Model correction method and device, electronic equipment and storage medium
By adjusting the pending parameters of the numerical model, and using wind tunnel experimental data to optimize the training set, the accuracy of the infrared interference bomb prediction model is improved, the problem of low prediction accuracy in the existing technology is solved, and more efficient state prediction is achieved.
Patent Information
- Application Number
- CN202411020195.6
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-07-29
- Publication Date
- 2024-11-22
AI Technical Summary
In the prior art, the state prediction accuracy of infrared interference bombs is not high, mainly due to the deviation of the numerical model calculation results, which leads to inaccurate prediction results of the neural network model.
By adjusting the pending parameters in the numerical model and using the initial wind tunnel experimental data, the difference between the calculation results of the numerical model and the experimental data is smaller than the difference threshold, thereby generating an optimized training set, and optimizing the pre-trained interference bomb prediction model is obtained. Closer to actual predictive models.
It improves the accuracy of infrared interference bomb state prediction, makes the prediction results closer to the actual situation, and enhances the anti-interference ability.
Smart Images

Figure CN119004959A_ABST
Abstract
Description
Technical Field
[0001] The present disclosure relates to the technical field of infrared jammer modeling, and in particular to a model correction method, device, electronic device and storage medium. Background Art
[0002] Infrared jammers are a type of military equipment that simulate or enhance the infrared characteristics of the target to mislead the enemy's infrared homing system, causing the missile to deviate from the real target, thereby protecting aircraft, ships, etc. from attacks by infrared-guided missiles.
[0003] The state prediction of infrared jammers is crucial to improving the combat effectiveness and anti-interference ability of infrared jammers. In related technologies, the state of infrared jammers is predicted through neural network models, but due to the small amount of measured data of infrared jammers, the calculation results of numerical models are usually used as numerical calculation samples, and the neural network is trained through a large number of numerical calculation samples.
[0004] However, the calculation results of the numerical model themselves are biased, resulting in low accuracy in the predictions of the neural network model. Summary of the invention
[0005] The present disclosure provides a model correction method, device, electronic device and storage medium, which can make the prediction results of the optimized jamming bomb prediction model closer to the actual situation.
[0006] The technical solution of the present disclosure is achieved as follows:
[0007] In a first aspect, the present disclosure provides a model correction method, which includes: adjusting the pending parameters in the numerical model according to the initial wind tunnel test data, so as to obtain the optimal pending parameters when the difference between the calculation result of the numerical model and the initial wind tunnel test data is less than the difference threshold; calculating a training set through the numerical model of the optimal pending parameters; optimizing and training the pre-trained jammer prediction model through the training set to obtain the target jammer prediction model.
[0008] In a second aspect, the present disclosure provides a model correction device, which includes: a model correction device, which includes: an adjustment part, a calculation part and a training part; the adjustment part is configured to adjust the pending parameters in the numerical model according to the initial wind tunnel test data, so as to obtain the optimal pending parameters when the difference between the calculation result of the numerical model and the initial wind tunnel test data is less than a difference threshold; the calculation part is configured to calculate a training set through the numerical model of the optimal pending parameters; the training part is configured to optimize the training of the pre-trained jammer prediction model through the training set to obtain the target jammer prediction model.
[0009] In a third aspect, the present disclosure provides an electronic device comprising a processor, a memory, and a program or instruction stored in the memory and executable on the processor, wherein the program or instruction, when executed by the processor, implements the steps of the model correction method described in the first aspect.
[0010] In a fourth aspect, the present disclosure provides a computer-readable storage medium storing a program or instruction, which, when executed by a processor, implements the steps of the model correction method described in the first aspect.
[0011] In a fifth aspect, the present disclosure provides a computer program product, wherein the computer program product includes a computer program or instructions, and when the computer program product runs on a processor, the processor executes the computer program or instructions to implement the steps of the model correction method described in the first aspect.
