CN111695608A - Data expansion method for preserving original sample distribution characteristics - Google Patents

Abstract

The invention discloses a data expansion method for preserving original sample distribution characteristics, and belongs to the technical field of micro-system board power supplies and electrical systems. The invention solves the problems of small test data volume and few fitting processing methods of the micro-system power panel. The method effectively reserves and extracts the distribution characteristics of the original data sample, fits the variation condition of the micro-system power panel under the nuclear irradiation stress, can be used as an effective and visual analysis basis for the variation of the micro-system power panel under the nuclear irradiation stress, and provides a statistical processing method better.

Description

Data expansion method for preserving original sample distribution characteristics

Technical Field

The invention belongs to the technical field of micro-system board power supplies and electrical systems, and particularly relates to an algorithm for expanding and processing nuclear radiation stress result data volume samples under the condition based on original data distribution characteristics.

Background

With the rapid development of power electronic technology, typical circuit boards have been widely applied to the fields of household appliances, vehicles, ships and warships, aerospace and the like, bring convenience to countless fields, and play a great role.

The change conditions of the micro-system power panel under different stress conditions are different, the service life and other parameters under different environments are different, the existing method mainly solves the problem of a change processing method of the power panel under common stress, such as high temperature, high humidity, strong current and other environments, and research on the nuclear irradiation environment is less.

The existing traditional method aiming at small sample size expansion mainly comprises mathematical statistical methods such as Bootstap resampling and the like, the principle is that a repeated sampling technology is adopted to extract a certain amount of original samples, a confidence interval and the like can be constructed through variance estimation, the application range of resampling is further extended, the method is a method for processing the resampling of original data values to further estimate true distribution, but the consideration on sample size distribution characteristics is less, and the problem is almost ignored by many existing traditional statistical methods.

Disclosure of Invention

Aiming at the defects of the background technology, the invention solves the problem of small test data result quantity of the micro-system power panel in the irradiation environment on the premise of keeping the distribution characteristics of the original data, and effectively enlarges and processes the original sample quantity data.

The technical scheme of the invention is as follows: a method for expanding data while preserving distribution characteristics of an original sample, the method comprising:

step 1: acquiring microchip voltage data in an irradiation stress environment, and determining a data failure threshold value according to the electrical characteristics of the microchip in the irradiation stress environment;

step 2: screening the microchip voltage data under the irradiation stress environment obtained in the step 1 according to the data failure threshold determined in the step 1;

and step 3: and 2, obtaining a limited data sample set H after screening in the step 2, and expanding the sample data to N according to the range distribution characteristics of all limited data in the H.

Further, the specific method of step 3 is as follows:

step 3.1: and (3) arranging the data from small to large in a limited data sample set H, wherein the arranged set is as follows: h₀＝{a₀，a₁，...a_n，a_n+1| n > 0 }; wherein (a)₀，...a_n+1) Representing the subsample data;

step 3.2: set H according to step 3.1₀Is prepared from H₀Difference of neutron sample:

d₁＝a₁-a₀；d₂＝a₂-a₁；...d_n+1＝a_n+1-a_n

whereby d is_iCalculation formula, expressed as d_i＝a_i-a_i-1，1≤i≤n+1；

Step 3.3: according to the result of step 3.2, the probability P of falling into each range section is calculated_i：

P_i＝W_i/∑W

Wherein the content of the first and second substances,

∑W＝W₁+W₂+W₃+...+W_n+1，W_i＝D/d_i

wherein the content of the first and second substances,

D＝∑d＝d₁+d₂+d₃+...+d_n+1；

step 3.4: collecting the sample data according to the interval probability of the step 3.3₀Dividing into corresponding probability sample interval sets Q, Q { (0, P)₁)，(P₁，P₂)，(P₂，P₃)，...，(P_n，P_n+1)}；

Step 3.5: produced in the (0,1) rangeGenerating random number Y₁；

If Y ∈ (0, P)₁) Then at H₀Of (0, a)₁) Generating X in the sub-sample interval₁∈(0，a₁)；

If Y ∈ (P)₁，P₂) Then at H₀A (a) of₁，a₂) Generating X in the sub-sample interval₂∈(a₁，a₂)；

……

If Y ∈ (P)_n，P_n+1) Then at H₀A (a) of_n，a_n+1) Generating X in the sub-sample interval_N∈(a_n，a_n+1)；

The generation of the random number Y in the range of (0,1) is repeated according to the method of step 3.5₂，Y₃… …, until the sample data is expanded to N.

The invention has the beneficial effects that: the problems of small test data volume and few fitting processing methods of the micro-system power panel are solved. The method effectively reserves and extracts the distribution characteristics of the original data sample, fits the variation condition of the micro-system power panel under the nuclear irradiation stress, can be used as an effective and visual analysis basis for the variation of the micro-system power panel under the nuclear irradiation stress, and provides a statistical processing method better.

Drawings

FIG. 1 is a schematic diagram of the principle of screening and rejecting sample size according to the present invention.

FIG. 2 is a comparison of the original sample set and the expanded sample set according to the present invention.

FIG. 3 is a schematic view of step 3.4 of the present invention.

Fig. 4 is a schematic flow chart of the core algorithm principle of the present invention.

FIG. 5 is a diagram of the fitting results of the Matlab Toolbox software package of the present invention.

