions include whether hot rolling, cold rolling deformation, whether solid solution, whether quenching, aging temperature and aging time. Since the hot rolling temperature of the samples in the database is 1000℃, the solid solution treatment condition is 1000℃ for one hour, the experimental parameters are constant for the model and do not help the model training; in addition, hot rolling has little effect on the performance of the alloy, so the hot rolling temperature, deformation and solid solution temperature and time are not used more specifically.

[0065] In feature construction, the composition, cold rolling deformation, aging temperature and aging time use actual numerical values as input model encoding; and whether hot rolling, solid solution and quenching use one-hot encoding, that is, "yes" is marked as 1 and "no" is marked as 0. The DSC curve is divided into two stages of heating and cooling, each with 112 points, in order to construct the generative model.

[0066] The shape memory alloy database used to construct the adiabatic temperature change prediction model is also accumulated from the experiments of the research group, and the database includes 32 types of Ti-Ni-based shape memory alloys with different compositions and different processes, corresponding DSC curves, and a total of 340 adiabatic temperature change data tested under different temperatures and strain conditions for each alloy.

[0067] The adiabatic temperature change database mainly contains Ti-Ni-based shape memory alloy data in the cold-rolled state. In this embodiment, the feature information extracted from the DSC curve and the temperature and strain of the thermoelastic test conditions are used as input, and the corresponding adiabatic temperature change data are used as output. All DSC curves are processed into 3915 points for subsequent feature extraction. The test temperature conditions of the accumulated adiabatic temperature change data are basically higher than the austenite transformation end temperature Af point, because the superelasticity of the shape memory alloy occurs above the Af temperature; the test strain conditions are concentrated in 1%~8%.

[0068] Step two, based on the above database, the encoder of the conditional variational autoencoder (CVAE) generation model learns the distribution of high-dimensional temperature-induced martensitic phase change data in the low-dimensional hidden space, while the low-dimensional composition process data are used as constraint conditions and are fused with the hidden space distribution. Through sampling in the hidden space, the sampled data are processed through the decoder of the generation model to generate temperature-induced martensitic phase change data (DSC data) under specific composition and process conditions.

[0069] In this embodiment, the composition and process parameters are used as the constraint conditions of the CVAE generation model, and the temperature-induced martensitic phase change data are used as the generation model data. The generation model is trained to output the phase change enthalpy data (DSC curve) under specific composition and process parameter conditions.

[0070] Step three, based on the above generated large amount of DSC data, the synchronous extrusion wavelet transform is combined with the variational autoencoder to automatically extract the DSC curve features, and the low-dimensional phase change features are obtained.

[0071] Synchronous extrusion wavelet transform is an effective tool in signal processing. The widely used wavelet transform converts the signal into a time-frequency graph, but its resolution cannot be optimal. Therefore, on the basis of continuous wavelet transform, the complex wavelet coefficients are rearranged in the frequency domain to maintain their reversibility and improve the time-frequency resolution. The method of synchronous extrusion wavelet transform is obtained through strict mathematical derivation. The specific method includes the following steps: first, the signal is subjected to continuous wavelet transform to obtain wavelet coefficients; second, the instantaneous frequency is obtained from the wavelet coefficients; third, after the rearrangement method, only the frequency axis is obtained to obtain the recombined frequency spectrum estimate, thereby obtaining a time-frequency spectrum with more concentrated energy.

[0072] Different from the traditional autoencoder which describes the potential energy space by numerical value, the variational autoencoder describes the potential energy space by probability, which shows great application in data generation. Similar to the autoencoder, it consists of two parts of encoder and decoder, and the variational autoencoder uses two neural networks to model two probability density distributions. One model is used for conversion inference on the original input data to generate the variation probability distribution of the hidden variable, which is called inference network; the other model is used for restoration network to generate a probability distribution approximating the original data, which is called generation network. Specifically, the variational autoencoder represents the latent feature of a given input as a probability distribution, and when decoding from the latent state, a vector is randomly sampled from each latent state distribution as the input of the decoder model. Through the above described encoding and decoding process, a continuous and smooth representation of the latent variable is basically achieved. In order to enable the decoder model to accurately reconstruct the input, values adjacent to each other in the potential energy space should correspond to very similar reconstructions. In order to improve the performance of the model, a pre-trained model ResNet50 is used as the encoder and decoder of VAE.

[0073] The specific operation of combining the synchronous extrusion wavelet transform and the variational autoencoder features is as follows: first, the DSC curve is processed by using the synchronous extrusion wavelet transform to generate a time-frequency graph, then the time-frequency graph is input into the variational autoencoder, and the hidden layer variable in the variational autoencoder is taken as the feature. The feature extraction process is shown in Figure 2 The number of features is consistent with the wavelet scattering transform method, and finally a 5-dimensional feature vector is obtained.