[0012] In a sixth aspect, the present disclosure provides a chip, comprising a processor and a communication interface, wherein the communication interface is coupled to the processor, and the processor is used to run a program or instruction to implement the model correction method as described in the first aspect.
[0013] The present disclosure provides a model correction method, which adjusts the undetermined parameters in the numerical model according to the initial wind tunnel test data, so as to obtain the optimal undetermined parameters when the difference between the calculation result of the numerical model and the initial wind tunnel test data is less than the difference threshold; the training set is calculated through the numerical model of the optimal undetermined parameters; the pre-trained jammer prediction model is optimized and trained through the training set to obtain the target jammer prediction model. Since the data in the training set is optimized through the wind tunnel test data, it is closer to the wind tunnel test data, and the wind tunnel test data is closer to the real state data of the infrared jammer, the state predicted by the model obtained by this training is closer to the actual state and has higher accuracy. BRIEF DESCRIPTION OF THE DRAWINGS
[0014] Figure 1 One of the flow charts of the model correction method provided by the present disclosure;
[0015] Figure 2 The second flowchart of the model correction method provided by the present disclosure;
[0016] Figure 3 The third flowchart of the model correction method provided by the present disclosure;
[0017] Figure 4 The fourth flowchart of the model correction method provided by the present disclosure;
[0018] Figure 5 A structural block diagram of a model correction device provided by the present disclosure;
[0019] Figure 6 A schematic diagram of the hardware structure of an electronic device provided by the present invention. DETAILED DESCRIPTION
[0020] The following will be combined with the drawings in the present disclosure to clearly describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are part of the embodiments of the present application, rather than all of the embodiments. All other embodiments obtained by ordinary technicians in this field based on the embodiments in the present application belong to the scope of protection of the present application.
[0021] Infrared jammers are a type of military equipment used primarily to counter enemy infrared-guided weapons, such as infrared-guided missiles. They attract the attention of infrared-guided weapons by simulating or enhancing the infrared radiation characteristics of the target (by burning to produce high-temperature flames to simulate the infrared radiation characteristics of the target), thereby protecting the real target from attack. The research focus of infrared jammers is to improve their aerodynamic characteristics and infrared radiation characteristics to adapt to the complex and changing battlefield environment. The characteristics of infrared jammers, such as infrared radiation intensity, dispersion area, and time characteristics, need to be analyzed in detail to ensure that they can effectively counter infrared-guided weapons.
[0022] Simulating infrared jammers to predict their state information involves aerodynamics, flight mechanics, thermodynamics, and materials science. Aerodynamic characteristics simulation: The movement of infrared jammers in the air is affected by factors such as air resistance, Mach number, and gravity. By establishing an aerodynamic model, the flight trajectory and speed change of the jammer can be simulated, which requires consideration of parameters including angle of attack, missile shape, flight speed, and altitude. Radiation characteristics simulation: The radiation characteristics of infrared jammers are crucial to their jamming effect. Factors such as fuel combustion characteristics, radiation spectrum, and interaction with the atmosphere need to be considered during simulation. Dynamic characteristics simulation: The movement of infrared jammers is nonlinear, and its dynamic characteristics simulation requires solving a set of equations through numerical methods to predict its acceleration, attitude, and attitude change rate. In addition, the simulation of infrared jammers also needs to consider the coupling effects of multiple physical fields such as aerodynamic heating, structural deformation, and fuel consumption, which increases the complexity of the simulation.
[0023] In the related technology, the state information of infrared jammers is usually calculated through numerical simulation. However, since the simulation of infrared jammers involves the coupling of multiple physical fields, the computational complexity of solving the equations composed of various field equations in the numerical simulation is high, resulting in low computational efficiency and high computational cost. In addition, the computational theory for solving the equations is not yet mature, and the equations are prone to be ill-conditioned and often have no solution under many conditions.
[0024] In order to solve the above problems, in the related technology, the numerical calculation samples obtained by numerical simulation calculations are used as training sets, and the neural network model is trained by the training set composed of a large number of numerical calculation samples, so that the neural network model can learn the time step characteristics of the infrared jamming bomb under the coupling of multiple physical fields.
