METHOD OF DETERMINING EARTHQUAKE EVENT AND RELATED EARTHQUAKE DETECTING SYSTEM

Abstract

A method of determining an earthquake event for an earthquake detecting system includes retrieving at least one initial wave characteristic related to each earthquake data among a plurality of earthquake data; utilizing a support vector classification (SVC) method to establish an earthquake determination model according to the initial wave characteristic; and determining whether new earthquake data belong to an earthquake event or a non-earthquake event according to the earthquake determination model when the new earthquake data are received.

Description

BACKGROUND OF THE INVENTION

1. Field of the Invention

The present invention relates to a method of determining an earthquake event and a related earthquake detecting system, and more particularly, to a method of determining whether received earthquake data belong to an earthquake event or a non-earthquake event according to an initial wave characteristic of an earthquake, and a related earthquake detecting system.

2. Description of the Prior Art

Due to advances in seismology, digital communication, automatic processing and algorithms, earthquake early warning (EEW) techniques have emerged in recent years. EEW techniques can be divided into two groups: regional warning and on-site warning. Since the regional warning leverages information of several earthquake stations next to the epicenter, the accuracy of earthquake parameter estimation of a regional warning system is usually higher than an on-site warning system. For regions close to the epicenter, however, the seismic intensity is much higher than in regions outside; hence, the lead time allows the regional warning system to estimate earthquake parameters before the arrival of destructive waves is limited. The on-site warning system can immediately provide earthquake parameter estimation to achieve rapid early warning since only the information at the target site is required, especially for regions close to the epicenter.

Recent studies show that the on-site warning system may be triggered due to vibration signals caused by non-earthquake events, which may result in false alarms at the station. The classification between true earthquake and non-earthquake events thereby becomes an important issue. A conventional approach to solve this problem applies two earthquake sensors installed at different locations as a double check procedure. This method is costly and the earthquake stations are also difficult to be installed and maintained. Thus, a more effective and cost saving method that can enhance the accuracy of the on-site warning system is required.

SUMMARY OF THE INVENTION

It is therefore an objective of the present invention to provide a method of determining an earthquake event and a related earthquake detecting system capable of determining whether received earthquake data belong to an earthquake event or a non-earthquake event via an earthquake determination model established by utilizing a support vector classification (SVC) method according to the initial wave characteristic of earthquakes.

The present invention discloses a method of determining an earthquake event for an earthquake detecting system. The method comprises retrieving at least one initial wave characteristic related to each earthquake data among a plurality of earthquake data; utilizing an SVC method to establish an earthquake determination model according to the initial wave characteristic; and determining whether new earthquake data belong to an earthquake event or a non-earthquake event according to the earthquake determination model when the new earthquake data are received.

The present invention further discloses an earthquake detecting system. The earthquake detecting system comprises an earthquake detecting module, for retrieving at least one initial wave characteristic related to each earthquake data among a plurality of earthquake data; a computation module, for utilizing an SVC method to establish an earthquake determination model according to the initial wave characteristic; and an earthquake determination module, for determining whether new earthquake data belong to an earthquake event or a non-earthquake event according to the earthquake determination model when the earthquake detecting module receives the new earthquake data.

These and other objectives of the present invention will no doubt become obvious to those of ordinary skill in the art after reading the following detailed description of the preferred embodiment that is illustrated in the various figures and drawings.

BRIEF DESCRIPTION OF THE DRAWINGS

FIG. 1 is a schematic diagram of an earthquake detecting system according to an embodiment of the present invention.

FIG. 2 is a schematic diagram of earthquake data belonging to earthquake events determined via an earthquake determination model according to an embodiment of the present invention.

FIG. 3 is a schematic diagram of earthquake data belonging to non-earthquake events determined via the earthquake determination model according to an embodiment of the present invention.

FIG. 4 is a schematic diagram of an earthquake determination process according to an embodiment of the present invention.

DETAILED DESCRIPTION

Please refer to FIG. 1, which is a schematic diagram of an earthquake detecting system 10 according to an embodiment of the present invention. The earthquake detecting system 10 includes an earthquake detecting module 102, a computation module 104 and an earthquake determination module 106. The earthquake detecting module 102 is utilized for retrieving at least one initial wave characteristic related to each earthquake data among a plurality of earthquake data Q₁-Q_l. The initial wave characteristic may be any physical quantity related to the initial waves, such as speed, acceleration or displacement. The initial wave characteristic can provide an earthquake early warning (EEW) before the primary waves arrive. In an embodiment, the initial wave characteristic includes the cumulative absolute velocity (CAV), integral of absolute velocity (IAV) and integral of absolute displacement (IAD) of the ground motion detected within a period of time after the initial waves arrive. The initial wave characteristic may be utilized as a threshold for determining whether an earthquake occurs. For example, an earthquake may occur when the CAV of the ground motion detected by the earthquake detecting system 10 exceeds a threshold value. In an embodiment, for the sake of immediacy of earthquake detection and to save on numbers of ground motion sensors, only the vertical component of the ground motion may be applied. When the initial waves arrive, the measured characteristics of the vertical ground motion are regarded as the initial wave characteristics, and these characteristics may include the CAV, IAV and IAD. In another embodiment, ground motion components along other directions may also be adopted as the initial wave characteristics. Considering the integrity of the statistic earthquake data, multiple ground motion sensors may also be utilized for simultaneously measuring the ground motion in multiple directions as the initial wave characteristics.

