CN111695230B - Neural network space mapping multi-physical modeling method for microwave passive device - Google Patents
Neural network space mapping multi-physical modeling method for microwave passive device Download PDFInfo
- Publication number
- CN111695230B CN111695230B CN201911402424.XA CN201911402424A CN111695230B CN 111695230 B CN111695230 B CN 111695230B CN 201911402424 A CN201911402424 A CN 201911402424A CN 111695230 B CN111695230 B CN 111695230B
- Authority
- CN
- China
- Prior art keywords
- mapping
- model
- physical
- neural network
- output
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
- 238000013507 mapping Methods 0.000 title claims abstract description 91
- 238000013528 artificial neural network Methods 0.000 title claims abstract description 42
- 238000000034 method Methods 0.000 title claims abstract description 22
- 238000012549 training Methods 0.000 claims description 41
- 238000012360 testing method Methods 0.000 claims description 15
- 238000004088 simulation Methods 0.000 claims description 4
- 230000004044 response Effects 0.000 claims description 3
- 101100269885 Arabidopsis thaliana ANN1 gene Proteins 0.000 claims description 2
- 101100269886 Arabidopsis thaliana ANN2 gene Proteins 0.000 claims description 2
- 238000013461 design Methods 0.000 abstract description 3
- 210000002569 neuron Anatomy 0.000 description 2
- 238000011160 research Methods 0.000 description 2
- 230000008859 change Effects 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 230000006866 deterioration Effects 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 238000010586 diagram Methods 0.000 description 1
- 230000005672 electromagnetic field Effects 0.000 description 1
- 230000005686 electrostatic field Effects 0.000 description 1
- 230000003993 interaction Effects 0.000 description 1
- 239000000463 material Substances 0.000 description 1
- 238000005259 measurement Methods 0.000 description 1
- 238000005457 optimization Methods 0.000 description 1
- 230000000704 physical effect Effects 0.000 description 1
- 230000008569 process Effects 0.000 description 1
- 230000009467 reduction Effects 0.000 description 1
Images
Landscapes
- Measurement Of Resistance Or Impedance (AREA)
Abstract
The invention belongs to the field of modeling of microwave circuits and devices, and provides a neural network space mapping multi-physical modeling method for microwave passive devices. The method adds a mapping network at the input end and the output end of an empirical model, the multi-physical domain input variable of the passive device is mapped to an electromagnetic domain coarse model after being input into a neural network, and the output of the electromagnetic domain coarse model is matched with device modeling data after being further optimized through output mapping. The invention improves modeling precision, reduces modeling data volume, shortens design period, and provides efficient and accurate prediction for multiple physical characteristics of devices.
Description
Technical Field
The invention belongs to the field of modeling of microwave circuits and devices, and relates to a neural network space mapping multi-physical modeling method for microwave passive devices.
Background
The performance requirements of the passive devices in the current market are higher and higher, the application occasions are more and more complex, and the performance of the passive devices is influenced by a plurality of physical parameters such as temperature, humidity, stress change and the like besides the structure and material parameters of the passive devices. These multiple physical parameter changes will cause the electromagnetic parameter drift of the device, and deviation occurs, resulting in the reduction of the working efficiency of the device, the deterioration of the working stability, and the influence on the performance of the device. Interactions between multiple physical domains are critical to accurate system analysis. The design of the passive device relates to electromagnetic analysis and the influence of multiple physical fields such as thermodynamics, electrostatic fields and the like, multiple physical parameters are introduced when the performance of the device is modeled, a multiple physical model of the device is built, and the multiple physical properties of the microwave device can be accurately described.
The neural network space mapping technology is one of knowledge-based neural networks, and can improve the accuracy of an existing empirical analysis model. The neural network space mapping model has the advantages of high operation speed and good compatibility of the coarse model, and improves the accuracy of the coarse model. The research results obtained by the multi-physical modeling method for the microwave passive device at present have application limitation on the condition that the multi-physical characteristics of the device are complex or the coarse model precision is insufficient. These results cannot be directly applied to passive device modeling involving multiple physical characteristics. Although the existing device modeling technology is mature, more modeling methods cannot meet the requirements of high precision and high speed at the same time, and a designed model and an actual measured value still have a certain gap, so that the research on the passive device multi-physical modeling method still remains to be researched.
Therefore, the invention aims to provide a neural network space mapping multi-physical modeling method for microwave passive devices, which improves modeling accuracy, reduces modeling data volume, shortens design period and provides efficient and accurate prediction for the multi-physical characteristics of the devices.
