Method, device and equipment for lightweighting key materials of hydrogen fuel cell BOP for aircraft

Abstract

The present invention provides a method, device and equipment for lightweighting key materials of hydrogen fuel cell BOP for aircraft. The method comprises training each key material prediction model according to a first material training data set of the hydrogen fuel cell BOP to predict the performance of the hydrogen fuel cell BOP; determining a stand-by key material prediction model from multiple key material prediction models according to a first performance evaluation parameter of each key material prediction model; selecting a target key material prediction model according to a second performance evaluation parameter of the stand-by key material prediction model; predicting actual performance data of the hydrogen fuel cell BOP according to the target key material prediction model and lightweight design data, comparing the actual performance data with the performance requirement data of the hydrogen fuel cell BOP, and outputting a verification result of the lightweight design of the key materials of the hydrogen fuel cell BOP.

Description

Technical Field

[0001] The present application relates to the fields of hydrogen fuel cells and artificial intelligence technology, and in particular to a method, device and equipment for lightweighting key materials of a hydrogen fuel cell BOP for aircraft. Background Art

[0002] As aircraft transform towards a more environmentally friendly and efficient energy system, hydrogen fuel cells, as a clean energy technology, have become a key component of aircraft power systems due to their high energy density and zero emission characteristics. The performance and reliability of hydrogen fuel cell systems depend largely on their BOP (Balancing of Plant) key materials, including but not limited to bipolar plates, seals, cooling system components, etc.

[0003] Traditional BOP material selection often focuses on ensuring the system's mechanical strength and durability, but this usually leads to an increase in overall weight, affecting the energy efficiency and range of the aircraft. Lightweighting is an important development direction for the aviation industry, which not only improves the fuel efficiency of aircraft, but also reduces emissions and enhances economy. At present, the lightweighting of key materials for hydrogen fuel cell BOP faces multiple challenges. Lightweight materials must reduce material usage while maintaining structural strength and stiffness to reduce overall weight. Existing design methods often rely on experience and trial and error, and lack accurate predictive models to guide material selection and design optimization.

[0004] Therefore, it is necessary to develop a new method to use artificial intelligence technology, especially machine learning algorithms, to predict and optimize the performance of key materials in hydrogen fuel cell BOP to achieve the goal of lightweighting. Summary of the invention

[0005] The embodiments of the present application provide a method, device and equipment for lightweighting key materials of hydrogen fuel cell BOP for aircraft, which can use artificial intelligence technology to accurately predict the performance of key materials of hydrogen fuel cell BOP after lightweight design, achieve lightweighting goals and improve lightweighting efficiency and accuracy.

[0006] In one aspect, a method for lightweighting a key material of a hydrogen fuel cell BOP for an aircraft is provided, the method comprising:

[0007] Acquire a first material training data set of a hydrogen fuel cell BOP, wherein the first material training data set includes training data of a plurality of hydrogen fuel cell BOPs, wherein the training data includes material property characteristics, structural design characteristics, and material performance characteristics of key materials of the hydrogen fuel cell BOP, wherein the material performance characteristics include performance data of the influence of the key materials on the hydrogen fuel cell BOP obtained through experiments;

[0008] Acquire multiple key material prediction models, train each of the key material prediction models according to the first material training data set to predict the performance of the hydrogen fuel cell BOP, and obtain predicted performance information of the hydrogen fuel cell BOP output by each of the key material prediction models;

[0009] Calculating first performance evaluation parameters of a plurality of the key material prediction models according to the predicted performance information of the hydrogen fuel cell BOP output by each of the key material prediction models and the material performance characteristics; determining a key material prediction model to be used from a plurality of key material prediction models according to the first performance evaluation parameter of each of the key material prediction models, wherein the first performance evaluation parameter is used to measure the distance quantification between key material training data;

[0010] Acquire a second material training data set for the hydrogen fuel cell BOP, where the second material training data set is a training data set for key materials of the hydrogen fuel cell BOP in different test environments;

[0011] Training the key material prediction model for standby use according to the second material training data set to calculate a second performance evaluation parameter of the key material prediction model for standby use on the second material training data set, wherein the second performance evaluation parameter is used to measure the stability of the prediction result of the key material prediction model for standby use;

[0012] Selecting a target key material prediction model according to a second performance evaluation parameter of the key material prediction model to be used;

[0013] Obtaining lightweight design data of key materials of hydrogen fuel cell BOP, wherein the lightweight design data includes material property characteristics and structural design characteristics of the key materials after lightweighting;

[0014] The actual performance data of the hydrogen fuel cell BOP is predicted based on the target key material prediction model and the lightweight design data, the actual performance data is compared with the performance requirement data of the hydrogen fuel cell BOP for aircraft, and the verification result of the lightweight design of the key materials of the hydrogen fuel cell BOP is output based on the comparison result.

[0015] On the other hand, a device for lightweighting key materials of a hydrogen fuel cell BOP for an aircraft is provided, the device comprising:

[0016] A first data acquisition module is used to acquire a first material training data set of a hydrogen fuel cell BOP, wherein the first material training data set includes training data of a plurality of hydrogen fuel cell BOPs, wherein the training data includes material property characteristics, structural design characteristics, and material performance characteristics of key materials of the hydrogen fuel cell BOP, wherein the material performance characteristics include performance data of the influence of the key materials on the hydrogen fuel cell BOP obtained through experiments;

[0017] A first training module is used to obtain multiple key material prediction models, train each of the key material prediction models according to the first material training data set to predict the performance of the hydrogen fuel cell BOP, and obtain the predicted performance information of the hydrogen fuel cell BOP output by each of the key material prediction models;

[0018] A first performance parameter module is used to calculate a first performance evaluation parameter of a plurality of the key material prediction models according to the predicted performance information of the hydrogen fuel cell BOP output by each of the key material prediction models and the material performance characteristics; determine a stand-by key material prediction model from a plurality of key material prediction models according to the first performance evaluation parameter of each of the key material prediction models, wherein the first performance evaluation parameter is used to measure the distance quantification between key material training data;

[0019] A second data acquisition module is used to acquire a second material training data set of the hydrogen fuel cell BOP, where the second material training data set is a training data set of key materials of the hydrogen fuel cell BOP in different test environments;

[0020] a second training module, used for training the key material prediction model to be used according to the second material training data set to calculate a second performance evaluation parameter of the key material prediction model to be used on the second material training data set, wherein the second performance evaluation parameter is used to measure the stability of the prediction result of the key material prediction model to be used;

[0021] A model selection module, configured to select a target key material prediction model according to a second performance evaluation parameter of the key material prediction model to be used;

[0022] A lightweight module is used to obtain lightweight design data of key materials of hydrogen fuel cell BOP, wherein the lightweight design data includes material property characteristics and structural design characteristics of lightweight key materials; predict actual performance data of hydrogen fuel cell BOP according to the target key material prediction model and the lightweight design data, compare the actual performance data with the performance requirement data of hydrogen fuel cell BOP for aircraft, and output the verification result of lightweight design of key materials of hydrogen fuel cell BOP according to the comparison result:

[0023] On the other hand, a computer device is provided, which includes a processor and a memory, wherein the memory is used to store at least one computer program, and the at least one computer program is loaded and executed by the processor to implement the key material lightweighting method in the embodiment of the present application.

