CN105806400A - Intelligent method and system for monitoring hydrant's safety state - Google Patents

Abstract

The invention discloses an intelligent method and a system for monitoring a hydrant's safety state. The method is performed through the following steps: 101) acquiring the measuring information of all sensors and conducting time synchronization processing to all acquired measuring information, wherein the measuring information includes water pressure inside a hydrant well, the inclination angle of a well cover, or environmental brightness inside the well; 102) adopting principal component analysis to degrade the dimensions of the measuring information after time synchronization and extracting the characteristic vectors of the dimension degraded space in the same process, according to the values of the characteristic vectors, picking a plurality of the characteristic vectors as the characteristic vectors for current hydrant's safety state judgment; 103) utilizing historical data to learn the relationship between the characteristic vectors and the safety state so as to establish a machine learning model; and 104) based on the characteristic vectors processed in step 102) and by combining with the SVM recognition function built in step 103), identifying the current safety sate of the hydrant.

Description

A kind of fire hydrant safe condition intelligent monitoring method and system

Technical field

The invention belongs to information technology, automatically control and electronic technology field, it is specifically related to realize the safe condition of current fire hydrant is judged by structure fire hydrant safe condition intelligent recognition model, thus Added Management personnel carry out scientificlly and effectively supervision and safeguard, it is ensured that the availability of fire-fighting system under emergency rating.

Background technology

Municipal fire bolt is the important component part of urban fire control system, bears and provides water source to meet the important mission of delivery of water for fire-fighting lance and fire fighting truck rapidly.But at present to the main mode adopting regular visit of the supervision of fire-fighting system, lack the regulatory measure of scientific and efficient, the phenomenons such as fire fighting device vitals is stolen, fire protection pipeline damaged leaks, fire fighting hydraulic pressure is abnormal happen occasionally, and greatly reduce the relief efficiency under the anomalous events such as fire.Accurately judge the safe condition that fire hydrant is current, find the various security risk hidden danger that fire hydrant exists in time, to ensureing that people's security of the lives and property is significant.Meanwhile, accurately identify the safe condition of current fire hydrant, and carry out personnel according to safe class and patrol and examine scheduling etc., to the supervision level improving fire-fighting system, improve work efficiency also significant.

At present differentiating of the safe condition of fire-fighting system is judged mainly through the mode setting simple threshold, exist safe condition identification fail to report, rate of false alarm high, safe condition divides the problems such as classification is coarse, how to carry out safe condition identification accurately, the intelligent regulatory level improving fire hydrant is significant.

Summary of the invention

Embodiments provide a kind of fire hydrant safe condition intelligent monitoring method and system, the mapping relations between input feature vector vector fire hydrant safe condition in a variety of situations can be taken into full account in model training process and then improve the precision of monitoring state, it is possible to meeting the index request failed to report and report by mistake in fire hydrant safe condition alarm procedure.

In order to overcome above-mentioned difficult point, the embodiment of the invention discloses following technical scheme:

First aspect, it is provided that a kind of fire hydrant safe condition intelligent monitoring method, described method comprises:

Step 101) gather the metrical information of each sensor, all metrical informations collected are carried out time synchronizing；

Wherein, described metrical information includes: hydraulic pressure, well lid inclination angle or well environment brightness in fire control well；

Step 102) adopt principal component analysis that the metrical information after synchronization process is carried out dimension-reduction treatment, extract the characteristic vector of dimension reduction space simultaneously；

Size according to characteristic vector character pair value, chooses several characteristic vectors as the current characteristic vector that fire hydrant safe condition is differentiated；

Step 103) utilize historical data that the relation between characteristic vector and safe condition is learnt, set up machine learning model；

In principal component analysis process, the characteristic vector of dimension reduction space is as training characteristics collection, using corresponding safe condition as output object set, adopts SVM machine learning algorithm to obtain SVM recognition function；

Step 104) based on step 102) in process after characteristic vector, integrating step 103) SVM that sets up identifies equation, identifies the safe condition of current fire hydrant.

In conjunction with above-mentioned first aspect, in the implementation that the first is possible, described step 102) comprise further:

Step 102-1) each sensor information in fire control well is acquired, it is assumed that form P dimension and gather data vector；Wherein, the information of collection includes hydraulic pressure in fire control well, well lid inclination angle or well environment brightness；

Step 102-2) for the data acquisition in certain period of time K time, and construct K × P dimension acquisition matrix U, it is designated as:

$U=\left[\begin{array}{cccc}{u}_{11}& u_{12}& \mathrm{...}& u_{1P}\\ u_{21}& u_{22}& \mathrm{...}& u_{2P}\\ .& .& .& .\\ .& .& .& .\\ .& .& .& .\\ u_{K1}& u_{K2}& \mathrm{...}& u_{KP}\end{array}\right]$

Step 102-3) matrix U asked its covariance matrix S, and the eigenvalue of solution matrix S and characteristic vector；

After solving obtain eigenvalue and especially vector be respectively labeled as: λ₁,λ₂,…,λ_P, and, η₁,η₂,…,η_P；

Step 2-4) by the characteristic vector of front m maximum eigenvalue characteristic of correspondence vector combination composition safe condition identification；

Wherein, m is the minima that eigenvalue sum is corresponding more than 80% with the ratio of total characteristic value.