Disclosure of Invention
The invention aims to overcome the defects of the prior art and provides a neural network space mapping multi-physical modeling method for a microwave passive device. The method adds a mapping network at the input end and the output end of the coarse model, the multi-physical domain input variable of the passive device acts on the electromagnetic domain coarse model after input mapping, and the output of the electromagnetic domain coarse model is matched with the multi-physical output of the device after further optimization of output mapping.
A neural network space mapping multi-physical modeling method for microwave passive devices comprises the following steps:
step 1: selecting and defining geometric parameters x of multiple physical models g Multiple physical parameters x of multiple physical models m And the data range of the frequency f of the multiple physical models, and performing multiple physical simulations to generate training and testing samples of the multiple physical models;
step 2: determining and defining the geometric parameter x of the electromagnetic domain coarse model according to step 1 gc And electromagnetic domain coarse model frequency f c And performing electromagnetic simulation to generate training and test samples of the coarse model. In order to ensure the accuracy of the whole multi-physical model, the geometric parameter x of the electromagnetic domain coarse model gc The range should be slightly larger than the geometric parameter x in the whole multi-physical model g ；
Step 3: establishing a coarse model, training the coarse model by using the training data obtained in the step 2 by using a neural network technology, and testing the coarse model by using the test data obtained in the step 2 until the training error and the test precision of the coarse model are reachedMeets the requirement, and the weight value w in the coarse model * Fixing;
step 4: initializing the input mapping network and the output mapping network, and optimizing the internal weight variable w of the input mapping 1 To the point of
Optimizing the internal weight variable w of the output map 2 To->
Realization of x gc ＝x g 、f c =f and y=y c The accuracy of the whole model is not reduced after the mapping network is added;
step 5: building a preliminary neural network space mapping multi-physical model by using the rough model obtained in the step 2 and the input mapping unit mapping obtained in the step 4, training the preliminary neural network space mapping model by using the multi-physical training data obtained in the step 1, and optimizing weight parameters of the input mapping module
To->
Mapping between the multi-physical domain input signals and the electromagnetic domain input signals is realized;
step 6: adding the output mapping unit mapping obtained in the step 4 on the preliminary neural network space mapping multi-physical model built in the step 5, and fixing the mapping unit mapping
Training the unit mapping of the output mapping module by using the multi-physical training data obtained in the step 1, and optimizing the weight parameter of the output mapping module>
To->
Let the output y of the coarse model c Consistent with the response y of the modeled device;
step 7: further training the whole multi-physical model by using the multi-physical training data obtained in the step 1, and simultaneously adjusting the weight value of the input neural network
To->
Adjusting the weight value of the output neural network +.>
To->
Until the built multi-physical model can accurately represent the characteristics of the passive device.
In the step 3 of the invention, the trained coarse model formula is as follows
y c ＝g ANN (x gc ，f c ，w * ) (1)
Wherein g ANN (. Cndot.) is a coarse model neural network mapping formula, w * A vector representing all weight parameters contained in the coarse model network.
In step 5 of the present invention, the trained input mapping network formula is
Wherein f ANN1 (. Cndot.) is the input neural network mapping formula, x g And x m Is the input of the input map, x gc And f c Is the output of the input map and,
a vector representing all weight parameters contained in the mapping network.
In step 6 of the present invention, the trained output mapping network formula is
Wherein f ANN2 (. Cndot.) is the output mapped neural network formula, representing y and y c The relationship between the two.
Representing all internal weight variables of the mapped neural network.
The neural network space mapping multi-physical modeling method provided by the invention does not need internal structure information of the microwave passive device, and has the advantages of high model precision, strong robustness, less required modeling data and short modeling time. When the multi-physical characteristics of the microwave device are complex or the precision of the coarse model is low, the two mapping networks are mutually adjusted, and the model can accurately reflect the multi-physical characteristics of the device.
Drawings
FIG. 1 is a block diagram of the structure of the present invention;
FIG. 2 is a flow chart of multi-physical modeling of microwave passive devices in accordance with an embodiment of the invention;
FIG. 3 is a graph of sample data versus model output characteristics for an embodiment of the invention;
FIG. 4 is a graph of sample data versus model output characteristics for an embodiment of the invention.
Detailed Description
For the purpose of making the objects, technical solutions and advantages of the present invention more apparent, embodiments of the present invention will be described in detail below with reference to the accompanying drawings.