[0025] However, due to the deviation of the calculation samples obtained by numerical simulation, the accuracy of the prediction results of the neural network model trained by this method is not good. The experimental data obtained by wind tunnel experiments are closer to the actual situation and more reliable than the calculation results obtained by numerical simulation. However, the cost of wind tunnel experiments is high, and the number of experiments and data sampling points that can be performed are limited.
[0026] Therefore, the present disclosure aims to provide a model correction method for correcting the training set of a neural network model using wind tunnel test data, so that the state information of the infrared jammer predicted by the trained model is more consistent with the actual situation and the prediction result is more accurate.
[0027] like Figure 1 As shown, a model correction method is provided in the present disclosure, and the method may include the following steps S101 to S103.
[0028] In step S101, the undetermined parameters in the numerical model are adjusted according to the initial wind tunnel test data, so as to obtain the optimal undetermined parameters when the difference between the calculation result of the numerical model and the initial wind tunnel test data is less than a difference threshold.
[0029] The wind tunnel test data is the state information of the infrared jammer at the measurement point measured under certain wind tunnel test conditions. The numerical model is used to calculate the state information of the infrared jammer in the calculation domain according to relevant equations (including fluid mechanics equations, combustion equations, and heat transfer equations). The calculation domain includes the measurement points of the wind tunnel test, and the calculation results are the results of the same measurement points in the calculation domain as those in the wind tunnel test calculated by the numerical model.
[0030] Optionally, the numerical model includes a turbulence model, a combustion model, and a heat transfer model. The turbulence model corresponds to the fluid mechanics equation, the combustion model corresponds to the combustion model, and the heat transfer model corresponds to the heat transfer equation. When each model calculates the state information of the jammer in the calculation domain, some parameters need to be set based on experience, that is, undetermined parameters, such as: turbulent viscosity coefficient and turbulent kinetic energy dissipation rate in the turbulence model, chemical reaction rate constant, Weber combustion rule, etc. in the combustion model, thermal conductivity, specific heat capacity, convection heat transfer coefficient and other parameters in the heat transfer model.
[0031] According to the initial wind tunnel test data, the pending parameters in the numerical model are adjusted. Specifically, a preset adjustment step size can be set for each parameter, and each pending parameter is adjusted step by step. After the adjustment, the calculation result is obtained by calculating again through the numerical model, and the difference between the calculation result and the wind tunnel test data is detected; if the difference is less than the difference threshold, the various pending coefficients are adjusted again and calculated again through the numerical model until the difference between the calculation result and the wind tunnel test data is less than the difference threshold.
[0032] The difference between the calculated result and the wind tunnel test data can be measured by using mean square error, cross entropy loss, mean absolute error, Euclidean distance, Manhattan distance, etc. The specific method is determined according to actual needs and is not specifically limited in the present disclosure.
[0033] In the present disclosure, the strategy for adjusting the pending parameters is not specifically limited, such as: starting with coarse adjustment, adjusting some or all of the pending parameters by multiple steps each time, and when the difference is less than the preset difference but greater than or equal to the difference threshold, starting fine adjustment, adjusting some or all of the pending parameters by one step each time, and repeating the adjustment until the difference is less than the difference threshold.
[0034] In step S102, a training set is calculated using a numerical model of the optimal parameters to be determined.
[0035] Optionally, the training set may also include wind tunnel test data. That is, the training set is composed of wind tunnel test data and numerical calculation samples. Specifically, the proportion of wind tunnel test data and numerical calculation samples is determined according to actual needs. If the wind tunnel test data is insufficient, the wind tunnel test data can be copied and used in multiple copies.
[0036] In step S103, the pre-trained jammer prediction model is optimized and trained through the training set to obtain a target jammer prediction model.