The computation module 104 may utilize a support vector classification (SVC) method to establish an earthquake determination model according to the initial wave characteristic obtained by the earthquake detecting module 102. According to this earthquake determination model, the earthquake determination module 106 may determine whether new earthquake data QN belong to an earthquake event or a non-earthquake event when the earthquake detecting module 102 receives the new earthquake data QN. Through the SVC method, the accuracy of the on-site earthquake warning system determining earthquake events may be enhanced, which preserves the immediacy of the on-site earthquake warning system while preventing erroneous determination caused by non-earthquake events.

According to the SVC method, the initial wave characteristic related to the earthquake data Q₁-Q_l may correspond to a plurality of vectors x₁-x_l, wherein each vector x_i may correspond to a target value y_i and y_iε{1,−1}. If the earthquake data Q_i corresponding to the vector x_i are caused by an earthquake event, the corresponding target value may be set to y_i=1; if the earthquake data Q_i corresponding to the vector x_i are caused by a non-earthquake event, the corresponding target value may be set to y_i=−1. The information of whether the earthquake data Q₁-Q_l are caused by the earthquake events or non-earthquake events is obtained from statistical results of the historical data. The SVC method retrieves the initial wave characteristic corresponding to the earthquake events or non-earthquake events based on the historical data, and generates the earthquake determination model. When the new earthquake data QN is received, the SVC method can thereby determine whether the initial wave characteristic of the new earthquake data QN is closer to the initial wave characteristic of the earthquake event or that of the non-earthquake event in the historical data according to the earthquake determination model, in order to determine whether the new earthquake data QN are caused by the earthquake event or the non-earthquake event.

According to the SVC method, the vectors x₁-x_l may be mapped into a high-dimensional characteristic space H, and classified in the high-dimensional characteristic space H. The computation module 104 may determine whether the new earthquake data QN belong to the earthquake event or the non-earthquake event by defining the earthquake determination model as the following function ƒ(x):

$f\left(x\right)=\mathrm{sgn}\left[\sum _{i,j=1}^ly_i{\alpha }_iK\left(x_i,x_j\right)+b\right]$

wherein K is a kernel function corresponding to the high-dimensional characteristic space H and K(x_i,x_j)≡φ(x_i)^Tφ(x_j), and α₁-α_l and b are both coefficients derived from the vectors x₁-x_l and the target values y₁-y_l corresponding to the earthquake data Q₁-Q_l according to the SVC method. In the SVC method, the parameters α₁-α_l and b may be obtained by solving the following primal problem:

$\underset{w,b,\xi }{\min }\left[\frac{1}{2}w^Tw+C\sum _{i=1}^l{\xi }_i\right]$

subjected to

y _i(w^Tφ(x_i)+b)≧1−ξ_i, ξ_i≧0, i=1, . . . ,l

wherein w is a vector in the high-dimensional characteristic space H, ξ₁-ξ_l and b are variables of the primal problem, and φ is a function mapping the vectors x₁-x_l to the high-dimensional characteristic space H. The primal problem has a dual form as follows:

$\underset{\alpha }{\min }\left[\frac{1}{2}{\alpha }^TQ\alpha -e^T\alpha \right]$

subjected to

y ^Tα=0, 0≦α_i≦C, i=1, . . . ,l

wherein e is a unit vector, C>0 and is an upper limit of α_i, Q is an l×l positive semi-definite matrix, and Q_ij≡y_iy_jK(x_i,x_j). According to the above equations, the vector corresponding to the initial wave characteristic of the inputted new earthquake data QN may be calculated, and two types of results ƒ(x)=1 or ƒ(x)=−1 may be obtained. When ƒ(x)=1, the new earthquake data QN may belong to an earthquake event; when ƒ(x)=−1, the new earthquake data QN may belong to a non-earthquake event.

In other words, based on the initial wave characteristics related to the earthquake data Q₁-Q_l in the historical data, the computation module 104 may calculate the parameters α₁-α_l and b via the above equations according to the SVC method, in order to establish the earthquake determination model ƒ(x). When the earthquake detecting module 102 detects a new earthquake data QN, the initial wave characteristic of the new earthquake data QN may be converted into a specific vector, and the earthquake determination module 106 may calculate that ƒ(x)=1 or −1 according to this specific vector and the earthquake determination model ƒ(x), in order to determine whether the new earthquake data QN is caused by an earthquake event or a non-earthquake event. This prevents ground motions generated by a non-earthquake event from being erroneously attributed to an earthquake.

