CN104198932A - High voltage circuit breaker machinery property online monitoring system and fault diagnosis method - Google Patents

Abstract

The invention discloses a high voltage circuit breaker machinery property online monitoring system and a fault diagnosis method. The fault diagnosis method replaces a traditional offline overhaul mode, and can accurately judge running status information of a circuit breaker. The high voltage circuit breaker machinery property online monitoring system comprises a linear displacement sensor, a vibration sensor, a signal conditioning circuit, an A/D (analog to digital) conversion module, an optoelectronic isolation circuit, a drive isolation circuit, a switch opening/ closing electric control module, a DSP (digital signal processor), a human and computer interface module, a data storage module, a system time alignment module and a communication module, wherein the linear displacement sensor and the vibration sensor are respectively connected with the signal conditioning circuit, an output end of the signal conditioning circuit is connected with the A/D conversion module, an output end of the A/D conversion module, the optoelectronic isolation circuit, the drive isolation circuit, the human and computer interface module, the system time alignment module, the data storage module and the communication module are connected with the DSP, an output end of the drive isolation circuit is connected with the switch opening/ closing electric control module, and the communication module is connected with an upper computer.

Description

Mechanical characteristic of high-voltage circuit breaker on-line monitoring system and method for diagnosing faults

Technical field

The present invention relates to a kind of primary cut-out, relate in particular to a kind of mechanical characteristic of high-voltage circuit breaker on-line monitoring system and method for diagnosing faults, belong to high-tension switch gear technical field.

Background technology

Isolating switch, as one of most important primary equipment in electric system, is also the highest equipment of proportion in all primary equipments, and its safe operation is played huge supporting role to national economic development.Traditional overhaul of line breaker mode adopts regular off-line test method, and this just needs isolating switch out of service, can obtain the status data that isolating switch is relevant.This method has obvious deficiency, needs the isolating switch of Awaiting Overhaul out of service, has reduced overhaul efficiency.For example, some the High Voltage Circuit Breaker Conditions are good, but want regularly out of service maintenance, and fault appearred in some isolating switchs, but because being less than turn(a)round " in spite of illness " move, thereby buried serious hidden danger for the safe operation of electrical network.For addressing the above problem, breaker mechanic property on-line monitoring and method for diagnosing faults arise at the historic moment, and constantly improve.

The patent documentation that application number is " 201320124521.9 " discloses a kind of breaker mechanic property on-line monitoring instrument, it adopts angular displacement sensor to be arranged on breaker main shaft, by gathering the angular displacement of breaker main shaft, calculate the mechanical features parameter that obtains isolating switch, the method has obvious deficiency, because the angular displacement of breaker main shaft is very little, not only be difficult to gather accurately, and the stochastic error of each collection is larger, repeatable poor, in addition, because it only uses angular displacement sensor image data, cannot complete complicated fault diagnosis and location, thereby actual monitoring effect is unsatisfactory.

Summary of the invention

The object of the present invention is to provide a kind of mechanical characteristic of high-voltage circuit breaker on-line monitoring system and method for diagnosing faults, substitute traditional off-line maintenance pattern, can accurately judge the High Voltage Circuit Breaker Condition information.

Object of the present invention is achieved by the following technical programs:

A kind of mechanical characteristic of high-voltage circuit breaker on-line monitoring system, comprise host computer 14, power module 12, lower computer system 15, described lower computer system 15 comprise linear displacement transducer 1, vibration transducer 2, signal conditioning circuit 3, A/D modular converter 4, photoelectric isolating circuit 5, driving isolation circuit 6, minute/combined floodgate electric control module 7, dsp processor 8, human-computer interface module 9, data memory module 10, system to time module 11, communication module 13, described linear displacement transducer 1 gathers the displacement signal of isolating switch driving stem, described linear displacement transducer 1, vibration transducer 2 is connected with signal conditioning circuit 3 respectively, the output terminal of described signal conditioning circuit 3 is connected with A/D modular converter 4, the output terminal of A/D modular converter 4, photoelectric isolating circuit 5, driving isolation circuit 6, human-computer interface module 9, system to time module 11, data memory module 10, communication module 13 is connected with dsp processor 8, the output terminal of described driving isolation circuit 6 is connected with divide/combined floodgate electric control module 7, communication module 13 is connected with host computer 14, power module 12 is connected with lower computer system 15 each several parts, for its power supply.