[0074] Step four, based on step four, the 5-dimensional feature vector obtained by using the automatic feature extraction method is combined with the process parameters as the input of the prediction model, and the corresponding adiabatic temperature change is taken as the output. The artificial neural network model is trained to build the mapping relationship between the DSC data and the process parameters and the adiabatic temperature change, so as to predict the adiabatic temperature change corresponding to any service condition in the service space under the given DSC data, thereby realizing the rapid prediction of the pyroshock performance.

[0075] In step four, the artificial neural network adopts a fully connected neural network, including an input layer, a hidden layer and an output layer. Since the experimental data is limited, in order to avoid overfitting of the model, a Dropout layer is added after the hidden layer. Dropout can be used as a trick to train deep neural networks. In each training batch, by ignoring half of the feature detectors (letting half of the hidden layer nodes be 0), the overfitting phenomenon can be obviously reduced. This way can reduce the interaction between feature detectors (hidden layer nodes), and the detector interaction refers to the fact that some detectors depend on other detectors to function. In a popular way, a certain neuron is stopped working with a certain probability p, so that the model generalization is stronger because it does not rely too much on some local features. In the present application, p is 0.2. The neural network designed in the present application includes three hidden layers, and the number of nodes is 16, 32 and 16 respectively, and the output layer is the adiabatic temperature change value.

[0076] Step five, predicting the optimal elastic-thermal performance of the material according to the trained performance prediction model.

[0077] In order to verify the effectiveness of the method of the present application, the original data set is randomly divided into a training set and a test set according to an 8:2 ratio, the training set data is used for model training, and the test set data is used for model testing. The prediction results of the six models on the training set and the test set data are shown in FIGS. 3(a)-3(f), and the blue points are the training set data and the red points are the test set data. In the figure, the horizontal coordinate is the actual value obtained by experiment, and the vertical coordinate is the predicted value obtained by the model. The closer the data point is to the diagonal line, the higher the accuracy of the model.

[0078] In this embodiment, the fitting effect of the features extracted by the synchronous extrusion wavelet transform combined variational autoencoder on six models, including extreme gradient boosting decision tree (XGBR), random forest (RF), gradient boosting regression (GBR), AdaBoost, CatBoost and fully connected artificial neural network (ANN), is compared. The intuitive comparison between the model prediction value and the actual value is shown in FIGS. 3(a)-3(f). As can be seen from the figure, the data set of the present application has an overfitting phenomenon on the XGBR, GBR and CatBoost models. The random forest model and the AdaBoost model perform well on the training set, but have a large prediction deviation for the test set. According to the results of the test set and the training set, the fully connected artificial neural network model performs best. The fully connected artificial neural network model fitted by the features extracted by the synchronous extrusion wavelet transform combined variational autoencoder is finally obtained, and the effect of the model for predicting the elastic-thermal effect of the shape memory alloy is best.

[0079] In order to verify the prediction performance of the model in unknown service space, taking the Ti48Ni45Cu5Hf2 alloy (Hf2A400T2) sample after cold rolling 40% and heat treatment at 400 DEG C for 2h as an example, the model can generate a three-dimensional adiabatic temperature change surface with test condition temperature and strain as X and Y axes, as shown in Fig. 4 (a). Through the projection of the surface on the temperature axis and the strain axis (i.e. Fig. 4 (b) and Fig. 4 (c)), it can be known that under the same strain test condition, the adiabatic temperature change increases first and then decreases with the increase of temperature; under the same temperature test condition, the adiabatic temperature change gradually increases with the increase of strain, which is also consistent with the trend obtained by experiment. By using the latent heat-adiabatic temperature change model, researchers can predict the elastocaloric performance of the shape memory alloy system based on the DSC data which is cheap and easy to obtain, obtain the prediction value of the adiabatic temperature change under any experimental condition, judge the elastocaloric refrigeration potential of the alloy, and have certain guiding significance for the development of elastocaloric materials.

[0080] In another aspect, the present application also provides a system for realizing the performance prediction method for the phase change refrigeration material, comprising:

[0081] A database construction module is used to construct a shape memory alloy database about the composition, process condition, temperature-induced martensitic phase transition enthalpy change data and adiabatic temperature change of the phase change refrigeration material;

[0082] A first model establishment module is used to establish a composition process-adiabatic temperature change generation model by using the shape memory alloy database and a machine learning model;

[0083] A data set acquisition module is used to acquire a phase transition enthalpy change data set by using the composition process-adiabatic temperature change generation model;

[0084] A feature extraction module is used to extract the DSC curve features in the phase transition enthalpy change data set to obtain low-dimensional phase change features;

[0085] A second model establishment module is used to construct an elastocaloric performance prediction model by using an artificial neural network model based on the low-dimensional phase change features and in combination with process parameters;

[0086] A prediction module is used to obtain the corresponding adiabatic temperature change under the condition of given DSC data by using the elastocaloric performance prediction model under the condition of given DSC data, so as to realize the prediction of the elastocaloric performance of the phase change refrigeration material.