[0024] The embodiment of the present application provides a method for lightweighting key materials of a hydrogen fuel cell BOP for an aircraft, which includes obtaining a first material training data set of the hydrogen fuel cell BOP; obtaining multiple key material prediction models, training each of the key material prediction models according to the first material training data set to predict the performance of the hydrogen fuel cell BOP, and obtaining predicted performance information of the hydrogen fuel cell BOP output by each key material prediction model; calculating first performance evaluation parameters of multiple key material prediction models according to the predicted performance information of the hydrogen fuel cell BOP output by each key material prediction model and the material performance characteristics; determining a key material prediction model to be used from multiple key material prediction models according to the first performance evaluation parameters of each key material prediction model. Key material prediction model; obtain a second material training data set of hydrogen fuel cell BOP; train the key material prediction model to be used according to the second material training data set to calculate the second performance evaluation parameter of the key material prediction model to be used on the second material training data set; select a target key material prediction model according to the second performance evaluation parameter of the key material prediction model to be used; predict the actual performance data of hydrogen fuel cell BOP according to the target key material prediction model and lightweight design data, compare the actual performance data with the performance requirement data of hydrogen fuel cell BOP for aircraft, and output the verification result of lightweight design of key materials of hydrogen fuel cell BOP according to the comparison result. This solution can use artificial intelligence technology, especially machine learning, to accurately predict the performance of key materials of hydrogen fuel cell BOP after lightweight design, achieve lightweight goals, and improve lightweight efficiency and accuracy. BRIEF DESCRIPTION OF THE DRAWINGS

[0025] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0026] Figure 1 It is a schematic diagram of the implementation environment of a method for lightweighting key materials of a hydrogen fuel cell BOP for an aircraft provided in an embodiment of the present application;

[0027] Figure 2This is a flow chart of a method for lightweighting key materials of a hydrogen fuel cell BOP for an aircraft provided according to an embodiment of the present application;

[0028] Figure 3 is a flow chart of performance evaluation parameter calculation provided according to an embodiment of the present application;

[0029] Figure 4 It is a structural schematic diagram of a key material lightweight device for a hydrogen fuel cell BOP for an aircraft provided according to an embodiment of the present application;

[0030] Figure 5 It is a structural diagram of a server provided according to an embodiment of the present application. DETAILED DESCRIPTION

[0031] In order to make the objectives, technical solutions and advantages of the present application clearer, the implementation methods of the present application will be further described in detail below with reference to the accompanying drawings.

[0032] In this application, the terms "first", "second", etc. are used to distinguish identical or similar items with basically the same effects and functions. It should be understood that there is no logical or temporal dependency between "first", "second", and "nth", nor is there any limitation on quantity and execution order.

[0033] In the present application, the term "at least one" means one or more, and the term "plurality" means two or more.

[0034] It should be noted that in the embodiments of the present application, data related to the thermal conductivity, composition information, structural information, etc. of thermoelectric materials are involved. When the above embodiments of the present application are applied to products or technologies, relevant licenses or consents need to be obtained, and the collection, use and processing of relevant data need to comply with relevant laws, regulations and standards of relevant countries and regions.

[0035] The following explains the terms used in this application.

[0036] The field of artificial intelligence (AI) is dedicated to simulating and expanding human intelligence, involving theories, methods, technologies and application systems for perceiving the environment, acquiring knowledge and applying knowledge to achieve the best results. As a branch of computer science, AI aims to understand the nature of intelligence and create machines that can simulate human intelligent responses, with perception, reasoning and decision-making capabilities.

[0037] AI technology is an interdisciplinary field that covers both hardware and software technologies. Hardware includes sensors, AI chips, etc., while software includes computer vision, speech processing, natural language processing, and machine learning / deep learning. Among them, computer vision (CV) technology enables machines to "see", that is, to use cameras and computers instead of human eyes to perform recognition, measurement, and image processing to obtain information from images or multi-dimensional data. Computer vision technology includes image processing, recognition, semantic understanding, retrieval, OCR, video processing, semantic understanding, content / behavior recognition, three-dimensional reconstruction, 3D technology, virtual reality, augmented reality, simultaneous positioning and map construction, as well as biometric recognition technologies such as face recognition and fingerprint recognition.

[0038] Aircraft hydrogen fuel cell BOP (Balancing of Plant) key materials refer to those inactive components that are critical to the performance of hydrogen fuel cell systems. Although these materials and components do not directly participate in the electrochemical reaction, they play a vital role in ensuring the stable operation, performance optimization and long-term durability of the entire battery system. The following are some of the main BOP key materials:

[0039] Bipolar plates: Bipolar plates are key components in hydrogen fuel cells that separate individual cells and conduct current. They are usually made of lightweight, highly conductive and corrosion-resistant materials such as graphite composites, metal alloys (such as stainless steel or titanium alloys), etc.

[0040] Seals: Seals are used to prevent gas leakage and water vapor crossover, and they must be able to withstand the chemical environment and temperature changes inside the battery. Commonly used sealing materials include fluororubber, silicone rubber and other high-performance polymers.

[0041] Gas Diffusion Layer (GDL): The gas diffusion layer is located above the catalyst layer. Its main function is to evenly distribute the reactant gases and remove the generated water. GDL is usually made of carbon fiber paper or carbon fiber cloth, which have good gas permeability and electrical conductivity.

[0042] Membrane electrode assembly (MEA): MEA is the core part of hydrogen fuel cells, consisting of a proton exchange membrane (PEM), a catalyst layer and a gas diffusion layer. Among them, the proton exchange membrane is a key material that allows protons to pass through while blocking electrons, and is usually made of perfluorosulfonic acid polymer.

[0043] Cooling system materials: Since hydrogen fuel cells generate heat during operation, an effective cooling system is required to maintain a constant operating temperature. The cooling system may include heat exchangers, coolant pipes, etc. Commonly used materials include aluminum alloys, stainless steel, etc.

[0044] Electrical connectors: including cables, connectors and contactors, etc., which must be able to withstand high currents and voltages while having good corrosion resistance. Commonly used materials include copper alloys, gold-plated or silver-plated contact surfaces.

[0045] Control and sensor systems: Various sensors and actuators used to monitor and control battery performance, such as temperature sensors, pressure sensors, flow sensors, etc., their material selection depends on specific application requirements.

[0046] Insulation: Insulating materials such as fiberglass, ceramic, etc. used to ensure electrical safety and prevent short circuits.

[0047] The selection and design of these key BOP materials have a significant impact on the performance, life and cost of hydrogen fuel cells. Therefore, in high-performance applications such as aircraft, the requirements for these materials are particularly stringent.

[0048] Traditional BOP material selection often focuses on ensuring the mechanical strength and durability of the system, but this usually leads to an increase in overall weight, affecting the energy efficiency and range of the aircraft. Lightweighting is an important development direction of the aviation industry. It can not only improve the fuel efficiency of aircraft, but also reduce emissions and enhance economy. At present, the lightweighting of key materials for hydrogen fuel cell BOP faces multiple challenges. Lightweight materials must reduce the amount of material used while maintaining structural strength and stiffness to reduce the overall weight. Existing design methods often rely on experience and trial and error, and lack accurate predictive models to guide material selection and design optimization. With the development of artificial intelligence, especially machine learning technology, it is possible to achieve lightweight goals by utilizing artificial intelligence technology, especially machine learning algorithms, to predict and optimize the performance of key materials for hydrogen fuel cell BOP.