A method for diagnosing faults for mechanical characteristic of high-voltage circuit breaker on-line monitoring system, the method comprises the following steps:

(1) establish isolating switch and have n kind state, comprise state 1, state 2 ..., state m-1, state m, state m+1 ..., state n, separating brake under collection said n kind state and the vibration data sample of combined floodgate, vibration data under different conditions is divided into two groups, be separating brake vibration data sample group and combined floodgate vibration data sample group, the proper vector of two grouping vibration signals under each state is extracted in employing based on interval wavelet bag analytic approach, obtain respectively separating brake proper vector group, combined floodgate proper vector group; The described method based on interval wavelet bag analytic approach extraction vibration signal characteristics vector is as follows:

A. vibration signal is carried out to the decomposition of j layer, obtain sequence S _{j, k}, calculate each segmentation energy:

$Q_{i(j,k)}={\∫}_{t_{i-1}}^{t_i}{\left|A_i\left(t\right)\right|}^2\mathrm{dt}$

Wherein, get j=2, k=0,1,2,3, i represents the sequence number of segmentation after signal decomposition, i=1,2,3,4; t _i-1, t _ibe respectively i piecewise interval end points, A _i(t) be the amplitude of i segmentation;

B. each segmentation energy of normalized, obtains normalized value:

${\mathrm{\ϵ}}_{i(j,k)}=\frac{Q_{i(j,k)}}{{\∫}_{t_0}^{t_4}{\left|A\left(t\right)\right|}^2\mathrm{dt}}$

C. according to the theory of information entropy, calculate each segmentation wavelet-packet energy entropy:

$H_{(j,k)}=-\underset{i=1}{\overset{4}{\mathrm{\Σ}}}{\mathrm{\ϵ}}_{i(j,k)}\lg {\mathrm{\ϵ}}_{i(j,k)}$

D. the wavelet-packet energy entropy of each segmentation is formed to vibration signal characteristics vector:

x＝[H ₂₀，H ₂₁，H ₂₂，H ₂₃]

(2) try to achieve separating brake vibration signal characteristics under n kind state vector, combined floodgate vibration signal characteristics vector, then adopt " a pair of all the other " strategy of support vector machine theory, ask circuit-breaker status decision function; When using separating brake vibration signal characteristics Vector Groups Training Support Vector Machines, obtain gate-dividing state decision function; When using combined floodgate vibration signal characteristics Vector Groups Training Support Vector Machines, obtain "on" position decision function; The acquiring method following steps of described state decision function:

A. the vibration signal characteristics vector under n kind state is set up to proper vector Sample Storehouse, using the vibration signal characteristics vector of state m as a class with+1 mark, remaining n-1 kind vibration signal characteristics vector with-1 mark, builds new sample set T={ (x as all the other classes ₁, y ₁), (x ₂, y ₂) ..., (x _l, y _l), y _i∈+1 ,-1}, wherein: i=1,2 ..., l, the number that l is sample, x _irepresentation feature vector sample, y _ipresentation class mark, value+1 or-1;

B. select kernel function K (x, x _i) make K (x, x _i)=Φ (x) Φ (x _i), this method kernel function K (x, x _i) be gaussian kernel function wherein, Φ is the mapping that theorem in Euclid space arrives Hilbert space, the width parameter that σ is function, the radial effect scope of control function;

C. in constraint condition $\underset{i=1}{\overset{l}{\mathrm{\Σ}}}{y}_i{\mathrm{\α}}_i=0$ With 0≤α _iunder≤C, solve $\max W\left(\mathrm{\α}\right)=\underset{i=1}{\overset{l}{\mathrm{\Σ}}}{\mathrm{\α}}_i-\frac{1}{2}\underset{i=1}{\overset{l}{\mathrm{\Σ}}}\underset{j=1}{\overset{l}{\mathrm{\Σ}}}{\mathrm{\α}}_i{\mathrm{\α}}_j{y}_i{y}_{j}K(x_i,x_j)$ Obtain optimum weight vector wherein, C is penalty factor;