In the earthquake detecting system 10, the earthquake detecting module 102, the computation module 104 and the earthquake determination module 106 may be located in different locations and may communicate with each other via a wired network or a wireless network. For example, the earthquake detecting module 102 may not be limited to a single earthquake station or earthquake detector; it may also include a plurality of earthquake stations or earthquake detectors located in different locations. The earthquake stations or earthquake detectors of the earthquake detecting module 102 may be installed in the regions where earthquakes frequently occur, so that new earthquake data QN can be rapidly detected. The computation module 104 may be located in the earthquake information center for obtaining the massive earthquake data Q₁-Q_l, in order to improve the accuracy of the earthquake determination model ƒ(x). The earthquake determination module 106 may be located in the warning center, so that an earthquake warning can be delivered immediately when determining that the new earthquake data QN belong to an earthquake event.

The above equations related to the SVC method are only one embodiment of the present invention, and should not be taken as limiting the scope of the present invention. Those skilled in the art may use other mathematical equations along with the SVC method to obtain equations capable of determining whether new earthquake data belong to an earthquake event.

As a further illustration of the effects of the above embodiments, the accuracy of the earthquake data and non-earthquake data determined by the earthquake determination model ƒ(x) may be obtained via simulations. The earthquake data Q₁-Q_l utilized by the computation module 104 for training the earthquake determination model ƒ(x) should include both earthquake data belonging to earthquake events and earthquake data belonging to non-earthquake events. Those earthquake data belonging to earthquake events may be data of actual earthquakes obtained from historical records of, for example, the Taiwan Central Weather Bureau, or may be data obtained from an earthquake early warning system (EEWS) where the data is known for an occurring earthquake. Those earthquake data belonging to non-earthquake events may be analogous earthquake data or non-earthquake data obtained from the EEWS, or may be data obtained from EEWS where the data is not regarded as an earthquake event by the Taiwan Central Weather Bureau. Please note that, in Taiwan, the Taiwan Central Weather Bureau has a complete arrangement of earthquake stations and earthquake detecting systems, and can thereby obtain high accuracy earthquake data. Earthquake data recorded by the Taiwan Central Weather Bureau may be considered to correspond to actual earthquake events.

In an illustration of the earthquake determination method, the earthquake detecting system 10 utilizes the vertical component of the acceleration of the earthquake data Q₁-Q_l within the first 3 seconds immediately after the corresponding initial wave arrives, and then integrates the data to generate speed and displacement information. These physical quantities may be regarded as the initial wave characteristics utilized for establishing the earthquake determination model ƒ(x). The earthquake detecting system 10 then obtains 91,142 records from earthquake data of the Taiwan Central Weather Bureau from Jul. 29, 1992 to Dec. 31, 2006, and collects several earthquake data from earthquake stations which have been verified to belong to both earthquake and non-earthquake events, including 54 and 6,743 records, respectively. There are a total of 91,196 records of data belonging to earthquake events and 6,743 records of data belonging to non-earthquake events to be verified by simulations of the earthquake determination model ƒ(x) established in the above embodiments.

Please refer to FIG. 2, which is a schematic diagram of earthquake data belonging to earthquake events determined via the earthquake determination model ƒ(x) according to an embodiment of the present invention. In order to obtain determination results of earthquakes having different intensities, this embodiment also includes the peak ground acceleration (PGA) related to the intensities of earthquakes for analysis. As shown in FIG. 2, according to the determination of the earthquake determination module 106, when the earthquake determination model ƒ(x) outputs ƒ(x)=1, the earthquake data is determined to be an earthquake event; when the earthquake determination model ƒ(x) outputs ƒ(x)=−1, the earthquake data is determined to be a non-earthquake event. Among the above 91,196 records of earthquake data belonging to earthquake events, there are 90,652 records (99.4035%) accurately determined to be earthquake events, and 544 records (0.59652%) erroneously determined to be non-earthquake events. The earthquake determination model ƒ(x) of the present invention can achieve a high accuracy rate of 99.4% for those earthquake data belonging to earthquake events. In addition, all earthquake data having a PGA greater than 200 gal can accurately be determined to be earthquake events via the earthquake determination model ƒ(x) of the present invention.

