Digital-real fusion test method for aerospace equipment under extreme working condition

Abstract

The invention discloses a number-real fusion test method for aerospace equipment under extreme working conditions, and belongs to the field of electronic engineering and computer science. Comprising the steps of designing an aerospace equipment digital modeling module, extracting actual features of aerospace equipment, constructing a digital model by using a digital modeling method, and completing iteration and verification of the digital model of the aerospace equipment based on related features and historical data of the aerospace equipment; a digital-real fusion test module under the extreme working condition is designed, dynamic generalization is carried out in an existing digital model of aerospace equipment, then, a dynamic model is optimized by using test data of the aerospace equipment under the extreme working condition, and finally, actual physical verification is carried out on the optimized model to form a digital model control system under the extreme working condition. A digital twinning technology is introduced, closed-loop feedback is established between the model and the equipment, and the operation condition of the equipment is adjusted. Safety of aerospace equipment can be guaranteed, the physical test cost is reduced, the test efficiency is improved, and the operation condition of the equipment is adjusted in real time.

Description

Technical Field

[0001] The present invention belongs to the fields of electronic engineering and computer science, and specifically relates to a digital-real fusion test method for aerospace equipment under extreme working conditions. Background Art

[0002] Aerospace equipment has extremely important strategic and technical significance in modern society. It not only plays a key role in the military field, but is also widely used in scientific research, civil communications, global security and other aspects. In the application process of aerospace equipment, it faces many problems such as high temperature, low temperature, high pressure, low pressure, vibration, shock wave, radiation, vacuum, etc. It is difficult to simulate extreme working conditions in the existing experimental environment, or it will consume a lot of resources and have a low return rate. At the same time, traditional test methods usually assume that the working conditions are stable, but in the actual environment, the working environment of aerospace equipment may face dynamic changes, and the results obtained by traditional methods may not be applicable. Summary of the invention

[0003] The technical problem to be solved by the present invention is to provide a method for digital-real fusion testing of aerospace equipment under extreme working conditions. The method covers the design of a digital modeling module of aerospace equipment and a digital-real fusion testing module under extreme working conditions, which can ensure the safety of aerospace equipment to a certain extent, reduce the cost of physical testing, improve the test efficiency, and adjust the operating conditions of the equipment in real time. To this end, the present invention discloses a method for digital-real fusion testing of aerospace equipment under extreme working conditions. Through the design of a digital modeling module of aerospace equipment and a digital-real fusion testing module under extreme working conditions, with the help of the digital-real fusion testing method, the safety of aerospace equipment can be guaranteed to a certain extent, the cost of physical testing can be reduced, the test efficiency can be improved, and the operating conditions of the equipment can be adjusted in real time.

[0004] The present invention solves the technical problem by adopting the following technical solution: a digital-real fusion test method for aerospace equipment under extreme working conditions, comprising:

[0005] Step (1), constructing a digital modeling module for aerospace equipment, using a digital modeling method to construct a digital model of aerospace equipment, and using relevant features and historical data of aerospace equipment to complete the iteration and verification of the digital model of aerospace equipment;

[0006] Step (2): construct a digital-physical fusion test module under extreme working conditions to dynamically generalize the constructed digital model, perform preliminary verification using historical data, and verify the feasibility of the model using actual physics.

[0007] The advantages of the present invention compared with the prior art are:

[0008] The actual characteristics of aerospace equipment are extracted by using the digital-real fusion experimental method, and the digital model is constructed by the digital modeling method. The digital model of aerospace equipment is iterated and verified based on the relevant characteristics and historical data of the aerospace equipment. On this premise, dynamic generalization is carried out in the digital model of the existing aerospace equipment, and preliminary data verification and actual physical verification are carried out on the optimization model to form a digital model control system under extreme working conditions, introduce digital twin technology, establish closed-loop feedback between the model and the equipment, and adjust the operating conditions of the equipment. To a certain extent, it can ensure the safety of aerospace equipment, reduce the cost of physical experiments, improve the efficiency of experiments, and adjust the operating conditions of equipment in real time. BRIEF DESCRIPTION OF THE DRAWINGS

[0009] Figure 1 This is a flowchart of a method for digital-real fusion testing of aerospace equipment under extreme working conditions according to the present invention. DETAILED DESCRIPTION

[0010] In order to make the purpose, technical scheme and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention. In addition, the technical features involved in each embodiment of the present invention described below can be combined with each other as long as they do not conflict with each other. To achieve the above-mentioned purpose, the present invention adopts the following technical scheme.

