CN115877100A - Method for predicting damage effect of amplitude limiter based on machine learning - Google Patents
Method for predicting damage effect of amplitude limiter based on machine learning Download PDFInfo
- Publication number
- CN115877100A CN115877100A CN202211511622.1A CN202211511622A CN115877100A CN 115877100 A CN115877100 A CN 115877100A CN 202211511622 A CN202211511622 A CN 202211511622A CN 115877100 A CN115877100 A CN 115877100A
- Authority
- CN
- China
- Prior art keywords
- amplitude
- amplitude limiter
- electromagnetic pulse
- electromagnetic
- pulse
- 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.)
- Pending
Links
Images
Landscapes
- Testing Electric Properties And Detecting Electric Faults (AREA)
Abstract
The invention discloses a method for predicting the damage effect of an amplitude limiter based on machine learning, which comprises the following steps: dividing the electromagnetic pulse into a broadband electromagnetic pulse and a narrowband electromagnetic pulse; extracting amplitude, pulse width, repetition frequency, application time and rise time as characteristics for the broadband electromagnetic pulse; extracting amplitude, pulse width, frequency, repetition frequency and application time as characteristics for the narrow-band electromagnetic pulse; changing electromagnetic pulse, collecting amplitude limiting output power and insertion loss of the output end of the amplitude limiter, comparing the amplitude limiting output power and the insertion loss with a reference value, and considering that the amplitude limiter is damaged if the amplitude limiter exceeds an error allowable range; constructing a data set by taking the characteristics as input and whether the damage is output; and dividing the data set according to the type of the amplitude limiter and the electromagnetic pulse classification, constructing and training a neural network for each subdata set, and finally predicting the damage effect of the amplitude limiter. The method can predict the damage effect of limiters of different types and different models in different electromagnetic pulse environments.
Description
Technical Field
The invention belongs to the technical field of electromagnetic effect analysis, and particularly relates to a method for predicting damage effect of an amplitude limiter based on machine learning.
Background
Under the action of electromagnetic pulse, the main energy coupling channel of the electronic equipment is a front door, namely, pulse energy is injected into the radio-frequency front-end component of the electronic equipment through an antenna. The amplitude limiter serves as an important protection module of the radio frequency front end of the electronic equipment, the amplitude of output voltage can be limited within a certain range, and rear-end devices are protected from being damaged. However, under the action of electromagnetic pulses with extremely high energy, the limiter also faces risks of degradation of limiting capability, burning and the like, and the limiting function is lost, so that the rear-end electronic device of the electronic equipment is damaged.
Therefore, effect prediction aiming at the amplitude limiter is an important basis for developing the electromagnetic effect analysis of the radio frequency front end of the electronic equipment. The traditional method is to calculate the output voltage and the insertion loss of the amplitude limiter under the action of electromagnetic pulse through simulation software, compare the output voltage and the insertion loss with the value during normal working and judge whether the amplitude limiter is damaged or not, however, the simulation has the characteristics of difficulty in accurate modeling and long time consumption. Taking modeling of a PIN limiter as an example, a detailed parameter set which is difficult to master except for manufacturers, such as device structure parameters, doping concentration distribution and the like, needs to be input; in addition, each time any parameter such as frequency, pulse width, repetition frequency, amplitude, accumulated action time and the like of the electromagnetic pulse is changed, simulation calculation needs to be carried out again, and time is consumed. Therefore, a fast and effective method for predicting the damage effect of the limiter is urgently needed to be established, so that support is provided for the analysis of the radio frequency front end effect of the electronic equipment, and a basis is provided for protection and reinforcement.
Disclosure of Invention
The invention aims to provide a method for predicting the damage effect of an amplitude limiter based on machine learning, which can predict the damage effect of amplitude limiters of different types and different models in different electromagnetic pulse environments.