[0049] In one embodiment, for ease of understanding, please refer to Figure 1 , Figure 1It is a schematic diagram of the implementation environment of a method for lightweighting a key material of a hydrogen fuel cell BOP for aircraft provided in an embodiment of the present application. The method of the embodiment of the present application is applied to a lightweight design system for a key material of a hydrogen fuel cell BOP for aircraft. The lightweight design system for a key material of a hydrogen fuel cell BOP for aircraft includes: a lightweight prediction platform, a material performance testing device, and a terminal device; wherein the lightweight prediction platform can be an independent physical server, or a server cluster or distributed system composed of multiple physical servers, or a cloud server that provides basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communications, middleware services, domain name services, security services, content distribution networks (Content Delivery Network, CDN), and big data and artificial intelligence platforms. The terminal can be a smart phone, a tablet computer, a laptop, a desktop computer, a smart speaker, a smart watch, etc., but is not limited thereto. The material performance testing device is to provide a simulation environment to obtain the performance data of the key material on the hydrogen fuel cell BOP through experiments. The terminal and the server can be directly or indirectly connected by wired or wireless communication, and the embodiment of the present application is not limited here.

[0050] Among them, the material performance testing equipment collects performance indicators and other data through sensors of the aircraft hydrogen fuel cell system, and obtains performance data on the impact of key materials on the hydrogen fuel cell BOP based on simulation environment testing.

[0051] The lightweight prediction platform is used to: obtain a first material training data set of a hydrogen fuel cell BOP, the first material training data set includes training data of multiple hydrogen fuel cell BOPs, the training data includes material property characteristics, structural design characteristics and material performance characteristics of key materials of the hydrogen fuel cell BOP; obtain multiple key material prediction models, train each of the key material prediction models according to the first material training data set to predict the performance of the hydrogen fuel cell BOP, and obtain the predicted performance information of the hydrogen fuel cell BOP output by each predicted key material prediction model; calculate the first performance evaluation parameters of the multiple key material prediction models according to the predicted performance information of the hydrogen fuel cell BOP output by each key material prediction model and the material performance characteristics; determine the key material prediction model to be used from the multiple key material prediction models according to the first performance evaluation parameters of each key material prediction model, the first performance evaluation parameter is used to measure the distance quantification between different types of key material training data; obtain a second material training data set of the hydrogen fuel cell BOP, the first The first material training data set and the second material training data set are training data sets of key materials of hydrogen fuel cell BOP in different test environments; the prediction model of the key material to be used is trained according to the second material training data set to calculate the second performance evaluation parameter of the prediction model of the key material to be used on the second material training data set, and the second performance evaluation parameter is used to measure the stability of the prediction result of the prediction model of the key material to be used; the target key material prediction model is selected according to the second performance evaluation parameter of the prediction model of the key material to be used; the lightweight design data of the key material of the hydrogen fuel cell BOP is obtained, and the lightweight design data includes the material property characteristics and structural design characteristics after the key material is lightweighted; the actual performance data of the hydrogen fuel cell BOP is predicted according to the target key material prediction model and the lightweight design data, and the actual performance data is compared with the performance requirement data of the hydrogen fuel cell BOP for aircraft, and the verification result of the lightweight design of the key material of the hydrogen fuel cell BOP is output according to the comparison result. In one embodiment, the verification result can be sent to the terminal device so that the tester can adjust the lightweight design scheme, and finally achieve the lightweight goal of the key material of the hydrogen fuel cell BOP for aircraft.

[0052] The method for lightweighting the key materials of the hydrogen fuel cell BOP for aircraft provided in the embodiment of the present application can be executed by a lightweight prediction platform. In some embodiments, the lightweight prediction platform can be composed of a computer device as a terminal or a server. The following first takes the computer device as an example to introduce the method for lightweighting the key materials of the hydrogen fuel cell BOP for aircraft provided in the embodiment of the present application. Figure 2 This is a flowchart of a method for lightweighting key materials of a hydrogen fuel cell BOP for aircraft according to an embodiment of the present application, see Figure 2 In the present application, the lightweight prediction platform is used as an example to illustrate the material screening method. The material screening method includes the following steps:

[0053] 201. Obtain a first material training data set of a hydrogen fuel cell BOP, wherein the first material training data set includes training data of multiple hydrogen fuel cell BOPs, the training data includes material property characteristics, structural design characteristics, and material performance characteristics of key materials of the hydrogen fuel cell BOP, and the material performance characteristics include performance data of the influence of the key materials on the hydrogen fuel cell BOP obtained through experiments.

[0054] The first material training data set includes at least two training data, which are training data of hydrogen fuel cell BOP, including material property characteristics, structural design characteristics and material performance characteristics. It can be obtained from the data set platform or constructed based on historical data.

[0055] Among them, the material property characteristics of the key materials of the hydrogen fuel cell BOP include data characterizing the characteristics in terms of physics, chemistry and mechanics. In one embodiment, the lightweight prediction platform can obtain the material property characteristics characterizing the physical properties of the key material, and the physical properties include the number of electrons, atomic radius and valence electron number of the key material elements. In one embodiment, the key material can be a compound composed of at least one element. The chemical data includes chemical composition information, wherein the composition information may include at least one element information of the number of electrons, atomic mass, period number, family number, atomic radius, electronegativity, first ionization energy, density, melting point, boiling point, specific heat capacity, number of electron shells and number of valence electrons of at least one element constituting the key material. The thermal conductivity of the key material is the lattice thermal conductivity obtained by high-throughput calculation. The specific characteristics can be selected according to actual needs.

[0056] Structural design features characterize the structural properties of key materials, including data such as size, weight, geometry, layout, and connection method of key materials corresponding to the aircraft hydrogen fuel cell BOP.

[0057] Among them, the material performance characteristics characterize the performance data of key materials affecting the hydrogen fuel cell BOP, including: mechanical strength and stiffness, durability and fatigue life, thermal conductivity and thermal expansion coefficient, electrical conductivity, corrosion resistance, chemical resistance, sealing performance, lightweight, environmental adaptability and other aspects of material performance characteristics.

[0058] In practical applications, the performance of hydrogen fuel cell BOPs is affected by a variety of material performance characteristics of key materials, including mechanical strength and stiffness, durability and fatigue life, thermal conductivity and thermal expansion coefficient, electrical conductivity, corrosion resistance, chemical resistance, sealing performance, lightweight, environmental adaptability, thermal stability and temperature resistance, anti-aging performance, fracture toughness, wear resistance, electromagnetic compatibility, and machinability. Mechanical strength and stiffness characterize the ability of materials to resist deformation and fracture, and durability and fatigue life show the performance degradation of materials under repeated or continuous stress. Thermal conductivity and thermal expansion coefficient affect the temperature control of the battery, while electrical conductivity is particularly important for conductive materials. Corrosion resistance and chemical resistance are related to the stability of materials in a specific chemical environment, and sealing performance is aimed at the ability of sealing materials to prevent gas or liquid leakage. Lightweight affects the overall weight of the aircraft, environmental adaptability involves the performance changes of materials under different environmental conditions, and thermal stability and temperature resistance characterize the performance retention of materials at high or low temperatures. Anti-aging performance describes the performance degradation of materials in long-term use, and fracture toughness and wear resistance characterize the ability of materials to resist crack propagation and wear, respectively. Electromagnetic compatibility affects the performance of materials in electromagnetic fields, while machinability involves the ease and cost of material processing. The data on these material performance characteristics are obtained through laboratory tests, field tests and simulation calculations, providing an important reference for the design, material selection and optimization of hydrogen fuel cell BOP systems, and helping to improve the overall performance and reliability of the system.