As shown in fig. 2, in the neural network space mapping multi-physical modeling method for microwave passive devices according to the present invention, first, sample data for model training is obtained. For coarse model, let x gc Representing geometric parameters of a coarse model (electromagnetic domain), f c Representing the frequency of the coarse model, the input sample of the coarse model is x c ＝[x gc ，f c ] T . Let x g Representing the geometric parameters of the multiple physical models. For multiple thingsProblem management, including not only geometric parameter x g Also includes other physical parameters, x m Defined as other physical domain parameters, and the frequency parameter is defined as f. The input sample of the multiple physical models is x= [ x ] g ，x m ，f] T Outputting the sample data as S parameter, i.e. x c ＝[x gc ，f c ] T . The sample data may be obtained by means of an actual measurement device or simulation software.
As shown in fig. 2, the model employs a third-order training method. The first stage trains the electromagnetic domain coarse model. After electromagnetic domain coarse model training is complete, the weights in the coarse model are fixed, available to represent the electromagnetic response of the device, and are ready to be used as a priori knowledge for multi-physical model development. The second stage performs initialization training on the two mapping networks. The purpose of initializing the unit maps is to provide a good initial value for the mapped neural network before training them. The third stage is multi-physical domain training. The training data of this step is input-output samples of the multiple physical models. After the three-stage training process is finished, the built multi-physical model can accurately represent the characteristics of the microwave device.
The structure of the invention mainly comprises three parts: input mapping, output mapping, and coarse model. Building a model shown in fig. 1, training the coarse model in fig. 1 by using electromagnetic domain coarse model sample data, and adjusting weight parameters w of the coarse model * Fixing w after the training error and the testing precision of the coarse model meet the requirements * 。
The neural network structure shown in fig. 1 is built, and the input and output mapping network structure is set. The input mapping network adopts a 3-layer perceptron structure, and the input signal is [ x ] g ，x m ，f] T The output signal is [ x ] gc ，f c ] T . The output mapping network adopts a 3-layer perceptron structure, and the input signal is y c The output signal is y. To ensure that loading the input mapping network does not degrade the accuracy of the coarse model, the input mapping network and the output mapping network are unitized. Adjusting weight value w in input mapping network 1 To the point of
Let [ x ] g ，f] T ＝[x gc ，f c ] T The method comprises the steps of carrying out a first treatment on the surface of the Adjusting weight value w in output mapping network 2 To->
Let y=y c 。
Building a preliminary neural network space mapping multi-physical model by using the rough model and the input mapping unit mapping, training the unit mapping of the input mapping module by using the multi-physical sample data, and optimizing the weight parameters of the input mapping module
To->
Mapping the parameters of multiple physical domains to the electromagnetic domains, and converting the multiple physical modeling problem into an electromagnetic field modeling problem. If the test error does not meet the precision requirement, training with training data or adjusting the input mapping network structure is continued, the number of hidden layer neurons is changed, and training is resumed. If the test error meets the precision requirement, stopping training, otherwise, performing output mapping network adjustment.
Adding output mapping unit mapping on the preliminary neural network space mapping multi-physical model, fixing
Training the unit mapping of the output mapping module with the multi-physical sample data, optimizing the weight parameter of the output mapping module>
To->
And adjusting the output signal of the coarse model, and reducing the difference between the final output of the model and the output characteristic of the modeled device. If the test error does not meet the precision requirement, training with the training data or adjusting the output mapping network structure is continued to changeChanging the number of hidden layer neurons and retraining. If the test error meets the accuracy requirement, stopping training, otherwise fine tuning +.>
To->
Fine tuning->
To->
Until the error requirement is met.
Fig. 3 and 4 are graphs of the output characteristic curve of the model and the sample data established by the modeling method of the present invention, and it can be seen that the output curve of the model is consistent with the sample data.