[0011] The present invention relates to a method for digital-physical fusion testing of aerospace equipment under extreme working conditions, including the design of a digital modeling module for aerospace equipment and a digital-physical fusion testing module under extreme working conditions, which can ensure the safety of aerospace equipment to a certain extent, reduce the cost of physical testing, improve the testing efficiency, and adjust the operating conditions of the equipment in real time.

[0012] The flowchart of the present invention is as follows Figure 1 As shown, taking the blade disk of an aerospace aircraft engine as an example, the specific implementation method is as follows:

[0013] Figure 1 The project includes the construction of digital modeling modules for aerospace equipment and digital-physical fusion test modules under extreme working conditions;

[0014] Step (1) construct the digital-real fusion module of aerospace equipment, which is specifically implemented as follows;

[0015] The actual physical characteristics of the blade disk are extracted, and the digital model of the blade disk is constructed using the digital modeling method. The blade disk module can be divided into three functional modules, including the blade module, the wheel disk module and the connection module in the blade disk. The blade module is further subdivided into two submodules, which are the blade crown and the blade airway surface. The wheel disk module has one submodule, which is the main body of the disk. The connection module includes two submodules, which are the hub and the blade root area. The physical characteristics of the material properties, thermodynamic properties and dynamic properties of each submodule are extracted. The physical characteristics of the above digital model include the material properties of the blade disk made of materials such as titanium alloy and nickel-based high-temperature alloy; the vibration characteristics of the blade disk under operating conditions; the mechanical characteristics of the blade disk that is subjected to high temperature and mechanical loads at the same time during operation, and the coupling effect of thermal stress and mechanical stress is significant. Then, the mathematical characteristics of the physical characteristics of the module are modeled, and the material strength equation, dynamic equation, vibration mode equation, stress distribution formula and control equation in the operation of the module are extracted. Finally, the physical characteristics and mathematical characteristics of each module are integrated to form an overall digital model. For digital modeling methods, we mainly adopt data-driven digital modeling methods, use data to quickly build models, and can handle nonlinear and multivariable coupling problems. We use neural network and principal component analysis methods to build digital models.

[0016] The iteration and verification of the digital model of the blade disk is completed based on the inherent relevant characteristics and historical data of the blade disk. The inherent relevant characteristics are the physical size and material properties of the blade disk, that is, the relevant characteristics of the blade disk are used as the properties of the digital model of the blade disk; the interface and data input of each equipment in the historical data are used as the input of the digital model, and the output in the historical data is compared with the output of the digital model. If the average accuracy of the state curve corresponding to the above mathematical and physical characteristics simulated by the digital model at normal temperature and pressure is within the design allowable range, the average error is less than 5%, and the peak error is less than 10%, the correctness of the digital model can be determined, otherwise, the digital model of the blade disk is corrected;

[0017] Step (2) constructs a digital-real fusion test module under extreme working conditions, which is specifically implemented as follows;

[0018] The constructed leaf disk digital model is generalized and optimized, that is, the established digital model is generalized using causal learning, transfer learning and adaptive algorithms to obtain a digital dynamic model. The specific algorithm implementation can be shown as follows.

[0019] First, the causal relationship between the model input features and output results is analyzed through causal learning methods, thereby eliminating irrelevant variables or interfering features, improving the generalization ability of the model, and constructing a causal graph:

[0020] ,

[0021] Among them, the node V represents the variable, and the edge E represents the causal relationship. The causal relationship can be expressed as temperature corresponding to thermal stress, rotation speed corresponding to centrifugal force, and pressure corresponding to mechanical stress. Then construct the causal equation and use the instrumental variable method to eliminate pseudo-correlated variables:

[0022] ,

[0023] Among them, Y is the output variable, which is the stress distribution, X is the input feature, which is temperature, pressure, and rotation speed. is the error term, is the causal effect coefficient. Only the variables with direct causal relationship with the target output are selected, and the digital model is trained through deep learning neural network to obtain a model that satisfies causality.