In conjunction with above-mentioned first aspect, with in the implementation that the first is possible, in the implementation that the second is possible, described step 103) comprise further:

Step 103-1) it is some classes by fire hydrant state demarcation, based on historical data, SVM model is trained；Wherein, described fire hydrant state categories comprises: fire manhole cover accounts for pressure, fire manhole cover moves, fire manhole cover abnormal start-up, fire protection pipeline breakage are leaked or fire fighting hydraulic pressure is abnormal；

The input feature vector collection of SVM model: X=[x₁,x₂,…,x_N]；Wherein x_i, i=1,2 ..., N is i-th input feature vector vector；

The output object set of SVM model: f=[y₁,y₂,…,y_N]；Wherein y_i, i=1,2 ..., N is the fire hydrant safe condition that i-th input feature vector vector is corresponding；

Wherein: $x_i=\left[\begin{array}{c}{\mathrm{\η}}_1\\ {\mathrm{\η}}_2\\ .\\ .\\ .\\ {\mathrm{\η}}_m\end{array}\right],i=1,2,...,N;$

Step 103-2) adopt step 103-1) obtained training sample set pair support vector machine is trained, utilize input feature vector collection and output object set that training sample set provides, SVM is trained as follows, obtains the SVM recognition function for judging safe condition:

Step 103-2-1) utilize Novel Algorithm, solve optimization problem:

$\underset{\mathrm{\α},{\mathrm{\α}}^{*}}{\max }\{-\mathrm{\ϵ}\underset{m=1}{\overset{N}{\Σ}}({{\mathrm{\α}}_m}^{*}+{\mathrm{\α}}_m)+\underset{m=1}{\overset{N}{\Σ}}{f}_m({{\mathrm{\α}}_m}^{*}-{\mathrm{\α}}_m)-\frac{1}{2}\underset{m=1}{\overset{N}{\Σ}}\underset{n=1}{\overset{N}{\Σ}}{({{\mathrm{\α}}_m}^{*}-{\mathrm{\α}}_m)}^{T}K(X_m,X_n)({{\mathrm{\α}}_n}^{*}-{\mathrm{\α}}_n)\}$

s.t.

$\underset{m=1}{\overset{N}{\Σ}}({\mathrm{\α}}_m-{{\mathrm{\α}}_m}^{*})=0,0\≤{\mathrm{\α}}_m,{{\mathrm{\α}}_m}^{*}\≤C,m=1,2,...,N$

Wherein, ε is a given parameter value, α_m ^*、α_mFor the parameter that training is asked for, f_mFor the m-th target output value of training output collection, X_mFor the m-th sample of input feature vector collection, K (X_m,X_n) for RBF kernel function, its form is:

$K(X_m,X_n)=\exp (-\frac{\left|\right|X_m-X_n|{|}^2}{2{\mathrm{\γ}}^2})$

Wherein, γ is gaussian kernel function width parameter；

Step 103-2-2) trained after, setting up SVM function is:

${\hat{y}}_i\left(x\right)=\underset{m=1}{\overset{N}{\Σ}}({{\mathrm{\α}}_m}^{*}-{\mathrm{\α}}_m)K(X_m,X)+b$

Wherein, b is the threshold value that training obtains, α_m ^*And α_mFor the parameter that training is asked for；X is given sample input feature vector vector to be predicted；Output valve for recognition function.

Second aspect, it is provided that a kind of fire hydrant safe condition system, described system comprises:

Fire hydrant monitoring terminal, for gathering the metrical information of each sensor, and carries out synchronization process to the information gathered；Wherein, described status information comprises: hydraulic pressure, well lid inclination angle or well environment brightness in fire control well；

State decision process device is fastened in fire-fighting, is used for:

Adopting principal component analysis technology, the metrical information received is carried out dimension-reduction treatment, extracts the characteristic vector of dimension reduction space simultaneously, the size according to characteristic vector character pair value, selected part characteristic vector is as the current characteristic vector that fire hydrant safe condition is differentiated；

Utilize historical data that the relation between characteristic vector and safe condition is learnt, set up machine learning model；In principal component analysis process, the characteristic vector of dimension reduction space is as training characteristics collection, using corresponding safe condition as output object set, adopts SVM machine learning algorithm to obtain SVM recognition function；

Based on described characteristic vector, identify equation in conjunction with the SVM set up, identify the safe condition of current fire hydrant.