Detailed Description

Fig. 1 is a schematic diagram of a principle of screening and rejecting a sample size, in an embodiment, a micro power supply TPS54328DDA chip is selected as an object, the power supply outputs a voltage of 5V, and in a large amount of output voltages, a voltage value within a range where a variation range is more obvious is screened, and a result voltage value where other variations are not obvious is not used as processing data.

Through preprocessing, the voltage data table 1 of the micro power supply chip

Data in a range with obvious variation is selected through screening and elimination to serve as an original sample set H, after samples of a limited number of H sets are subjected to sampling screening, the samples are subjected to extended sampling to N groups according to the range distribution characteristics according to the algorithm provided by the invention, and the attached figure 4 is a flow schematic diagram of a core algorithm.

In the limited number of H groups of data, the data are arranged in order of magnitude, and the set after arrangement is H₀＝{a₀，a₁，...a_n，a_n+1|n＞0}，(a₀，...a_n+1) Representing the subsample data; dividing the subsamples into probability bins, and calculating the probability of each range bin according to steps S32 and S33 as P_i＝W_i/∑ W, wherein ∑ W ═ W₁+W₂+W₃+...+W_n+1∑ W ≈ 4062 in the present embodiment, the block interval probability may be divided into corresponding probability interval sample sets Q, Q { (0, P) according to the original data sample set₁)，(P₁，P₂)，(P₂，P₃)，...，(P_n，P_n+1) Where i is 5;

generating (m +1) random numbers Y within a range of (0,1) per probability subsample interval_m+1(m > 0), and judging Y_m+1Corresponding to (m +1) random numbers X generated in the original sample size interval H_m+1(m > 0), the method diagram is shown in FIG. 3, which shows the corresponding generation method diagram of the probability set and the original sample set.

The schematic diagram of the comparison between the expanded sample data size and the original sample set after processing and sample expansion is shown in fig. 2, and it is obvious that the data size greatly expands the number of sub-samples in the sample set under the processing based on the distribution characteristics of the original sample.

And (3) expanding a sampling result set N, performing degradation processing on all data subjected to re-expanding sampling, performing curve fitting in a Matlab Toolbox-based software package, wherein the result after fitting is shown in FIG. 5, and the curve result has a certain reference value for the research of the micro-system power panel in the nuclear radiation stress environment, so that the problem of small data volume is solved well, and the conventional sample volume expanding method is improved.

TABLE 1 preprocessed original sample data set

No. 1 board	No. 2 board	No. 3 board	No. 5 board	No. 6 board	No. 10 board
						4.973V	4.960V	4.992V	5.130V	5.027V	4.969V
4.976V	0.053V	4.992V	5.129V	5.045V	4.971V
						4.989V	0.048V	4.994V	5.139V	5.050V	4.976V
0.074V	0.052V	4.813V	5.156V	4.890V	0.230V
						0.045V	0.055V	4.790V	0.048V	4.882V	0.221V

Claims (1)

1. A method for expanding data while preserving distribution characteristics of an original sample, the method comprising:

step 1: acquiring microchip voltage data in an irradiation stress environment, and determining a data failure threshold value according to the electrical characteristics of the microchip in the irradiation stress environment;

step 2: screening the microchip voltage data under the irradiation stress environment obtained in the step 1 according to the data failure threshold determined in the step 1;

and step 3: obtaining a limited data sample set H after screening in the step 2, and expanding all limited data in the H to sample data N according to the range distribution characteristics;

the specific method of the step 3 comprises the following steps:

step 3.1: a limited set of data samples H, combining the dataArranging from small to large, wherein the set after arrangement is as follows: h₀＝{a₀,a₁,...a_n,a_n+1| n > 0 }; wherein (a)₀,...a_n+1) Representing the subsample data;

step 3.2: set H according to step 3.1₀Is prepared from H₀Difference of neutron sample:

d₁＝a₁-a₀；d₂＝a₂-a₁；...d_n+1＝a_n+1-a_n

whereby d is_iCalculation formula, expressed as d_i＝a_i-a_i-1，1≤i≤n+1；

Step 3.3: according to the result of step 3.2, the probability P of falling into each range section is calculated_i：

P_i＝W_i/∑W

Wherein the content of the first and second substances,

∑W＝W₁+W₂+W₃+...+W_n+1，W_i＝D/d_i

wherein the content of the first and second substances,

D＝∑d＝d₁+d₂+d₃+...+d_n+1；

step 3.4: collecting the sample data according to the interval probability of the step 3.3₀Dividing into corresponding probability sample interval sets Q, Q { (0, P)₁),(P₁,P₂),(P₂,P₃),...,(P_n,P_n+1)}；

Step 3.5: generating a random number Y in the range of (0,1)₁；

If Y ∈ (0, P)₁) Then at H₀Of (0, a)₁) Generating X in the sub-sample interval₁∈(0,a₁)；

If Y ∈ (P)₁,P₂) Then at H₀A (a) of₁,a₂) Generating X in the sub-sample interval₂∈(a₁,a₂)；

……

If Y ∈ (P)_n,P_n+1) Then at H₀A (a) of_n,a_n+1) Generating X in the sub-sample interval_N∈(a_n,a_n+1)；

The generation of the random number Y in the range of (0,1) is repeated according to the method of step 3.5₂，Y₃… …, until the sample data is expanded to N.

Images