CN105019482A - Method and system for on-line monitoring of stability of tunnel suspension fan foundation - Google Patents

Abstract

The invention relates to a method and system for on-line monitoring of the stability of a tunnel suspension fan foundation, and belongs to the technical field of stability monitoring of suspension equipment. The method includes the steps that an acceleration sensor is firmly installed on a fan pre-embedded steel plate, vibration information of a pre-embedded steel plate foundation is collected, and thus the accelerated speed information, the speed information and the displacement information of the pre-embedded steel plate foundation are acquired so as to comprehensively judge the damaged condition of the foundation; if the pre-embedded steel plate foundation is damaged, warning information is given out; and a health diagnosis model of the suspension fan foundation of a highway tunnel is established based on the manifold learning dimensionality simplifying and artificial intelligence pattern recognition theory, and the welding health condition between the pre-embedded steel plate and a pre-embedded steel bar is judged. According to the method and system, equipment operation is not affected in the monitoring process at all; the system principle for implementing the method is simple, operation is easy and convenient, test results are visual and reliable, and the requirement for guaranteeing safe operation of the highway tunnel is met.

Description

A kind of for tunnel stability of foundation of fan suspended on-line monitoring method and system

Technical field

The invention belongs to suspension equipment STABILITY MONITORING technical field, be specifically related to a kind of for tunnel stability of foundation of fan suspended on-line monitoring method and system.

Background technology

Along with developing rapidly of highway in China traffic, highway tunnel is built in a large number.To the end of the year 2013, ten thousand high pointes broken through in highway in China tunnel.The highway tunnel of growing up generally is provided with mechanical ventilator, and wherein more than 95% have employed suspension type longitudinal ventilation with jetblower, so the stability on the jet blower basis hung obtains the great attention of people gradually.

As shown in Figure 1, the mounting means of jet blower is generally that steel plate and embedded bar are welded to connect first at tunnel vault pre-embedded steel slab, then by assembling support welding on steel plate.Because blower fan is heavier, and be in operation and can produce certain vibrations, unavoidably harmful effect is produced to stability of foundation, so be necessary regularly or online to examine (prison) survey to the stability on built-in fitting basis.

Existing detection method is mainly periodic detection, generally has following three kinds of methods: (1) does anti-pulling test; (2) nondestructive inspection (ultrasonic wave or magnetic powder inspection) is adopted; (3) method for testing vibration.

And the equal existing defects of existing various detection method: (1) does anti-pulling test, and because blower fan build is comparatively large, general diameter is all more than one meter, and under the mounted condition of blower fan, pullout tests operation is very difficult, and observation is also very difficult.If unloaded by blower fan and remake pullout tests, workload is comparatively large, and the test period is also long.(2) nondestructive inspection is adopted, main detection pre-embedded steel slab and the reliability be connected between mounting bracket, and in fact comparatively difficulty is detected to the reliability be connected between pre-embedded steel slab with embedded bar, and the loosening situation between embedded bar and concrete can not be detected.(3) method for testing vibration, equipment cost is higher and require that testing staff has rich experience, and the method is mainly used in the periodic detection of tunnel foundation of fan suspended.

Summary of the invention

Given this, the object of the present invention is to provide a kind of for tunnel stability of foundation of fan suspended on-line monitoring method and system, the long-term on-line monitoring of tunnel foundation of fan suspended can be realized.

An object of the present invention is to provide a kind of for tunnel stability of foundation of fan suspended on-line monitoring method, and the method mainly has following two steps:

Step 1) by the on-line monitoring on pre-embedded steel slab basis, to the pre-embedded steel slab impaired timely alarm in basis;

Step 2) by the vibration-testing on pre-embedded steel slab basis, judge the health status of welding between pre-embedded steel slab with embedded bar.

Further, described step 1) realize mainly through following technical scheme, concrete steps are as follows:

Step 1-1) acceleration transducer collection pre-embedded steel slab base acceleration;

Step 1-2) vibration velocity that integration obtains pre-embedded steel slab basis is carried out to acceleration; Quadratic integral is carried out to acceleration and obtains pre-embedded steel slab basic displacement;

Step 1-3) by pre-embedded steel slab vibration acceleration, speed and shift value comprehensive distinguishing built-in fitting steel plate basis damage situations, there is sudden change (pulse signal) in such as certain moment embedded board acceleration and speed, and displacement is greater than threshold values; Then judge that this pre-embedded steel slab basis is impaired, give a warning.

Further, described step 2) mainly through setting up foundation of fan suspended in road tunnel health diagnostic model, the health status of welding between pre-embedded steel slab with embedded bar is diagnosed, model mainly comprises training stage and diagnostic phases, and its concrete steps are as follows:

Step 2-1) under experimental conditions, open fan suspended, after fan suspended operating steadily, obtain the vibration data on pre-embedded steel slab basis under different health status;

Step 2-2) extract temporal signatures and the frequency domain character of fan suspended pre-embedded steel slab foundation vibration, structure higher-dimension hybrid domain feature set;

Step 2-3) adopt LLTSA algorithm to carry out Dimensionality Reduction to higher-dimension hybrid domain feature set, obtain training sample d D feature vectors;

Step 2-4) under test conditions, open fan suspended, after fan suspended operating steadily, obtain the vibration data on pre-embedded steel slab basis; Repeat step 2-2) to 2-3), obtain test sample book d D feature vectors;

Step 2-5) by training sample d D feature vectors and test sample book d D feature vectors input nearest neighbor classifier, carry out state classification decision-making.

Further, described higher-dimension hybrid domain feature set comprises 11 time domain charactreristic parameter p ₁~ p ₁₁with 13 frequency domain character parameter p ₁₂~ p ₂₄; Time domain charactreristic parameter p ₁and p ₃~ p ₅for reflecting the size of time-domain signal amplitude and energy, time domain charactreristic parameter p ₂and p ₆~ p ₁₁for reflecting the time series distribution situation of time-domain signal; Frequency domain character parameter p ₁₂the size of reflection frequency domain vibrational energy; p ₁₃~ p ₁₅, p ₁₇and p ₂₁~ p ₂₄characterize dispersion or the intensity of frequency spectrum; p ₁₆and p ₁₈~ p ₂₀the change of reflection main band position.

Further, described step 2-3) specifically comprise the following steps:

Step 2-3-1) PCA projection, by principal component analysis PCA, data set is mapped to main body subspace; Higher-dimension hybrid domain feature set is mapped as R ^d(d<m) the data set Y=[y in space ₁, y ₂..., y _n],

Y＝A ^TX _ORGH _N，

Wherein, A is transition matrix, X _oRGfor Noise Data collection, H _n=I-ee ^tmatrix centered by/N, I is unit matrix, e to be all elements be all 1 N dimensional vector;

Step 2-3-2) determine neighborhood, build K-neighbour and scheme to find data point x _ineighborhood; Construct the distance matrix of all data points with Euclidean distance, then analyze distance matrix searching data point x _i(i=1,2 ... N) k Neighbor Points (j=1,2 ..., k);

Step 2-3-3) extract local message, calculate X _ih _kthe matrix V that forms of d characteristic vector corresponding to d eigenvalue of maximum _i, h _k=I-ee ^t/ k;

Step 2-3-4) construct permutation matrix, by the cumulative structural matrix B in local,

$B(I_i,I_i)\&LeftArrow;B(I_i,I_i)+W_i{W}_i^T,i=1,2,...,N$

Initialize B=0, I _i={ i ₁..., i _krepresent x _ithe indexed set of k Neighbor Points,

Step 2-3-5) calculate mapping, calculate characteristic value and the characteristic vector of following Generalized-grads Theory,

XH _NBH _NXα＝λXH _NX ^Tα

Wherein, with eigenvalue λ ₁< λ ₂< ... < λ _dcharacteristic of correspondence vector solution is α ₁, α ₂..., α _d, then A _lLTSA=(α ₁, α ₂..., α _d); Therefore transition matrix A=A _pCAa _lLTSA, then X → Y=A ^tx _oRGh _n, A _pCArepresent the transformation matrix of PCA.

Further, described step 2-5) specifically comprise the following steps:

Step 2-5-1) KNNC uses COS distance, as shown in the formula:

$Sim\left(v_1,v_2\right)=\frac{v_1\&CenterDot;v_2}{\left|\right|v_1\left|{|}_2\right|\left|v_2\right|{|}_2}=\frac{\underset{l=1}{\overset{d}{\&Sigma;}}{v}_{1l}\&times;v_{2l}}{\sqrt{\underset{l=1}{\overset{d}{\&Sigma;}}{v}_{1l}^2}\sqrt{\underset{l=1}{\overset{d}{\&Sigma;}}{v}_{2l}^2}}$

Wherein, d represents sample vector v ₁, v ₂dimension size;

Step 2-5-2) by using each neighborhood sample (training sample) and test sample book d ₀similitude come to the weighting of neighborhood class, as shown in the formula:

$score(v_0,C_i)=\underset{v_j\&Element;KNN\left(v_0\right)}{\&Sigma;}Sim(v_0,v_j)\mathrm{\&delta;}(v_j,C_i)$

Wherein, KNN (v ₀) represent test sample book v ₀k' nearest-neighbor collection, δ (v _j, C _i) represent neighborhood sample v _jabout classification C _igeneric attribute;

$\mathrm{\&delta;}(v_j,C_i)=\left\{\begin{array}{cc}1& v_j\&Element;C_i\\ 0& v_j\&NotElement;C_i\end{array}\right.$

Step 2-5-3) KNNC categorised decision rule is:

$\begin{array}{c}C=\arg {\max }_{C_i}\left(score\left(v_0,C_i\right)\right)\\ =\arg {\max }_{C_i}\left(\underset{v_j\&Element;KNN\left(v_0\right)}{\&Sigma;}Sim(v_0,v_j)\mathrm{\&delta;}(v_j,C_i)\right)\end{array}$

To test sample book v ₀all neighborhood class weights sums compare, classification C when wherein making class weights sum maximum is test sample book v ₀ownership class.

Two of object of the present invention is to provide a kind of for tunnel stability of foundation of fan suspended on-line monitoring system, and this system realizes mainly through following technical scheme:

A kind of for tunnel stability of foundation of fan suspended on-line monitoring system, this system comprises acceleration transducer, data acquisition instrument, communication system, supervisory control comuter; Described acceleration transducer is arranged on pre-embedded steel slab, for gathering the vibration information of pre-embedded steel slab; Data acquisition instrument is connected with acceleration transducer, vibration information is transferred to data acquisition instrument by acceleration transducer, data acquisition instrument to after vibration information process through communications system transmission to supervisory control comuter, supervisory control comuter carries out analyzing and processing to vibration information, judge that whether pre-embedded steel slab basis is impaired, the health status on monitoring pre-embedded steel slab basis.

Further, described communication system is wired or wireless communication system.

Owing to have employed technique scheme, the present invention has following advantage:

The present invention can hang or the reason such as inner solder joint is impaired and the abnormal conditions that are drawn out because blower fan is wiped by vehicle on on-line monitoring built-in fitting basis for a long time, and alarm; Can on-line checkingi pre-embedded steel slab and the health status of welding between embedded bar.Observation process does not affect the operation of equipment completely, and the system principle realizing the method is simple, easy and simple to handle, result of the test is intuitive and reliable, meets the needs ensureing that Road Tunnel Safety is run.

Accompanying drawing explanation

In order to make the object, technical solutions and advantages of the present invention clearly, below in conjunction with accompanying drawing, the present invention is described in further detail, wherein:

Fig. 1 is jet blower scheme of installation;

Fig. 2 is system construction drawing of the present invention;

Fig. 3 is the model framework chart of the method for the invention.

Detailed description of the invention

Below in conjunction with accompanying drawing, the preferred embodiments of the present invention are described in detail.

The invention provides a kind of for tunnel stability of foundation of fan suspended on-line monitoring method, specifically comprise the following steps:

Step 1) by the on-line monitoring on pre-embedded steel slab basis, judge that whether pre-embedded steel slab basis is impaired, pre-embedded steel slab is hung or the reason such as inner solder joint is with the passing of time impaired and the timely alarm of abnormal conditions that is drawn out because blower fan is wiped by vehicle.

Step 1-1) acceleration transducer collection pre-embedded steel slab base acceleration;

Step 1-2) integration is carried out to pre-embedded steel slab base acceleration obtain its velocity amplitude; Carry out the shift value that quadratic integral obtains pre-embedded steel slab basis;

Step 1-3) according to embedded board acceleration, speed and displacement information comprehensive distinguishing embedded board basis damage situations.Under confirming that built-in fitting basis is stable healthy condition, install native system, under encouraging (embedded board is without forced vibration) situation without external environment, acceleration transducer friction information, namely filtered noise post-acceleration, speed and displacement are all zero.If there is sudden change (pulse type signal) in the acceleration in a certain moment and speed, and displacement is greater than threshold values (as 2mm); Then judge that this pre-embedded steel slab basis is impaired, give a warning, overhaul.

Step 2) by the vibration-testing on pre-embedded steel slab basis, judge the health status of welding between pre-embedded steel slab with embedded bar.

For judging embedded board and the health status of welding between embedded bar, theory is identified based on popular study dimension abbreviation and artificial intelligence pattern, set up foundation of fan suspended in road tunnel health diagnostic model as shown in Figure 3, mainly comprise characteristic parameter extraction, huge feature set yojan and state recognition three aspects.The enforcement of model is mainly divided into two stages: training stage and diagnostic phases.Training stage carries out under experimental conditions, and main extraction blower foundation embedded board and the different health status of welding between embedded bar (fastening and loosening) feature, identify basis for diagnostic phases provides; Diagnostic phases is the field diagnostic stage, and concrete implementation step is as follows:

Step 2-1) under experimental conditions, open fan suspended, after fan suspended operating steadily, obtain the vibration data on pre-embedded steel slab basis under different health status.

Step 2-2) extract temporal signatures and the frequency domain character of fan suspended pre-embedded steel slab foundation vibration, structure higher-dimension hybrid domain feature set.

Step 2-3) adopt LLTSA algorithm to carry out Dimensionality Reduction to higher-dimension hybrid domain feature set, obtain training sample d D feature vectors.For d sets bound, namely require d _min≤ d≤d _max, (1≤d _min≤ d _max≤ n and d _min, d _max∈ Z ⁺).Best d value is obtained by iteration optimizing, and the d value making this loosening state recognition model reach the highest accuracy of identification is exactly best d value.

Step 2-4) under test conditions, open fan suspended, after fan suspended operating steadily, obtain the vibration data on pre-embedded steel slab basis; Repeat step 2-2) to 2-3), obtain test sample book d D feature vectors.

Step 2-5) by training sample d D feature vectors and test sample book d D feature vectors input nearest neighbor classifier, carry out state classification decision-making, obtain the health status of welding between embedded board with embedded bar.

1 characteristic parameter

In order to obtain the more information accurately reflecting blower fan built-in fitting state comprehensively, this method comprehensive utilization time domain and frequency domain character parameter, have selected 11 time domain charactreristic parameter (p ₁~ p ₁₁) and 13 frequency domain character parameter (p ₁₂~ p ₂₄) be configured to the essential characteristic of higher-dimension hybrid domain feature set as blower fan built-in fitting state.The structure of 11 time domain charactreristic parameters and 13 frequency domain character parameters as shown in Table 1 and Table 2.

Table 1 time domain charactreristic parameter

Note: in formula, x (n) is time-domain signal sequence, n=1,2 ..., N, N are sample points.Time domain charactreristic parameter p ₁and p ₃~ p ₅the size of reflection time-domain signal amplitude and energy; p ₂and p ₆~ p ₁₁the time series distribution situation of reflection time-domain signal.

Table 2 frequency domain character parameter

Note: s (k) is the frequency spectrum of signal x (n), k=1,2 ..., K, K are spectral line number, f _kit is the frequency values of kth bar spectral line; Frequency domain character parameter p ₁₂the size of reflection frequency domain vibrational energy; p ₁₃~ p ₁₅, p ₁₇and p ₂₁~ p ₂₄characterize dispersion or the intensity of frequency spectrum; p ₁₆and p ₁₈~ p ₂₀the change of reflection main band position.

2 feature set Dimensionality Reduction methods

For carrying out effective yojan to the higher-dimension hybrid domain feature set containing redundancy, interfere information, to obtain that dimension is low, sensitiveness is high and the principal character vector that classification error rate is little, this method adopts linear local tangent space alignment (LLTSA) manifold learning, for the state recognition of blower fan built-in fitting provides the rapidity Dimensionality Reduction means that well sort feature is strong simultaneously.The concrete implementation step of the method is as follows:

1) PCA projection

Huge feature set Dimensionality Reduction can be expressed as finds transition matrix A by R ^mthere is in space the Noise Data collection X of N number of point _oRG(fault sample collection) is mapped as R ^d(d<m) the data set Y=[y in space ₁, y ₂..., y _n], namely

Y＝A ^TX _ORGH _N

Wherein H _n=I-ee ^tmatrix centered by/N, I is unit matrix, e to be all elements be all 1 N dimensional vector.Y is just X _oRGpotential d ties up non-linearity manifold.

X in practical problem _oRGh _nx _oRG ^tnormal is singular matrix, and this quantity (i.e. the fault sample number of this paper) coming from data point is far smaller than the dimension (i.e. the data length of this paper fault sample signal) of data.Use principal component analysis (PCA) data set to be mapped to main body subspace and can overcome matrix X _oRGh _nx _oRG ^tsingularity, PCA pretreatment also can realize noise reduction in addition.X is below used to represent PCA subspace data set.Use A _pCArepresent the transformation matrix of PCA.

2) neighborhood is determined

Data point x is found by building stability good K-neighbour (KNN) figure _ineighborhood, namely first construct the distance matrix of all data points with Euclidean distance, then analyze distance matrix find data point x _i(i=1,2 ... N) k Neighbor Points $x_{i_j}(j=1,2,...,k).$

3) local message is extracted

Calculate by the matrix V that forms of d characteristic vector corresponding to d eigenvalue of maximum _i.Wherein H _k=I-ee ^t/ k.

4) permutation matrix is constructed

As follows by the cumulative structural matrix B in local:

$B(I_i,I_i)\&LeftArrow;B(I_i,I_i)+W_i{W}_i^T,(i=1,2,...,N)$

Initialize B=0, in formula, I _i={ i ₁..., i _krepresent x _ithe indexed set of k Neighbor Points,

5) mapping is calculated

Calculate characteristic value and the characteristic vector of following Generalized-grads Theory

XH _NBH _NXα＝λXH _NX ^Tα

With eigenvalue λ ₁< λ ₂< ... < λ _dcharacteristic of correspondence vector solution is α ₁, α ₂..., α _d, then A _lLTSA=(α ₁, α ₂..., α _d).Therefore transition matrix is as follows: A=A _pCAa _lLTSA, then X → Y=A ^tx _oRGh _n.

3 state recognition algorithms

Feature set Dimensionality Reduction export sort feature good low-dimensional data collection be all not easily allow people understand and accept, very abstract characteristic vector form, for intuitively expressing the recognition result of welding health state between embedded board and embedded bar, the characteristic vector after dimensionality reduction and the mapping relations between health status pattern must be set up by mode identification technology.This method adopts nearest neighbor classifier (KNNC) mode identification method for this reason, for the identification of blower fan built-in fitting health status provides improvement, efficient, stable mode identification method.

Nearest neighbor classifier specific algorithm is as follows: for a unknown sample v ₀classify, KNNC uses the class label of K' nearest-neighbor in training sample to predict v ₀class ownership.For effectively measuring similitude, KNNC uses following COS distance:

$Sim\left(v_1,v_2\right)=\frac{v_1\&CenterDot;v_2}{\left|\right|v_1\left|{|}_2\right|\left|v_2\right|{|}_2}=\frac{\underset{l=1}{\overset{d}{\&Sigma;}}{v}_{1l}\&times;v_{2l}}{\sqrt{\underset{l=1}{\overset{d}{\&Sigma;}}{v}_{1l}^2}\sqrt{\underset{l=1}{\overset{d}{\&Sigma;}}{v}_{2l}^2}}$

Wherein d represents sample vector v ₁, v ₂dimension size.

By using each neighborhood sample (training sample) and v ₀similitude come to the weighting of neighborhood class, as shown in the formula:

$score(v_0,C_i)=\underset{v_j\&Element;KNN\left(v_0\right)}{\&Sigma;}Sim(v_0,v_j)\mathrm{\&delta;}(v_j,C_i)$

Wherein KNN (v ₀) represent sample v ₀k' nearest-neighbor collection.δ (v _j, C _i) represent neighborhood sample v _jabout classification C _igeneric attribute, that is:

$\mathrm{\&delta;}(v_j,C_i)=\left\{\begin{array}{cc}1& v_j\&Element;C_i\\ 0& v_j\&NotElement;C_i\end{array}\right.$

Therefore, KNNC categorised decision rule is:

$\begin{array}{c}C=\arg {\max }_{C_i}\left(score\left(v_0,C_i\right)\right)\\ =\arg {\max }_{C_i}\left(\underset{v_j\&Element;KNN\left(v_0\right)}{\&Sigma;}Sim(v_0,v_j)\mathrm{\&delta;}(v_j,C_i)\right)\end{array}$

Namely to v ₀all neighborhood class weights sums compare, classification C when wherein making class weights sum maximum is v ₀ownership class.

A kind of for tunnel stability of foundation of fan suspended on-line monitoring system, this system comprises acceleration transducer, data acquisition instrument, communication system, supervisory control comuter; Described acceleration transducer is arranged on pre-embedded steel slab, for gathering the vibration information of pre-embedded steel slab; Data acquisition instrument is connected with acceleration transducer, vibration information is transferred to data acquisition instrument by acceleration transducer, data acquisition instrument to after vibration information process through communications system transmission to supervisory control comuter, supervisory control comuter carries out analyzing and processing to vibration information, judge that whether pre-embedded steel slab basis is impaired, the health status on monitoring pre-embedded steel slab basis.Communication system can be wired or wireless communication system.

What finally illustrate is, above preferred embodiment is only in order to illustrate technical scheme of the present invention and unrestricted, although by above preferred embodiment to invention has been detailed description, but those skilled in the art are to be understood that, various change can be made to it in the form and details, and not depart from claims of the present invention limited range.

Claims (9)

1., for a tunnel stability of foundation of fan suspended on-line monitoring method, it is characterized in that: the method comprises the following steps:

Step 1), by the on-line monitoring on pre-embedded steel slab basis, judges that whether pre-embedded steel slab basis is impaired, if impaired, gives a warning;

Step 2) by the vibration-testing on pre-embedded steel slab basis, judge the health status of welding between pre-embedded steel slab with embedded bar.

2. one according to claim 1 is used for tunnel stability of foundation of fan suspended on-line monitoring method, it is characterized in that:

Described step 1) specifically comprises the following steps:

Step 1-1) acceleration transducer collection pre-embedded steel slab base acceleration;

Step 1-2) an integration acquisition pre-embedded steel slab basal rate is carried out to pre-embedded steel slab base acceleration; Quadratic integral is carried out to pre-embedded steel slab base acceleration and obtains pre-embedded steel slab basic displacement;

Step 1-3) by pre-embedded steel slab base acceleration, speed and shift value comprehensive distinguishing built-in fitting steel plate basis damage situations, if impaired, give a warning.

3. one according to claim 2 is used for tunnel stability of foundation of fan suspended on-line monitoring method, it is characterized in that:

Described step 1-3) when certain moment embedded board acceleration and speed appearance sudden change, and displacement is greater than threshold values; Then judge that this pre-embedded steel slab basis is impaired.

4. one according to claim 1 is used for tunnel stability of foundation of fan suspended on-line monitoring method, it is characterized in that:

Described step 2) by setting up foundation of fan suspended in road tunnel health diagnostic model, the health status of welding between pre-embedded steel slab with embedded bar to be diagnosed, model comprises training stage and diagnostic phases, specifically comprises the following steps:

Step 2-1) under experimental conditions, open fan suspended, after fan suspended operating steadily, obtain the vibration data on pre-embedded steel slab basis under different health status;

Step 2-2) extract temporal signatures and the frequency domain character of fan suspended pre-embedded steel slab foundation vibration, structure higher-dimension hybrid domain feature set;

Step 2-3) adopt LLTSA algorithm to carry out Dimensionality Reduction to higher-dimension hybrid domain feature set, obtain training sample d D feature vectors;

Step 2-4) under test conditions, open fan suspended, after fan suspended operating steadily, obtain the vibration data on pre-embedded steel slab basis; Repeat step 2-2) to 2-3), obtain test sample book d D feature vectors;

Step 2-5) by training sample d D feature vectors and test sample book d D feature vectors input nearest neighbor classifier, carry out state classification decision-making.

5. one according to claim 3 is used for tunnel stability of foundation of fan suspended on-line monitoring method, it is characterized in that:

Described higher-dimension hybrid domain feature set comprises 11 time domain charactreristic parameter p ₁~ p ₁₁with 13 frequency domain character parameter p ₁₂~ p ₂₄; Time domain charactreristic parameter p ₁and p ₃~ p ₅for reflecting the size of time-domain signal amplitude and energy, time domain charactreristic parameter p ₂and p ₆~ p ₁₁for reflecting the time series distribution situation of time-domain signal; Frequency domain character parameter p ₁₂the size of reflection frequency domain vibrational energy; p ₁₃~ p ₁₅, p ₁₇and p ₂₁~ p ₂₄characterize dispersion or the intensity of frequency spectrum; p ₁₆and p ₁₈~ p ₂₀the change of reflection main band position.

6. one according to claim 3 is used for tunnel stability of foundation of fan suspended on-line monitoring method, it is characterized in that:

Described step 2-3) specifically comprise the following steps:

Step 2-3-1) PCA projection, by principal component analysis PCA, data set is mapped to main body subspace; Higher-dimension hybrid domain feature set is mapped as R ^d(d<m) the data set Y=[y in space ₁, y ₂..., y _n],

Y=A ^TX _ORGH _N

Wherein, A is transition matrix, X _oRGfor Noise Data collection, H _n=I-ee ^tmatrix centered by/N, I is unit matrix, e to be all elements be all 1 N dimensional vector;

Step 2-3-2) determine neighborhood, build K-neighbour and scheme to find data point x _ineighborhood; Construct the distance matrix of all data points with Euclidean distance, then analyze distance matrix searching data point x _i(i=1,2 ... N) k Neighbor Points $x_{i_j}(j=1,2,...,k);$

Step 2-3-3) extract local message, calculate X _ih _kthe matrix V that forms of d characteristic vector corresponding to d eigenvalue of maximum _i, h _k=I-ee ^t/ k;

Step 2-3-4) construct permutation matrix, by the cumulative structural matrix B in local,

$B(I_i,I_j)\&LeftArrow;B(I_i,I_j)+W_i{W_i}^T,i=1,2,...,N$

Initialize B=0, I _i={ i ₁..., i _krepresent x _ithe indexed set of k Neighbor Points, i=1,2 ..., N;

Step 2-3-5) calculate mapping, calculate characteristic value and the characteristic vector of following Generalized-grads Theory,

XH _NBH _NXα=λXH _NX ^Tα

Wherein, with eigenvalue λ ₁< λ ₂< ... < λ _dcharacteristic of correspondence vector solution is α ₁, α ₂..., α _d, then A _lLTSA=(α ₁, α ₂..., α _d); Therefore transition matrix A=A _pCAa _lLTSA, then X → Y=A ^tx _oRGh _n, A _pCArepresent the transformation matrix of PCA.

7. one according to claim 3 is used for tunnel stability of foundation of fan suspended on-line monitoring method, it is characterized in that:

Described step 2-5) specifically comprise the following steps:

Step 2-5-1) KNNC uses COS distance, as shown in the formula:

$Sim(v_1,v_2)=\frac{v_1\&CenterDot;v_2}{\left|\right|v_1\left|{|}_2\right|\left|v_2\right|{|}_2}=\frac{\underset{l=1}{\overset{d}{\mathrm{\&Sigma;}}}{v}_{1l}\&times;v_{2l}}{\sqrt{\underset{l=1}{\overset{d}{\mathrm{\&Sigma;}}}{v}_{1l}^2}\sqrt{\underset{l=1}{\overset{d}{\mathrm{\&Sigma;}}}{v}_{2l}^2}}$

Wherein, d represents sample vector v ₁, v ₂dimension size;

Step 2-5-2) by using each training sample and test sample book v ₀similitude come to the weighting of neighborhood class, as shown in the formula:

$score(v_0,C_i)=\underset{v_j\&Element;KNN\left(v_0\right)}{\&Sigma;}Sim(v_0,v_j)\mathrm{\&delta;}(v_j,C_i)$

Wherein, KNN (v ₀) represent test sample book v ₀k' nearest-neighbor collection; δ (v _j, C _i) represent neighborhood sample v _jabout classification C _igeneric attribute,

$\mathrm{\&delta;}(v_j,C_i)=\left\{\begin{array}{cc}1& v_j\&Element;C_i\\ 0& v_j\&NotElement;C_i\end{array}\right.$

Step 2-5-3) KNNC categorised decision rule is:

$\begin{array}{cc}C=\arg & {\max }_{C_i}\left(score\right(v_0,C_i))\\ =\arg & {\max }_{C_i}\left(\underset{v_j\&Element;KNN\left(v_0\right)}{\mathrm{\&Sigma;}}\mathrm{si}m\right(v_0,v_j)\mathrm{\&delta;}(v_j,C_i))\end{array}$

To test sample book v ₀all neighborhood class weights sums compare, classification C when wherein making class weights sum maximum is test sample book v ₀ownership class.

8. for a tunnel stability of foundation of fan suspended on-line monitoring system, it is characterized in that: this system comprises acceleration transducer, data acquisition instrument, communication system, supervisory control comuter; Described acceleration transducer is arranged on pre-embedded steel slab, for gathering the vibration information of pre-embedded steel slab; Data acquisition instrument is connected with acceleration transducer, vibration information is transferred to data acquisition instrument by acceleration transducer, data acquisition instrument to after vibration information process through communications system transmission to supervisory control comuter, supervisory control comuter carries out analyzing and processing to vibration information, judge that whether pre-embedded steel slab basis is impaired, the health status on monitoring pre-embedded steel slab basis.

9. one according to claim 8 is used for tunnel stability of foundation of fan suspended on-line monitoring system, it is characterized in that:

Described communication system is wired or wireless communication system.