CN105093285A - Artificial intelligence earthquake judgment method and earthquake detection system - Google Patents

Abstract

The invention discloses an earthquake detection system, and an earthquake judgment method used for the earthquake detection system. The method comprises: extracting at least one primary wave characteristic relevant to each piece of earthquake data from a plurality of pieces of earthquake data; utilizing a support vector classification (SVC) method to establish an earthquake judgment model according to the at least one primary wave characteristic; and determining whether a new piece of earthquake data belongs to an earthquake event or a non earthquake event according to the earthquake judgment model when receiving the new piece of earthquake data. The method can improve the accuracy of an on-site earthquake warning system, and furthermore improve earthquake warning effects.

Description

Artificial intelligence earthquake determination methods and earthquake detecting system

Technical field

The present invention relates to a kind of earthquake determination methods and earthquake detecting system, particularly relate to a kind of can according to first earthquake determination methods and the relative earthquake detecting system thereof reaching wave characteristic to judge received geological data and belong to seismic events or non-seismic events of earthquake.

Background technology

In recent years, due to the progress of the science and technology such as seismology, digital communication, automatic business processing and algorithm, the development of early earthquake early warning (EarthquakeEarlyWarning, EEW) technology is ripe gradually.According to the difference of used earthquake information, early earthquake early warning technology can be categorized as following two kinds of modes: domain type early warning (regionalwarning) and now type early warning (on-sitewarning).In general, because domain type earthquake early-warning system can detect the information at station to carry out estimating of seismologic parameter with reference to earthquake several near epicenter simultaneously, therefore, compared to existing ground type earthquake early-warning system, domain type earthquake early-warning system has higher accuracy.But, because the region near epicenter can produce higher earthquake degree usually, and near with the distance of epicenter, therefore before destructive seismic wave arrives, the time that domain type earthquake early-warning system can precompute seismologic parameter discreet value is very limited.On the other hand, because existing ground type earthquake early-warning system is only with reference to the information at single earthquake detecting station, therefore can seismologic parameter be provided more quickly to estimate, the position particularly near epicenter, early warning more rapidly can be reached.

Recent research finds, now type earthquake early-warning system may the signal that causes by some non-seismic events trigger, thus produce the earthquake warning of mistake.Therefore, the resolution of true earthquake and non-seismic events has become a major issue.A kind of existing settling mode is that employing two seismic sensors are placed on diverse location, respectively to carry out duplicate acknowledgment.But aforesaid way not only cost is higher, also improve the difficulty that earthquake detecting station builds and safeguards.Therefore, be necessary in fact to propose the more effective and cost-effective method of one, to improve the accuracy of existing ground type earthquake early-warning system, and then lifting earthquake pre-warning effect.

Summary of the invention

Therefore, namely fundamental purpose of the present invention is that providing a kind of can reach wave characteristic according to the first of earthquake, and the earthquake judgment models utilizing support vector classification (SupportVectorClassification, SVC) method to set up judges that received geological data belongs to earthquake determination methods and the relative earthquake detecting system thereof of seismic events or non-seismic events.

The present invention discloses a kind of earthquake determination methods, and for an earthquake detecting system, the method is included in multiple geological data, and taking-up is relevant at least one of each geological data and just reaches wave characteristic; Reach wave characteristic at the beginning of at least one according to this, utilize a support vector classification to set up an earthquake judgment models; And when the new geological data of reception one, according to this earthquake judgment models, judge that this new geological data belongs to a seismic events or a non-seismic events.

The present invention also discloses a kind of earthquake detecting system, comprises an earthquake detecting module, is used in multiple geological data, and taking-up is relevant at least one of each geological data and just reaches wave characteristic; One computing module, reaches wave characteristic at the beginning of at least one according to this, utilizes a support vector classification to set up an earthquake judgment models; And an earthquake judge module, be used for, when this earthquake detecting module receives a new geological data, according to this earthquake judgment models, judging that this new geological data belongs to a seismic events or a non-seismic events.

Accompanying drawing explanation

Fig. 1 is the schematic diagram of the embodiment of the present invention one earthquake detecting system.

Fig. 2 is the schematic diagram that the geological data belonging to seismic events is undertaken by earthquake judgment models of the present invention judging.

Fig. 3 is the schematic diagram that the geological data belonging to non-seismic events is undertaken by earthquake judgment models of the present invention judging.

Fig. 4 is the schematic diagram that the embodiment of the present invention one earthquake judges flow process.

Wherein, description of reference numerals is as follows:

10 earthquake detecting systems

102 earthquake detecting modules

104 computing modules

106 earthquake judge modules

Q ₁~ Q _lgeological data

The new geological data of QN

F (x) earthquake judgment models

40 flow processs

400 ~ 408 steps

Embodiment

Please refer to Fig. 1, Fig. 1 is the schematic diagram of the embodiment of the present invention one earthquake detecting system 10.As shown in Figure 1, earthquake detecting system 10 comprises earthquake detecting module 102, computing module 104 and an earthquake judge module 106.Earthquake detecting module 102 can be used at multiple geological data Q ₁~ Q _lin, taking-up is relevant at least one of each geological data and just reaches wave characteristic.Reaching wave characteristic at the beginning of this and can be the physical quantity reaching ripple at the beginning of any being relevant to, as speed, acceleration or displacement etc., its object is to the feature by just reaching ripple, before principal earthquake ripple arrives, realize early warning.In one embodiment, just reach wave characteristic be included in just reach ripple arrive after in a period of time detect the accumulation absolute speed (CumulativeAbsoluteVelocity of earthquake motion, CAV), absolute speed integration (IntegralofAbsoluteVelocity, and absolute displacement integration (IntegralofAbsoluteDisplacement, IAD) IAV).Just reach wave characteristic and can be used as the critical value judging whether earthquake occurs, for example, when the accumulation absolute speed of the earthquake motion that earthquake detecting system 10 detects is more than a critical value, the earthquake that namely expressed possibility occurs.In one embodiment, the real-time detected for asking earthquake also saves the number of seismic sensor, only can adopt the vertical component of earthquake motion, with when just to reach ripple and arrive, measure earthquake motion in the various characteristics with earth's surface vertical direction as just reaching wave characteristic, as accumulation absolute speed, absolute speed integration and absolute displacement integration etc.But, in another embodiment, the component of earthquake motion in other direction also can be adopted as just reaching wave characteristic, and be not limited thereto.In addition, for asking the integrality of earthquake statistics data, also can using multiple seismic sensor, measuring the component of earthquake motion in multiple directions as just reaching wave characteristic simultaneously.

Please continue to refer to Fig. 1.Computing module 104 can reach wave characteristic according to first acquired by earthquake detecting module 102, utilizes support vector classification (SupportVectorClassification, a SVC) method to set up an earthquake judgment models.According to this earthquake judgment models, when earthquake detecting module 102 receives a new geological data QN, earthquake judge module 106 can judge that new geological data QN belongs to a seismic events or a non-seismic events.By support vector classification, the accuracy that existing ground type earthquake early-warning system judges seismic events can be improved.Thus, except can possessing the real-time of existing ground type earthquake early-warning system, non-seismic events also can be avoided to cause the erroneous judgement of earthquake.

Specifically, in support vector classification, geological data Q is relevant to ₁~ Q _lthe first wave characteristic that reaches may correspond to multiple vector x ₁~ x _l, wherein, each vector x _iall may correspond to a desired value y _iand y _i∈ { 1 ,-1}.If correspond to x _igeological data Q _iby seismic events caused time, its desired value can be set as y _i=1; If correspond to x _igeological data Q _iby non-seismic events caused time, its desired value can be set as y _i=-1.In general, about geological data Q ₁~ Q _lby seismic events or non-seismic events to cause be the statistics coming from historical data, support vector classification be namely according to historical data take out corresponding to seismic events or non-seismic events to just reach wave characteristic, an earthquake judgment models is produced by training (training), and then when receiving new geological data QN, what judge new geological data QN according to earthquake judgment models just reaches wave characteristic compared with reaching wave characteristic close to seismic events in historical data first or the first of non-seismic events reaches wave characteristic, and then judge that new geological data QN is caused by seismic events or non-seismic events.

According to support vector classification, vector x ₁~ x _lcan project a high-dimensional feature space H, and classify at high-dimensional feature space H, and computing module 104 definable earthquake judgment models is function f (x), is used for judging that new geological data QN belongs to seismic events or non-seismic events:

$f\left(x\right)=\mathrm{sgn}[\underset{i,j=1}{\overset{l}{\mathrm{\Σ}}}{y}_i{\mathrm{\α}}_{i}K(x_i,x_j)+b]$

Wherein, K corresponds to the kernel function of high-dimensional feature space H and K (x _i, x _j) ≡ φ (x _i) ^tφ (x _j).α ₁~ α _iand b is constant, can according to support vector classification, by corresponding to geological data Q ₁~ Q _ivector x ₁~ x _land desired value y ₁~ y _lcalculate and obtain.In support vector method, parameter alpha ₁~ α _land b obtains by solving following objective function:

$\underset{w,b,\mathrm{\ξ}}{\min }[\frac{1}{2}{w}^{T}w+C\underset{i=1}{\overset{l}{\mathrm{\Σ}}}{\mathrm{\ξ}}_i]$

Be limited to

y _i(w ^Tφ(x _i)+b)≥1-ξ _i，ξ _i≥0，i＝1，...，l

Wherein, w is a vector in high-dimensional feature space H, ξ ₁~ ξ _land b is the variable of objective function, and φ is by vector x ₁~ x _lcorrespond to a function of high-dimensional feature space H.The antithesis pattern of this objective function is as follows:

$\underset{\mathrm{\α}}{\min }[\frac{1}{2}{\mathrm{\α}}^T\mathrm{Q\α}-e^T\mathrm{\α}]$

Be limited to

y ^Tα＝0，0≤α _i≤C，i＝1，...，l

Wherein, e is a vector of unit length, C > 0 and be α _ithe upper limit, Q is l × l positive semidefinite matrix and Q _ij≡ y _iy _jk (x _i, x _j).According to above-mentioned formula, input new geological data QN first reach corresponding to wave characteristic to vector calculate, f (x)=1 or two kinds, f (x)=-1 result can be obtained, when f (x)=1, represent new geological data QN and belong to seismic events, when f (x)=-1, represent new geological data QN and belong to non-seismic events.

In other words, according to being relevant to geological data Q in historical data ₁~ Q _ljust reach wave characteristic, computing module according to support vector classification, can go out parameter alpha by above formulae discovery ₁~ α _land b, and then set up earthquake judgment models f (x).When earthquake detecting module 102 detects new geological data QN, the first wave characteristic that reaches of new geological data QN can be converted to specific vector, earthquake judge module 106 can according to this specific vector and earthquake judgment models f (x), calculate f (x)=1 or-1, and then judge that new geological data QN caused by a seismic events or by a non-seismic events, be mistaken for earthquake with the earthquake motion avoiding non-seismic events to produce.

It should be noted that in earthquake detecting system 10, earthquake detecting module 102, computing module 104 and earthquake judge module 106 may lay respectively at different regions, and by wired or wireless network delivery information.For example, earthquake detecting module 102 is not limited to single earthquake detecting station or earthquake arrangement for detecting, and it also may comprise the combination that station or earthquake arrangement for detecting are detected in multiple earthquake being positioned at different location.In general, the earthquake detecting station in earthquake detecting module 102 or earthquake arrangement for detecting can be arranged on the area be subject to earthquakes, to detect new geological data QN rapidly.Computing module 104 may be positioned at geological data center to obtain a large amount of geological data Q ₁~ Q _l, to increase the precision of earthquake judgment models f (x).Earthquake judge module 106 may be positioned at alarm center, so as judge make new advances geological data QN belong to a seismic events time, determine whether issue earthquake warning immediately.

In addition, the above-mentioned formula being relevant to support vector classification is only the one in the middle of the numerous embodiment of the present invention, is not used to limit category of the present invention.Those skilled in the art is when using other mathematical formulaes and the support vector classification and solve and can judge whether new geological data belongs to the formula of seismic events of arranging in pairs or groups.

For further illustrating effect of the present invention, obtain by simulation analysis the accuracy that earthquake judgment models f (x) judges geological data and Non-seismology data.In general, computing module 104 be used for train earthquake judgment models f (x) geological data Q ₁~ Q _lthe geological data belonging to seismic events and the geological data belonging to non-seismic events must be comprised simultaneously.The geological data belonging to seismic events can come from the data of the actual generation earthquake that Central Weather Bureau of the Republic of China records, or by the acquired data also having confirmed as actual generation earthquake of an earthquake early warning system (EarthquakeEarlyWarningSystem, EEWS).The geological data belonging to non-seismic events then may comprise the class geological data or Non-seismology data that earthquake early warning system obtains, or comes from the geological data of earthquake early warning system but be not considered as the related data of seismic events by Central Weather Bureau of the Republic of China.It should be noted that, in Taiwan, because Central Weather Bureau of the Republic of China has complete earthquake detecting station arrangement and earthquake detecting system, the geological data that accuracy is high can be obtained, therefore, the geological data of Central Weather Bureau of Republic of China record all can be considered the actual situation having seismic events to occur.

In one embodiment, earthquake detecting system 10 adopts geological data Q ₁~ Q _lin, the first vertical component reaching ripple and occur the acceleration of latter first 3 seconds of each geological data, then carried out integration generation speed and displacement information, and using these physical quantitys as just reaching wave characteristic, and then set up earthquake judgment models f (x).Then, 91142 data are altogether obtained July 29 1992 Christian era to 31 days Dec in 2006 of Christian era by the seismologic record of Central Weather Bureau of the Republic of China, regather the geological data having confirmed to belong to seismic events and non-seismic events in several earthquake detecting station, be respectively 54 and 6743.Therefore, always have the data that 91196 data belonging to seismic events and 6743 belong to non-seismic events, simplation verification is carried out in earthquake judgment models f (x) of being set up by the present invention.

Please refer to Fig. 2, Fig. 2 is the schematic diagram that the geological data belonging to seismic events is judged by earthquake judgment models f (x) of the present invention carrying out.In order to obtain the judged result of the earthquake of different earthquake degree (Intensity, I), this embodiment is arranged in pairs or groups simultaneously and is relevant to maximum acceleration surface (PeakGroundAcceleration, the PGA) data analysis of earthquake degree size.As shown in Figure 2, according to the judgement of earthquake judge module 106, when earthquake judgment models f (x) exports f (x)=1, represent geological data and be judged as seismic events; When earthquake judgment models f (x) exports f (x)=-1, represent geological data and be judged as non-seismic events.Belong in the geological data of seismic events at above-mentioned 91196, always have 90652 (about 99.4035%) geological datas and be correctly judged as seismic events, and 544 (about 0.59652%) geological datas are wrongly judged as non-seismic events.Therefore, for the geological data belonging to seismic events, earthquake judgment models f (x) of the present invention has the accuracy up to 99.4%.In addition, for all geological datas of maximum acceleration surface more than 200 gals (gal), earthquake judgment models f (x) of the present invention all can correctly judge as seismic events.

Please refer to Fig. 3, Fig. 3 is the schematic diagram that the geological data belonging to non-seismic events is judged by earthquake judgment models f (x) of the present invention carrying out.Similarly, this embodiment maximum earth's surface acceleration information of also having arranged in pairs or groups is analyzed.As shown in Figure 3, according to the judgement of earthquake judge module 106, when earthquake judgment models f (x) exports f (x)=1, represent geological data and be judged as seismic events; When earthquake judgment models f (x) exports f (x)=-1, represent geological data and be judged as non-seismic events.Belong in the geological data of non-seismic events at above-mentioned 6743, always have 6196 (about 91.8879%) geological datas and be correctly judged as non-seismic events, and 547 (about 8.1121%) geological datas are wrongly judged as seismic events.Therefore, for the geological data belonging to non-seismic events, earthquake judgment models f (x) of the present invention has the accuracy rate of about 91.89%.In other words, for the geological data of 6743 non-seismic events that the past records at earthquake detecting station, earthquake judgment models f (x) of the present invention can judge that the geological data of wherein 91.89% belongs to non-seismic events, and then avoids the mistake of earthquake warning to send out.

According to above-mentioned judged result, even if only there is the acceleration vertical component of latter first 3 seconds by just reaching ripple, earthquake judgment models f (x) of the present invention still can reach high accuracy.Owing to only needing to detect the ground motion parameter of vertically apparent bearing, the seismic sensor negligible amounts that earthquake detecting module 102 uses, and then reduce earthquake detecting station build cost.On the other hand, new earthquake information to change and the vector obtained still can be used for the training of earthquake judgment models f (x), to improve the accuracy of earthquake judgment models f (x).

The above-mentioned running about earthquake detecting system 10 can be summarized as an earthquake further and judge flow process 40, as shown in Figure 4.Earthquake judges that flow process 40 comprises the following steps:

Step 400: start.

Step 402: at geological data Q ₁~ Q _lin, taking-up is relevant at least one of each geological data and just reaches wave characteristic.

Step 404: according to just reaching wave characteristic, utilizes support vector classification to set up earthquake judgment models f (x).

Step 406: as the new geological data QN of reception one, according to earthquake judgment models f (x), judge that new geological data QN belongs to a seismic events or a non-seismic events.

Step 408: terminate.

Earthquake judges that the Detailed Operation mode of flow process 40 and change with reference to aforementioned, can be not repeated herein.

In the prior art, now type earthquake early-warning system may the vibration signal that causes by some non-seismic events trigger, thus produce the earthquake warning of mistake.If use two seismic sensors to be placed on diverse location respectively to carry out the mode of duplicate acknowledgment, not only cost is higher, also improves the difficulty that earthquake detecting station builds and safeguards.In comparison, the earthquake judgment models that earthquake determination methods of the present invention and earthquake detecting system can utilize support vector classification to set up judges that received new geological data belongs to seismic events or non-seismic events.According to the simulation result of historical data, when reaching the vertical component of the ground motion parameter of ripple at the beginning of only using, the geological data belonging to seismic events is judged, the accuracy rate of 99.4% can be reached, the geological data belonging to non-seismic events is judged, the accuracy rate of 91.89% can be reached.Thus, the present invention, except the earthquake that can reach high-accuracy judges, also has the advantage of low cost simultaneously.

The foregoing is only the preferred embodiments of the present invention, be not limited to the present invention, for a person skilled in the art, the present invention can have various modifications and variations.Within the spirit and principles in the present invention all, any amendment done, equivalent replacement, improvement etc., all should be included within protection scope of the present invention.

Claims (18)

1. an earthquake determination methods, for an earthquake detecting system, this earthquake determination methods comprises:

In multiple geological data, taking-up is relevant at least one of each geological data and just reaches wave characteristic;

Reach wave characteristic at the beginning of at least one according to this, utilize a support vector classification to set up an earthquake judgment models; And

When the new geological data of reception one, according to this earthquake judgment models, judge that this new geological data belongs to a seismic events or a non-seismic events.

2. earthquake determination methods as claimed in claim 1, is characterized in that, this at least one wave characteristic that just reaches comprises the vertical component reaching the earthquake motion detected when ripple arrives at the beginning of.

3. earthquake determination methods as claimed in claim 2, is characterized in that, this is at least one just reaches wave characteristic and comprise an accumulation absolute speed of this earthquake motion, an absolute speed integration and an absolute displacement integration.

4. earthquake determination methods as claimed in claim 1, it is characterized in that, this at least one wave characteristic that just reaches being relevant to the plurality of geological data corresponds to multiple vector, and wherein each vector corresponds to a geological data in the plurality of geological data, and corresponds to a desired value.

5. earthquake determination methods as claimed in claim 4, is characterized in that, when receiving this new geological data, according to this earthquake judgment models, judges that the step that this new geological data belongs to this seismic events or this non-seismic events uses lower array function to judge:

$\mathrm{sgn}[\underset{i,j=1}{\overset{l}{\mathrm{\Σ}}}{y}_i{\mathrm{\α}}_{i}K(x_i,x_j)+b]$

Wherein, x ₁~ x _lfor the plurality of vector, y ₁~ y _lbe respectively and correspond to the plurality of vector x ₁~ x _lin this desired value of each vector, wherein y _i∈ { 1 ,-1}, α ₁~ α _land b is respectively according to this support vector classification, the constant pushed away by the plurality of vector, this desired value and the plurality of geological data, and K is the kernel function corresponding to a high-dimensional feature space.

6. earthquake determination methods as claimed in claim 5, is characterized in that, when this new geological data belongs to this seismic events, this function exports 1, and when this new geological data belongs to this non-seismic events, this function exports-1.

7. earthquake determination methods as claimed in claim 5, is characterized in that, α ₁~ α _land b is that this objective function is as follows according to solving an objective function and obtaining:

$\underset{w,b,\mathrm{\ξ}}{\min }[\frac{1}{2}{w}^{T}w+C\underset{i=1}{\overset{l}{\mathrm{\Σ}}}{\mathrm{\ξ}}_i]$

Be limited to

y _i(w ^Tφ(x _i)+b)≥1-ξ _i，ξ _i≥0，i=1，...，l

Wherein, w is a vector in this high-dimensional feature space, ξ ₁~ ξ _land b is the variable of this objective function, and φ is by the plurality of vector x ₁~ x _lcorrespond to a function of this high-dimensional feature space.

8. earthquake determination methods as claimed in claim 7, it is characterized in that, it is as follows that this objective function is converted to an antithesis pattern:

$\underset{\mathrm{\α}}{\min }[\frac{1}{2}{\mathrm{\α}}^T\mathrm{Q\α}-e^T\mathrm{\α}]$

Be limited to

y ^Tα＝0，0≤α _i≤C，i＝1，...，l

Wherein, e is a vector of unit length, C > 0 and be α _ithe upper limit, Q be l × l positive semidefinite matrix and

Q _ij≡y _iy _jK(x _i，x _j)。

9. earthquake determination methods as claimed in claim 8, it is characterized in that, this kernel function is expressed as K (x _i, x _j) ≡ φ (x _i) ^tφ (x _j).

10. an earthquake detecting system, comprising:

One earthquake detecting module, is used in multiple geological data, and taking-up is relevant at least one of each geological data and just reaches wave characteristic;

One computing module, reaches wave characteristic at the beginning of at least one according to this, utilizes a support vector classification to set up an earthquake judgment models; And

One earthquake judge module, is used for, when this earthquake detecting module receives a new geological data, according to this earthquake judgment models, judging that this new geological data belongs to a seismic events or a non-seismic events.

11. earthquake detecting systems as claimed in claim 10, is characterized in that, this at least one wave characteristic that just reaches comprises the vertical component reaching the earthquake motion that this earthquake detecting module detects when ripple arrives at the beginning of.

12. earthquake detecting systems as claimed in claim 11, is characterized in that, this is at least one just reaches wave characteristic and comprise an accumulation absolute speed of this earthquake motion, an absolute speed integration and an absolute displacement integration.

13. earthquake detecting systems as claimed in claim 10, it is characterized in that, this at least one wave characteristic that just reaches being relevant to the plurality of geological data corresponds to multiple vector, and wherein each vector corresponds to a geological data in the plurality of geological data, and corresponds to a desired value.

14. earthquake detecting systems as claimed in claim 13, is characterized in that, this earthquake judge module uses lower array function to judge that this new geological data belongs to this seismic events or this non-seismic events:

$\mathrm{sgn}[\underset{i,j=1}{\overset{l}{\mathrm{\Σ}}}{y}_i{\mathrm{\α}}_{i}K(x_i,x_j)+b]$

Wherein, x ₁~ x _lfor the plurality of vector, y ₁~ y _lbe respectively and correspond to the plurality of vector x ₁~ x _lin this desired value of each vector, wherein y _i∈ { 1 ,-1}, α ₁~ α _land b is respectively according to this support vector classification, the constant pushed away by the plurality of vector, this desired value and the plurality of geological data, and K is the kernel function corresponding to a high-dimensional feature space.

15. earthquake detecting systems as claimed in claim 14, is characterized in that, when this new geological data belongs to this seismic events, this function exports 1, and when this new geological data belongs to this non-seismic events, this function exports-1.

16. earthquake detecting systems as claimed in claim 14, is characterized in that, α ₁~ α _land b is that this objective function is as follows according to solving an objective function and obtaining:

$\underset{w,b,\mathrm{\ξ}}{\min }[\frac{1}{2}{w}^{T}w+C\underset{i=1}{\overset{l}{\mathrm{\Σ}}}{\mathrm{\ξ}}_i]$

Be limited to

y _i(w ^Tφ(x _i)+b)≥1-ξ _i，ξ _i≥0，i＝1，...，l

Wherein, w is a vector in this high-dimensional feature space, ξ ₁~ ξ _land b is the variable of this objective function, and φ is by the plurality of vector x ₁~ x _lcorrespond to a function of this high-dimensional feature space.

17. earthquake detecting systems as claimed in claim 16, it is characterized in that, it is as follows that this objective function is converted to an antithesis pattern:

$\underset{\mathrm{\α}}{\min }[\frac{1}{2}{\mathrm{\α}}^T\mathrm{Q\α}-e^T\mathrm{\α}]$

Be limited to

y ^Tα＝0，0≤α _i≤C，i＝1，...，l

Wherein, e is a vector of unit length, C > 0 and be α _ithe upper limit, Q be l × l positive semidefinite matrix and

Q _ij≡y _iy _jK(x _i，x _j)。

18. earthquake detecting systems as claimed in claim 17, it is characterized in that, this kernel function is expressed as K (x _i, x _j) ≡ φ (x _i) ^tφ (x _j).