CN102256269B - Detection-information-fusion-based wireless sensor network deterministic deployment method - Google Patents

Abstract

The invention discloses a kind of wireless sensor network certainty dispositions methods based on detection information fusion, detect Target Signal Strength x using sensor device, target can be detected by multiple sensor devices simultaneously; According to the Target Signal Strength and target signature parameter expression detected

Using minimum dispersion linear unbiased estimator method, the estimated value of target signature parameter θ is calculated

And parameter estimating error

Then, the largest interval distance dmax between adjacent sensors node is calculated; Finally, calculating the sensor node quantity for needing to dispose and specific placement location according to dmax. The present invention utilizes the Information Cooperation processing capacity between sensor node, based on parameter estimation theories, under the conditions of given detection probability, the sensor node quantity for detection can be largely reduced, improve detection efficiency, it reduces wireless sensor network and implements the cost effectively monitored, be suitable for the fields such as automation equipment in factory condition monitoring and fault diagnosis and urban traffic safety monitoring.

Description

A kind of wireless sensor network certainty dispositions method based on detecting information fusion

Technical field

The present invention relates to a kind of wireless sensor network disposition method, be specifically related to a kind of wireless sensor network certainty dispositions method based on detecting information fusion, can be applied to the fields such as battlefield surroundings monitoring, automation equipment in factory state-detection and diagnosis and urban traffic safety monitoring.

Background technology

Utilize wireless sensor network implementation environment parameter or equipment condition monitoring, in fields such as battlefield surroundings monitoring, automation equipment in factory detection and diagnosis, have a wide range of applications.In network, how sensor node is disposed, and is being related to accuracy and the monitoring efficiency of monitoring, is also being related to and is implementing cost and the cost that monitoring spends simultaneously, and therefore, effectively deployment way and method seem particularly important.

In prior art, wireless sensor network node is disposed and is comprised random placement mode and certainty deployment way, random placement is mainly used in that staff cannot approach or dangerous monitoring of environmental, such as virgin forest and nuclear radiation region etc., certainty deployment is mainly applicable to automatic factory and shipyard etc. and is convenient to the place that staff approaches and installs.At present, certainty deployment way has occupied leading role in wireless sensor network practical engineering application.

Wireless sensor network node is disposed, and need to guarantee that the target in monitored area can be monitored on the one hand, needs on the other hand to reduce the sensor node quantity of disposing, to reduce network monitor cost as far as possible.Traditional wireless sensor network disposition method, conventionally from physics angle of coverage, take geometrical analysis as means, determines and needs the sensor node quantity of deployment and concrete placement location.The method is not considered the Information Cooperation disposal ability between sensor node, belongs to conservative type method, based on the method, calculates and needs the sensor node quantity of deployment more, has increased the cost price of network monitor.In addition, owing to disposing the object of wireless sensor network, be to realize the Real-Time Monitoring of monitored area internal object, dispositions method based on geometrical analysis, only can determine sensor node quantity and concrete placement location, cannot be applied to analysis and judgement that whether target exists.

Dispose wireless sensor network node, another method is from information angle of coverage.Be different from physics and cover, information covers mainly from cooperative information process angle, by merging the detection information of different sensors node, improves the sensing range of sensor node, thereby reduces the sensor node quantity for monitoring, and reduces network monitor cost.The wireless sensor network node covering based on information is disposed, relate to sensor senses mode, the many aspects such as parameter Estimation and data fusion, the method proposing at present need to adopt the application of random placement mode mainly for forest fire monitoring and nuclear pollution area monitoring etc., for automation equipment monitoring and the monitoring of shipyard container etc., can carry out the application of certainty deployment, the sensor node quantity that how to confirm need to be disposed and the concrete placement location of sensor node, especially the spacing distance between how to confirm sensor node, still lack at present effective ways.

Summary of the invention

Goal of the invention of the present invention is to provide a kind of wireless sensor network certainty dispositions method, make full use of cooperative information disposal ability between sensor node, reduce the sensor node quantity that reaches the required deployment of given detection probability, thereby reduce the cost that utilizes wireless sensor network implementing monitoring.Another object of the present invention is that said method is detected and diagnosis for target monitoring and equipment fault.