In conjunction with above-mentioned second aspect, in the implementation that the first is possible, described fire-fighting is fastened state decision process device and is comprised further:

First processing module, for for the measurement data acquisition in certain period of time K time, structure K × P dimension acquisition matrix U, being designated as:

$U=\left[\begin{array}{cccc}{u}_{11}& u_{12}& \mathrm{...}& u_{1P}\\ u_{21}& u_{22}& \mathrm{...}& u_{2P}\\ .& .& .& .\\ .& .& .& .\\ .& .& .& .\\ u_{K1}& u_{K2}& \mathrm{...}& u_{KP}\end{array}\right]$

Matrix U is asked its covariance matrix S, and the eigenvalue of solution matrix S and characteristic vector；

The eigenvalue obtained after solving is: λ₁,λ₂,…,λ_P, the characteristic vector obtained is: η₁,η₂,…,η_P

Second processing module, constitutes the characteristic vector of Subsequent secure state recognition for the eigenvalue characteristic of correspondence vector combination maximum by front m；

It is designated as: $x=\left[\begin{array}{c}{\mathrm{\η}}_1\\ {\mathrm{\η}}_2\\ .\\ .\\ .\\ {\mathrm{\η}}_m\end{array}\right].$

In conjunction with above-mentioned second aspect, and/or the first possible implementation, in the implementation that the second is possible, described fire-fighting is fastened state decision process device and is also comprised:

3rd processing module, for by the state demarcation of fire hydrant being: various states such as fire manhole cover accounts for pressure, fire manhole cover moves, fire manhole cover abnormal start-up, fire protection pipeline breakage are leaked, fire fighting hydraulic pressure is abnormal, is trained SVM model based on historical data；

Input feature vector collection: X=[x₁,x₂,…,x_N]；Wherein x_i, i=1,2 ..., N is i-th input feature vector vector；

Output object set: f=[y₁,y₂,…,y_N]；Wherein y_i, i=1,2 ..., N is the fire hydrant safe condition that i-th input feature vector vector is corresponding；

Wherein: $x_i=\left[\begin{array}{c}{\mathrm{\η}}_1\\ {\mathrm{\η}}_2\\ .\\ .\\ .\\ {\mathrm{\η}}_m\end{array}\right],i=1,2,...,N$

Fourth processing module, is trained for the training sample set pair support vector machine that adopts the first processing module obtained, utilizes input feature vector collection and output object set that training sample set provides, to SVM by being trained and then obtaining SVM recognition function.

In conjunction with above-mentioned second aspect, and the implementation that the second is possible, in the implementation that the third is possible, described fourth processing module comprises further:

Processing unit, is used for utilizing Novel Algorithm, solves optimization problem:

$\underset{\mathrm{\α},{\mathrm{\α}}^{*}}{\max }\{-\mathrm{\ϵ}\underset{m=1}{\overset{N}{\Σ}}({{\mathrm{\α}}_m}^{*}+{\mathrm{\α}}_m)+\underset{m=1}{\overset{N}{\Σ}}{f}_m({{\mathrm{\α}}_m}^{*}-{\mathrm{\α}}_m)-\frac{1}{2}\underset{m=1}{\overset{N}{\Σ}}\underset{n=1}{\overset{N}{\Σ}}{({{\mathrm{\α}}_m}^{*}-{\mathrm{\α}}_m)}^{T}K(X_m,X_n)({{\mathrm{\α}}_n}^{*}-{\mathrm{\α}}_n)\}$

s.t.

$\underset{m=1}{\overset{N}{\Σ}}({\mathrm{\α}}_m-{{\mathrm{\α}}_m}^{*})=0,0\≤{\mathrm{\α}}_m,{{\mathrm{\α}}_m}^{*}\≤C,m=1,2,...,N$

Wherein, ε is a given parameter value, α m^*, α m be the parameter asked for of training, f_mFor the m-th target output value of training output collection, X_mFor the m-th sample of input feature vector collection, K (X_m,X_n) for RBF kernel function, its form is:

$K(X_m,X_n)=\exp (-\frac{\left|\right|X_m-X_n|{|}^2}{2{\mathrm{\γ}}^2})$

Wherein, γ is gaussian kernel function width parameter；

Function sets up unit, and for after training completes, setting up SVM function is:

${\hat{y}}_i\left(x\right)=\underset{m=1}{\overset{N}{\Σ}}({{\mathrm{\α}}_m}^{*}-{\mathrm{\α}}_m)K(X_m,X)+b$

Wherein: b is the threshold value that training obtains, α_m ^*、α_mFor the parameter that training is asked for；X is given sample input feature vector vector to be predicted；Output valve for recognition function.

Compared with prior art, it is an advantage of the current invention that:

The present invention is mainly through comprehensively utilizing the metrical information of the various sensors of monitoring terminal, by input data are carried out dimensionality reduction thus removing noise and various interference effect, form metastable input feature vector vector, again through the mapping relations adopted between the different safe condition of SVM machine learning method structure and characteristic vector, set up corresponding SVM model of cognition, due in training process, the mapping relations between input feature vector vector fire hydrant safe condition in a variety of situations can be taken into full account, therefore the method that the present invention proposes has higher precision, disclosure satisfy that in alarm procedure and fail to report, the index request of wrong report.

Accompanying drawing explanation

Fig. 1 is the method flow diagram that safe condition intellectual monitoring is fastened in fire-fighting provided by the invention.