[0024] Secondly, we construct feature maps based on the variables selected by causal learning, and align the data distribution of the source domain and the target domain using the maximum mean difference (MMD) method. Then we train the migration model, and finally fine-tune the parameters according to the following formula:

[0025] ,

[0026] is the parameter set of the model, k is the current number of iterations, Represents the learning rate, which controls the step size of each gradient update. Represents the model parameters The calculated loss function gradient is used to update the parameters.

[0027] Finally, the migrated model is dynamically optimized based on the real-time error function and adjusted according to the following formula:

[0028] ,

[0029] is the real-time error function. In addition, for non-neural network parameters, Kalman filtering is used for optimization. First, state prediction is performed. The following is the state prediction function:

[0030] ,

[0031] According to the current status The predicted state at the next moment represents the parameter estimation at the k+1th moment. A is the state transfer matrix, which represents the evolution law of the parameter state from moment k to moment k+1. B is the control input matrix, which is used to describe the impact of external control on the state. is the control input, which represents the adjustment of the state by external effects (such as the influence of sensor data or feedback data).

[0032] Then the state is updated according to the prediction function. The state update function is as follows:

[0033] ,

[0034] The updated model parameters are corrected by combining the predicted state and measurement errors. is the Kalman gain, which weighs the impact of the predicted value and the measured value. The larger it is, the greater the impact on the measured value. is the real observation value in the target domain (such as the real-time stress value measured by the sensor). C is the measurement matrix, which is used to map the predicted state to the observation space.

[0035] Dynamically adjust the variable relationship in the causal graph according to environmental data to respond to real-time environmental changes and ensure the effectiveness of transfer learning results in the target domain. For example, when high temperature causes the elastic modulus of the material to decrease, recalculate thermal stress and mechanical stress.

[0036] A preliminary data verification is carried out on the digital dynamic model, that is, the input of the historical data based on extreme working conditions in the original blade disk is used as the input of the digital dynamic model, and the output data under extreme working conditions is compared with the output data of the digital dynamic model. If the accuracy of the output of the digital dynamic model meets the allowable range of the blade disk design, the overall average error is less than 5%, and the peak error is less than 10%, then the correctness of the digital dynamic model is confirmed. Otherwise, the digital dynamic model needs to be recalibrated.

[0037] The digital dynamic model is physically verified, that is, an intelligent detection method is introduced on the blade disk. This intelligent detection method realizes the real-time monitoring and data processing functions of the blade disk operation status in an all-round way by constructing a multimodal sensor network and edge computing method. The multimodal sensor network combines a variety of sensing technologies, uses environmental sensors to monitor external conditions, including wind speed, humidity, temperature, and pressure, and uses vibration sensors to measure the mechanical vibration characteristics of aerospace equipment and capture anomalies. These sensors are networked through the high-bandwidth, low-latency TSN time-sensitive network communication protocol to achieve data interconnection and synchronous collection. Through the edge computing method, real-time data processing is performed directly on the sensor end, and the low-latency fast Fourier transform FFT is used to perform spectrum analysis on high-frequency data such as vibration, temperature, and acceleration to eliminate noise or outliers, reduce delays and increase feedback speed. The dynamic data is input into the trained model, and the output data is compared in real time to avoid the randomness and inaccuracy of historical data. If the accuracy of the output data is within the allowable range of the aerospace equipment design, the overall average error is less than 5%, and the peak error is less than 10%, the correctness of the digital dynamic model is verified;