[0037] Specifically, after setting the undetermined parameters in the numerical model to the optimal undetermined parameters, multiple numerical calculation samples are calculated, and the corrected multiple numerical calculation samples are used as training sets to optimize the jammer prediction model. Since the data in the training set is optimized by wind tunnel test data, it is closer to wind tunnel test data, and wind tunnel test data is closer to the real state data of infrared jammers, the state predicted by the model obtained by training is closer to the actual state and has higher accuracy.
[0038] Since the jammer prediction model has been trained with a large number of samples, in order to reduce the amount of calculation required for each iteration, thereby speeding up the training speed. In some embodiments of the present disclosure, the jammer prediction model includes: an encoding layer and a decoding layer; Figure 1 ,like Figure 2 As shown, the method also includes the following step S104.
[0039] In step S104, in the process of optimizing the training jammer prediction model through the training set, the network parameters of the encoding layer are fixed and the network parameters of the decoding layer are opened, so that only the network parameters of the decoding layer are updated during the training process.
[0040] Specifically, the network parameters of the encoding layer of the jammer prediction model are fixed, decoding is opened for training, and the parameters of the encoding layer are set to non-trainable. Configure the training process, including loss function, optimizer, etc. Since only the parameters of the decoding layer will be updated, the optimizer needs to point to the parameters of the decoding layer. Train the model, perform forward propagation in the training loop, and predict the state information of the jammer in the next time step; use the loss function to evaluate the difference between the predicted output and the actual value; perform the backpropagation algorithm only on the parameters of the decoding layer to calculate the gradient; use the optimizer to update the parameters of the decoding layer and repeat the training.
[0041] In this way, the encoding layer, as a feature extractor, has been trained on big data, and the decoding layer is fine-tuned for specific tasks. Since the parameters of the encoding layer do not need to be updated, the amount of calculation required for each iteration can be reduced, thereby speeding up the training. Fixed encoding layers can also reduce the complexity of the model and reduce the risk of overfitting.
[0042] In some embodiments of the present disclosure, in order to quickly obtain the optimal parameters to be determined, the numerical model can be optimized by an optimization method. Specifically, Figure 1 ,like Figure 3 As shown, in the above step S101, the undetermined parameters in the numerical model are adjusted according to the initial wind tunnel test data, so as to obtain the optimal undetermined parameters when the difference between the calculation result of the numerical model and the initial wind tunnel test data is less than the difference threshold, which can be achieved by the following step S101a.
[0043] In step S101a, based on the initial wind tunnel test data, the numerical model is optimized by an optimization method to obtain the optimal undetermined parameters of the numerical model.
[0044] Among them, the objective function of the optimization method is that the difference between the initial wind tunnel test data and the calculation results is less than the difference threshold, and the constraint condition is: the relevant equation corresponding to the numerical model.
[0045] Specifically, based on the initial wind tunnel test data, the numerical model is optimized by an optimization method, such as gradient descent method, particle swarm optimization, genetic algorithm or Bayesian optimization. The goal of the optimization process is to minimize the difference between the initial wind tunnel test data and the calculation results, so that it is less than a preset difference threshold, which can be measured in a variety of ways, such as mean square error, mean absolute error or Euclidean distance. The numerical model must follow the fluid mechanics equation, combustion equation and heat transfer equation. Select a suitable optimization algorithm to search for the optimal parameters. Before starting the optimization, it is necessary to provide an initial undetermined parameter for the model parameters. The initial undetermined parameter can be obtained based on experience, previous research or through sensitivity analysis. The optimization algorithm is executed to minimize the objective function by iteratively adjusting the undetermined parameters while satisfying all constraints. When the improvement of the objective function is less than a certain threshold, reaches a predetermined number of iterations or calculation time limit, the optimization process terminates; the optimal undetermined parameters of the optimized numerical model are verified to ensure that they not only meet the optimization criteria numerically, but are also physically reasonable.