Detailed description of the invention

Below in conjunction with accompanying drawing, the present invention is described further.

Embodiment 1

The fire hydrant safe condition intelligent monitoring method of present invention proposition and system, including: several big module of Big Dipper fire hydrant monitoring terminal, communications terminal, Beidou navigation Location Service Platform, alarm indicator terminal.Hydraulic pressure in Big Dipper position monitor terminal dynamic acquisition fire control well, well lid inclination angle, well environment brightness information, and carry out the pretreatment of data, pretreated data are sent to Beidou navigation Location Service Platform by Big Dipper position monitor communication terminal, Beidou navigation Location Service Platform is by extracting the pattern measurement information that monitoring terminal sends, utilize the supporting vector machine model set up, carrying out the identification of current safe state, recognition result is displayed by alarm indicator terminal module and result propelling movement processes.

Its feature specifically includes that

Step 1) timing of Big Dipper fire hydrant monitoring terminal or the metrical information of each sensor of Real-time Collection, these metrical informations include hydraulic pressure in fire control well, well lid inclination angle, well environment brightness etc., all metrical informations are by by, after time synchronized, being sent to Beidou navigation Location Service Platform；

Step 2) adopt principal component analysis technology, the metrical information that Big Dipper fire hydrant monitoring terminal is sent carries out dimension-reduction treatment, simultaneously using the Partial Feature vector of dimension reduction space as the current characteristic vector that fire hydrant safe condition is differentiated；

Step 2-1) each sensor information in fire control well is acquired by Big Dipper fire hydrant monitoring terminal, and the information of collection includes hydraulic pressure in fire control well, well lid inclination angle, well environment brightness etc., it is assumed that forms P dimension and gathers data vector；

Step 2-2) data in certain period of time are acquired K time, structure K × P ties up acquisition matrix U, is designated as:

$U=\left[\begin{array}{cccc}{u}_{11}& u_{12}& \mathrm{...}& u_{1P}\\ u_{21}& u_{22}& \mathrm{...}& u_{2P}\\ .& .& .& .\\ .& .& .& .\\ .& .& .& .\\ u_{K1}& u_{K2}& \mathrm{...}& u_{KP}\end{array}\right]$

Step 2-3) matrix U asked its covariance matrix S, and the eigenvalue of solution matrix S and characteristic vector；

The eigenvalue obtained after solving is: λ₁,λ₂,…,λ_P, characteristic vector is: η₁,η₂,…,η_P

Step 2-4) by the characteristic vector of front m maximum eigenvalue characteristic of correspondence vector combination composition Subsequent secure state recognition；

It is designated as: x=[η₁,η₂,…,η_m]^T, subscript T represents transposition；

Step 3) utilize historical data that the relation between characteristic vector and safe condition is learnt, set up machine learning model；In principal component analysis process, the characteristic vector of dimension reduction space is as training characteristics collection, using corresponding safe condition as output object set, adopts SVM machine learning algorithm to obtain SVM recognition function；

Step 3-1) state of fire hydrant is divided the various states such as fire manhole cover accounts for pressure, fire manhole cover moves, fire manhole cover abnormal start-up, fire protection pipeline breakage is leaked, fire fighting hydraulic pressure is abnormal, based on historical data, SVM model is trained；

Input feature vector collection: X=[x₁,x₂,…,x_N]；Wherein x_i, i=1,2 ..., N is i-th input feature vector vector；

Output object set: f=[y₁,y₂,…,y_N]；Wherein y_i, i=1,2 ..., N is the fire hydrant safe condition that i-th input feature vector vector is corresponding；

Wherein: $x_i=\left[\begin{array}{c}{\mathrm{\η}}_1\\ {\mathrm{\η}}_2\\ .\\ .\\ .\\ {\mathrm{\η}}_m\end{array}\right],i=1,2,...,N$

Step 3-2) adopt step 3-1) obtained training sample set pair support vector machine is trained, utilize input feature vector collection and output object set that training sample set provides, SVM is trained as follows, obtains the SVM recognition function for judging current mobile terminal position:

Step 3-2-1) utilize Novel Algorithm, solve optimization problem:

$\underset{\mathrm{\α},{\mathrm{\α}}^{*}}{\max }\{-\mathrm{\ϵ}\underset{m=1}{\overset{N}{\Σ}}({{\mathrm{\α}}_m}^{*}+{\mathrm{\α}}_m)+\underset{m=1}{\overset{N}{\Σ}}{f}_m({{\mathrm{\α}}_m}^{*}-{\mathrm{\α}}_m)-\frac{1}{2}\underset{m=1}{\overset{N}{\Σ}}\underset{n=1}{\overset{N}{\Σ}}{({{\mathrm{\α}}_m}^{*}-{\mathrm{\α}}_m)}^{T}K(X_m,X_n)({{\mathrm{\α}}_n}^{*}-{\mathrm{\α}}_n)\}$

s.t.