To achieve the above object of the invention, the technical solution used in the present invention is: a kind of wireless sensor network certainty dispositions method based on detecting information fusion, first measure largest interval between adjacent sensors node apart from d _max, the more definite sensor node quantity of deployment and concrete placement location of needing; Wherein, the largest interval between mensuration adjacent sensors node is apart from d _maxmethod be:

(1) N transducer is set in the place that need to dispose transducer, utilizes sensor device to detect echo signal intensity x _k, k=1,2, L, N, wherein, k represents k transducer, and N represents the quantity of sensor node in network, and N is more than or equal to 2 integer;

In general, the deployment of transducer need to be according to square or triangular shaped deployment, for guaranteeing that target can be detected, need to guarantee the spacing distance between sensor node, the sensor node quantity of disposing is associated with needing the size in the region of monitoring, and those skilled in the art can select the size of N accordingly;

(2) according to target signature parameter expression

k=1,2, L, K, wherein, α represents signal attenuation coefficient, K represents to detect the sensor node quantity of information fusion, d _krepresent the spacing distance between target and k sensor node, n _krepresent the noise of introducing when k sensor node detects, utilize optimum linearity without bias estimation (Best Linear Unbiased Estimator, BLUE), calculate the estimated value of target signature parameter θ

and parameter estimating error

wherein, ${\widehat{\mathrm{\θ}}}_K=\frac{{\mathrm{\Σ}}_{k=1}^K{d}_k^{-\mathrm{\α}}{\mathrm{\σ}}_k^{-2}{x}_k}{{\mathrm{\Σ}}_{k=1}^K{d}_k^{-2\mathrm{\α}}{\mathrm{\σ}}_k^{-2}},$ ${\widetilde{\mathrm{\θ}}}_K=\frac{{\mathrm{\Σ}}_{k=1}^K{d}_k^{-\mathrm{\α}}{\mathrm{\σ}}_k^{-2}{n}_k}{{\mathrm{\Σ}}_{k=1}^K{d}_k^{-2\mathrm{\α}}{\mathrm{\σ}}_k^{-2}},$ σ _krepresent to measure the standard variance size in noise;

Now the situation with α=2 and K=2 is specifically described.α=2, show square being inversely proportional to of echo signal strength retrogression and distance, and K=2, shows to utilize the detection information of three sensor nodes to carry out estimating target characteristic parameter θ, now, and target signature estimates of parameters

parameter estimating error

wherein, x ₁be the signal strength signal intensity that the 1st transducer receives target, n ₁be the 1st noise signal strength that transducer receives, d ₁be the spacing distance between the 1st transducer and target, σ ₁be the standard variance of the 1st noise signal that transducer receives, x ₂be the signal strength signal intensity that the 2nd transducer receives target, n ₂be the 2nd noise signal strength that transducer receives, d ₂be the spacing distance between the 2nd transducer and target, σ ₂it is the standard variance of the 2nd noise signal that transducer receives.

(3) utilize the maximum boundary value A of minimum detection probability E and parameter estimating error, according to formula

largest interval between calculating adjacent sensors node is apart from d _max;

Now with α=2 and K=2, parameter estimating error meets Gaussian Profile and noise and meets Gauss and describe with distributing.Due to noise n _i, i=1,2, L, N meets Gauss with distributing, and can suppose noise n _iaverage E{n _i}=0, i=1,2, L, N, n _ivariance

because parameter estimating error meets Gaussian Profile, according to parameter Estimation expression formula

by calculating, can further obtain

average $\mathrm{\&mu;}=E\left\{\widetilde{\mathrm{\&theta;}}\right\}=E\left\{\frac{d_1^{-2}{\mathrm{\&sigma;}}_1^{-2}{\mathrm{<figure-callout\; id="1"\; label="n"\; filenames="110830093613.png,110830093632.png"\; state="\{\{state\}\}">n</figure-callout>}}_1+d_2^{-2}{\mathrm{\&sigma;}}_2^{-2}{n}_2}{d_1^{-4}{\mathrm{\&sigma;}}_1^{-2}+d_2^{-4}{\mathrm{\&sigma;}}_2^{-2}}\right\}=0,$