Claims (4)
1. A neural network space mapping multi-physical modeling method for microwave passive devices comprises the following steps:
step 1: selecting and defining geometric parameters x of multiple physical models g Multiple physical parameters x of multiple physical models m And the data range of the frequency f of the multiple physical models, and performing multiple physical simulations to generate training and testing samples of the multiple physical models;
step 2: determining and defining the geometric parameter x of the electromagnetic domain coarse model according to step 1 gc And electromagnetic domain coarse model frequency f c Training and testing samples of the electromagnetic domain coarse model are generated by electromagnetic simulation, and in order to ensure the accuracy of the whole multi-physical model, the geometric parameter x of the electromagnetic domain coarse model is calculated gc The range is larger than the geometric parameter x in the integral multi-physical model g ；
Step 3: establishing a coarse model, training the coarse model by using the training data obtained in the step 2 by using a neural network technology, and testing the coarse model by using the test data obtained in the step 2 until the training error and the test precision of the coarse model meet the requirements, wherein the weight value w in the coarse model is the weight value w in the coarse model * Fixing;
step 4: initializing the input mapping network and the output mapping network, and optimizing the internal weight variable w of the input mapping 1 To the point of
Optimizing the internal weight variable w of the output map 2 To->
Realization of x gc ＝x g 、f c =f and y=y c The accuracy of the whole model is not reduced after the mapping network is added;
step 5: building a preliminary neural network space mapping multi-physical model by using the rough model obtained in the step 2 and the input mapping unit mapping obtained in the step 4, training the preliminary neural network space mapping model by using the multi-physical training data obtained in the step 1, and optimizing weight parameters of the input mapping module
To->
Mapping between the multi-physical domain input signals and the electromagnetic domain input signals is realized;
step 6: adding the output mapping unit mapping obtained in the step 4 on the preliminary neural network space mapping multi-physical model built in the step 5, and fixing the mapping unit mapping
Training the unit mapping of the output mapping module by using the multi-physical training data obtained in the step 1, and optimizing the weight parameter of the output mapping module>
To->
Let the output y of the coarse model c Consistent with the response y of the modeled device;
step (a)7: further training the whole multi-physical model by using the multi-physical training data obtained in the step 1, and simultaneously adjusting the weight value of the input neural network
To->
Adjusting the weight value of the output neural network +.>
To->
Until the built multi-physical model can accurately represent the characteristics of the passive device.
2. The neural network spatial mapping multi-physical modeling method for microwave passive devices according to claim 1, wherein in step 3, the trained coarse model formula is:
y c ＝g ANN (x gc ，f c ，w * ) (1)
wherein g ANN (. Cndot.) is a coarse model neural network mapping formula, w * A vector representing all weight parameters contained in the coarse model network.
3. A neural network spatial mapping multi-physical modeling method for microwave passive devices according to claim 1, wherein in step 5, the neural network is used to describe a nonlinear relationship between signals of the coarse model and the modeled device:
wherein f ANN1 (. Cndot.) is the input neural network mapping formula, x g And x m Is the input of the input map, x gc And f c Is the output of the input map and,
a vector representing all weight parameters contained in the mapping network.
4. The method of claim 1, wherein in step 6, the neural network is used to describe nonlinear relationships between the output signals of the coarse model and the output signals of the modeled device, respectively:
wherein f ANN2 (. Cndot.) is the output mapped neural network formula, representing y and y c The relationship between the two is that,
representing all internal weight variables of the mapped neural network. />
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201911402424.XA CN111695230B (en) | 2019-12-31 | 2019-12-31 | Neural network space mapping multi-physical modeling method for microwave passive device |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201911402424.XA CN111695230B (en) | 2019-12-31 | 2019-12-31 | Neural network space mapping multi-physical modeling method for microwave passive device |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN111695230A CN111695230A (en) | 2020-09-22 |
| CN111695230B true CN111695230B (en) | 2023-05-02 |
Family
ID=72476220
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201911402424.XA Active CN111695230B (en) | 2019-12-31 | 2019-12-31 | Neural network space mapping multi-physical modeling method for microwave passive device |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN111695230B (en) |
Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US5553195A (en) * | 1993-09-30 | 1996-09-03 | U.S. Philips Corporation | Dynamic neural net |