[0059] In one embodiment, to improve the efficiency and accuracy of obtaining training data sets, the lightweight prediction platform can construct training data in the following manner:

[0060] For each key material of the hydrogen fuel cell BOP, construct material property characteristics of the key material according to the material stoichiometric ratio and physical and mechanical properties of the elements of the key material;

[0061] For each key material of the hydrogen fuel cell BOP, construct the structural design features of the key material according to the structural properties of the key material;

[0062] For each key material of a hydrogen fuel cell BOP, the material performance characteristics of the key material are constructed based on the performance data of the key material on the hydrogen fuel cell BOP obtained through experiments;

[0063] The material property characteristics of key materials of multiple hydrogen fuel cell BOPs are integrated to obtain a first material training data set.

[0064] The material stoichiometric ratio is the ratio of each element in the chemical composition of the material, and the physical and mechanical properties are data characterizing the physical and mechanical characteristics of the material, such as strength, etc. In one embodiment, the material stoichiometric ratio may include a stoichiometric ratio, which refers to the molar ratio between reactants and products calculated according to a chemical reaction stoichiometric equation.

[0065] In the field of materials science, the performance of key materials is determined by the material stoichiometric ratio and physical and mechanical properties of its elements. The chemical composition, including the ratio of each element, determines the basic chemical properties and stability of the material, while the phase composition involves the ratio of different phases in the material, such as solid solution, intermetallic compound, carbide, etc., which directly affects the microstructure and macroscopic properties of the material. The impurity content, even in trace amounts, may significantly affect the strength, toughness and corrosion resistance of the material. In terms of physical and mechanical properties, density affects the weight and application range of the material, especially in aircraft materials; elastic modulus and tensile strength determine the stiffness and load-bearing capacity of the material; yield strength marks the starting point of plastic deformation of the material; hardness and fracture toughness are the key to measuring the wear resistance and crack propagation resistance of the material; fatigue strength and impact toughness show the performance of the material under cyclic loading and impact load. In one embodiment, the first material training data set is constructed by the lightweight prediction platform based on the composition information, physical structure characteristics, mechanical strength, stiffness, thermal conductivity, and lightweight material performance characteristics of the key materials of the hydrogen fuel cell BOP.

[0066] In some embodiments, for each key material of a hydrogen fuel cell BOP, the material property characteristics of the key material are constructed according to the material stoichiometric ratio and physical and mechanical properties of the elements of the key material. The lightweight prediction platform includes: for any key material, the lightweight prediction platform can determine the material property characteristics of the key material according to the material stoichiometric ratio (stoichiometric ratio, which refers to the molar ratio between reactants and products calculated according to the chemical reaction stoichiometric equation) and physical and mechanical properties of the elements constituting the key material. The lightweight prediction platform can process the material property characteristics of multiple key materials to obtain a sample data set. Among them, the lightweight prediction platform can process the material property characteristics in a manner that includes screening and normalization. The material property characteristics may include data of multiple dimensions, and the multiple dimensions include at least one of the number of electrons, atomic mass, period number, family number, atomic radius, electronegativity, first ionization energy, density, melting point, boiling point, specific heat capacity, number of electron shells, and number of valence electrons. By constructing a first material training data set based on the composition information and physical and mechanical property characteristics of key materials, the training data in the first material training data set can be used as training data for the model. Since the composition information of various key materials contains rich data, the generalization ability of the trained model can be improved.

[0067] 202. Acquire multiple key material prediction models, train each of the key material prediction models according to the first material training data set to predict the performance of the hydrogen fuel cell BOP, and obtain predicted performance information of the hydrogen fuel cell BOP output by each key material prediction model.

[0068] Among them, multiple key material prediction models are prediction models built with different machine learning algorithms. When building a prediction model for the performance of key materials of hydrogen fuel cell BOP, a variety of machine learning algorithms can be used, each of which has its own unique advantages and applicable scenarios. Linear regression models are suitable for preliminary analysis and simple prediction tasks, especially when the data is linearly related. Logistic regression focuses on binary classification problems and predicts the probability of a specific event. Decision trees classify data through a series of rules, while random forests improve the accuracy and robustness of predictions by integrating multiple decision trees. Support vector machines (SVMs) are suitable for classification problems in high-dimensional spaces with their maximum margin classifier characteristics. Neural networks and deep learning models can handle complex nonlinear problems and make predictions by learning the mapping relationship between input and output, which is particularly suitable for large-scale data sets and complex feature learning. Gradient boosting machines (GBM) and extreme gradient boosting (XGBoost) improve prediction performance by building multiple weak prediction models and gradually optimizing their errors, and are suitable for various complex prediction problems. Bayesian networks are based on probabilistic graph models, describing the probabilistic relationship between variables, and are suitable for problems that need to deal with uncertainty and dependencies between variables. Combining these algorithms, we can select and adjust them according to specific application requirements, data characteristics and prediction goals to achieve the best prediction results. In practical applications, we may need to try multiple algorithms and evaluate their performance through methods such as cross-validation to select the most suitable model.

[0069] In one embodiment, for any prediction model, the lightweight prediction platform can adopt a supervised learning method to perform multiple iterations of training on the constructed prediction model based on the training data set. During each iteration of training, the candidate prediction model of this iteration processes the material property characteristics, structural design characteristics and other characteristics of the training data to obtain the predicted performance information of the key material, and then adjusts the model parameters of the candidate prediction model of this iteration based on the difference between the predicted performance information and the actual material performance characteristics of the key material. Among them, the candidate prediction model is used to output the performance prediction results of the key material.

[0070] In some embodiments, training each of the key material prediction models according to the first key material training data set to predict the performance of the hydrogen fuel cell BOP includes:

[0071] For any deep learning algorithm, the model parameters of the key material prediction model corresponding to the deep learning algorithm are determined according to the attribute information of the first material training data set; according to the first material training data set and the model parameters of the key material prediction model, the key material prediction model is trained to predict the performance of the hydrogen fuel cell BOP. For example, for any machine learning algorithm, the lightweight prediction platform can determine the hyperparameters of the prediction model corresponding to the machine learning algorithm based on the training data set, and then train the candidate prediction model based on the training data set. Among them, the hyperparameters of the prediction model include the number of training times, the number of dimensions of the input features, and the evaluation index, etc., which are not limited in the embodiments of the present application.

[0072] 203. Calculate first performance evaluation parameters of multiple key material prediction models based on the predicted performance information of the hydrogen fuel cell BOP output by each key material prediction model and the material performance characteristics; determine a stand-by key material prediction model from multiple key material prediction models based on the first performance evaluation parameter of each key material prediction model, wherein the first performance evaluation parameter is used to quantify the distance between the key material training data.

[0073] Among them, the first performance evaluation parameter is a quantitative indicator calculated based on the training data of the model, and the quantitative indicator can characterize the classification ability of the prediction model, that is, the ability to represent the same type of sample training data as similar results, and / or to represent different types of sample training data as distant results. In other words, the quantitative indicator can measure the degree of alienation between samples of different categories and the degree of similarity between samples of the same category.