variance ${\widetilde{\mathrm{\&sigma;}}}^2\frac{1}{d_1^{-4}+d_2^{-4}}{\mathrm{\&sigma;}}^2.$ According to formula $p_r\left(\right|\widetilde{\mathrm{\&theta;}}|\&le;A)\&GreaterEqual;\mathrm{\&epsiv;},$ With season $Q\left(x\right)=\frac{1}{\sqrt{2\mathrm{\&pi;}}}{\&Integral;}_x^{\&infin;}\exp (-\frac{{\mathrm{<figure-callout\; id="2"\; label="t"\; filenames="110830093613.png,110830093620.png"\; state="\{\{state\}\}">t</figure-callout>}}^2}{2})\mathrm{dt},$ Can obtain $Q\left(\frac{A}{\stackrel{\&OverBar;}{\mathrm{\&sigma;}}}\right)\&le;\frac{1-\mathrm{\&epsiv;}}{2}.$ When A=σ, by calculating, can obtain:

when ε=95%, can further obtain:

by looking into gaussian distribution table, can obtain: this constraints all needs to meet to any point in monitored area, and in conjunction with concrete deployment way, such as triangle is disposed, the largest interval between expression formula calculating and definite two sensor nodes is apart from d thus _max.

The method of the sensor node quantity that definite needs are disposed and concrete placement location is:

First by parallel lines, monitored area is divided into strip region, the distance between adjacent two parallel lines

by 60 °, 90 ° or 120 ° of all parallel lines rotations, in monitored area, meet at respectively a s with former each parallel lines ₁, s ₂, L, s _n, N is for needing the quantity of the sensor node of deployment, intersection point s ₁, s ₂, L, s _nit is exactly the concrete placement location of node in sensor network.

Generally can adopt and turn clockwise, when rotation 60 ° or 120 °, for equilateral triangle is disposed, when half-twist, be that square is disposed.

The application of above-mentioned wireless sensor network node certainty dispositions method in target monitoring and equipment fault detection and diagnosis, in environment to be monitored or equipment to be detected, adopt said method to determine the concrete placement location of sensor node, utilize transducer to detect echo signal intensity x, calculate the estimated value of target signature parameter θ simultaneously

and parameter estimating error

if

when

time, judge that target exists or equipment fault occurs, otherwise, when

time, judge that target does not exist or equipment fault-free occurs, wherein, θ ₀for the judgment threshold of setting in advance.

Because technique scheme is used, the present invention compared with prior art has following advantages:

1. the present invention is owing to having utilized cooperative information disposal ability between sensor node, expanded the sensing range of sensor node, at cooperative information, process on basis and calculate the spacing distance between sensor node, can reduce in large quantities the quantity of the sensor node for monitoring, thereby reduce the cost of network monitor.

2. owing to having adopted optimum linearity without bias estimation (Best Linear Unbiased Estimator, BLUE), estimate target signature parameter, therefore to meeting the monitoring system of linear conditions, can obtain optimal estimation value, accuracy in detection is higher, and detection efficiency is also higher.

3. the present invention is owing to adopting detection information fusion mode to carry out network node deployment, after the concrete placement location of calculating and definite sensor node, the judgement that the present invention can directly utilize target signature parameter Estimation result to realize whether monitored area internal object is existed is identified with automatic, reaches the object of wireless sensor network Real-Time Monitoring.

Accompanying drawing explanation

Fig. 1 is equilateral triangular portion management side formula schematic diagram in embodiment mono-;

Fig. 2 is three kinds of situations of equilateral triangular portion management side formula in embodiment mono-;

Fig. 3 is the geometrical analysis schematic diagram of equilateral triangular portion management side formula in embodiment mono-;

Fig. 4 is detection probability P in embodiment mono- _tand detect the variation relation between the sensor node quantity n merging;

Fig. 5 be in embodiment mono-in equilateral triangular portion management side formula detection probability during Case 2 distribute;

Fig. 6 is square deployment way schematic diagram in embodiment bis-;

Fig. 7 is three kinds of situations of square deployment way in embodiment bis-;

Fig. 8 is the geometrical analysis schematic diagram of square deployment way in embodiment bis-;

Fig. 9 is detection probability P in embodiment bis- _sand detect the variation relation between the sensor node quantity n merging;

Figure 10 be in embodiment in square deployment way detection probability during Case 2 distribute.