$\underset{m=1}{\overset{N}{\Σ}}({\mathrm{\α}}_m-{{\mathrm{\α}}_m}^{*})=0,0\≤{\mathrm{\α}}_m,{{\mathrm{\α}}_m}^{*}\≤C,m=1,2,...,N$

Wherein, ε is a given parameter value, α_m ^*、α_mFor the parameter that training is asked for, f_mFor the m-th target output value of training output collection, X_mFor the m-th sample of input feature vector collection, K (X_m,X_n) for RBF kernel function, its form is:

$K(X_m,X_n)=\exp (-\frac{\left|\right|X_m-X_n|{|}^2}{2{\mathrm{\γ}}^2})$

Wherein, γ is gaussian kernel function width parameter；

Step 3-2-2), trained after, setting up SVM function is:

${\hat{y}}_i\left(x\right)=\underset{m=1}{\overset{N}{\Σ}}({{\mathrm{\α}}_m}^{*}-{\mathrm{\α}}_m)K(X_m,X)+b$

Wherein: b is the threshold value that training obtains, α_m ^*、α_mFor the parameter that training is asked for；X is given sample input feature vector vector to be predicted；Output valve for recognition function.

Step 4) based on step 2) in process after characteristic vector, integrating step 3) SVM that sets up identifies equation, identifies the safe condition of current fire hydrant；

For arbitrarily inputting x, recognition result is status=I, wherein:

I ∈ { fire manhole cover accounts for pressure, fire manhole cover moves, fire manhole cover abnormal start-up, fire protection pipeline leak, fire fighting hydraulic pressure abnormal }.

Embodiment 2

The present invention also provides for a kind of fire hydrant safe condition system, and described system comprises:

Fire hydrant monitoring terminal, for gathering the metrical information of each sensor, and carries out synchronization process to the information gathered；Wherein, described status information comprises: hydraulic pressure, well lid inclination angle or well environment brightness in fire control well；

State decision process device is fastened in fire-fighting, is used for:

Adopting principal component analysis technology, the metrical information received is carried out dimension-reduction treatment, extracts the characteristic vector of dimension reduction space simultaneously, the size according to characteristic vector character pair value, selected part characteristic vector is as the current characteristic vector that fire hydrant safe condition is differentiated；

Utilize historical data that the relation between characteristic vector and safe condition is learnt, set up machine learning model；In principal component analysis process, the characteristic vector of dimension reduction space is as training characteristics collection, using corresponding safe condition as output object set, adopts SVM machine learning algorithm to obtain SVM recognition function；

Based on described characteristic vector, in conjunction with the SVM recognition function set up, identify the safe condition of current fire hydrant.

For technique scheme, in the implementation that the first is possible, described fire-fighting is fastened state decision process device and is comprised further:

First processing module, for for the measurement data acquisition in certain period of time K time, structure K × P dimension acquisition matrix U, being designated as:

$U=\left[\begin{array}{cccc}{u}_{11}& u_{12}& \mathrm{...}& u_{1P}\\ u_{21}& u_{22}& \mathrm{...}& u_{2P}\\ .& .& .& .\\ .& .& .& .\\ .& .& .& .\\ u_{K1}& u_{K2}& \mathrm{...}& u_{KP}\end{array}\right]$

Matrix U is asked its covariance matrix S, and the eigenvalue of solution matrix S and characteristic vector；

The eigenvalue obtained after solving is: λ₁,λ₂,…,λ_P, the characteristic vector obtained is: η₁,η₂,…,η_P

Second processing module, constitutes the characteristic vector of Subsequent secure state recognition for the eigenvalue characteristic of correspondence vector combination maximum by front m；

It is designated as: x=[η₁,η₂,…,η_m]^T, subscript T represents transposition.

For technique scheme and the first implementation, in the implementation that the second is possible, described fire-fighting is fastened state decision process device and is also comprised:

3rd processing module, for by the state demarcation of fire hydrant being: various states such as fire manhole cover accounts for pressure, fire manhole cover moves, fire manhole cover abnormal start-up, fire protection pipeline breakage are leaked, fire fighting hydraulic pressure is abnormal, is trained SVM model based on historical data；

Input feature vector collection: X=[x₁,x₂,…,x_N]；Wherein x_i, i=1,2 ..., N is i-th input feature vector vector；

Output object set: f=[y₁,y₂,…,y_N]；Wherein y_i, i=1,2 ..., N is the fire hydrant safe condition that i-th input feature vector vector is corresponding；

Wherein: $x_i=\left[\begin{array}{c}{\mathrm{\η}}_1\\ {\mathrm{\η}}_2\\ .\\ .\\ .\\ {\mathrm{\η}}_m\end{array}\right],i=1,2,...,N$

Fourth processing module, is trained for the training sample set pair support vector machine that adopts the first processing module obtained, utilizes input feature vector collection and output object set that training sample set provides, to SVM by being trained and then obtaining SVM recognition function.