[0087] The present application can also provide an equipment comprising a processor and a memory, the memory is used to store computer executable programs, the processor reads part or all of the computer executable programs from the memory and executes, and the processor can realize the performance prediction method for the phase change refrigeration material when executing part or all of the computer executable programs.

[0088] In another aspect, the present application provides a computer readable storage medium, wherein a computer program is stored in the computer readable storage medium, and the computer program, when executed by a processor, enables the phase change refrigeration material performance prediction method to be implemented.

[0089] The computer device can be a notebook computer, a tablet computer, a desktop computer, a mobile phone or a workstation.

[0090] The processor can be a central processing unit (CPU), a digital signal processor (DSP), an application specific integrated circuit (ASIC) or a field programmable gate array (FPGA).

[0091] The memory can be an internal storage unit of a notebook computer, a tablet computer, a desktop computer, a mobile phone or a workstation, such as a memory or a hard disk, or can be an external storage unit, such as a mobile hard disk or a flash card.

[0092] The computer readable storage medium can include a computer storage medium and a communication medium. The computer storage medium includes volatile and non-volatile, removable and non-removable media implemented in any method or technology for storing information such as computer readable instructions, data structures, program modules or other data. The computer readable storage medium can include a read only memory (ROM), a random access memory (RAM), a solid state disk (SSD) or an optical disk. Among them, the random access memory can include a resistance random access memory (ReRAM) and a dynamic random access memory (DRAM).

[0093] Some parts of the present application not described in detail are known to those skilled in the art.

[0094] Those skilled in the art in the technical field should realize that the above embodiments are only used to illustrate the present application, and are not used as a limitation of the present application, and as long as the above described embodiments are changed and modified within the scope of the spirit of the present application, they will fall within the scope of the claims of the present application.

[0095] It should be pointed out finally that the above embodiments are only used to illustrate the technical solutions of the present application but not to limit it. Although the present application has been described in detail with reference to the above embodiments, it should be understood by those skilled in the art that the specific embodiments of the present application can be modified or replaced equivalently without departing from the spirit and scope of the present application, and any modification or equivalent replacement should be covered in the protection scope of the claims of the present application.

Claims

1. A method for predicting the performance of a phase change refrigeration material, comprising: The method comprises the following steps: constructing a shape memory alloy database about the composition, process condition, temperature-induced martensitic phase transition enthalpy change data and adiabatic temperature change of the phase change refrigeration material; the composition comprises elements contained in the phase change refrigeration material; the process condition comprises part or all of the following: whether hot rolling, cold rolling deformation amount, whether solid solution, whether quenching, aging treatment temperature and aging treatment time; using the shape memory alloy database and the machine learning model to establish a composition process-adiabatic temperature change generation model, specifically comprising the following steps: based on the shape memory alloy database, a variational conditional autoencoder generation model is used to establish a mapping mechanism from the composition and process condition to the temperature-induced martensitic phase transition enthalpy change data, so as to obtain the composition process-adiabatic temperature change generation model and realize the hidden space generation from the composition and process condition to the temperature-induced martensitic phase transition enthalpy change data; when the variational conditional autoencoder generation model is used to establish the mapping mechanism from the composition and process condition to the temperature-induced martensitic phase transition enthalpy change data to obtain the composition process-adiabatic temperature change generation model, the actual values of the composition, cold rolling deformation amount, aging treatment temperature and aging treatment time are used as the input of the variational conditional autoencoder for coding; whether hot rolling, whether solid solution and whether quenching are represented by one-hot encoding; by sampling in the hidden space, the data after sampling is decoded by the decoder of the composition process-adiabatic temperature change generation model to generate the temperature-induced martensitic phase transition data under specific composition and process condition, which is used as the phase transition enthalpy change data set; using the composition process-adiabatic temperature change generation model to obtain the phase transition enthalpy change data set; extracting the DSC curve features in the phase transition enthalpy change data set to obtain low-dimensional phase change features, specifically comprising the following steps: based on the phase transition enthalpy change data set, the DSC curve features are automatically extracted by using synchronous extrusion wavelet transform combined with variational autoencoder to obtain low-dimensional phase change features; based on the low-dimensional phase change features and process parameters, an artificial neural network model is used to construct a pyroelectric performance prediction model, specifically comprising the following steps: the low-dimensional phase change features and process parameters are used as the input of the artificial neural network model, the corresponding adiabatic temperature change is used as the output of the artificial neural network model, the artificial neural network model is trained, the mapping relationship between the DSC data and process condition and the adiabatic temperature change is constructed, and the trained artificial neural network model is used as the pyroelectric performance prediction model; under the condition of given DSC data, the corresponding adiabatic temperature change under the condition of given DSC data is obtained by using the pyroelectric performance prediction model, so as to realize the prediction of the pyroelectric performance of the phase change refrigeration material.