Embodiment

Below in conjunction with drawings and Examples, the invention will be further described:

Embodiment mono-: equilateral triangle deployment way

In wireless sensor network certainty deployment strategy, equilateral triangle deployment way is conventional deployment way.So-called equilateral triangle deployment way, refers to that sensor node arranges in equilateral triangle mode, as shown in Figure 1.In Fig. 1, the home position of three circles is the deployed position of sensor node, the sensing range of sensor node is to take the circle that radius is r centered by sensor node, if do not consider cooperative information disposal ability between sensor node (being that physics covers), when the distance d between target and at least one sensor node is not more than r, target just can be detected, otherwise can not be detected.In Fig. 1, border circular areas is the sensing range of single-sensor node, and the black region between circle is for processing the region that can perceive by cooperative information between sensor node.Adopt equilateral triangle deployment way, conventionally have three kinds of situations, as shown in Figure 2.Case1 represents that the sensing range of sensor node exists common factor situation, and the sensing range that Case2 represents sensor node is tangent situation just, and Case3 represents that the sensing range of sensor node does not exist common factor situation.Along with the spacing distance between sensor node becomes large, need to adopt cooperative information to process detection order target area and become large, as horizontal line area part in Fig. 2.

First the present embodiment utilizes sensor device to detect echo signal intensity x, then by optimum linearity, without bias estimation, calculates target signature parameter

and parameter estimating error

again according to parameter estimating error and in advance the largest interval between given minimum detection probability calculation sensor node apart from d _max, finally according to d _maxdetermine and need the sensor node quantity of deployment and concrete placement location.For equilateral triangular portion management side formula, specific implementation process is as follows:

First suppose that it is x that t moment k sensor node detects echo signal intensity _k(t), k=1,2, L, K, wherein K represents to detect the sensor node quantity of information fusion, and for simplicity, order is t=0 constantly.

Suppose that signal attenuation model expression formula is

k=1,2, L, N, wherein, α represents signal attenuation coefficient, N represents sensor node quantity, d _krepresent the spacing distance between target and k sensor node, n _krepresent to measure noise, θ represents target signature parameter, can judge whether target exists.For simplicity, suppose α=2, represent that signal is inversely proportional to square distance, measure noise n _kfor standardized normal distribution, average is μ=0, and variance is σ ²=1.

When detecting the sensor node quantity K=2 of information fusion, utilize optimum linearity without bias estimation, calculate the estimated value of target signature parameter θ and parameter estimating error

can obtain following result:

${\widehat{\mathrm{\&theta;}}}_2=\frac{d_1^{-2}{\mathrm{\&sigma;}}_1^{-2}{x}_1+d_2^{-2}{\mathrm{\&sigma;}}_2^{-2}{x}_2}{d_1^{-4}{\mathrm{\&sigma;}}_1^{-2}+d_1^{-4}{\mathrm{\&sigma;}}_1^{-2}}---\left(1\right)$

${\widetilde{\mathrm{\&theta;}}}_2=\frac{d_1^{-2}{\mathrm{\&sigma;}}_1^{-2}{\mathrm{<figure-callout\; id="1"\; label="n"\; filenames="110830093613.png,110830093632.png"\; state="\{\{state\}\}">n</figure-callout>}}_1+d_2^{-2}{\mathrm{\&sigma;}}_2^{-2}{n}_2}{d_1^{-4}{\mathrm{\&sigma;}}_1^{-2}+d_2^{-4}{\mathrm{\&sigma;}}_2^{-2}}---\left(2\right)$

Due to therefore the above results can be further converted to:

${\widehat{\mathrm{\&theta;}}}_2=\frac{d_1^{-2}{x}_1+d_2^{-2}{x}_2}{d_1^{-4}+d_1^{-4}}---\left(3\right)$

${\widetilde{\mathrm{\&theta;}}}_2=\frac{d_1^{-2}{\mathrm{<figure-callout\; id="1"\; label="n"\; filenames="110830093613.png,110830093632.png"\; state="\{\{state\}\}">n</figure-callout>}}_1+d_2^{-2}{n}_2}{d_1^{-4}+d_2^{-4}}---\left(4\right)$