[0074] In one embodiment, reference Figure 3 , the lightweight prediction platform can calculate the first performance evaluation parameter in the following way:

[0075] 2031. Obtain, according to the predicted performance information and the material performance characteristics, a first predicted distance between different types of training data and a second predicted distance between the same type of training data during training of the key material prediction model.

[0076] The first prediction distance represents or measures the similarity between different types of training data during training of the prediction model, and the second prediction distance represents or measures the degree of alienation or difference between the same type of training data of the prediction model.

[0077] In one embodiment, the cosine distance calculation method can be used to calculate the predicted distance; of course, in actual application, other methods can also be used to calculate the distance of training data, such as Euclidean distance, Manhattan distance, Chebyshev distance, etc., which is not limited in this embodiment.

[0078] 2032. Fuse the first prediction distance and the second prediction distance to obtain the first performance evaluation parameter.

[0079] There are many fusion processing methods. For example, in order to improve the accuracy of the performance evaluation parameters, the first weight corresponding to the same type of training data and the second weight corresponding to the different types of training data can be obtained;

[0080] A weighted summation process is performed on the first predicted distance and the second predicted distance according to the first weight and the second weight, and a result of the weighted summation process is used as a first performance evaluation parameter.

[0081] In one embodiment, in order to improve the prediction accuracy, the lightweight prediction platform can obtain the performance prediction results output by the prediction model, and then determine the inter-category distance and intra-category distance of each type of training data according to the label of each training data set of the first material (the label represents the material performance characteristics), and generate a split-combination ratio according to the ratio of the inter-category distance and the intra-category distance, and use the split-combination ratio as the first performance evaluation parameter.

[0082] In one embodiment, the median of the distance may be used instead of the mean distance to calculate the split-to-combination ratio. The above is only an example of the present application and does not constitute a limitation on the technical solution of the present application.

[0083] In the implementation of this application, after the lightweight prediction platform obtains the first performance evaluation parameter of each key material prediction model, the key material prediction model to be used is selected from it according to the first performance evaluation parameter. For example, in one embodiment, a prediction model whose first performance evaluation parameter is greater than the first parameter threshold can be selected as the key material prediction model to be used. The larger the first performance evaluation parameter, the greater the distance between the training data of the representation prediction model, indicating that the difference between different categories of training data is more obvious, and the more similar the same type of training data is, reflecting that the result of unsupervised representation learning meets the classification standard on a small training sample, and the more accurate the representation prediction classification result output by the representation prediction model is, the effectiveness and accuracy of the entire model training are guaranteed.

[0084] 204. Obtain a second material training data set of a hydrogen fuel cell BOP, where the first material training data set and the second material training data set are training data sets of key materials of the hydrogen fuel cell BOP in different test environments.

[0085] In one embodiment, the lightweight prediction platform can acquire or construct a second material training data set, where the second material training data set and the first material training data set are training data sets constructed from data of key materials of hydrogen fuel cell BOP collected in different test environments.

[0086] The different test environments here can be different temperature environments, humidity environments or pressure environments, etc. The second material training data set obtained under different test environments can be effectively distinguished from the first material training data set, so that more different data can be used to evaluate the prediction model of the key material to be used, thereby improving the accuracy of the second performance evaluation parameters, and then improving the accuracy of the subsequent acquisition of the target key material prediction model.

[0087] Similarly, the second material training data set may include training data of multiple hydrogen fuel cell BOPs, wherein the training data includes material property characteristics, structural design characteristics, and material performance characteristics of key materials of the hydrogen fuel cell BOPs. The construction and generation of the training data in the second material training data set may refer to the above introduction to the first material training data set.

[0088] For example, the lightweight prediction platform can construct the second material training data set in the following manner: for each key material of a hydrogen fuel cell BOP, the material property characteristics of the key material are constructed according to the material stoichiometric ratio and physical and mechanical properties of the elements of the key material; for each key material of a hydrogen fuel cell BOP, the structural design characteristics of the key material are constructed according to the structural properties of the key material; for each key material of a hydrogen fuel cell BOP, the material performance characteristics of the key material are constructed according to the performance data of the influence of the key material on the hydrogen fuel cell BOP obtained through experiments; the material property characteristics of the key materials of multiple hydrogen fuel cell BOPs are integrated to obtain the second material training data set.

[0089] 205. Train the stand-by key material prediction model according to the second material training data set to calculate a second performance evaluation parameter of the stand-by key material prediction model on the second material training data set, wherein the second performance evaluation parameter is used to measure the stability of the prediction result of the stand-by key material prediction model.

[0090] Among them, the second performance evaluation parameter is a quantitative indicator for measuring the stability of training data representation generated based on the similarity results corresponding to the prediction results of each training data in the second material training data set learned by the prediction model during the training process (that is, the similarity vector corresponding to the prediction result vector in the computer algorithm).

[0091] Specifically, in one embodiment, the lightweight prediction platform can extract the prediction results of the key material prediction model to be used for each second training data in the second material training data set; determine the sub-material training data set to which it belongs in the second material training data set according to the prediction results of the second training data, and the sub-material training data set includes multiple similar second training data; and generate a second performance evaluation parameter of the key material prediction model to be used on the second material training data set based on the sub-material training data set to which each second training data belongs.

[0092] In one embodiment, in order to improve the prediction accuracy and stability of the model, the lightweight prediction platform can obtain the reference model stability parameters of the training data in the corresponding sub-material training data set for each second training data; count the proportion of training data in the corresponding sub-material training data set whose reference model stability parameters are greater than a preset stability parameter threshold, and calculate the target model stability parameters corresponding to each second training data according to the training data proportion; calculate the second performance evaluation parameters of the stand-by key material prediction model on the second material training data set according to the target model stability parameters corresponding to each second training data.

[0093] Among them, the reference model stability parameter can be a quantitative indicator used to measure the stability of training data representation generated based on the similar results corresponding to the prediction results of each training data in the sub-material training data set of the second material training data set learned by the prediction model during the training process.

[0094] In one embodiment, the prediction model can be iteratively trained in rounds. In order to calculate the index measuring the stability of the training data sample representation, the lightweight prediction platform can obtain the prediction result vectors obtained by the prediction model in multiple iteration rounds for each sample in the training data sample set, and select the most similar M samples for each training data sample in the sub-material training data set according to the prediction result vectors of each sample in the training sample set learned in each iteration round, and add the M similar training data samples selected for the training data sample to the similar sample set corresponding to each sample in the subset and the iteration round, and generate the index corresponding to each training data sample in the sub-material training data set, such as the Jacquard index, for the multiple similar sample sets corresponding to each sample in the sub-material training data set, and use the Jacquard index as the reference model stability parameter. Wherein, M is a preset number, which can be set according to actual needs and can be set to a positive integer greater than 1.

[0095] Among them, the Jaccard Index, also known as the Jaccard Similarity Coefficient, is a statistic used to measure the similarity between two training sample sets. It is defined as the ratio of the size of the intersection of two training sets to the size of their union, and is usually used to describe the similarity or diversity between samples.

[0096] In one embodiment, after obtaining the reference model stability parameters of the training data, in order to improve the screening of models with strong stability, the proportion of training data in the sub-material training data set whose reference model stability parameters are greater than a preset stability parameter threshold can be counted, and the target model stability parameters corresponding to each second training data are calculated according to the training data proportion. In one embodiment, the training data proportion can be directly used as the target model stability parameter.