[0038] Build a digital model control system for blade disks, that is, build an extreme working condition system for blade disks, introduce the digital twin method, use a 3D scanner to scan the corresponding blade disk, convert the scanned point cloud modeling data format into FBX format, and transfer it to Unity3D software to build a simulation model, bind the data interface of the model with the corresponding data interface of the actual aerospace equipment, realize data transmission, including temperature and humidity in real conditions, and the timing working data of the equipment, and realize the data connection between the virtual model and the actual equipment. Bind the output data and command interface of the model with the data and command interface of the actual equipment to realize the data and command connection between the virtual model and the actual equipment, and control the working state of the equipment by controlling the output command of the model. The specific control process can be described as follows: take the digital dynamic model as the core algorithm module, establish a closed-loop feedback system between the virtual model and the actual equipment through real-time data update, and transfer the real-time data of the aerospace equipment under extreme working conditions into the dynamic digital model for prediction and analysis, and compare the degree of conformity with the timing working data of the equipment under actual conditions, so as to meet the curve average error degree of less than 5%. If the error of the predicted time series data is greater than 5%, and the error of the previous test is less than 5%, the actual time series error and the corresponding environmental data are re-input into the digital dynamic model, and the algorithm parameters are readjusted, including the learning rate and the hyperparameters related to the number of neural network layers, so that the error of the existing data is within the 5% range again. Then use the model to predict the working curve of the blade disk under the corresponding high temperature and high pressure conditions. If the working curve can ensure that the decline of the fatigue characteristic curve of the blade disk under long-term work is less than 20% of the normal curve, no data needs to be changed and the instruction is directly output. If the decline rate of the fatigue characteristic curve is greater than 20% at normal temperature and pressure, readjust the blade disk speed of the model, and perform fatigue curve prediction until it is less than 20%. The corresponding speed is output as an instruction to the corresponding actual equipment to complete the construction of the digital model control system of aerospace equipment.

[0039] In summary, the present invention discloses a digital-physical fusion test method for aerospace equipment under extreme working conditions, including the design of a digital modeling module for aerospace equipment and a digital-physical fusion test module under extreme working conditions, which can ensure the safety of aerospace equipment to a certain extent, reduce the cost of physical testing, improve test efficiency, and adjust the operating conditions of the equipment in real time.

[0040] The contents not described in detail in the specification of the present invention belong to the prior art known to the professional and technical personnel in this field.

[0041] The above is only a preferred embodiment of the present invention. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principle of the present invention. These improvements and modifications should also be regarded as the scope of protection of the present invention.

Claims

1. A digital-real fusion test method for aerospace equipment under extreme working conditions, characterized in that: include: Step (1), constructing a digital modeling module for aerospace equipment, using a digital modeling method to construct a digital model of aerospace equipment, and using relevant features and historical data of aerospace equipment to complete the iteration and verification of the digital model of aerospace equipment; Step (2): construct a digital-physical fusion test module under extreme working conditions to dynamically generalize the constructed digital model, perform preliminary verification using historical data, and verify the feasibility of the model using actual physics.

2. According to claim 1, a method for digital-real fusion testing of aerospace equipment under extreme working conditions is characterized in that: The specific implementation of the aerospace equipment digital modeling module in step (1) is as follows: Step (1.1) Extract the actual physical characteristics of the aerospace equipment and use the digital modeling method to build a digital model of the aerospace equipment; in the digital modeling, first divide the aerospace equipment into n functional modules, and each functional module is further subdivided into submodules, extract the physical characteristics of each submodule, conduct mathematical modeling for the physical characteristics of each submodule, and finally form an overall digital model by integrating the physical characteristics and mathematical characteristics of each module. For the digital modeling method, a data-driven digital modeling method is adopted, and the model is established using data to deal with nonlinear and multivariable coupling problems. The digital model is established using neural network and principal component analysis methods. Step (1.2) Iterate and verify the digital model of aerospace equipment based on the relevant characteristics and historical data of aerospace equipment, that is, take the relevant characteristics of aerospace equipment as the attributes of the digital model of aerospace equipment; use the interface and data input of each equipment in the historical data as the input of the digital model, and compare the output of each aerospace equipment in the historical data with the output of the digital model. If the average accuracy of the state curve corresponding to the mathematical and physical characteristics simulated by the digital model at normal temperature and pressure is within the design allowable range, the average error is less than m%, and the peak error is less than n%, where n>m>0, then the correctness of the digital model can be determined. Otherwise, the digital model of the aerospace equipment is corrected.