D. calculate $w^{*}=\underset{i=1}{\overset{l}{\mathrm{\Σ}}}{\mathrm{\α}}_i^{*}{y}_i{x}_i,$ Select α ^*a positive component meet calculate $b^{*}=y_j-\underset{i=1}{\overset{l}{\mathrm{\Σ}}}{y}_i{\mathrm{\α}}_i^{*}K(x_i\·x_j),$ Substitution f (x)=w ^*.x+b ^*, obtain state decision function $f\left(x\right)=\underset{i=1}{\overset{l}{\mathrm{\Σ}}}(y_i{\mathrm{\α}}_i^{*}K(x_i\·x)+b^{*}),$ Be designated as $f_m\left(x\right)=\underset{i=1}{\overset{l}{\mathrm{\Σ}}}(y_i^m{\mathrm{\α}}_i^{m^{*}}K(x_i^m\·x)+b^{m^{*}});$

For a, b, c, the d process of step (2), m traversal value 1,2,3 ..., n-1, n, can obtain n state decision function $f_m\left(x\right)=\underset{i=1}{\overset{l}{\mathrm{\Σ}}}(y_i^m{\mathrm{\α}}_i^{m^{*}}K(x_i^m\·x)+b^{m^{*}}),$ m＝1，2，3，…，n；

(3), for treating diagnostic sample x, first judge that x is separating brake vibration signal characteristics vector or combined floodgate vibration signal characteristics vector, then decision function corresponding to substitution.As state decision function f (x)=f _p(x) time, obtain maximal value, judgement treat that diagnostic sample x belongs to state p, wherein p ∈ 1,2,3 ..., n}.

Object of the present invention can also further realize by following technical measures:

Aforementioned mechanical characteristic of high-voltage circuit breaker on-line monitoring system, wherein linear displacement transducer 1 is direct slide conductive plastic potentiometer, and model is WDL-50-2, and independent linearity reaches 0.1%.

Aforementioned mechanical characteristic of high-voltage circuit breaker on-line monitoring system, wherein linear displacement transducer 1 is arranged on the support in circuit breaker electric magnet outside, by driving stem, drives semi-girder to move together with linear displacement transducer 1.

Aforementioned mechanical characteristic of high-voltage circuit breaker on-line monitoring system, wherein vibration transducer 2 is ZD24T type reluctance type vibration transducer.

Aforementioned mechanical characteristic of high-voltage circuit breaker on-line monitoring system, wherein vibration transducer 2 is arranged on the pedestal that isolating switch approaches vibration source.

Aforementioned mechanical characteristic of high-voltage circuit breaker on-line monitoring system, wherein system to time module 11 adopt X1226 time calendar clocks.

Compared with prior art, the invention has the beneficial effects as follows: utilize vibration transducer, accurately capture the vibration signal of isolating switch, based on interval wavelet bag analytic approach, extract the proper vector of vibration signal, input corresponding state decision function, thereby accurately judge the running status of primary cut-out.

Accompanying drawing explanation

Fig. 1 is on-line monitoring system structured flowchart;

Fig. 2 is linear displacement transducer installation site figure;

Fig. 3 is vibration transducer installation site figure;

Fig. 4 is method for diagnosing faults process flow diagram;

Fig. 5 is on-line monitoring system workflow diagram.

Embodiment

Below in conjunction with the drawings and specific embodiments, the invention will be further described.

As shown in Figure 1, on-line monitoring system of the present invention comprises: comprise host computer 14, power module 12, lower computer system 15.Lower computer system 15 comprise linear displacement transducer 1, vibration transducer 2, signal conditioning circuit 3, A/D modular converter 4, photoelectric isolating circuit 5, driving isolation circuit 6, minute/combined floodgate electric control module 7, dsp processor 8, human-computer interface module 9, data memory module 10, system to time module 11, communication module 13; Wherein:

Linear displacement transducer 1, vibration transducer 2 are arranged on breaker body, gather displacement signal and the shell vibration signal of isolating switch driving stem.

Signal conditioning circuit 3, for nursing one's health the output signal of each sensor, carries out filtering to signal, also comprises the convergent-divergent of signal voltage and translation, makes it to meet the voltage request of A/D modular converter 4.

A/D modular converter 4, for being converted to digital signal by the analog voltage signal of signal conditioning circuit 3 outputs.

Human-computer interface module 9, for showing isolating switch actuation time, also, for showing circuit-breaker status judged result, is convenient to maintainer and checks.

Dsp processor 8, for carrying out compiled control program in advance, complete that sensor signal processing, fault analysis, man-machine interface show, system to time, the function such as data storage, communication, can also, according to the instruction of host computer 14, complete corresponding divide/feed motion.