The technical scheme adopted by the invention is as follows:
a method for predicting the damage effect of a limiter based on machine learning comprises the following steps:
dividing the electromagnetic pulse into a broadband electromagnetic pulse and a narrowband electromagnetic pulse; extracting amplitude, pulse width, repetition frequency, application time and rise time as characteristics for the broadband electromagnetic pulse; extracting amplitude, pulse width, frequency, repetition frequency and application time as characteristics for the narrow-band electromagnetic pulse;
collecting amplitude limiting output power and insertion loss of an amplitude limiter in a normal working state without the action of electromagnetic pulses, and using the amplitude limiter as a reference for judging whether the amplitude limiter is damaged; changing electromagnetic pulse, collecting amplitude limiting output power and insertion loss of the output end of the amplitude limiter, comparing the amplitude limiting output power and the insertion loss with a reference value, and if the amplitude limiter exceeds an error allowable range, determining that the amplitude limiter is damaged, otherwise, determining that the amplitude limiter is not damaged;
taking the characteristics of electromagnetic pulses as input and whether the amplitude limiter is damaged as output to construct a data set; and dividing the data set according to the type of the amplitude limiter and the electromagnetic pulse classification, constructing and training a neural network for each subdata set, and finally predicting the damage effect of the amplitude limiter.
Furthermore, the electromagnetic pulses are classified into broadband electromagnetic pulses and narrowband electromagnetic pulses according to the signal frequency bandwidth.
Furthermore, the bandwidth of more than 25% is broadband electromagnetic pulse, and the bandwidth of less than 10% is narrow-band electromagnetic pulse.
Furthermore, before the data set is constructed, the collected data is preprocessed, and the preprocessing comprises data cleaning, abnormal value elimination and data standardization.
Further, the neural network is a fully connected neural network.
Compared with the prior art, the invention has the following advantages and beneficial effects:
the invention provides a method for predicting the damage effect of an amplitude limiter under the action of electromagnetic pulses based on a machine learning algorithm, which divides the electromagnetic pulses into broadband electromagnetic pulses and narrowband electromagnetic pulses, extracts amplitude, pulse width, repetition frequency, application time and rise time as characteristics for the broadband electromagnetic pulses, extracts amplitude, pulse width, frequency, repetition frequency and application time as characteristics for the narrowband electromagnetic pulses, establishes a machine learning model for rapidly predicting the damage effect of the amplitude limiter based on historical data, solves the problems that the common simulation method is difficult to accurately fit and consumes a long time, and provides a foundation for effect analysis and protection reinforcement of the radio frequency front end of electronic equipment under the electromagnetic pulses.
Drawings
FIG. 1 is a flow chart of the electromagnetic effect data acquisition of the limiter of the present invention under the action of electromagnetic pulses;
FIG. 2 is a schematic diagram of data set classification according to the present invention.
Detailed Description
In order to make the objects, technical solutions and advantages of the present invention more apparent, the present invention is described in further detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention. In addition, the technical features involved in the respective embodiments of the present invention described below may be combined with each other as long as they do not conflict with each other.
The invention provides a method for predicting the damage effect of an amplitude limiter based on machine learning, which has the following technical key points:
(1) Electromagnetic pulse characteristic modeling and electromagnetic effect data set acquisition and construction of amplitude limiter under electromagnetic pulse action
Comprehensively considering the type and the characteristics of the electromagnetic pulse, dividing the electromagnetic pulse into a broadband electromagnetic pulse and a narrowband electromagnetic pulse, and respectively extracting characteristics to construct a model; the method comprises the steps of obtaining electromagnetic effect data of an amplitude limiter under the action of electromagnetic pulses by changing test conditions, completing data preprocessing work through the steps of data cleaning, abnormal value elimination, data standardization and the like, matching whether the amplitude limiter corresponding to preprocessed electromagnetic pulse characteristic data is damaged or not, and constructing a data set for algorithm learning and training.