| DE102006011831A1 (en) * | 2006-03-10 | 2007-09-13 | Iav Gmbh Ingenieurgesellschaft Auto Und Verkehr | Model structure for non-linear system, has demodulator that is arranged between model input and non-parametric model, and splitting input signal according to frequency rates of demodulator |
| WO2017000557A1 (en) * | 2015-06-30 | 2017-01-05 | 东南大学 | Traffic prediction-based base station hibernation method in heterogeneous network |
| CN106777620A (en) * | 2016-12-05 | 2017-05-31 | 天津工业大学 | A kind of neutral net space reflection modeling method for power transistor |
| CN106777621A (en) * | 2016-12-05 | 2017-05-31 | 天津工业大学 | A kind of neutral net space reflection modeling method for packaged transistor |
| CN108549767A (en) * | 2018-04-13 | 2018-09-18 | 天津工业大学 | A kind of neural network space reflection modeling method for big signal power transistor |
| CN109284541A (en) * | 2018-08-31 | 2019-01-29 | 天津工业大学 | A kind of more Method of Physical Modeling of neural network for microwave passive component |
-
2019
- 2019-12-31 CN CN201911402424.XA patent/CN111695230B/en active Active
Patent Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US5553195A (en) * | 1993-09-30 | 1996-09-03 | U.S. Philips Corporation | Dynamic neural net |
| DE102006011831A1 (en) * | 2006-03-10 | 2007-09-13 | Iav Gmbh Ingenieurgesellschaft Auto Und Verkehr | Model structure for non-linear system, has demodulator that is arranged between model input and non-parametric model, and splitting input signal according to frequency rates of demodulator |
| WO2017000557A1 (en) * | 2015-06-30 | 2017-01-05 | 东南大学 | Traffic prediction-based base station hibernation method in heterogeneous network |
| CN106777620A (en) * | 2016-12-05 | 2017-05-31 | 天津工业大学 | A kind of neutral net space reflection modeling method for power transistor |
| CN106777621A (en) * | 2016-12-05 | 2017-05-31 | 天津工业大学 | A kind of neutral net space reflection modeling method for packaged transistor |
| CN108549767A (en) * | 2018-04-13 | 2018-09-18 | 天津工业大学 | A kind of neural network space reflection modeling method for big signal power transistor |
| CN109284541A (en) * | 2018-08-31 | 2019-01-29 | 天津工业大学 | A kind of more Method of Physical Modeling of neural network for microwave passive component |
Non-Patent Citations (2)
| Title |
|---|
| 史伟光 ; 杜江霞 ; 闫淑霞 ; 张晨光 ; .基于射频识别技术的实践教学平台建设方法研究.考试周刊.2017,(第36期),全文. * |
| 温佳 ; 刘明威 ; 崔军 ; 闫淑霞 ; .基于自适应阈值的干涉高光谱图像稀疏重建.光电工程.2019,(第06期),全文. * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN111695230A (en) | 2020-09-22 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109255160B (en) | Neural network-based unit delay prediction method and unit delay sensitivity calculation method | |
| CN105843073B (en) | A kind of wing structure aeroelastic stability analysis method not knowing depression of order based on aerodynamic force | |
| CN107784180B (en) | Design method of time-varying convex quadratic programming solver | |
| JP6404909B2 (en) | How to calculate the output model of a technical system | |
| CN109284541A (en) | A kind of more Method of Physical Modeling of neural network for microwave passive component | |
| CN108549767B (en) | Neural network space mapping modeling method for large-signal power transistor | |
| CN112580588B (en) | Intelligent flutter signal identification method based on empirical mode decomposition | |
| CN106777620A (en) | A kind of neutral net space reflection modeling method for power transistor | |
| CN112380774B (en) | Dynamic modeling method and system based on residual echo state network | |
| CN115688288B (en) | Aircraft pneumatic parameter identification method and device, computer equipment and storage medium | |
| CN112086100B (en) | Quantization error entropy based urban noise identification method of multilayer random neural network | |
| CN109033021A (en) | A kind of linear equation solver design method for joining convergence neural network based on change | |
| CN115062668A (en) | Harmonic parameter detection method and system based on RAdam optimization width learning | |
| CN111339642A (en) | Simulation model calibration method, system, readable medium and device | |
| CN115270239A (en) | Bridge reliability prediction method based on dynamic characteristics and intelligent algorithm response surface method | |
| CN110008914A (en) | A kind of pattern recognition system neural network based and recognition methods | |
| CN110362881B (en) | Microwave power device nonlinear model method based on extreme learning machine | |
| CN117709488B (en) | Dam seepage prediction method based on RUN-XGBoost | |
| CN111695230B (en) | Neural network space mapping multi-physical modeling method for microwave passive device | |
| CN117910128A (en) | Airfoil design method and device based on depth operator network | |
| CN111460734A (en) | Microwave device automatic modeling method combining advanced adaptive sampling and artificial neural network | |
| CN103778281A (en) | GaAs pHEMT pipe core nonlinear model parameter extraction method | |
| CN116992806A (en) | Automatic optimization method for large-scale analog integrated circuit based on self-attention mechanism | |
| CN107957685B (en) | Neurodynamics method for solving noise-containing time-varying problem | |
| CN114692529B (en) | CFD high-dimensional response uncertainty quantification method and device, and computer equipment |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |