CN102855346B - Electromagnetic compatibility forecasting method based on function link neural network - Google Patents

Electromagnetic compatibility forecasting method based on function link neural network Download PDF

Info

Publication number
CN102855346B
CN102855346B CN201210257617.2A CN201210257617A CN102855346B CN 102855346 B CN102855346 B CN 102855346B CN 201210257617 A CN201210257617 A CN 201210257617A CN 102855346 B CN102855346 B CN 102855346B
Authority
CN
China
Prior art keywords
neural network
function chain
chain neural
output
function
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
Application number
CN201210257617.2A
Other languages
Chinese (zh)
Other versions
CN102855346A (en
Inventor
沈文
邓辉
侯功
王玮
奚后玮
吴军民
张刚
黄在朝
黄辉
刘川
吴鹏
陈磊
于海
虞跃
姚启桂
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
State Grid Corp of China SGCC
China Electric Power Research Institute Co Ltd CEPRI
Global Energy Interconnection Research Institute
Original Assignee
State Grid Corp of China SGCC
China Electric Power Research Institute Co Ltd CEPRI
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by State Grid Corp of China SGCC, China Electric Power Research Institute Co Ltd CEPRI filed Critical State Grid Corp of China SGCC
Priority to CN201210257617.2A priority Critical patent/CN102855346B/en
Publication of CN102855346A publication Critical patent/CN102855346A/en
Application granted granted Critical
Publication of CN102855346B publication Critical patent/CN102855346B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Landscapes

  • Feedback Control In General (AREA)

Abstract

The invention provides an electromagnetic compatibility forecasting method based on a function link neural network. The method comprises the following steps of: (1) taking n times of signal radiation strength statistical data as an initial input' (2) expanding one-dimensional data of the initial input to be two-dimensional data to obtain an input m of a two-dimensional function link neural network; (3) calculating a weight value of the function link neural network to obtain a forecasting function f(y'); and (4) according to the forecasting function, forecasting the electromagnetic compatibility between electrical elements. By the electromagnetic compatibility forecasting method based on the function link neural network, the dimension of the input part is expanded, and the conventional nonlinear problem is converted into a linear problem, so that the shortcomings of the conventional gradient search algorithm can be well overcome; and the problems of relatively complicated and troublesome multi-layer calculation of the BP neural network are solved.