For making the estimated value of target signature parameter θ

there is confidence level, must make

evaluated error

meet some requirements.By (4), can be calculated

average

and variance result is as follows:

$\widetilde{\mathrm{\&mu;}}=E\left\{{\widetilde{\mathrm{\&theta;}}}_2\right\}=E\left\{\frac{d_1^{-2}{\mathrm{<figure-callout\; id="1"\; label="n"\; filenames="110830093613.png,110830093632.png"\; state="\{\{state\}\}">n</figure-callout>}}_1+d_2^{-2}{n}_2}{d_1^{-4}+d_2^{-4}}\right\}=0---\left(5\right)$

${\widetilde{\mathrm{\&sigma;}}}^2=E\left\{{\widetilde{\mathrm{\&theta;}}}_2^2\right\}=E\left\{{\left(\frac{d_1^{-2}{\mathrm{<figure-callout\; id="1"\; label="n"\; filenames="110830093613.png,110830093632.png"\; state="\{\{state\}\}">n</figure-callout>}}_1+d_2^{-2}{n}_2}{d_1^{-4}+d_2^{-4}}\right)}^2\right\}=\frac{1}{d_1^{-4}+d_2^{-4}}---\left(6\right)$

For any point in monitored area, all need to meet minimum detection Probability Condition, i.e. parameter estimating error

the probability that is less than or equal to certain given constant A is more than or equal to certain given constant ε in advance, and mathematic(al) representation is as follows:

$p_r\left(\right|\widetilde{\mathrm{\&theta;}}|\&le;A)\&GreaterEqual;\mathrm{\&epsiv;}---\left(7\right)$

Wherein, ε represents minimum detection probability, and A represents parameter estimating error

the maximum boundary value allowing.Suppose

meeting average is with variance be gaussian Profile, order $Q\left(x\right)=\frac{1}{\sqrt{2\mathrm{\&pi;}}}{\&Integral;}_x^{\&infin;}\exp (-\frac{{\mathrm{<figure-callout\; id="2"\; label="t"\; filenames="110830093613.png,110830093620.png"\; state="\{\{state\}\}">t</figure-callout>}}^2}{2})\mathrm{dt},$ By (7), can be obtained:

$1-2Q\left(\frac{A}{\widetilde{\mathrm{\&sigma;}}}\right)\&GreaterEqual;\mathrm{\&epsiv;}---\left(8\right)$

After abbreviation, obtain:

$Q\left(\frac{A}{\widetilde{\mathrm{\&sigma;}}}\right)\&le;\frac{1-\mathrm{\&epsiv;}}{2}---\left(9\right)$

When A=σ, by calculating, can obtain:

when ε=95%, can further obtain:

by looking into gaussian distribution table, can obtain:

this constraints all needs to meet to any point in monitored area, to equilateral triangular portion management side formula, and can be by the largest interval between two sensor nodes of calculative determination apart from d _max, computational process is as follows:

As shown in Figure 3, A, B, the deployed position that C is sensor node, with A, B, the triangle that C is summit is equilateral triangle, the length of side of supposing equilateral triangle is d, i.e. spacing distance between sensor node.For the inner any point D of equilateral triangle ABC, all need the detection probability that meets this point to be not less than prior set-point.When K=2, adopt the detection information of two sensor nodes to merge, because parameter estimating error and target and the internodal distance of detecting sensor are inversely proportional to, for improving, check accuracy, need to utilize two nearest sensor node detection information of distance objective to merge.Without loss of generality, as shown in Figure 3, utilize the detection information of 2 of B and C to merge, as can be known from the above analysis, the inner any point of △ ABC, all will meet following condition:

$Q\left(\sqrt{d_1^{-4}+d_2^{-4}}\right)\&le;\frac{1-\mathrm{\&epsiv;}}{2}---\left(10\right)$

When ε=95%, by looking into normal distyribution function table, can obtain:

$\sqrt{d_1^{-4}+d_2^{-4}}\&GreaterEqual;1.96---\left(11\right)$

Utilize optimal method, can know, when

time,

obtain minimum value, now, the spacing distance d between sensor node obtains maximum d _max=3.03, when the spacing distance between two sensor nodes is greater than 3.03, cannot guarantee that in monitored area, arbitrfary point all can accurately be detected.Now, three perception radiuses are 1, the sensor node of disposing in equilateral triangle mode, if do not adopt detection information fusion, the region area size that can monitor is about S=9.42, and adopt two sensor detection information amalgamation modes, the region area size that can monitor is about S=11.75, monitored area size has approximately improved 24.7%, as can be seen here, utilization is carried out sensor network nodes deployment based on detecting information fusion mode, can effectively improve surveyed area, for given monitored area, can effectively reduce the sensor node quantity that needs deployment, thereby reduce network monitor cost.

When K=3, adopt the detection information of three sensor nodes to merge, utilize the analysis similar to K=2, by calculating the largest interval can obtain between sensor node apart from d _max=3.26, adopt three sensor detection information amalgamation modes, the region area size that can monitor is about S=12.46, and monitored area size has approximately improved 32.3%.In equilateral triangle deployment way, when K=2, monitored area size has approximately improved 24.7%, and when K=3, monitored area size has approximately improved 32.3%, compare two kinds of situations, can find: by increase, detect the quantity of the sensor node of information fusion, can increase monitored area size, for given monitored area, can reduce and reach the sensor node quantity that given detection probability need to be disposed, thereby reduce the cost of monitoring.

Increase the sensor node quantity that detects information fusion, can improve detection probability, but not linear between the two, as shown in Figure 4.Owing to detecting information fusion, first need exchange data information between sensor node, need to spend communication cost, in Practical Project, selecting the detection information of 2～4 sensor nodes to merge is reasonable selection.

In equilateral triangle deployment way, in monitored area, the detection probability of diverse location is also different, detection probability distribution situation when Fig. 5 represents Case2 in equilateral triangle deployment way, therefrom can see, the center of equilateral triangle is the position of detection probability minimum.

Embodiment bis-: square deployment way

Similar to equilateral triangle deployment way, square deployment way requires sensor node to arrange with square mode, as shown in Figure 6.Three kinds of situations of the same existence of square deployment way, as shown in Figure 7.Case1 represents that the sensing range of sensor node exists common factor situation, and the sensing range that Case2 represents sensor node is tangent situation just, and Case3 represents that the sensing range of sensor node does not exist common factor situation.Along with the spacing distance between sensor node becomes large, need to adopt cooperative information to process detection order target area and become large, as horizontal line area part in Fig. 7.

When detecting the sensor node quantity K=2 of information fusion, utilize optimum linearity without bias estimation, calculate the estimated value of target signature parameter θ

and parameter estimating error

can obtain following result:

${\widehat{\mathrm{\&theta;}}}_2=\frac{d_1^{-2}{\mathrm{\&sigma;}}_1^{-2}{\mathrm{<figure-callout\; id="1"\; label="x"\; filenames="110830093613.png,110830093632.png"\; state="\{\{state\}\}">x</figure-callout>}}_1+d_2^{-2}{\mathrm{\&sigma;}}_2^{-2}{x}_2}{d_1^{-4}{\mathrm{\&sigma;}}_1^{-2}+d_1^{-4}{\mathrm{\&sigma;}}_1^{-2}}---\left(12\right)$

${\widetilde{\mathrm{\&theta;}}}_2=\frac{d_1^{-2}{\mathrm{\&sigma;}}_1^{-2}{\mathrm{<figure-callout\; id="1"\; label="n"\; filenames="110830093613.png,110830093632.png"\; state="\{\{state\}\}">n</figure-callout>}}_1+d_2^{-2}{\mathrm{\&sigma;}}_2^{-2}{n}_2}{d_1^{-4}{\mathrm{\&sigma;}}_1^{-2}+d_2^{-4}{\mathrm{\&sigma;}}_2^{-2}}---\left(13\right)$

Due to

therefore the above results can be further converted to:

${\widehat{\mathrm{\&theta;}}}_2=\frac{d_1^{-2}{\mathrm{<figure-callout\; id="1"\; label="x"\; filenames="110830093613.png,110830093632.png"\; state="\{\{state\}\}">x</figure-callout>}}_1+d_2^{-2}{x}_2}{d_1^{-4}+d_1^{-4}}---\left(14\right)$