System to time module 11, when with 8 pairs of dsp processors.Because the DSP time adopts the interrupt mode timing that adds up, long-play can cause cumulative errors to increase, this module is selected special-purpose high precision clock calendar chip X1226, with dsp processor 8 integral points to time, guarantee accurately to record the moment that each event occurs, and accurately according to the time, carry out the instruction that host computer issues.

Communication module 13, adopt RS-485 communication mode, can grow apart from receiving and transmitting signal, lower computer system 15 can be sent out data mutually with host computer 14, lower computer system 15 is uploaded sensor raw data and result, and host computer 14 can send command information to lower computer system 15.

Data memory module 10, for program code stored, can also store the High Voltage Circuit Breaker Condition data, while needing, can transfer at any time.

Photoelectricity is isolated 5 circuit, adopts TLP521-2 opto-coupler chip, for isolating the voltage of auxiliary contact, prevents overtension, damages dsp chip.

Driving isolation 6 circuit, adopt relay isolation method, for strengthening the driving force of dsp processor, drive divide/combined floodgate electric control module.

Power module 12, employing alternating current 220V turns direct current+12V ,+5V (AC-DC), direct current+5V turn direct current+3.3V ,+1.8V (DC-DC) design, vibration transducer 2 employing+12V power supplies, the opto-coupler chip of linear displacement transducer 1, signal conditioning circuit 3, A/D modular converter 4, photoelectric isolating circuit 5, be+5V of driving isolation circuit 6 power supply, dsp processor chip 8 employing+3.3VFlash voltages and+1.8V core voltage power supply, use PS767D301 high precision electro source chip complete+5V～+ 3.3V ,+5V～+ 1.8V level conversion.

As shown in Figure 2, on-line monitoring system of the present invention adopts linear displacement transducer 1 to realize picking up displacement signal, linear displacement transducer 1 is arranged on the support 201 on electromagnet side, when driving stem 203 drives semi-girder 202 to move together with linear displacement transducer moving lever, the output voltage of displacement transducer will change.This mounting means of linear displacement transducer 1, neither affects the original inner structure of isolating switch, has improved again displacement detecting precision.

As shown in Figure 3, vibration transducer 2 is arranged on circuit breaker base shell, and this mounting means of vibration transducer can be caught the vibration signal of isolating switch accurately, and comprises abundant circuit-breaker status information.

As shown in Figure 4, the process that represents circuit breaker failure diagnostic method, vibration signal while first gathering isolating switch different conditions, as normal condition, pedestal loosened screw, case for circuit breaker crackle, operating mechanism bite, minute/power switch is under-voltage, then utilize wavelet packet analysis method to extract the proper vector of vibration signal, with proper vector training acquisition state decision function.Sampling feature vectors substitution decision function to be diagnosed is calculated, can obtain condition diagnosing result, concrete method is as follows: determine the n kind Status Type that will diagnose, comprise state 1, state 2 ..., state m-1, state m, state m+1 ..., state n, obtain separating brake and combined floodgate vibration data sample under said n kind state, the separating brake under different conditions and combined floodgate vibration data sample are divided into two groupings.The proper vector of two grouping vibration signals under each state is extracted in employing based on interval wavelet bag analytic approach, obtain respectively separating brake proper vector group, combined floodgate proper vector group.Wherein, based on interval wavelet bag analytic approach, extract the method following steps of vibration signal characteristics vector:

A. vibration signal is carried out to the decomposition of j layer, obtain sequence S _{j, k}, calculate each segmentation energy:

$Q_{i(j,k)}={\∫}_{t_{i-1}}^{t_i}{\left|A_i\left(t\right)\right|}^2\mathrm{dt}$

Wherein: get j=2, k=0,1,2,3, i represents the sequence number of segmentation after signal decomposition, i=1,2,3,4; t _i-1, t _ibe respectively i piecewise interval end points, A _i(t) be the amplitude of i segmentation.