(2) Network structure design for predicting damage effect of amplitude limiter based on machine learning
The method is used for predicting whether the amplitude limiter is damaged essentially by utilizing the advantages of a fully-connected neural network in machine learning on nonlinear fitting capability and algorithm robustness to predict whether the amplitude limiter is damaged under the action of electromagnetic pulses, and mainly comprises the steps of data set division, network input and output node design, network hyper-parameter selection, network training and prediction.
The method for predicting the damage effect of the amplitude limiter based on machine learning comprises the following steps:
1. electromagnetic pulse characteristic modeling and electromagnetic effect data set acquisition and construction of amplitude limiter under electromagnetic pulse effect
The electromagnetic effect data acquisition process of the amplitude limiter under the action of the electromagnetic pulse is shown in figure 1. Electromagnetic pulses can be generally divided into broadband electromagnetic pulses (with a bandwidth greater than 25%) and narrowband electromagnetic pulses (with a bandwidth less than 10%) according to the signal frequency bandwidth.
The broadband electromagnetic pulse has the characteristics of steep rising front (ps level), narrow pulse width (ns-ps level) and wide radio frequency spectrum, and the amplitude, the pulse width, the repetition frequency, the application time and the rising time of the broadband electromagnetic pulse are extracted as characteristics. The method is characterized in that a data set is constructed based on damage test data of broadband electromagnetic pulses to the amplitude limiter, and amplitude limiting output power and insertion loss of the amplitude limiter in a normal working state without the action of the electromagnetic pulses are firstly tested and used as a reference for judging whether the amplitude limiter is damaged or not. And then starting a broadband electromagnetic pulse source, collecting the amplitude, pulse width, repetition frequency, application time and rising time values of the broadband electromagnetic pulse at the input end of the amplitude limiter, collecting the amplitude limiting output power and insertion loss of the output end of the amplitude limiter, comparing the amplitude limiting output power and the insertion loss with a reference value, and considering that the amplitude limiter is damaged if the amplitude limiter exceeds an error allowable range, otherwise, not damaging the amplitude limiter. Changing the test condition, changing the amplitude, the pulse width, the repetition frequency, the application time and the rise time, acquiring the amplitude, the pulse width, the repetition frequency, the application time and the rise time of the broadband electromagnetic pulse at the input end of the amplitude limiter again, acquiring the amplitude limiting output power and the insertion loss of the output end of the amplitude limiter, comparing the amplitude limiting output power and the insertion loss with the reference value, and judging whether the amplitude limiter is damaged or not until the data quantity meeting the requirement is acquired.
For narrow-band electromagnetic pulses, the present invention extracts amplitude, pulse width, frequency, repetition frequency, and application time as features. The method is characterized in that a data set is constructed based on damage test data of narrow-band electromagnetic pulses to the amplitude limiter, and amplitude limiting output power and insertion loss of the amplitude limiter in a normal working state without the action of the electromagnetic pulses are firstly tested and used as a reference for judging whether the amplitude limiter is damaged or not. And then starting a narrow-band electromagnetic pulse signal source, collecting the amplitude, pulse width, frequency, repetition frequency and application time value of the narrow-band electromagnetic pulse at the input end of the amplitude limiter, collecting the amplitude limiting output power and insertion loss of the output end of the amplitude limiter, comparing the amplitude limiting output power and insertion loss with a reference value, considering that the amplitude limiter is damaged if the amplitude limiter exceeds the error allowable range, and otherwise, judging that the amplitude limiter is not damaged. And changing the test conditions, changing the amplitude, the pulse width, the frequency, the repetition frequency and the application time, acquiring the amplitude, the pulse width, the frequency, the repetition frequency and the application time of the narrow-band electromagnetic pulse at the input end of the amplitude limiter, acquiring the amplitude limiting output power and the insertion loss of the output end of the amplitude limiter, comparing the amplitude limiting output power and the insertion loss with the reference value, and judging whether the amplitude limiter is damaged or not until the data quantity meeting the requirements is acquired.