${\widetilde{\mathrm{\&theta;}}}_2=\frac{d_1^{-2}{\mathrm{<figure-callout\; id="1"\; label="n"\; filenames="110830093613.png,110830093632.png"\; state="\{\{state\}\}">n</figure-callout>}}_1+d_2^{-2}{n}_2}{d_1^{-4}+d_2^{-4}}---\left(15\right)$

For making the estimated value of target signature parameter θ

there is confidence level, must make

evaluated error meet some requirements.By (4), can be calculated average and variance

result is as follows:

$\widetilde{\mathrm{\&mu;}}=E\left\{{\widetilde{\mathrm{\&theta;}}}_2\right\}=E\left\{\frac{d_1^{-2}{\mathrm{<figure-callout\; id="1"\; label="n"\; filenames="110830093613.png,110830093632.png"\; state="\{\{state\}\}">n</figure-callout>}}_1+d_2^{-2}{n}_2}{d_1^{-4}+d_2^{-4}}\right\}=0---\left(16\right)$

${\widetilde{\mathrm{\&sigma;}}}^2=E\left\{{\widetilde{\mathrm{\&theta;}}}_2^2\right\}=E\left\{{\left(\frac{d_1^{-2}{\mathrm{<figure-callout\; id="1"\; label="n"\; filenames="110830093613.png,110830093632.png"\; state="\{\{state\}\}">n</figure-callout>}}_1+d_2^{-2}{n}_2}{d_1^{-4}+d_2^{-4}}\right)}^2\right\}=\frac{1}{d_1^{-4}+d_2^{-4}}---\left(17\right)$

For any point in monitored area, all need to meet minimum detection Probability Condition, i.e. parameter estimating error

the probability that is less than or equal to certain given constant A is more than or equal to certain given constant ε in advance, and mathematic(al) representation is as follows:

$p_r\left(\right|\widetilde{\mathrm{\&theta;}}|\&le;A)\&GreaterEqual;\mathrm{\&epsiv;}---\left(18\right)$

Wherein, ε represents minimum detection probability, and A represents parameter estimating error

the maximum boundary value allowing.Suppose meeting average is

with variance be gaussian Profile, order $Q\left(x\right)=\frac{1}{\sqrt{2\mathrm{\&pi;}}}{\&Integral;}_x^{\&infin;}\exp (-\frac{{\mathrm{<figure-callout\; id="2"\; label="t"\; filenames="110830093613.png,110830093620.png"\; state="\{\{state\}\}">t</figure-callout>}}^2}{2})\mathrm{dt},$ By (7), can be obtained:

$1-2Q\left(\frac{A}{\widetilde{\mathrm{\&sigma;}}}\right)\&GreaterEqual;\mathrm{\&epsiv;}---\left(19\right)$

After abbreviation, obtain:

$Q\left(\frac{A}{\widetilde{\mathrm{\&sigma;}}}\right)\&le;\frac{1-\mathrm{\&epsiv;}}{2}---\left(20\right)$

When A=σ, by calculating, can obtain:

when ε=95%, can further obtain:

by looking into gaussian distribution table, can obtain:

this constraints all needs to meet to any point in monitored area, aligns square deployment way, can be by the largest interval between two sensor nodes of calculative determination apart from d _max, computational process is as follows:

As shown in Figure 8, A, B, C, the deployed position that D is sensor node, with A, B, C, D is that summit forms square, supposes that the foursquare length of side is d, i.e. spacing distance between sensor node.For the inner any point E of square ABCD, all need the detection probability that meets this point to be not less than prior set-point.When K=2, adopt the detection information of two sensor nodes to merge, because parameter estimating error and target and the internodal distance of detecting sensor are inversely proportional to, for improving, check accuracy, need to utilize two nearest sensor node detection information of distance objective to merge.Without loss of generality, as shown in Figure 8, utilize the detection information of 2 of C and D to merge, as can be known from the above analysis, the inner any point of △ OCD, all will meet following condition:

$Q\left(\sqrt{d_1^{-4}+d_2^{-4}}\right)\&le;\frac{1-\mathrm{\&epsiv;}}{2}---\left(21\right)$