[0046] Preferably, the optimization method is an adjoint optimization method. The adjoint method is particularly suitable for highly nonlinear and multi-scale characteristics. The adjoint optimization method can very efficiently calculate the gradient of the objective function to the undetermined parameters of the numerical model. This is because the solution of the adjoint equation (adjoint vector) can be directly used to calculate the gradient, avoiding the numerical differentiation of each parameter separately. When calibrating the undetermined parameters of the numerical model based on experimental data, the adjoint method can quickly identify which parameters have the greatest impact on the objective function, thereby reducing the necessary number of experiments and experimental costs.
[0047] In order to enable the final trimmed jammer prediction model to perform accurate prediction, in some embodiments of the present disclosure, Figure 1 ,like Figure 4 As shown, the above step S103 optimizes and trains the pre-trained jammer prediction model through the training set to obtain the target jammer prediction model. The method also includes the following steps S105 to S108.
[0048] In step S105, the prediction result of the target jammer prediction model is tested by a test set to see whether it meets the quantitative evaluation index.
[0049] The test set may be composed of wind tunnel test data and calculation results of representative numerical models. The quantitative evaluation index may be that the number of iterations is greater than a number threshold, or the prediction accuracy is greater than an accuracy threshold.
[0050] Optionally, the quantitative evaluation indicators include: flow field reconstruction error, physical control equation reconstruction error, strong discontinuity region capture error;
[0051] Among them, the flow field reconstruction error indicates that the difference between the predicted result and the expected result is less than the error threshold; the physical control equation reconstruction error indicates that the degree to which the predicted result satisfies the physical control equation is greater than the degree threshold; the strong discontinuity area capture error indicates that the accuracy of the predicted result in the area where the flow field changes violently is greater than the accuracy threshold.
[0052] Specifically, the flow field reconstruction error includes: pressure field, velocity field, density field, temperature field reconstruction error; the physical control equation reconstruction error includes: the mass equation, momentum equation and energy equation formed after the fluid particles are discretized in the Lagrangian system, as well as the reconstruction error of the equation corresponding to the combustion and heat transfer phenomenon. The flow field reconstruction error can be obtained by weighted summing the errors of each field; the specific calculation method of the physical control equation reconstruction error is to substitute the physical quantity corresponding to each fluid particle at the relevant time step into the equation, and then statistically sum the residual terms formed. The statistical summation can be calculated by the mean square error or root mean square error; the capture error of the strong discontinuity area is obtained by using multi-level wavelet transform. This method decomposes and learns the sudden high-frequency signal in the flow field, and then improves the capture accuracy of the area with drastic changes in the flow field (such as shock waves).
[0053] If the prediction result does not meet the quantitative evaluation index, step S105 is executed.
[0054] In step S106, based on the prediction results, the sampling locations of new sample points of the wind tunnel experiment are determined by using the Kriging model.
[0055] Due to the high cost of wind tunnel experiments, if all planned experimental points are tested at one time, the data from the experimental points may not be able to effectively guide the training of the model. Therefore, active design is achieved through the Kriging model. According to the difference between the predicted results and the expected results, some experimental points that can minimize the uncertainty of the entire spatial prediction and those that can provide more new information are selected. This allows the model to be tuned with as few wind tunnel experimental points as possible.
[0056] In step S107, new wind tunnel test data is acquired according to the sampling position of the new sample point.
[0057] In step S108, the initial wind tunnel test data is updated according to the new wind tunnel test data.
[0058] Specifically, the initial wind tunnel test data is updated to the new wind tunnel test data, and steps S101 to S108 are repeatedly performed until a corrected jammer prediction model is obtained whose prediction results meet the quantitative evaluation index, so that the prediction accuracy of the corrected jammer prediction model finally obtained meets the standard.