B. each segmentation energy of normalized, obtains normalized value:

${\mathrm{\ϵ}}_{i(j,k)}=\frac{Q_{i(j,k)}}{{\∫}_{t_0}^{t_4}{\left|A\left(t\right)\right|}^2\mathrm{dt}}$

C. according to the theory of information entropy, calculate each segmentation wavelet-packet energy entropy:

$H_{(j,k)}=-\underset{i=1}{\overset{4}{\mathrm{\Σ}}}{\mathrm{\ϵ}}_{i(j,k)}\lg {\mathrm{\ϵ}}_{i(j,k)}$

D. the wavelet-packet energy entropy of each segmentation is formed to vibration signal characteristics vector:

x＝[H ₂₀，H ₂₁，H ₂₂，H ₂₃]

Try to achieve separating brake vibration signal characteristics under n kind state vector, combined floodgate vibration signal characteristics vector, then adopt " a pair of all the other " strategy of support vector machine theory, ask circuit-breaker status decision function; When using separating brake vibration signal characteristics Vector Groups Training Support Vector Machines, obtain gate-dividing state decision function; When using combined floodgate vibration signal characteristics Vector Groups Training Support Vector Machines, obtain "on" position decision function; The acquiring method following steps of described state decision function:

A. by n kind state (state 1, state 2 ... state m-1, state m, state m+1 ... state n) separating brake/combined floodgate proper vector under is set up proper vector Sample Storehouse, using state m separating brake/combined floodgate proper vector as a class with+1 mark, remaining n-1 kind state separating brake/combined floodgate proper vector (state 1, state 2 ..., state m-1, state m+1 ..., state n) as all the other classes with-1 mark, build new sample set T={ (x ₁, y ₁), (x ₂, y ₂) ..., (x _l, y _l), y _i∈+1 ,-1}, wherein: i=1,2 ..., l, the number that l is sample, x _irepresentation feature vector sample, y _ipresentation class mark, value+1 or-1;

B. select suitable kernel function K (x, x _i)=Φ (x) Φ (x _i), select gaussian kernel function here wherein, Φ is the mapping that theorem in Euclid space arrives Hilbert space, the width parameter that σ is function, the radial effect scope of control function;

C. in constraint condition $\underset{i=1}{\overset{l}{\mathrm{\Σ}}}{y}_i{\mathrm{\α}}_i=0$ With 0≤α _iunder≤C, solve $\max W\left(\mathrm{\α}\right)=\underset{i=1}{\overset{l}{\mathrm{\Σ}}}{\mathrm{\α}}_i-\frac{1}{2}\underset{i=1}{\overset{l}{\mathrm{\Σ}}}\underset{j=1}{\overset{l}{\mathrm{\Σ}}}{\mathrm{\α}}_i{\mathrm{\α}}_j{y}_i{y}_{j}K(x_i,x_j)$ Obtain optimum weight vector wherein, C is penalty factor;

D. calculate $w^{*}=\underset{i=1}{\overset{l}{\mathrm{\Σ}}}{\mathrm{\α}}_i^{*}{y}_i{x}_i,$ Select α ^*a positive component meet calculate $b^{*}=y_j-\underset{i=1}{\overset{l}{\mathrm{\Σ}}}{y}_i{\mathrm{\α}}_i^{*}K(x_i\·x_j),$ Substitution f (x)=w ^*.x+b ^*, obtain state decision function $f\left(x\right)=\underset{i=1}{\overset{l}{\mathrm{\Σ}}}(y_i{\mathrm{\α}}_i^{*}K(x_i\·x)+b^{*}),$ Be designated as $f_m\left(x\right)=\underset{i=1}{\overset{l}{\mathrm{\Σ}}}(y_i^m{\mathrm{\α}}_i^{m^{*}}K(x_i^m\·x)+b^{m^{*}}).$

For above-mentioned A, B, C, D process, m traversal value 1,2,3 ..., n-1, n, can obtain n state decision function, note by abridging and be $f_m\left(x\right)=\underset{i=1}{\overset{l}{\mathrm{\Σ}}}(y_i^m{\mathrm{\α}}_i^{m^{*}}K(x_i^m\·x)+b^{m^{*}}),$ m＝1，2，3，…，n。

For treating diagnostic sample x, first according to auxiliary contacts of circuit breaker judgement x, belong to separating brake proper vector or combined floodgate proper vector, when detecting when often driving auxiliary contact and being high level, isolating switch carries out feed motion, and judgement treats that diagnostic sample x is combined floodgate proper vector; When detecting when often driving auxiliary contact and being low level, isolating switch carries out separating brake action, and judgement treats that diagnostic sample x is separating brake proper vector; Then by n state decision function treating that diagnostic sample x substitution is corresponding m=1,2,3 ..., n.As state decision function f (x)=f _p(x) time, obtain maximal value, judgement sample x belongs to state p, wherein: p ∈ 1,2,3 ..., n}.