2. The method for predicting the performance of a phase change refrigeration material according to claim 1, wherein Based on the phase transition enthalpy change data set, the DSC curve features are automatically extracted by using synchronous extrusion wavelet transform combined with variational autoencoder, comprising: the DSC curve is processed by using synchronous extrusion wavelet transform to generate a time-frequency graph, and then the time-frequency graph is input into the variational autoencoder, and the hidden layer variables in the variational autoencoder are taken as the low-dimensional phase change features.

3. The method for predicting the performance of a phase change refrigeration material according to claim 1, wherein The artificial neural network model adopts a full connection neural network, and comprises an input layer, a hidden layer and an output layer; wherein, a Dropout layer is added after the hidden layer, and the Dropout layer is selected as a trick for training the full connection neural network; three hidden layers are included, and the node numbers of the three hidden layers are 16, 32 and 16 respectively; and the output layer is an adiabatic temperature change value; In each training batch, half of the hidden layer node values are set to 0 to ignore half of the feature detectors.

4. A phase change refrigerant material performance prediction system, comprising: Comprise: A database construction module is configured to construct a shape memory alloy database about a phase change refrigeration material, including a composition of the phase change refrigeration material, a process condition, temperature-induced martensitic phase transformation enthalpy data and an adiabatic temperature change; the composition includes elements contained in the phase change refrigeration material; the process condition includes part or all of whether hot rolling, cold rolling deformation, whether solid solution, whether quenching, aging treatment temperature and aging treatment time; A first model establishing module is configured to establish a composition process-adiabatic temperature change generation model by using the shape memory alloy database and a machine learning model, and specifically includes the following processes: based on the shape memory alloy database, a variational conditional autoencoder generation model is used to establish a mapping mechanism from the composition and the process condition to the temperature-induced martensitic phase transformation enthalpy data, to obtain the composition process-adiabatic temperature change generation model, and to realize hidden space generation from the composition and the process condition to the temperature-induced martensitic phase transformation enthalpy data; when the variational conditional autoencoder generation model is used to establish the mapping mechanism from the composition and the process condition to the temperature-induced martensitic phase transformation enthalpy data to obtain the composition process-adiabatic temperature change generation model, the actual values of the composition, the cold rolling deformation, the aging treatment temperature and the aging treatment time are used as input codes of the variational conditional autoencoder; whether hot rolling, whether solid solution and whether quenching are represented by one-hot encoding; by sampling in the hidden space, the sampled data are decoded by the decoder of the composition process-adiabatic temperature change generation model to generate temperature-induced martensitic phase transformation data under specific composition and process conditions, and are used as the phase transformation enthalpy data set; A data set acquisition module is configured to acquire the phase transformation enthalpy data set by using the composition process-adiabatic temperature change generation model; A feature extraction module is configured to extract DSC curve features in the phase transformation enthalpy data set to obtain low-dimensional phase change features, and specifically includes the following processes: based on the phase transformation enthalpy data set, synchronous squeezing wavelet transform combined with a variational autoencoder is used to automatically extract DSC curve features to obtain low-dimensional phase change features; A second model establishing module is configured to construct a pyroelectric performance prediction model by using an artificial neural network model based on the low-dimensional phase change features and process parameters, and specifically includes the following processes: the low-dimensional phase change features and the process parameters are used as inputs of the artificial neural network model, and the corresponding adiabatic temperature change is used as an output of the artificial neural network model, the artificial neural network model is trained, a mapping relationship between the DSC data and the process condition and the adiabatic temperature change is constructed, and the trained artificial neural network model is used as the pyroelectric performance prediction model. The prediction module is configured to utilize the elastic-thermal performance prediction model to obtain the corresponding adiabatic temperature change under the condition of the given DSC data, so as to realize the prediction of the elastic-thermal performance of the phase change refrigeration material.

5. An electronic device, comprising: The method comprises: one or more processors; a storage device having stored thereon one or more programs; when the one or more programs are executed by the one or more processors, the one or more processors realize the phase change refrigeration material performance prediction method according to any one of claims 1 to 4.

6. A storage medium, characterized by a computer program is stored thereon, and the computer program is executed by a processor to realize the phase change refrigeration material performance prediction method according to any one of claims 1 to 4.

Images