[0059] Figure 5FIG. 5 is a structural block diagram of a model correction device 500 shown in the present disclosure, such as Figure 5 As shown, it includes: an adjustment part 501, a calculation part 502 and a training part 503;
[0060] The adjustment part 501 is configured to adjust the undetermined parameters in the numerical model according to the initial wind tunnel test data, so as to obtain the optimal undetermined parameters when the difference between the calculation result of the numerical model and the initial wind tunnel test data is less than a difference threshold;
[0061] The calculation part 502 is configured to calculate the training set through the numerical model of the optimal undetermined parameter;
[0062] The training part 503 is configured to optimize the pre-trained jammer prediction model through the training set to obtain the target jammer prediction model.
[0063] In some embodiments of the present disclosure, the jammer prediction model includes: a coding layer and a decoding layer; the training part 503 is also configured to fix the network parameters of the coding layer and open the network parameters of the decoding layer in the process of optimizing the training of the jammer prediction model through the training set, so as to only update the network parameters of the decoding layer during the training process.
[0064] In some embodiments of the present disclosure, the adjustment part 501 is specifically configured to optimize the numerical model through an optimization method based on the initial wind tunnel test data to obtain the optimal unknown parameters of the numerical model. The objective function of the optimization method is that the difference between the initial wind tunnel test data and the calculation results is less than the difference threshold, and the constraint condition is: the relevant equation corresponding to the numerical model.
[0065] In some embodiments of the present disclosure, the optimization method is a companion optimization method.
[0066] In some embodiments of the present disclosure, the device also includes: a testing part, a determining part, an acquiring part and an executing part; the testing part is configured to optimize the training of the pre-trained jammer prediction model through the training set to obtain the target jammer prediction model, and then the testing part is configured to test whether the prediction result of the target jammer prediction model meets the quantitative evaluation index through the test set; the determining part is configured to determine the new sample point sampling position of the wind tunnel experiment through the Kriging model according to the prediction result when the prediction result does not meet the quantitative evaluation index; the acquiring part is configured to acquire new wind tunnel experiment data according to the new sample point sampling position; the updating part updates the initial wind tunnel experiment data according to the new wind tunnel experiment data until a revised jammer prediction model is obtained whose prediction result meets the quantitative evaluation index.
[0067] In some embodiments of the present disclosure, the quantitative evaluation indicators include: flow field reconstruction error, physical control equation reconstruction error, strong discontinuity region capture error;
[0068] Among them, the flow field reconstruction error indicates that the difference between the predicted result and the expected result is less than the error threshold; the physical control equation reconstruction error indicates that the degree to which the predicted result satisfies the physical control equation is greater than the degree threshold; the strong discontinuity area capture error indicates that the accuracy of the predicted result in the area where the flow field changes violently is greater than the accuracy threshold.
[0069] It should be noted that the above-mentioned model correction device can be the electronic device in the above-mentioned method embodiment of the present application, or it can be a functional module and / or functional entity in the electronic device that can realize the function of the device embodiment, and the embodiment of the present application is not limited.
[0070] In the embodiments of the present application, each module can implement the model correction method provided in the above method embodiments and can achieve the same technical effect. To avoid repetition, it will not be repeated here.
[0071] Please refer to Figure 6 , which shows a structural block diagram of an electronic device provided by an exemplary embodiment of the present disclosure. In some examples, the electronic device may be at least one of a smart phone, a smart watch, a desktop computer, a laptop, a virtual reality terminal, an augmented reality terminal, a wireless terminal, and a laptop computer. The electronic device has a communication function and can access a wired network or a wireless network. The electronic device may generally refer to one of a plurality of terminals, and those skilled in the art may know that the number of the above terminals may be more or less. It can be understood that the electronic device undertakes the calculation and processing work of the technical solution of the present disclosure, and the present disclosure does not limit this.
[0072] like Figure 6 As shown, the electronic device in the present disclosure may include one or more of the following components: a processor 610 and a memory 620 .
[0073] Optionally, the processor 610 uses various interfaces and lines to connect various parts of the entire electronic device, and executes various functions of the electronic device and processes data by running or executing instructions, programs, code sets or instruction sets stored in the memory 620, and calling data stored in the memory 620. Optionally, the processor 610 can be implemented in at least one hardware form of digital signal processing (DSP), field-programmable gate array (FPGA), and programmable logic array (PLA). The processor 610 can integrate one or a combination of a central processing unit (CPU), a graphics processing unit (GPU), a neural network processor (NPU), and a baseband chip. Among them, the CPU mainly processes the operating system, user interface, and application programs; the GPU is responsible for rendering and drawing the content that needs to be displayed on the touch display; the NPU is used to implement artificial intelligence (AI) functions; and the baseband chip is used to process wireless communications. It is understandable that the above baseband chip may not be integrated into the processor 610, but may be implemented by a separate chip.
[0074] The memory 620 may include a random access memory (RAM) or a read-only memory (ROM). Optionally, the memory 620 includes a non-transitory computer-readable storage medium. The memory 620 may be used to store instructions, programs, codes, code sets, or instruction sets. The memory 620 may include a program storage area and a data storage area, wherein the program storage area may store instructions for implementing an operating system, instructions for at least one function (such as a touch function, a sound playback function, an image playback function, etc.), instructions for implementing the above various method embodiments, etc.; the data storage area may store data created according to the use of the electronic device, etc.
[0075] In addition, those skilled in the art will understand that the structure of the electronic device shown in the above drawings does not constitute a limitation on the electronic device, and the electronic device may include more or fewer components than shown, or combine certain components, or arrange the components differently. For example, the electronic device also includes a display screen, a camera assembly, a microphone, a speaker, a radio frequency circuit, an input unit, a sensor (such as an acceleration sensor, an angular velocity sensor, a light sensor, etc.), an audio circuit, a WiFi module, a power supply, a Bluetooth module and other components, which will not be described in detail here.
[0076] The present disclosure also provides a computer-readable storage medium storing at least one instruction, wherein the at least one instruction is used to be executed by a processor to implement the model correction method described in the above embodiments.
[0077] The present disclosure also provides a computer program product, which includes computer instructions stored in a computer-readable storage medium; a processor of an electronic device reads the computer instructions from the computer-readable storage medium, and the processor executes the computer instructions, so that the electronic device executes to implement the model correction method described in the above-mentioned embodiments.
[0078] An embodiment of the present application further provides a chip, which includes a processor and a communication interface, wherein the communication interface is coupled to the processor, and the processor is used to run programs or instructions to implement the various processes of the above-mentioned model correction method embodiment, and can achieve the same technical effect. To avoid repetition, it will not be repeated here.
[0079] It should be understood that the chip mentioned in the embodiments of the present application can also be called a system-level chip, a system chip, a chip system or a system-on-chip chip, etc.
[0080] In the several embodiments provided in the present disclosure, it should be understood that the disclosed systems, devices, servers and methods can be implemented in other ways. For example, the device embodiments described above are only schematic. For example, the division of the units is only a logical function division. There may be other division methods in actual implementation, such as 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.
[0081] 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.
[0082] In addition, each functional unit in each embodiment of the present disclosure 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.
[0083] 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 is essentially 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, and the computer software product is stored in a storage medium, including a number of instructions to enable 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 method described in each embodiment of the present application. The aforementioned storage medium includes: various media that can store program codes, such as a USB flash drive, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a disk or an optical disk.
[0084] Those skilled in the art should be aware that in one or more of the above examples, the functions described in the present disclosure can be implemented using hardware, software, firmware, or any combination thereof. When implemented using software, these functions can be stored in a computer-readable medium or transmitted as one or more instructions or codes on a computer-readable medium. Computer-readable media include computer storage media and communication media, wherein communication media include any media that facilitates the transmission of computer programs from one place to another. Storage media can be any available media that can be accessed by a general or special-purpose computer.
[0085] It should be noted that the technical solutions described in the present disclosure can be combined arbitrarily without conflict.
[0086] The above description is only a specific implementation mode of the present invention, but the protection scope of the present invention is not limited thereto. Any technician familiar with the technical field can easily think of changes or substitutions within the technical scope disclosed by the present invention, which should be covered by the protection scope of the present invention.