As shown in Figure 5, when the present invention works, after system power supply powers on, first executive system initialization, then upgrades man-machine interface information, when system clock integral point, read to time module time information renovation system time; If receive divide/reclosing command of host computer, to drive minute/combined floodgate electric control module is carried out instruction, each sensor starts sampling, then first displacement signal is adopted to simple Digital smooth filtering, calculate the divide/parameters such as closing speed, if there is parameter out-of-limit, just the out-of-limit data of correspondence are sent to host computer, upgrade man-machine interface simultaneously; Extraction vibration signal characteristics vector, result is inputted corresponding state decision function and is calculated, and analyzes circuit-breaker status information, if circuit breaker failure sends to host computer by failure message so, and preserves failure message.

In addition to the implementation, the present invention can also have other embodiments, and all employings are equal to the technical scheme of replacement or equivalent transformation formation, all drop in the protection domain of requirement of the present invention.

Claims (7)

1. a mechanical characteristic of high-voltage circuit breaker on-line monitoring system, it is characterized in that, comprise host computer (14), power module (12), lower computer system (15), described lower computer system (15) comprises linear displacement transducer (1), vibration transducer (2), signal conditioning circuit (3), A/D modular converter (4), photoelectric isolating circuit (5), driving isolation circuit (6), minute/combined floodgate electric control module (7), dsp processor (8), human-computer interface module (9), data memory module (10), system to time module (11), communication module (13), described linear displacement transducer (1) gathers the displacement signal of isolating switch driving stem, described linear displacement transducer (1), vibration transducer (2) is connected with signal conditioning circuit (3) respectively, the output terminal of described signal conditioning circuit (3) is connected with A/D modular converter (4), the output terminal of A/D modular converter (4), photoelectric isolating circuit (5), driving isolation circuit (6), human-computer interface module (9), system to time module (11), data memory module (10), communication module (13) is connected with dsp processor (8), the output terminal of described driving isolation circuit (6) is connected with divide/combined floodgate electric control module (7), communication module (13) is connected with host computer (14), power module (12) is connected with lower computer system (15) each several part, for its power supply.

2. mechanical characteristic of high-voltage circuit breaker on-line monitoring system according to claim 1, is characterized in that, described linear displacement transducer (1) is direct slide conductive plastic potentiometer, and model is WDL-50-2, and independent linearity reaches 0.1%.

3. mechanical characteristic of high-voltage circuit breaker on-line monitoring system according to claim 1 and 2, it is characterized in that, described linear displacement transducer (1) is arranged on the support in electromagnet of high-voltage circuit breaker outside, by driving stem, drives semi-girder to move together with linear displacement transducer (1).

4. mechanical characteristic of high-voltage circuit breaker on-line monitoring system according to claim 1, is characterized in that, described vibration transducer (2) is ZD24T type reluctance type vibration transducer.

5. according to the mechanical characteristic of high-voltage circuit breaker on-line monitoring system described in claim 1 or 4, it is characterized in that, described vibration transducer (2) is arranged on the pedestal that isolating switch approaches vibration source.

6. mechanical characteristic of high-voltage circuit breaker on-line monitoring system according to claim 1, is characterized in that, described system to time module (11) adopt X1226 time calendar clock.

7. a method for diagnosing faults for mechanical characteristic of high-voltage circuit breaker on-line monitoring system according to claim 1, is characterized in that, the method comprises the following steps:

(1) establish isolating switch and have n kind state, comprise state 1, state 2 ..., state m-1, state m, state m+1 ..., state n, separating brake under collection said n kind state and the vibration data sample of combined floodgate, vibration data under different conditions is divided into two groups, be separating brake vibration data sample group and combined floodgate vibration data sample group, the proper vector of two grouping vibration signals under each state is extracted in employing based on interval wavelet bag analytic approach, obtain respectively separating brake proper vector group, combined floodgate proper vector group; The described method based on interval wavelet bag analytic approach extraction vibration signal characteristics vector is as follows:

A. vibration signal is carried out to the decomposition of j layer, obtain sequence S _{j, k}, calculate each segmentation energy:

$Q_{i(j,k)}={\∫}_{t_{i-1}}^{t_i}{\left|A_i\left(t\right)\right|}^2\mathrm{dt}$

Wherein, get j=2, k=0,1,2,3, i represents the sequence number of segmentation after signal decomposition, i=1,2,3,4; t _i-1, t _ibe respectively i piecewise interval end points, A _i(t) be the amplitude of i segmentation;

B. each segmentation energy of normalized, obtains normalized value:

${\mathrm{\ϵ}}_{i(j,k)}=\frac{Q_{i(j,k)}}{{\∫}_{t_0}^{t_4}{\left|A\left(t\right)\right|}^2\mathrm{dt}}$

C. according to the theory of information entropy, calculate each segmentation wavelet-packet energy entropy:

$H_{(j,k)}=-\underset{i=1}{\overset{4}{\mathrm{\Σ}}}{\mathrm{\ϵ}}_{i(j,k)}\lg {\mathrm{\ϵ}}_{i(j,k)}$

D. the wavelet-packet energy entropy of each segmentation is formed to vibration signal characteristics vector:

x＝[H ₂₀，H ₂₁，H ₂₂，H ₂₃]

(2) try to achieve separating brake vibration signal characteristics under n kind state vector, combined floodgate vibration signal characteristics vector, then adopt " a pair of all the other " strategy of support vector machine theory, ask circuit-breaker status decision function; When using separating brake vibration signal characteristics Vector Groups Training Support Vector Machines, obtain gate-dividing state decision function; When using combined floodgate vibration signal characteristics Vector Groups Training Support Vector Machines, obtain "on" position decision function; The acquiring method following steps of described state decision function:

A. the vibration signal characteristics vector under n kind state is set up to proper vector Sample Storehouse, using the vibration signal characteristics vector of state m as a class with+1 mark, remaining n-1 kind vibration signal characteristics vector with-1 mark, builds new sample set T={ (x as all the other classes ₁, y ₁), (x ₂, y ₂) ..., (x _l, y _l), y _i∈+1 ,-1}, wherein: i=1,2 ..., l, the number that l is sample, x _irepresentation feature vector sample, y _ipresentation class mark, value+1 or-1;

B. select kernel function K (x, x _i) make K (x, x _i)=Φ (x) Φ (x _i), this method kernel function K (x, x _i) be gaussian kernel function wherein, Φ is the mapping that theorem in Euclid space arrives Hilbert space, the width parameter that σ is function, the radial effect scope of control function;

C. in constraint condition $\underset{i=1}{\overset{l}{\mathrm{\Σ}}}{y}_i{\mathrm{\α}}_i=0$ With 0≤α _iunder≤C, solve $\max W\left(\mathrm{\α}\right)=\underset{i=1}{\overset{l}{\mathrm{\Σ}}}{\mathrm{\α}}_i-\frac{1}{2}\underset{i=1}{\overset{l}{\mathrm{\Σ}}}\underset{j=1}{\overset{l}{\mathrm{\Σ}}}{\mathrm{\α}}_i{\mathrm{\α}}_j{y}_i{y}_{j}K(x_i,x_j)$ Obtain optimum weight vector wherein, C is penalty factor;

D. calculate $w^{*}=\underset{i=1}{\overset{l}{\mathrm{\Σ}}}{\mathrm{\α}}_i^{*}{y}_i{x}_i,$ Select α ^*a positive component meet calculate $b^{*}=y_j-\underset{i=1}{\overset{l}{\mathrm{\Σ}}}{y}_i{\mathrm{\α}}_i^{*}K(x_i\·x_j),$ Substitution f (x)=w ^*.x+b ^*, obtain state decision function $f\left(x\right)=\underset{i=1}{\overset{l}{\mathrm{\Σ}}}(y_i{\mathrm{\α}}_i^{*}K(x_i\·x)+b^{*}),$ Be designated as $f_m\left(x\right)=\underset{i=1}{\overset{l}{\mathrm{\Σ}}}(y_i^m{\mathrm{\α}}_i^{m^{*}}K(x_i^m\·x)+b^{m^{*}});$

For a, b, c, the d process of step (2), m traversal value 1,2,3 ..., n-1, n, can obtain n state decision function $f_m\left(x\right)=\underset{i=1}{\overset{l}{\mathrm{\Σ}}}(y_i^m{\mathrm{\α}}_i^{m^{*}}K(x_i^m\·x)+b^{m^{*}}),$ m＝1，2，3，…，n；

(3), for treating diagnostic sample x, first judge that x is separating brake vibration signal characteristics vector or combined floodgate vibration signal characteristics vector, then decision function corresponding to substitution; As state decision function f (x)=f _p(x) time, obtain maximal value, judgement treat that diagnostic sample x belongs to state p, wherein p ∈ 1,2,3 ..., n}.