Description

Electromagnetic compatibility Forecasting Methodology based on function chain neural network
Technical field
The invention belongs to the Electro Magnetic Compatibility field of high-speed digital circuit, be specifically related to a kind of electromagnetic compatibility Forecasting Methodology based on function chain neural network.
Background technology
Along with frequency of operation and the packaging density of device on high speed circuit board improve constantly, the operating voltage continuous decrease in circuit, so just causes circuit more and more lower to the tolerance of electromagnetic noise, and Electro Magnetic Compatibility just becomes the major issue that affects high speed circuit performance.Impact for fear of electromagnetic noise on circuit, when design High Speed PCB Board, designer need to consider the Electro Magnetic Compatibility of circuit simultaneously.
Electromagnetic compatibility (Electromagnetic Compatibility) refers under specific electromagnetic environment, the ability of mutual co-ordination between electronic devices and components.The power supply noise that on pcb board, chip switch produces is main electromagnet source, in order to accurately calculate this power supply noise, is necessary the Electro Magnetic Compatibility of circuit board to predict.
Electromagnetic compatibility prediction can be High-speed Board Design provides theoretical direction.At present, both at home and abroad electromagnetic compatibility theory is conducted in-depth research, set up various analyses and prediction models: the electromagnetic compatibility Forecasting Methodology based on BP neural network; Based on the electromagnetic compatibility Forecasting Methodology of ant group neural network, and the electromagnetic compatibility Forecasting Methodology based on Speed Controlling Based on Improving BP Neural Network etc.
The method of using neural network to carry out electromagnetic compatibility prediction at present faces following defect:
Very responsive to initial weight, BP is the rapid adjustment network Weight algorithm based on gradient search, but in nonlinear problem, gradient search algorithm exists certain defect.Although have polyalgorithm to be optimized weight, still can make neural network converge under certain condition local minimum.
Summary of the invention
For overcoming above-mentioned defect, the invention provides a kind of electromagnetic compatibility Forecasting Methodology based on function chain neural network, dimension expansion is carried out in importation, script nonlinear problem can be transformed into linear problem, therefore can be fine the deficiency that exists of solution gradient search algorithm, avoided the problem of the more loaded down with trivial details and trouble of the multilayer computing of BP neural network simultaneously.
For achieving the above object, the invention provides a kind of electromagnetic compatibility Forecasting Methodology based on function chain neural network, each electric elements on PCB are carried out to emc testing, its improvements are, described method comprises the steps:
(1). using n signal radiation intensity statistics data as initial input;
(2). the one-dimensional data to initial input expands to two dimension, obtains the input m of a two-dimentional function chain neural network;
(3). computing function chain neural network weights, obtain anticipation function f (y ');
(4). according to anticipation function, predict the Electro Magnetic Compatibility between each electric elements.
In optimal technical scheme provided by the invention, in described step 1, n signal radiation intensity statistics data are the signal radiation intensity statistics data between upper each electric elements of PCB.
In the second optimal technical scheme provided by the invention, in described step 2, in two-dimensional space, describe originate mode and be: x 1..., x n, x 1x 2..., x n-1x n.
In the 3rd optimal technical scheme provided by the invention, in described step 3, the step of the calculating based on function chain neural network is as follows: the output of computing function chain neural network, and adjust neuron weights according to this output.
In the 4th optimal technical scheme provided by the invention, the comprising the steps: of described step 3
(3-1). the output of computing function chain neural network;
(3-2). judge whether to adjust neuron weights, if needed, neuron weights are adjusted, and returned to step 3-1; Otherwise output function chain neural network anticipation function f (y ').
In the 5th optimal technical scheme provided by the invention, in described step 3-1, following parameter is arranged: Shiftable window, the signal radiation intensity statistics number of times that described Shiftable window covers is n; Two-dimensional function chain neural network be input as m, the Output rusults of getting after Shiftable window is y '.
In the 6th optimal technical scheme provided by the invention, the formula of computing function chain neural network output y ' is as follows:
y ′ = 1 1 + e θ - wx t
Wherein, the biasing that θ is y ', wx t=w 1x 1+ ...+w nx n+ w n+1x 1x 2+ ...+w n+n (n-1)/2x n-1x n, w i, i=1,2,3...n represents weight.The physical significance of y please be described.
In the 7th optimal technical scheme provided by the invention, in described step 3-2, the step that judges whether to adjust neuron weights is as follows: | y '-y| > k, adjust neuron weights, until | y '-y|≤k; Wherein, y represents current demand signal actual emanations intensity, and k represents error precision, gives the random value in (0,1) interval to each neuron connection weights of network.
Compared with the prior art, a kind of electromagnetic compatibility Forecasting Methodology based on function chain neural network provided by the invention, solved initial weight very responsive, BP is the rapid adjustment network Weight algorithm based on gradient search, but in nonlinear problem, there is the problem of certain defect in gradient search algorithm; Dimension expansion is carried out in importation, script nonlinear problem can be transformed into linear problem, therefore can be fine the deficiency that exists of solution gradient search algorithm, avoided the problem of the more loaded down with trivial details and trouble of the multilayer computing of BP neural network simultaneously.
Accompanying drawing explanation
Fig. 1 is the general schematic view of the electromagnetic compatibility Forecasting Methodology based on function chain neural network.
Fig. 2 is the embodiment schematic diagram of the electromagnetic compatibility Forecasting Methodology based on function chain neural network.
Fig. 3 is the logical schematic of computing function chain neural network output.
Embodiment
A kind of electromagnetic compatibility Forecasting Methodology based on function chain neural network as shown in Figure 1, 2, comprises the following steps:
(1). using the signal radiation intensity statistics data of n time as training sample as initial input pattern;
(2). before the statistics of signal radiation intensity is carried out the expansion (x of dimension as initial input 1, x 2... x n) → x 1..., x n, x 1x 2..., x n-1x n, obtain the input of a two-dimentional function chain neural network;
(3). computing function chain neural network weights, obtain anticipation function f (y ');
(4). according to anticipation function, predict the Electro Magnetic Compatibility between each electric elements.
Extract electromagnetic compatibility parameter as initial input pattern, strengthen initial input pattern, the input pattern of one dimension is expanded to two dimension upper, in two-dimensional space, describe originate mode x 1..., x n, x 1x 2..., x n-1x n, form the expression of enhancement mode.
It is as shown in Figure 3, described that by function chain neural network forecasting process, to obtain the concrete steps of Output rusults as follows:
(3-1). arrange one movably window size be n, the signal radiation intensity statistics number of times that described movably window covers is n;
(3-2). n the signal radiation intensity level obtaining carried out to dimension and expand the initial input m that obtains function chain neural network, get Shiftable window numerical value afterwards for output y;
(3-3). the input layer number of function chain neural network is m, and output layer node is 1;
(3-4). step-up error precision k, gives the random value in (0,1) interval to each neuron connection weights of network;
(3-5). the output of computing function chain neural network
Figure BDA00001922811400041
the biasing that wherein θ is y ',
Wx t=w 1x 1+ ...+w nx n+ w n+1x 1x 2+ ...+w n+n (n-1)/2x n-1x n, w i, i=1,2,3...n represents weight;
If (3-6). | y '-y| > k, adjusts neuron weight w i, computing function chain neural network output again, until | y '-y|≤k, the function chain neural network anticipation function f (y ') of the weights that are optimized;
(3-7). according to f (y '), the numerical value after Shiftable window is predicted for exporting.
Need statement, content of the present invention and embodiment are intended to prove the practical application of technical scheme provided by the present invention, should not be construed as limiting the scope of the present invention.Those skilled in the art inspired by the spirit and principles of the present invention, can do various modifications, be equal to and replace or improve.But in the protection domain that these changes or modification are all awaited the reply in application.

Claims (3)

1. the electromagnetic compatibility Forecasting Methodology based on function chain neural network, carries out emc testing to each electric elements on PCB, it is characterized in that, described method comprises the steps:
(1). using n signal radiation intensity statistics data as initial input;
(2). the one-dimensional data to initial input expands to two dimension, obtains the input m of a two-dimentional function chain neural network;
(3). computing function chain neural network weights, obtain anticipation function f (y');
(4). according to anticipation function, predict the Electro Magnetic Compatibility between each electric elements;
In described step (3), the step of the calculating based on function chain neural network is as follows: the output of computing function chain neural network, and adjust neuron weights according to this output;
Described step (3) comprises the steps:
(3-1). the output of computing function chain neural network;
(3-2). judge whether to adjust neuron weights, if needed, neuron weights are adjusted, and returned to step (3-1); Otherwise output function chain neural network anticipation function f (y');
In described step (3-1), following parameter is arranged: Shiftable window, the signal radiation intensity statistics number of times that described Shiftable window covers is n; Two-dimensional function chain neural network be input as m, the Output rusults of getting after Shiftable window is y ';
The formula of computing function chain neural network output y ' is as follows:
y ′ = 1 1 + e θ - wx t
Wherein, the biasing that θ is y', wx t=w 1x 1+ ...+w nx n+ w n+1x 1x 2+ ...+w n+n (n-1)/2x n-1x n, w i, i=1,2,3...n represents weight;
In described step (3-2), the step that judges whether to adjust neuron weights is as follows: | y'-y| > k, adjust neuron weights, until | y'-y|≤k; Wherein, y represents current demand signal actual emanations intensity, and k represents error precision, gives the random value in (0,1) interval to each neuron connection weights of network.
2. method according to claim 1, is characterized in that, in described step (1), n signal radiation intensity statistics data are the signal radiation intensity statistics data between upper each electric elements of PCB.
3. method according to claim 1, is characterized in that, in described step (2), describes originate mode be in two-dimensional space: x 1..., x n, x 1x 2..., x n-1x n.
CN201210257617.2A 2012-07-24 2012-07-24 Electromagnetic compatibility forecasting method based on function link neural network Active CN102855346B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201210257617.2A CN102855346B (en) 2012-07-24 2012-07-24 Electromagnetic compatibility forecasting method based on function link neural network

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201210257617.2A CN102855346B (en) 2012-07-24 2012-07-24 Electromagnetic compatibility forecasting method based on function link neural network

Publications (2)

Publication Number Publication Date
CN102855346A CN102855346A (en) 2013-01-02
CN102855346B true CN102855346B (en) 2014-04-23

Family

ID=47401932

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201210257617.2A Active CN102855346B (en) 2012-07-24 2012-07-24 Electromagnetic compatibility forecasting method based on function link neural network

Country Status (1)

Country Link
CN (1) CN102855346B (en)

Families Citing this family (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN106777612B (en) * 2016-12-02 2020-12-01 全球能源互联网研究院有限公司 Method and device for establishing PCB type prediction model and PCB design
CN109858360B (en) * 2018-12-28 2020-09-18 华中科技大学 Dynamic management method and system for electromagnetic compatibility of fleet