For the implementation that technique scheme and/or the second are possible, in the implementation that the third is possible, described fourth processing module comprises further:

Processing unit, is used for utilizing Novel Algorithm, solves optimization problem:

$\underset{\mathrm{\α},{\mathrm{\α}}^{*}}{\max }\{-\mathrm{\ϵ}\underset{m=1}{\overset{N}{\Σ}}({{\mathrm{\α}}_m}^{*}+{\mathrm{\α}}_m)+\underset{m=1}{\overset{N}{\Σ}}{f}_m({{\mathrm{\α}}_m}^{*}-{\mathrm{\α}}_m)-\frac{1}{2}\underset{m=1}{\overset{N}{\Σ}}\underset{n=1}{\overset{N}{\Σ}}{({{\mathrm{\α}}_m}^{*}-{\mathrm{\α}}_m)}^{T}K(X_m,X_n)({{\mathrm{\α}}_n}^{*}-{\mathrm{\α}}_n)\}$

s.t.

$\underset{m=1}{\overset{N}{\Σ}}({\mathrm{\α}}_m-{{\mathrm{\α}}_m}^{*})=0,0\≤{\mathrm{\α}}_m,{{\mathrm{\α}}_m}^{*}\≤C,m=1,2,...,N$

Wherein, ε is a given parameter value, α_m ^*、α_mFor the parameter that training is asked for, f_mFor the m-th target output value of training output collection, X_mFor the m-th sample of input feature vector collection, K (X_m,X_n) for RBF kernel function, its form is:

$K(X_m,X_n)=\exp (-\frac{\left|\right|X_m-X_n|{|}^2}{2{\mathrm{\γ}}^2})$

Wherein, γ is gaussian kernel function width parameter；

Function sets up unit, and for after training completes, setting up SVM function is:

${\hat{y}}_i\left(x\right)=\underset{m=1}{\overset{N}{\Σ}}({{\mathrm{\α}}_m}^{*}-{\mathrm{\α}}_m)K(X_m,X)+b$

Wherein: b is the threshold value that training obtains, α_m ^*、α_mFor the parameter that training is asked for；X is given sample input feature vector vector to be predicted；Output valve for recognition function.

The particular hardware that technique scheme relates to comprises Big Dipper fire hydrant intelligent terminal and Beidou navigation Location Service Platform, and the concrete function of these hardware entities is as follows:

Big Dipper fire hydrant intelligent terminal hardware mainly includes in hydraulic pressure sensor, well lid obliquity sensor, well environment brightness sensor, well in wireless communication transmissions module, Big Dipper well inner position module, well the part compositions such as mobile communication module.

Relevant parameter in well regularly or is acquired by hydraulic pressure sensor, well lid obliquity sensor, well environment brightness sensor according to rule set in advance, wherein hydraulic pressure information is transmitted by wireless communication module in well, other sensor informations are directly acquired by respective data acquisition module, all collection data are carried out integrated reception by Big Dipper well inner position module, and by, after the pretreatment such as time alignment, carrying out data transmission by mobile communication module in well to Big Dipper navigation position service platform.

The technical characteristics that technique scheme adopts comprises: (1) adopts the mode of radio communication to gather hydraulic pressure information, meets adverse circumstances demand in well, it is to avoid wiring；(2) hydraulic pressure sensor, well lid obliquity sensor, well environment brightness sensing data collected after complete the integrated of all data by transmission to Big Dipper well inner position module；(3) Big Dipper locating module carries out the pretreatment such as the time alignment of data, sends data to Beidou navigation Location Service Platform by mobile communication module.

Beidou navigation Location Service Platform in above-described embodiment is for receiving the transmission data of different terminals, and the fire hydrant safe condition of each terminal is differentiated, it is achieved unified monitoring and the management and control to all fire-fighting systems in whole region.

It should be noted last that, above example is only in order to illustrate technical scheme and unrestricted.Although the present invention being described in detail with reference to embodiment, it will be understood by those within the art that, technical scheme being modified or equivalent replacement, without departure from the spirit and scope of technical solution of the present invention, it all should be encompassed in the middle of scope of the presently claimed invention.

Claims (7)

1. a fire hydrant safe condition intelligent monitoring method, described method comprises:

Step 101) gather the metrical information of each sensor, all metrical informations collected are carried out time synchronizing；

Wherein, described metrical information includes: hydraulic pressure, well lid inclination angle or well environment brightness in fire control well；

Step 102) adopt principal component analysis that the metrical information after synchronization process is carried out dimension-reduction treatment, extract the characteristic vector of dimension reduction space simultaneously；

Size according to characteristic vector character pair value, chooses several characteristic vectors as the current characteristic vector that fire hydrant safe condition is differentiated；

Step 103) utilize historical data that the relation between characteristic vector and safe condition is learnt, set up machine learning model；

In principal component analysis process, the characteristic vector of dimension reduction space is as training characteristics collection, using corresponding safe condition as output object set, adopts SVM machine learning algorithm to obtain SVM recognition function；

Step 104) based on step 102) in process after characteristic vector, integrating step 103) the SVM recognition function set up, identify the safe condition of current fire hydrant.