When ε=95%, by looking into normal distyribution function table, can obtain:

$\sqrt{d_1^{-4}+d_2^{-4}}\&GreaterEqual;1.96---\left(22\right)$

Utilize optimal method, can know, when

time,

obtain minimum value, now, the spacing distance d between sensor node obtains maximum d _max=2.20, when the spacing distance between two sensor nodes is greater than 2.20, cannot guarantee that in monitored area, arbitrfary point all can accurately be detected.Now, three perception radiuses are 1, the sensor node of disposing in equilateral triangle mode, if do not adopt detection information fusion, the region area size that four sensor nodes can be monitored is about S=12.56, and adopt two sensor detection information amalgamation modes, the region area size that can monitor is about S=17.40, monitored area size has approximately improved 38.5%, as can be seen here, utilization is carried out sensor network nodes deployment based on detecting information fusion mode, can effectively improve surveyed area, for given monitored area, can effectively reduce the sensor node quantity that needs deployment, thereby reduce network monitor cost.

When K=3, adopt the detection information of three sensor nodes to merge, utilize the analysis similar to K=2, by calculating the largest interval can obtain between sensor node apart from d _max=2.4, adopt three sensor detection information amalgamation modes, the region area size that can monitor is about S=12.56, and monitored area size has approximately improved 45.8%.In square deployment way, when K=2, monitored area size has approximately improved 38.5%, and when K=3, monitored area size has approximately improved 45.8%, compare two kinds of situations, can find: by increase, detect the quantity of the sensor node of information fusion, can increase monitored area size, for given monitored area, can reduce and reach the sensor node quantity that given detection probability need to be disposed, thereby reduce the cost of monitoring.

Increase the sensor node quantity that detects information fusion, can improve detection probability, but not linear between the two, as shown in Figure 9.Owing to detecting information fusion, first need exchange data information between sensor node, need to spend communication cost, in Practical Project, selecting the detection information of 2～4 sensor nodes to merge is reasonable selection.

In square deployment way, in monitored area, the detection probability of diverse location is also different, and detection probability distribution situation when Figure 10 represents Case 2 in square deployment way therefrom can see, foursquare center is the position of detection probability minimum.

Claims (2)

1. the wireless sensor network certainty dispositions method based on detecting information fusion, is characterized in that: first measure the largest interval distance between adjacent sensors node

, the more definite sensor node quantity of deployment and concrete placement location of needing; Wherein, measure the largest interval distance between adjacent sensors node

method be:

(1) N transducer is set in the place that need to dispose transducer, utilizes sensor device to detect echo signal intensity

,

, wherein, k represents k transducer, and N represents the quantity of sensor node in network, and N is more than or equal to 2 integer;

(2) according to target signature parameter expression

,

, wherein,

represent signal attenuation coefficient, K represents to detect the sensor node quantity of information fusion, d _krepresent the spacing distance between target and k sensor node, n _krepresent the noise of introducing when k sensor node detects, utilize optimum linearity without bias estimation, calculate target signature parameter

estimated value

and parameter estimating error

, wherein,

, , σ _krepresent to measure the standard variance size in noise;

(3) utilize minimum detection probability

with the maximum boundary value A of parameter estimating error, according to formula

, calculate the largest interval distance between adjacent sensors node

;

The method of the sensor node quantity that definite needs are disposed and concrete placement location is:

First by parallel lines, monitored area is divided into strip region, the distance between adjacent two parallel lines

, by 60 °, 90 ° or 120 ° of all parallel lines rotations, in monitored area, meet at a little respectively with former each parallel lines , N is for needing the quantity of the sensor node of deployment, intersection point it is exactly the concrete placement location of node in sensor network.

2. the application of wireless sensor network node certainty dispositions method in target monitoring and equipment fault detection and diagnosis described in claim 1, it is characterized in that: in environment to be monitored or equipment to be detected, described in employing claim 1, method is determined the concrete placement location of sensor node, utilize transducer to detect echo signal intensity x, calculate target signature parameter simultaneously

estimated value

and parameter estimating error

if,

, when

time, judge that target exists or equipment fault occurs, otherwise, when

time, judge that target does not exist or equipment fault-free occurs, wherein,

for the judgment threshold of setting in advance.

Images