[0097] In one embodiment, in order to further accurately determine the stability parameters, after obtaining the target model stability parameters corresponding to each second training data, the target model stability parameters corresponding to each second training data may be enhanced by summing or evaluating the calculation results, and the calculation results are used as the second performance evaluation parameters.

[0098] It should be understood that in an effective model learning and training process, as the prediction model training deepens, the similar sample sets corresponding to each sample in the sub-material training data set should show high correlation and have a trend of gradually fixing. In other words, the similar sample sets corresponding to each training data sample no longer change a lot. According to this situation, it can be determined whether the training of the prediction model tends to be stable based on the size of the Jacquard index within a preset time period. For the samples in the second sample subset, if its Jacquard index is greater than the Jacquard index threshold, it indicates that the similar sample set of the sample is almost the same before and after the iteration. If the Jacquard index of samples exceeding the preset proportion is greater than the Jacquard index threshold, it indicates that the prediction model tends to be stable.

[0099] In each of the embodiments of the present application, the size of the Jacquard index within a preset time period in the sub-material training data set can be calculated, and then the proportion of the number of samples in the sub-training data corresponding to each training data sample whose Jacquard index is greater than a preset threshold is counted, and the sample number ratio is used as the target model stability parameter corresponding to the training data sample. Finally, the sample number ratio is weighted and summed to obtain a parameter value between 0 and 1, which is the second performance evaluation parameter, which is used to measure the stability of the prediction model during the training process.

[0100] 206. Select a target key material prediction model according to the second performance evaluation parameter of the stand-by key material prediction model.

[0101] After obtaining the second performance evaluation parameter of each prediction model, the lightweight prediction platform can select the target key material prediction model according to the second performance evaluation parameter of the standby key material prediction model. For example, in one embodiment, the model with the largest parameter value can be selected as the target key material prediction model, that is, the model with strong stability and high prediction classification accuracy can be selected as the target key material prediction model.

[0102] In one embodiment, in order to improve the accuracy and stability of material performance prediction, specifically, the stand-by key material prediction models can be sorted according to the second performance evaluation parameter, and the model with the largest parameter value of the second performance evaluation parameter and a parameter value greater than a preset abnormal performance threshold is selected as the target key material prediction model.

[0103] Among them, the preset abnormal performance threshold can represent the threshold of model stability abnormality, and the size can be set according to actual needs.

[0104] 207. Obtain lightweight design data of key materials of the hydrogen fuel cell BOP, wherein the lightweight design data includes material property characteristics and structural design characteristics of the key materials after lightweighting.

[0105] In practical applications, in order to make lightweight design for key materials of hydrogen fuel cell BOP, designers can design lightweight data of key materials on the terminal and send it to the lightweight prediction platform through the terminal device for performance prediction.

[0106] 208. Predict actual performance data of the hydrogen fuel cell BOP based on the target key material prediction model and the lightweight design data, compare the actual performance data with the performance requirement data of the hydrogen fuel cell BOP for aircraft, and output the verification result of the lightweight design of the key materials of the hydrogen fuel cell BOP based on the comparison result.

[0107] After the lightweight prediction platform obtains data, the trained target key material prediction model can be used to predict the actual performance of the hydrogen fuel cell BOP based on the lightweight design data, such as physical properties (strength, force, etc.), chemical properties (heat resistance, etc.), etc., and then compare the actual performance data with the performance requirement data of the aircraft hydrogen fuel cell BOP. If it meets the performance requirements of the aircraft hydrogen fuel cell BOP, the lightweight design verification is passed, otherwise, the output is failed. After obtaining the verification results, they can be fed back to the terminal device.

[0108] As can be seen from the above, in an embodiment of the present application, a first material training data set of a hydrogen fuel cell BOP is obtained, the first material training data set includes training data of multiple hydrogen fuel cell BOPs, the training data includes material property characteristics, structural design characteristics and material performance characteristics of key materials of the hydrogen fuel cell BOP, and the material performance characteristics include performance data of the influence of the key materials on the hydrogen fuel cell BOP obtained through experiments; multiple key material prediction models are obtained, and each of the key material prediction models is trained according to the first material training data set to predict the performance of the hydrogen fuel cell BOP, and the predicted performance information of the hydrogen fuel cell BOP output by each predicted key material prediction model is obtained; the first performance evaluation parameters of the multiple key material prediction models are calculated according to the predicted performance information of the hydrogen fuel cell BOP output by each key material prediction model and the material performance characteristics; the key material prediction model to be used is determined from the multiple key material prediction models according to the first performance evaluation parameters of each key material prediction model, and the first performance evaluation parameter is used to quantify the distance between the key material training data; the hydrogen fuel cell BOP is obtained. a second material training data set for a hydrogen fuel cell BOP, wherein the first material training data set and the second material training data set are training data sets for key materials of the hydrogen fuel cell BOP in different test environments; the stand-by key material prediction model is trained according to the second material training data set to calculate a second performance evaluation parameter of the stand-by key material prediction model on the second material training data set, wherein the second performance evaluation parameter is used to measure the stability of the prediction result of the stand-by key material prediction model; a target key material prediction model is selected according to the second performance evaluation parameter of the stand-by key material prediction model; lightweight design data of key materials of the hydrogen fuel cell BOP are obtained, wherein the lightweight design data includes material property characteristics and structural design characteristics of the lightweighted key materials; actual performance data of the hydrogen fuel cell BOP is predicted according to the target key material prediction model and the lightweight design data, the actual performance data is compared with the performance requirement data of the aircraft hydrogen fuel cell BOP, and the verification result of the lightweight design of the key materials of the hydrogen fuel cell BOP is output according to the comparison result. This solution can be based on pre-training of multiple key material prediction models, and select a prediction model with high accuracy and stability as the key material prediction model for lightweight design through the first and second performance evaluation parameters of multiple key material prediction models, thereby improving the accuracy of key material performance prediction and further improving the reliability of lightweight design of key materials of hydrogen fuel cell BOP.

[0109] Figure 4 This is a block diagram of a key material lightweight device for a hydrogen fuel cell BOP for an aircraft provided according to an embodiment of the present application. The device is used to implement the above-mentioned key material lightweight method, see Figure 4, the device comprises:

[0110] A first data acquisition module 401 is used to acquire a first material training data set of a hydrogen fuel cell BOP, wherein the first material training data set includes training data of a plurality of hydrogen fuel cell BOPs, wherein the training data includes material property characteristics, structural design characteristics, and material performance characteristics of key materials of the hydrogen fuel cell BOP, wherein the material performance characteristics include performance data of the influence of the key materials on the hydrogen fuel cell BOP obtained through experiments;

[0111] A first training module 402 is used to obtain a plurality of key material prediction models, train each of the key material prediction models according to the first material training data set to predict the performance of the hydrogen fuel cell BOP, and obtain the predicted performance information of the hydrogen fuel cell BOP output by each predicted key material prediction model;

[0112] A first performance parameter module 403 is used to calculate the first performance evaluation parameters of multiple key material prediction models according to the predicted performance information of the hydrogen fuel cell BOP output by each key material prediction model and the material performance characteristics; determine the key material prediction model to be used from multiple key material prediction models according to the first performance evaluation parameter of each key material prediction model, wherein the first performance evaluation parameter is used to measure the distance quantification between the key material training data;

[0113] The second data acquisition module 404 is used to acquire a second material training data set of the hydrogen fuel cell BOP, where the second material training data set is a training data set of key materials of the hydrogen fuel cell BOP in different test environments;

[0114] A second training module 405 is used to train the standby key material prediction model according to the second material training data set to calculate a second performance evaluation parameter of the standby key material prediction model on the second material training data set, wherein the second performance evaluation parameter is used to measure the stability of the prediction result of the standby key material prediction model;

[0115] A model selection module 406, configured to select a target key material prediction model according to the second performance evaluation parameter of the key material prediction model to be used;

[0116] The lightweight module 407 is used to obtain lightweight design data of key materials of the hydrogen fuel cell BOP, wherein the lightweight design data includes material property characteristics and structural design characteristics of the key materials after lightweighting; predict actual performance data of the hydrogen fuel cell BOP according to the target key material prediction model and the lightweight design data, compare the actual performance data with the performance requirement data of the hydrogen fuel cell BOP for aircraft, and output the verification result of the lightweight design of the key materials of the hydrogen fuel cell BOP according to the comparison result.