2. fire hydrant safe condition intelligent monitoring method according to claim 1, it is characterised in that described step 102) comprise further:

Step 102-1) assume that each sensor information in the fire control well gathered forms P dimension and gathers data vector；Wherein, the information of collection includes hydraulic pressure in fire control well, well lid inclination angle or well environment brightness；

Step 102-2) for the data acquisition in certain period of time K time, and construct K × P dimension acquisition matrix U, it is designated as:

$U=\left[\begin{array}{cccc}{u}_{11}& u_{12}& ...& u_{1P}\\ u_{21}& u_{22}& ...& u_{2P}\\ .& .& .& .\\ .& .& .& .\\ .& .& .& .\\ u_{K1}& u_{K2}& ...& u_{KP}\end{array}\right]$

Step 102-3) matrix U asked its covariance matrix S, and the eigenvalue of solution matrix S and characteristic vector；

After solving obtain eigenvalue and especially vector be respectively labeled as: λ₁,λ₂,…,λ_P, and, η₁,η₂,…,η_P；

Step 102-4) by the characteristic vector of front m maximum eigenvalue characteristic of correspondence vector combination composition safe condition identification；

Wherein, m is the minima that eigenvalue sum is corresponding more than 80% with the ratio of total characteristic value.

3. fire hydrant safe condition intelligent monitoring method according to claim 2, it is characterised in that described step 103) comprise further:

Step 103-1) it is some classes by fire hydrant state demarcation, based on historical data, SVM model is trained；Wherein, described fire hydrant state categories comprises: fire manhole cover accounts for pressure, fire manhole cover moves, fire manhole cover abnormal start-up, fire protection pipeline breakage are leaked or fire fighting hydraulic pressure is abnormal；

The input feature vector collection of SVM model: X=[x₁,x₂,…,x_N]；Wherein x_i, i=1,2 ..., N is i-th input feature vector vector；

The output object set of SVM model: f=[y₁,y₂,…,y_N]；Wherein y_i, i=1,2 ..., N is the fire hydrant safe condition that i-th input feature vector vector is corresponding；

Wherein: $x_i=\left[\begin{array}{c}{\mathrm{\η}}_1\\ {\mathrm{\η}}_2\\ .\\ .\\ .\\ {\mathrm{\η}}_m\end{array}\right],i=1,2,...,N;$

Step 103-2) adopt step 103-1) obtained training sample set pair support vector machine is trained, utilize input feature vector collection and output object set that training sample set provides, SVM is trained as follows, obtains the SVM recognition function for judging safe condition:

Step 103-2-1) utilize Novel Algorithm, solve optimization problem:

$\begin{array}{c}\underset{\mathrm{\α},{\mathrm{\α}}^{*}}{\max }\{-\mathrm{\ϵ}\underset{m=1}{\overset{N}{\Σ}}({{\mathrm{\α}}_m}^{*}+{\mathrm{\α}}_m)+\underset{m=1}{\overset{N}{\Σ}}{f}_m({{\mathrm{\α}}_m}^{*}-{\mathrm{\α}}_m)-\frac{1}{2}\underset{m=1}{\overset{N}{\Σ}}\underset{n=1}{\overset{N}{\Σ}}{({{\mathrm{\α}}_m}^{*}-{\mathrm{\α}}_m)}^{T}K(X_m,X_n)({{\mathrm{\α}}_n}^{*}-{\mathrm{\α}}_n)\}\\ s.t.\\ \underset{m=1}{\overset{N}{\Σ}}({\mathrm{\α}}_m-{{\mathrm{\α}}_m}^{*})=0,0\≤{\mathrm{\α}}_m,{{\mathrm{\α}}_m}^{*}\≤C,m=1,2,...,N\end{array}$

Wherein, ε is a given parameter value, α_m ^*、α_mFor the parameter that training is asked for, f_mFor the m-th target output value of training output collection, X_mFor the m-th sample of input feature vector collection, K (X_m,X_n) for RBF kernel function, its form is:

$K(X_m,X_n)=\exp (-\frac{\left|\right|X_m-X_n|{|}^2}{2{\mathrm{\γ}}^2})$

Wherein, γ is gaussian kernel function width parameter；

Step 103-2-2) trained after, setting up SVM function is:

${\hat{y}}_i\left(x\right)=\underset{m=1}{\overset{N}{\Σ}}({{\mathrm{\α}}_m}^{*}-{\mathrm{\α}}_m)K(X_m,X)+b$

Wherein, b is the threshold value that training obtains, α_m ^*And α_mFor the parameter that training is asked for；X is given sample input feature vector vector to be predicted；Output valve for recognition function.

4. a fire hydrant safe condition system, it is characterised in that described system comprises:

Fire hydrant monitoring terminal, for gathering the metrical information of each sensor, and carries out synchronization process to the information gathered；Wherein, described status information comprises: hydraulic pressure, well lid inclination angle or well environment brightness in fire control well；

State decision process device is fastened in fire-fighting, is used for:

Adopting principal component analysis technology, the metrical information received is carried out dimension-reduction treatment, extracts the characteristic vector of dimension reduction space simultaneously, the size according to characteristic vector character pair value, selected part characteristic vector is as the current characteristic vector that fire hydrant safe condition is differentiated；

Utilize historical data that the relation between characteristic vector and safe condition is learnt, set up machine learning model；In principal component analysis process, the characteristic vector of dimension reduction space is as training characteristics collection, using corresponding safe condition as output object set, adopts SVM machine learning algorithm to obtain SVM recognition function；

Based on described characteristic vector, in conjunction with the SVM recognition function set up, identify the safe condition of current fire hydrant.