3. According to claim 2, a method for digital-real fusion testing of aerospace equipment under extreme working conditions is characterized in that: The specific implementation of the digital-real fusion test module under extreme working conditions in step (2) is as follows: Step (2.1) Generalize and optimize the digital model of aerospace equipment, that is, use causal learning, transfer learning and adaptive algorithms to generalize the established digital model to obtain a digital dynamic model. The specific algorithm is implemented as follows: First, the causal relationship between the model input features and output results is analyzed through causal learning methods, thereby eliminating irrelevant variables or interfering features, improving the generalization ability of the model, and constructing a causal graph; ， Among them, the node V represents the variable, the edge E represents the causal relationship, and then the causal equation is constructed, and the instrumental variable method is used to eliminate the pseudo-correlated variables; ， Among them, Y is the output variable, X is the input feature, is the error term, is the causal effect coefficient; only the variables with direct causal relationship with the target output are selected, and the digital model is trained through deep learning neural network to obtain a model that satisfies causality; Secondly, we construct feature maps based on the variables selected by causal learning, align the data distribution of the source domain and the target domain through the maximum mean difference method, and then train the migration model. Finally, we fine-tune the parameters according to the following formula: ， is the parameter set of the model, k is the current number of iterations, Represents the learning rate, which controls the step size of each gradient update. Represents the model parameters The calculated loss function gradient is used to update the parameters; Finally, the migrated model is dynamically optimized based on the real-time error function and adjusted according to the following formula: ， is the real-time error function; for non-neural network parameters, Kalman filtering is used for optimization. First, state prediction is performed. The following formula is the state prediction function: ， According to the current status The predicted state at the next moment represents the parameter estimation at the k+1th moment. A is the state transfer matrix, which represents the evolution law of the parameter state from moment k to moment k+1. B is the control input matrix, which is used to describe the influence of external control on the state. is the control input, which indicates the adjustment of the state by external action; Then the state is updated according to the prediction function; the state update function is shown as follows: ， The updated model parameters are corrected by combining the predicted state and measurement error. is the Kalman gain, is the true observation value in the target domain, and C is the measurement matrix, which is used to map the predicted state to the observation space; Dynamically adjust the variable relationships in the causal graph according to environmental data to cope with real-time environmental changes and ensure the effectiveness of transfer learning results in the target domain; Step (2.2) performs preliminary data verification on the digital dynamic model, that is, the input of the historical data based on extreme working conditions in the original aerospace equipment is used as the input of the digital dynamic model, and the output data under extreme working conditions is compared with the output data of the digital dynamic model. If the accuracy of the output of the digital dynamic model meets the allowable range of the aerospace equipment design, the overall average error is less than m%, and the peak error is less than n%, where n>m>0, then the correctness of the digital dynamic model is confirmed. Otherwise, the digital dynamic model is recalibrated; Step (2.3) Conduct actual physical verification of the digital dynamic model, that is, introduce intelligent detection methods on aerospace equipment. This intelligent detection method realizes all-round real-time monitoring and data processing of the operating status of aerospace equipment by constructing a multimodal sensor network and edge computing methods. The multimodal sensor network combines multiple sensing methods, and these sensors are networked through the high-bandwidth, low-latency TSN time-sensitive network communication protocol to achieve data interconnection and synchronous acquisition. Through the edge computing method, real-time data processing is performed directly on the sensor end. The low-latency fast Fourier transform FFT is used to perform spectrum analysis on high-frequency data such as vibration, temperature, and acceleration to eliminate noise or outliers, reduce delays and increase feedback speed. The dynamic data is input into the trained digital dynamic model, and the output data is compared in real time to avoid the randomness and inaccuracy of historical data. If the accuracy of the output data is within the allowable range of the aerospace equipment design, the overall average error is less than m%, and the peak error is less than n%, where n>m>0, the correctness of the digital dynamic model is verified. Step (2.4) builds a digital model system for aerospace equipment, that is, builds an extreme working condition system for aerospace equipment, introduces the digital twin method, uses a 3D scanner to scan the corresponding aerospace equipment, and uses the scanned point cloud modeling data to build a simulation model. The data interface of the simulation model is bound to the corresponding data interface of the actual aerospace equipment to realize data transmission, including temperature and humidity in real conditions, as well as the timing working data of the equipment, to realize the data connection between the virtual model and the actual equipment, and the output data and command interface of the simulation model is bound to the data and command interface of the actual equipment to realize the data and command connection between the virtual model and the actual equipment. The digital dynamic model is used as the algorithm module of the digital model system, and the working state of the equipment is controlled by controlling the output command of the simulation model. Through real-time data update, a closed-loop feedback system is established between the virtual model and the actual equipment to complete the construction of the digital model control system of aerospace equipment.

Images