The acquired data is preprocessed through steps of data cleaning, outlier elimination, data standardization and the like, and then a data set is constructed by the preprocessed data and is used for algorithm learning and training.
2. Network structure design for predicting damage effect of amplitude limiter based on machine learning
The constructed data set is classified according to the model of the amplitude limiter and the electromagnetic pulse environment, and a classification schematic diagram is shown in fig. 2. For each sub data set, a fully connected neural network is constructed and trained, and then it is predicted whether the slicer is impaired at a given input. Taking broadband electromagnetic pulse as an example, the construction and training key points of the fully-connected neural network are as follows:
(1) The characteristics of the broadband electromagnetic pulse comprise amplitude, pulse width, repetition frequency, application time and rise time, and after bias terms are considered, the number of corresponding input end nodes of the fully-connected neural network is 6;
(2) The network output is whether the amplitude limiter is damaged or not, the probability of damage or no damage is respectively represented by 2 nodes, the value range is [0,1], and the sum of the probabilities is 1;
(3) A hidden layer is arranged between an input layer and an output layer, the number of layers of the hidden layer and the number of nodes of each layer are hyper-parameters of the fully-connected neural network, and the selection of the hyper-parameters directly influences the performance of the network;
(4) Dividing the sub data set into a training set, a cross validation set and a test set, respectively training by using the training set aiming at a plurality of groups of preset hyper-parameters, then calculating prediction errors under different hyper-parameter combinations by using the cross validation set, and selecting the hyper-parameter which best appears on the cross validation set as a network structure parameter;
(5) And performing effect prediction on the test set by using the trained fully-connected neural network, calculating the error of the test set, and representing the performance of the fully-connected neural network on real data under the conditions of data balance and enough data quantity.
It should be noted that, according to implementation requirements, each step/component described in the present application can be divided into more steps/components, and two or more steps/components or partial operations of the steps/components can also be combined into a new step/component to achieve the purpose of the present invention.
It will be understood by those skilled in the art that the foregoing is merely a preferred embodiment of the present invention, and is not intended to limit the invention, and that any modification, equivalent replacement, or improvement made within the spirit and principle of the present invention should be included within the scope of the present invention.
Claims (5)
1. A method for predicting the damage effect of a limiter based on machine learning is characterized by comprising the following steps:
dividing the electromagnetic pulse into a broadband electromagnetic pulse and a narrowband electromagnetic pulse; extracting amplitude, pulse width, repetition frequency, application time and rise time as characteristics for the broadband electromagnetic pulse; extracting amplitude, pulse width, frequency, repetition frequency and application time as features for the narrow-band electromagnetic pulse;
collecting amplitude limiting output power and insertion loss of an amplitude limiter in a normal working state without the action of electromagnetic pulses, and using the amplitude limiter as a reference for judging whether the amplitude limiter is damaged; changing electromagnetic pulse, collecting amplitude limiting output power and insertion loss of the output end of the amplitude limiter, comparing the amplitude limiting output power and the insertion loss with a reference value, and if the amplitude limiter exceeds an error allowable range, determining that the amplitude limiter is damaged, otherwise, determining that the amplitude limiter is not damaged;
taking the characteristics of the electromagnetic pulse as input and whether the amplitude limiter is damaged or not as output to construct a data set; and dividing the data set according to the type of the amplitude limiter and the electromagnetic pulse classification, constructing and training a neural network for each subdata set, and finally predicting the damage effect of the amplitude limiter.
2. The method of predicting slicer impairment effects based on machine learning of claim 1, wherein the electromagnetic pulses are classified into wideband electromagnetic pulses and narrowband electromagnetic pulses according to signal frequency bandwidth.
3. The method for predicting slicer impairment effects based on machine learning of claim 2, wherein bandwidths greater than 25% are broadband electromagnetic pulses and bandwidths less than 10% are narrowband electromagnetic pulses.