5. fire hydrant safe condition system according to claim 4, it is characterised in that described fire-fighting is fastened state decision process device and comprised further:

First processing module, for being K time for the number of times of measurement data acquisition in certain period of time, structure K × P ties up acquisition matrix U, is designated as:

$U=\left[\begin{array}{cccc}{u}_{11}& u_{12}& ...& u_{1P}\\ u_{21}& u_{22}& ...& u_{2P}\\ .& .& .& .\\ .& .& .& .\\ .& .& .& .\\ u_{K1}& u_{K2}& ...& u_{KP}\end{array}\right]$

Matrix U is asked its covariance matrix S, and the eigenvalue of solution matrix S and characteristic vector；

The eigenvalue obtained after solving is: λ₁,λ₂,…,λ_P, the characteristic vector obtained is: η₁,η₂,…,η_P

Second processing module, constitutes the characteristic vector of Subsequent secure state recognition for the eigenvalue characteristic of correspondence vector combination maximum by front m；

It is designated as: $x=\left[\begin{array}{c}{\mathrm{\η}}_1\\ {\mathrm{\η}}_2\\ .\\ .\\ .\\ {\mathrm{\η}}_m\end{array}\right].$

6. the fire hydrant safe condition system according to claim 4 or 5, it is characterised in that described fire-fighting is fastened state decision process device and also comprised:

3rd processing module, for by the state demarcation of fire hydrant being: various states such as fire manhole cover accounts for pressure, fire manhole cover moves, fire manhole cover abnormal start-up, fire protection pipeline breakage are leaked, fire fighting hydraulic pressure is abnormal, is trained SVM model based on historical data；

Input feature vector collection: X=[x₁,x₂,…,x_N]；Wherein x_i, i=1,2 ..., N is i-th input feature vector vector；

Output object set: f=[y₁,y₂,…,y_N]；Wherein y_i, i=1,2 ..., N is the fire hydrant safe condition that i-th input feature vector vector is corresponding；

Wherein: $x_i=\left[\begin{array}{c}{\mathrm{\η}}_1\\ {\mathrm{\η}}_2\\ .\\ .\\ .\\ {\mathrm{\η}}_m\end{array}\right],i=1,2,...,N$

Fourth processing module, is trained for the training sample set pair support vector machine that adopts the first processing module obtained, utilizes input feature vector collection and output object set that training sample set provides, to SVM by being trained and then obtaining SVM recognition function.

7. fire hydrant safe condition system according to claim 6, it is characterised in that described fourth processing module comprises further:

Processing unit, is used for utilizing Novel Algorithm, solves optimization problem:

$\begin{array}{c}\underset{\mathrm{\α},{\mathrm{\α}}^{*}}{\max }\{-\mathrm{\ϵ}\underset{m=1}{\overset{N}{\Σ}}({{\mathrm{\α}}_m}^{*}+{\mathrm{\α}}_m)+\underset{m=1}{\overset{N}{\Σ}}{f}_m({{\mathrm{\α}}_m}^{*}-{\mathrm{\α}}_m)-\frac{1}{2}\underset{m=1}{\overset{N}{\Σ}}\underset{n=1}{\overset{N}{\Σ}}{({{\mathrm{\α}}_m}^{*}-{\mathrm{\α}}_m)}^{T}K(X_m,X_n)({{\mathrm{\α}}_n}^{*}-{\mathrm{\α}}_n)\}\\ s.t.\\ \underset{m=1}{\overset{N}{\Σ}}({\mathrm{\α}}_m-{{\mathrm{\α}}_m}^{*})=0,0\≤{\mathrm{\α}}_m,{{\mathrm{\α}}_m}^{*}\≤C,m=1,2,...,N\end{array}$

Wherein, ε is a given parameter value, α_m ^*、α_mFor the parameter that training is asked for, f_mFor the m-th target output value of training output collection, X_mFor the m-th sample of input feature vector collection, K (X_m,X_n) for RBF kernel function, its form is:

$K(X_m,X_n)=\exp (-\frac{\left|\right|X_m-X_n|{|}^2}{2{\mathrm{\γ}}^2})$

Wherein, γ is gaussian kernel function width parameter；

Function sets up unit, and for after training completes, setting up SVM function is:

${\hat{y}}_i\left(x\right)=\underset{m=1}{\overset{N}{\Σ}}({{\mathrm{\α}}_m}^{*}-{\mathrm{\α}}_m)K(X_m,X)+b$

Wherein: b is the threshold value that training obtains, α_m ^*、α_mFor the parameter that training is asked for；X is given sample input feature vector vector to be predicted；Output valve for recognition function.