[0117] In one embodiment, the first training module 402 is used to:

[0118] For any deep learning algorithm, determining model parameters of a key material prediction model corresponding to the deep learning algorithm according to the attribute information of the first material training data set;

[0119] The key material prediction model is trained according to the first material training data set and the model parameters of the key material prediction model to predict the performance of the hydrogen fuel cell BOP.

[0120] In one embodiment, the first data acquisition module 401 is further used to:

[0121] For each key material of the hydrogen fuel cell BOP, construct material property characteristics of the key material according to the material stoichiometric ratio and physical and mechanical properties of the elements of the key material;

[0122] For each key material of the hydrogen fuel cell BOP, construct the structural design features of the key material according to the structural properties of the key material;

[0123] For each key material of a hydrogen fuel cell BOP, constructing material performance characteristics of the key material according to performance data of the key material on the hydrogen fuel cell BOP obtained through experiments;

[0124] The material property characteristics of key materials of multiple hydrogen fuel cell BOPs are integrated to obtain a first material training data set.

[0125] In one embodiment, the first performance parameter module 403 is specifically used to:

[0126] Acquire, according to the predicted performance information and the material performance characteristics, a first predicted distance between different types of training data and a second predicted distance between the same type of training data during training of the key material prediction model;

[0127] The first prediction distance and the second prediction distance are fused to obtain the first performance evaluation parameter.

[0128] In one embodiment, the first performance parameter module 403 is specifically used to: obtain a first weight corresponding to the same type of training data and a second weight corresponding to different types of training data; perform weighted sum processing on the first predicted distance and the second predicted distance according to the first weight and the second weight, and use the result of the weighted sum processing as the first performance evaluation parameter.

[0129] In one embodiment, the second training module 405 is specifically configured to:

[0130] Extracting prediction results of the to-be-used key material prediction model training on each second training data in the second material training data set;

[0131] Determine a corresponding sub-material training data set in the second material training data set according to the prediction result of each second training data, wherein the sub-material training data set includes a plurality of similar second training data;

[0132] A second performance evaluation parameter of the key material prediction model to be used on the second material training data set is generated based on the sub-material training data set to which each of the second training data belongs.

[0133] In one embodiment, the second training module 405 is specifically configured to:

[0134] For each of the second training data, obtaining a reference model stability parameter of the training data in the corresponding sub-material training data set;

[0135] Counting the proportion of training data in which the reference model stability parameter is greater than a preset stability parameter threshold in the sub-material training data set, and calculating the target model stability parameter corresponding to each second training data according to the training data proportion;

[0136] A second performance evaluation parameter of the stand-by key material prediction model on the second material training data set is calculated according to the target model stability parameter corresponding to each of the second training data.

[0137] In one embodiment, the model selection module 406 is used to sort the standby key material prediction models according to the second performance evaluation parameter, and select the model with the largest parameter value of the second performance evaluation parameter and a parameter value greater than a preset abnormal performance threshold as the target key material prediction model.

[0138] An embodiment of the present application provides a solution for lightweighting key materials. The solution can be based on pre-training of multiple key material prediction models, and select a prediction model with high accuracy and stability as the key material prediction model for lightweight design through the first and second performance evaluation parameters of multiple key material prediction models, thereby improving the accuracy of key material performance prediction and further improving the reliability of lightweight design of key materials of hydrogen fuel cell BOP.

[0139] It should be noted that: the key material lightweighting device provided in the above embodiment only uses the division of the above functional modules as an example when performing key material lightweighting. In actual applications, the above functions can be assigned to different functional modules as needed, that is, the internal structure of the device can be divided into different functional modules to complete all or part of the functions described above. In addition, the key material lightweighting device provided in the above embodiment and the key material lightweighting method embodiment belong to the same concept. The specific implementation process is detailed in the method embodiment and will not be repeated here.

[0140] In the embodiments of the present application, the computer device can be configured as a terminal or a server. When the computer device is configured as a terminal, the terminal serves as the execution subject to implement the technical solution provided in the embodiments of the present application; when the computer device is configured as a server, the server serves as the execution subject to implement the technical solution provided in the embodiments of the present application; or, the technical solution provided in the present application is implemented through interaction between the terminal and the server, which is not limited in the embodiments of the present application.

[0141] Figure 5 It is a schematic diagram of the structure of a server (computer device) provided in an embodiment of the present application. The server 300 may have relatively large differences due to different configurations or performances, and may include one or more central processing units (CPU) 321 (for example, one or more processors) and memory 332, one or more storage media for storing application programs or data (for example, one or more mass storage devices). Among them, the memory 332 and the storage medium can be short-term storage or permanent storage. The program stored in the storage medium may include one or more modules (not shown in the figure), and each module may include a series of instruction operations on the server. Furthermore, the central processing unit 321 can be configured to communicate with the storage medium and execute a series of instruction operations in the storage medium on the server 300.

[0142] The steps performed by the server in the above embodiment can be based on the Figure 5The server structure shown. Those skilled in the art can clearly understand that for the convenience and brevity of description, the specific working processes of the above-described systems, devices and units can refer to the corresponding processes in the aforementioned method embodiments, and will not be repeated here.

[0143] In the several embodiments provided in the present application, it should be understood that the disclosed systems, devices and methods can be implemented in other ways. For example, the device embodiments described above are only schematic. For example, the division of units is only a logical function division. There may be other division methods in actual implementation. For example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be an indirect coupling or communication connection through some interfaces, devices or units, which can be electrical, mechanical or other forms.

[0144] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0145] In addition, each functional unit in each embodiment of the present application may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit. The above-mentioned integrated unit may be implemented in the form of hardware or in the form of software functional units.

[0146] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application, or the part that contributes to the prior art, or all or part of the technical solution can be embodied in the form of a software product, which is stored in a storage medium and includes several instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the various embodiments of the present application. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM), random access memory (RAM), disk or optical disk, etc., and other media that can store program codes.