4. The method for predicting the damage effect of the amplitude limiter based on the machine learning as claimed in claim 1, wherein the collected data is preprocessed before the data set is constructed, and the preprocessing comprises data cleaning, outlier rejection and data normalization.
5. The method for predicting the effect of slicer impairment based on machine learning of claim 1, wherein the neural network is a fully-connected neural network.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211511622.1A CN115877100A (en) | 2022-11-29 | 2022-11-29 | Method for predicting damage effect of amplitude limiter based on machine learning |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211511622.1A CN115877100A (en) | 2022-11-29 | 2022-11-29 | Method for predicting damage effect of amplitude limiter based on machine learning |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN115877100A true CN115877100A (en) | 2023-03-31 |
Family
ID=85764650
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202211511622.1A Pending CN115877100A (en) | 2022-11-29 | 2022-11-29 | Method for predicting damage effect of amplitude limiter based on machine learning |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN115877100A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116776736A (en) * | 2023-06-29 | 2023-09-19 | 中国人民解放军国防科技大学 | Diode structure prediction method based on feature extraction and random noise injection |
-
2022
- 2022-11-29 CN CN202211511622.1A patent/CN115877100A/en active Pending
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116776736A (en) * | 2023-06-29 | 2023-09-19 | 中国人民解放军国防科技大学 | Diode structure prediction method based on feature extraction and random noise injection |
| CN116776736B (en) * | 2023-06-29 | 2024-01-12 | 中国人民解放军国防科技大学 | Diode structure prediction method based on feature extraction and random noise injection |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN111539515B (en) | Complex equipment maintenance decision method based on fault prediction | |
| CN111459700B (en) | Equipment fault diagnosis method, diagnosis device, diagnosis equipment and storage medium | |
| CN112910695B (en) | Network fault prediction method based on global attention time domain convolution network | |
| CN109193650B (en) | Power grid weak point evaluation method based on high-dimensional random matrix theory | |
| CN108535635B (en) | EEMD and HMM based analog circuit intermittent fault diagnosis method | |
| CN110059845B (en) | Metering device clock error trend prediction method based on time sequence evolution gene model | |
| CN105300692A (en) | Bearing fault diagnosis and prediction method based on extended Kalman filtering algorithm | |
| CN111669123B (en) | Method and device for fault diagnosis of photovoltaic string | |
| CN109492790A (en) | Wind turbines health control method based on neural network and data mining | |
| CN107942994A (en) | A kind of satellite temperature control system method for diagnosing faults based on temperature curve feature | |
| CN115877100A (en) | Method for predicting damage effect of amplitude limiter based on machine learning | |
| CN108846511A (en) | A kind of defect of transformer equipment trend analysis based on regulation big data platform | |
| CN112379325A (en) | Fault diagnosis method and system for intelligent electric meter | |
| CN112507605A (en) | Power distribution network anomaly detection method based on AnoGAN | |
| CN114757282A (en) | Intelligent electric meter operation state evaluation method based on entropy weight method and random forest model | |
| CN114548493A (en) | Method and system for predicting current overload of electric energy meter | |
| CN114091549A (en) | Equipment fault diagnosis method based on deep residual error network | |
| CN103020346B (en) | Test method for physical design similarity of circuit | |
| CN116128690B (en) | Carbon emission cost value calculation method, device, equipment and medium | |
| CN107220712A (en) | A kind of method and system of change of current valve status monitoring and maintenance decision based on markoff process | |
| CN103400201A (en) | Method for solving state estimation problem taking maximum normal rate of measurement point as target | |
| Jia et al. | A comparison study of hidden Markov model and particle filtering method: Application to fault diagnosis for gearbox | |
| CN116431988A (en) | Resident trip activity time sequence generation method based on activity mode-Markov chain | |
| CN115640542A (en) | Intelligent electric energy meter state evaluation method and evaluation device based on Bayesian theory | |
| CN114298188A (en) | Intelligent analysis method and system for power equipment faults |
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 |