Claims
1. A model correction method, characterized in that: The method comprises: According to the initial wind tunnel test data, adjusting the undetermined parameters in the numerical model to obtain the optimal undetermined parameters when the difference between the calculation result of the numerical model and the initial wind tunnel test data is less than a difference threshold; By using the numerical model of the optimal undetermined parameter, a training set is calculated; The pre-trained jammer prediction model is optimized and trained through the training set to obtain a target jammer prediction model.
2. The method according to claim 1, characterized in that The jammer prediction model includes: a coding layer and a decoding layer; the method also includes: In the process of optimizing and training the jammer prediction model through the training set, the network parameters of the encoding layer are fixed and the network parameters of the decoding layer are opened, so that only the network parameters of the decoding layer are updated during the training process.
3. The method according to claim 1, characterized in that The step of adjusting the undetermined parameters in the numerical model according to the initial wind tunnel test data to obtain the optimal undetermined parameters when the difference between the calculation result of the numerical model and the initial wind tunnel test data is less than a difference threshold comprises: Based on the initial wind tunnel test data, the numerical model is optimized by an optimization method to obtain the optimal unknown parameters of the numerical model. The objective function of the optimization method is that the difference between the initial wind tunnel test data and the calculation result is less than a difference threshold, and the constraint condition is: the relevant equation corresponding to the numerical model.
4. The method according to claim 3, characterized in that The optimization method is a companion optimization method.
5. The method according to claim 1, characterized in that After optimizing and training the pre-trained jammer prediction model through the training set to obtain the target jammer prediction model, the method further includes: Testing whether the prediction results of the target jammer prediction model meet the quantitative evaluation index through the test set; In the case where the prediction result does not meet the quantitative evaluation index, determining the sampling location of new sample points for the wind tunnel experiment through the Kriging model according to the prediction result; Acquire new wind tunnel test data according to the sampling position of the new sample point; The initial wind tunnel test data is updated according to the new wind tunnel test data until a revised jammer prediction model is obtained whose prediction results meet the quantitative evaluation index.
6. The method according to claim 5, characterized in that The quantitative evaluation indicators include: flow field reconstruction error, physical control equation reconstruction error, and strong discontinuity region capture error; Among them, the flow field reconstruction error indicates that the difference between the predicted result and the expected result is less than the error threshold; the physical control equation reconstruction error indicates that the degree to which the predicted result satisfies the physical control equation is greater than the degree threshold; the strong discontinuity area capture error indicates that the accuracy of the predicted result in the area where the flow field changes violently is greater than the accuracy threshold.
7. A model correction device, characterized in that: The device comprises: an adjustment part, a calculation part and a training part; The adjustment part is configured to adjust the undetermined parameters in the numerical model according to the initial wind tunnel test data, so as to obtain the optimal undetermined parameters when the difference between the calculation result of the numerical model and the initial wind tunnel test data is less than a difference threshold; The calculation part is configured to calculate and obtain a training set through a numerical model of the optimal undetermined parameter; The training part is configured to optimize the pre-trained jammer prediction model through the training set to obtain a target jammer prediction model.
8. The device according to claim 7, characterized in that The jammer prediction model includes: a coding layer and a decoding layer; The training part is specifically configured to open the network parameters of the decoding layer during the process of optimizing and training the jammer prediction model through the training set, so that only the network parameters of the decoding layer are updated during the training process.
9. An electronic device, characterized in that: The method comprises a processor, a memory and a program or instruction stored in the memory and executable on the processor, wherein the program or instruction, when executed by the processor, implements the steps of the model correction method according to any one of claims 1 to 6.
10. A computer-readable storage medium, characterized in that: The readable storage medium stores a program or an instruction, and when the program or the instruction is executed by the processor, the steps of the model correction method according to any one of claims 1 to 6 are implemented.
Citation Information
Cited By
Method and equipment for generating decoy jamming bomb discrimination model
CN120852307A