[0147] The above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit them. Although the present application has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. A method for lightweighting key materials of hydrogen fuel cell BOP for aircraft, characterized in that: The method comprises: Acquire a first material training data set of a hydrogen fuel cell BOP, wherein the first material training data set includes training data of a plurality of hydrogen fuel cell BOPs, wherein the training data includes material property characteristics, structural design characteristics, and material performance characteristics of key materials of the hydrogen fuel cell BOP, wherein the material performance characteristics include performance data of the influence of the key materials on the hydrogen fuel cell BOP obtained through experiments; Acquire multiple key material prediction models, train each of the key material prediction models according to the first material training data set to predict the performance of the hydrogen fuel cell BOP, and obtain predicted performance information of the hydrogen fuel cell BOP output by each of the key material prediction models; Calculating first performance evaluation parameters of a plurality of the key material prediction models according to the predicted performance information of the hydrogen fuel cell BOP output by each of the key material prediction models and the material performance characteristics; determining a key material prediction model to be used from a plurality of key material prediction models according to the first performance evaluation parameter of each of the key material prediction models, wherein the first performance evaluation parameter is used to quantify the distance between key material training data; Acquire a second material training data set for a hydrogen fuel cell BOP, where the second material training data set is a training data set for key materials of the hydrogen fuel cell BOP in different test environments; Training the key material prediction model for standby use according to the second material training data set to calculate a second performance evaluation parameter of the key material prediction model for standby use on the second material training data set, wherein the second performance evaluation parameter is used to measure the stability of the prediction result of the key material prediction model for standby use; Selecting a target key material prediction model according to a second performance evaluation parameter of the key material prediction model to be used; Obtaining lightweight design data of key materials of hydrogen fuel cell BOP, wherein the lightweight design data includes material property characteristics and structural design characteristics of the key materials after lightweighting; The actual performance data of the hydrogen fuel cell BOP is predicted based on the target key material prediction model and the lightweight design data, the actual performance data is compared with the performance requirement data of the hydrogen fuel cell BOP for aircraft, and the verification result of the lightweight design of the key materials of the hydrogen fuel cell BOP is output based on the comparison result.

2. The method according to claim 1, characterized in that The step of training each of the key material prediction models according to the first material training data set to predict the performance of the hydrogen fuel cell BOP includes: For any deep learning algorithm, determining model parameters of a key material prediction model corresponding to the deep learning algorithm according to the attribute information of the first material training data set; The key material prediction model is trained according to the first material training data set and the model parameters of the key material prediction model to predict the performance of the hydrogen fuel cell BOP.

3. The method according to claim 2, characterized in that The method further comprises: For each key material of the hydrogen fuel cell BOP, construct material property characteristics of the key material according to the material stoichiometric ratio and physical and mechanical properties of the elements of the key material; For each key material of the hydrogen fuel cell BOP, construct the structural design features of the key material according to the structural properties of the key material; For each key material of a hydrogen fuel cell BOP, constructing material performance characteristics of the key material according to performance data of the key material on the hydrogen fuel cell BOP obtained through experiments; The material property characteristics of key materials of multiple hydrogen fuel cell BOPs are integrated to obtain the first material training data set.

4. The method according to claim 1, characterized in that: Calculating first performance evaluation parameters of a plurality of the key material prediction models according to the predicted performance information of the hydrogen fuel cell BOP output by each of the key material prediction models and the material performance characteristics, including: Acquire, according to the predicted performance information and the material performance characteristics, a first predicted distance between different types of training data and a second predicted distance between the same type of training data during training of the key material prediction model; The first prediction distance and the second prediction distance are fused to obtain the first performance evaluation parameter.

5. The method according to claim 4, characterized in that The first predicted distance and the second predicted distance are fused to obtain the first performance evaluation parameter, including: Obtaining first weights corresponding to the same type of training data and second weights corresponding to different types of training data; A weighted summation process is performed on the first predicted distance and the second predicted distance according to the first weight and the second weight, and a result of the weighted summation process is used as a first performance evaluation parameter.

6. The method according to any one of claims 1 to 5, characterized in that The key material prediction model to be used is trained according to the second material training data set to calculate a second performance evaluation parameter of the key material prediction model to be used on the second material training data set, including: Extracting the prediction result of each second training data of the to-be-used key material prediction model in the second material training data set; Determine, in the second material training data set, a sub-material training data set according to a prediction result of the second training data, wherein the sub-material training data set includes a plurality of similar second training data; A second performance evaluation parameter of the key material prediction model to be used on the second material training data set is generated based on the sub-material training data set to which each of the second training data belongs.

7. The method according to claim 6, characterized in that Generating a second performance evaluation parameter of the key material prediction model to be used on the second material training data set based on the sub-material training data set to which each of the second training data belongs includes: For each of the second training data, obtaining a reference model stability parameter of the training data in the corresponding sub-material training data set; Counting the proportion of training data whose reference model stability parameter is greater than a preset stability parameter threshold, and calculating the target model stability parameter corresponding to each second training data according to the training data proportion; A second performance evaluation parameter of the stand-by key material prediction model on the second material training data set is calculated according to the target model stability parameter corresponding to each of the second training data.

8. The method according to claim 1, characterized in that: Selecting a target key material prediction model according to the second performance evaluation parameter of the key material prediction model to be used includes: The standby key material prediction models are sorted according to the second performance evaluation parameter, and a model with the largest parameter value of the second performance evaluation parameter and a parameter value greater than a preset abnormal performance threshold is selected as the target key material prediction model.

9. A device for lightweighting key materials of hydrogen fuel cell BOP for aircraft, characterized in that: The device comprises: A first data acquisition module is used to acquire a first material training data set of a hydrogen fuel cell BOP, wherein the first material training data set includes training data of a plurality of hydrogen fuel cell BOPs, wherein the training data includes material property characteristics, structural design characteristics, and material performance characteristics of key materials of the hydrogen fuel cell BOP, wherein the material performance characteristics include performance data of the influence of the key materials on the hydrogen fuel cell BOP obtained through experiments; A first training module is used to obtain multiple key material prediction models, train each of the key material prediction models according to the first material training data set to predict the performance of the hydrogen fuel cell BOP, and obtain the predicted performance information of the hydrogen fuel cell BOP output by each of the key material prediction models; A first performance parameter module is used to calculate a first performance evaluation parameter of a plurality of the key material prediction models according to the predicted performance information of the hydrogen fuel cell BOP output by each of the key material prediction models and the material performance characteristics; and to determine a stand-by key material prediction model from a plurality of key material prediction models according to the first performance evaluation parameter of each of the key material prediction models, wherein the first performance evaluation parameter is used to measure the distance quantification between key material training data; A second data acquisition module is used to acquire a second material training data set of the hydrogen fuel cell BOP, where the second material training data set is a training data set of key materials of the hydrogen fuel cell BOP in different test environments; a second training module, used for training the key material prediction model to be used according to the second material training data set to calculate a second performance evaluation parameter of the key material prediction model to be used on the second material training data set, wherein the second performance evaluation parameter is used to measure the stability of the prediction result of the key material prediction model to be used; A model selection module, used for selecting a target key material prediction model according to a second performance evaluation parameter of the key material prediction model to be used; A lightweight module is used to obtain lightweight design data of key materials of a hydrogen fuel cell BOP, wherein the lightweight design data includes material property characteristics and structural design characteristics of the key materials after lightweighting; actual performance data of the hydrogen fuel cell BOP is predicted based on the target key material prediction model and the lightweight design data, and the actual performance data is compared with performance requirement data of the hydrogen fuel cell BOP for aircraft, and verification results of the lightweight design of the key materials of the hydrogen fuel cell BOP are output based on the comparison results.

10. A computer device, characterized in that: The computer device includes a processor and a memory, wherein the memory is used to store at least one computer program, and the at least one computer program is loaded by the processor and executes the key material lightweighting method described in any one of claims 1 to 8.

Images