Family Cites Families (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN101714184B (en) * 2009-11-20 2011-08-17 北京航空航天大学 Behavioral level modeling system for circuit board level electromagnetic compatible sensitivity

Also Published As

Publication number Publication date
CN102855346A (en) 2013-01-02

Similar Documents

Publication Publication Date Title
CN106886636A (en) A kind of accurate Forecasting Methodology of the worst power supply noise of high-speed circuit system
CN110442926A (en) IC Statistical Time Series Analysis Method under advanced technologies and low-voltage
CN106777620A (en) A kind of neutral net space reflection modeling method for power transistor
Ramahi et al. A simple finite-difference frequency-domain (FDFD) algorithm for analysis of switching noise in printed circuit boards and packages
CN102855346B (en) Electromagnetic compatibility forecasting method based on function link neural network
Zhu et al. A novel automatically designed EBG structure by improved GA for ultrawideband SSN mitigation of system in package
Shringarpure et al. Sensitivity analysis of a circuit model for power distribution network in a multilayered printed circuit board
CN104752245B (en) The numerical analysis method that high power pulse influences on FET amplifier performance
Yi et al. Skin-effect-incorporated transient simulation using the Laguerre-FDTD scheme
Müller et al. Non-uniform currents on vias and their effects in a parallel-plate environment
Jia et al. Time-domain shielding effectiveness analysis based on DGTD method accelerated by local time-stepping and parallel techniques
Choi et al. Near-field and far-field analyses of alternating impedance electromagnetic bandgap (AI-EBG) structure for mixed-signal applications
Jia et al. Slit-surface disturbance lattice EBG structure for simultaneously switching noise suppression in high-speed data acquisition system
Liu et al. Application of asymptotic waveform evaluation to eigenmode expansion method for analysis of simultaneous switching noise in printed circuit boards (PCBs)
CN111079357A (en) Electromagnetic interference time domain analysis method of microstrip line on PCB under action of electromagnetic wave
Wei et al. Extraction of equivalent network of arbitrarily shaped power-ground planes with narrow slots using a novel integral equation method
See et al. Full-Wave Modeling of RF Voltage Drop on Printed Circuit Interconnects
Wang et al. A Modeling Method for Common Mode Noise of ANPC Inverter System Based on ANN
Cryan et al. Electromagnetic Compatibility Analysis of Multilayer PCBs Using a Hybrid Finite Difference Time Domain (FDTD)-Partial Element Equivalent Circuit (PEEC) Method
Li et al. Application and Prospect of Artificial Intelligence Method in Signal Integrity Design of Microsystem
Li et al. [Retracted] Electronic Controller Automatic Test System Based on Intelligent Control Algorithm
Wu et al. Examining the Thermal Stress Effect of Electromagnetic Susceptibility of Automotive CAN Controller
Sartori EMC Spotlight
Zhang High-Speed Circuit Power Integrity Design Based on Impedance Characteristic Analysis
Yuan et al. A simulation-based method for EMR assessment of aviation electronic products

Legal Events

Date Code Title Description
C06 Publication
PB01 Publication
C10 Entry into substantive examination
SE01 Entry into force of request for substantive examination
C14 Grant of patent or utility model
GR01 Patent grant
C41 Transfer of patent application or patent right or utility model
TR01 Transfer of patent right

Effective date of registration: 20160425

Address after: 100192 Beijing city Haidian District Qinghe small Camp Road No. 15

Patentee after: China Electric Power Research Institute

Patentee after: State Grid Smart Grid Institute

Patentee after: State Grid Corporation of China

Address before: 100192 Beijing city Haidian District Qinghe small Camp Road No. 15

Patentee before: China Electric Power Research Institute

Patentee before: State Grid Corporation of China

C56 Change in the name or address of the patentee
CP01 Change in the name or title of a patent holder

Address after: 100192 Beijing city Haidian District Qinghe small Camp Road No. 15

Patentee after: China Electric Power Research Institute

Patentee after: GLOBAL ENERGY INTERCONNECTION RESEARCH INSTITUTE

Patentee after: State Grid Corporation of China

Address before: 100192 Beijing city Haidian District Qinghe small Camp Road No. 15

Patentee before: China Electric Power Research Institute

Patentee before: State Grid Smart Grid Institute

Patentee before: State Grid Corporation of China