Please refer to FIG. 3, which is a schematic diagram of earthquake data belonging to non-earthquake events determined via the earthquake determination model ƒ(x) according to an embodiment of the present invention. Similarly, this embodiment also includes the PGA data for analysis. As shown in FIG. 3, according to the determination of the earthquake determination module 106, when the earthquake determination model ƒ(x) outputs ƒ(x)=1, the earthquake data is determined to be an earthquake event; when the earthquake determination model ƒ(x) outputs ƒ(x)=−1, the earthquake data is determined to be a non-earthquake event. Among the above 6,743 records of earthquake data belonging to non-earthquake events, there are 6,196 records (91.8879%) accurately determined to be non-earthquake events, and 547 records (8.1121%) erroneously determined to be earthquake events. The earthquake determination model ƒ(x) of the present invention can achieve an accuracy rate of 91.89% for those earthquake data belonging to non-earthquake events. In other words, for those 6,743 records of earthquake data measured by the earthquake stations in response to non-earthquake events, the earthquake determination model ƒ(x) of the present invention can accurately determine that 91.89% of those earthquake data belong to non-earthquake events, and thereby prevent false alarms.

As shown in the above determination results, even if only the vertical component of the acceleration within the first 3 seconds immediately after the initial wave arrives is utilized, the earthquake determination model ƒ(x) of the present invention can still achieve high accuracy. Since only the vertical ground motion parameters are required, the earthquake detecting module 102 may use fewer numbers of ground motion sensors, which reduces the cost involved for the earthquake stations. The vectors converted from the new earthquake information may still be utilized for training the earthquake determination model ƒ(x), to thereby enhance the accuracy of the earthquake determination model ƒ(x).

The abovementioned operations related to the earthquake detecting system 10 can further be summarized into an earthquake determination process 40, as shown in FIG. 4. The earthquake determination process 40 includes the following steps:

Step 400: Start.

Step 402: Retrieve at least one initial wave characteristic related to each earthquake data among the earthquake data Q₁-Q_l.

Step 404: Utilize an SVC method to establish the earthquake determination model ƒ(x) according to the initial wave characteristic.

Step 406: Determine whether new earthquake data QN belong to an earthquake event or a non-earthquake event according to the earthquake determination model ƒ(x) when the new earthquake data QN are received.

Step 408: End.

Detailed operations and variations of the earthquake determination process 40 are illustrated above, and will not be repeated herein.

In the prior art, an on-site warning system may be triggered by vibration signals caused by non-earthquake events, which may result in false alarms at the station. If two earthquake sensors are installed at different locations as a double check procedure, both the cost and difficulty in installing and maintaining the earthquake stations may increase. In comparison, the earthquake determination method and earthquake detecting system of the present invention utilize the earthquake determination model established according to the SVC method to determine whether received new earthquake data belong to an earthquake event or a non-earthquake event. According to the simulation results of historical data where only the vertical component of the ground motion parameters of the initial waves is utilized, the determination for the earthquake data belonging to earthquake events achieves an accuracy rate of 99.4%, and the determination for the earthquake data belonging to non-earthquake events achieves an accuracy rate of 91.89%. The present invention therefore not only provides a higher accuracy earthquake determination, but also achieves the benefit of low cost.

Those skilled in the art will readily observe that numerous modifications and alterations of the device and method may be made while retaining the teachings of the invention. Accordingly, the above disclosure should be construed as limited only by the metes and bounds of the appended claims.

Claims

1. A method of determining an earthquake event for an earthquake detecting system, the method comprising: retrieving at least one initial wave characteristic related to each earthquake data among a plurality of earthquake data;utilizing a support vector classification (SVC) method to establish an earthquake determination model according to the initial wave characteristic; anddetermining whether new earthquake data belong to an earthquake event or a non-earthquake event according to the earthquake determination model when the new earthquake data are received.

2. The method of claim 1, wherein the initial wave characteristic comprises a vertical component of a ground motion detected when an initial wave arrives.

3. The method of claim 2, wherein the initial wave characteristic comprises a cumulative absolute velocity (CAV), an integral of absolute velocity (IAV) and an integral of absolute displacement (IAD) of the ground motion.

4. The method of claim 1, wherein the initial wave characteristic related to the plurality of earthquake data corresponds to a plurality of vectors, each of which corresponds to one of the plurality of earthquake data and a target value.

5. The method of claim 4, wherein the step of determining whether the new earthquake data belong to the earthquake event or the non-earthquake event according to the earthquake determination model when the new earthquake data are received utilizes a function for determination:

6. The method of claim 5, wherein the function outputs 1 when the new earthquake data belong to the earthquake event and outputs −1 when the new earthquake data belong to the non-earthquake event.

7. The method of claim 5, wherein α1-αl and b are obtained by solving a primal problem:

8. The method of claim 7, wherein the primal problem is converted into a dual form:

9. The method of claim 8, wherein the kernel function is denoted by K(xi,xj)≡φ(xi)Tφ(xj).

10. An earthquake detecting system, comprising: an earthquake detecting module, for retrieving at least one initial wave characteristic related to each earthquake data among a plurality of earthquake data;a computation module, for utilizing a support vector classification (SVC) method to establish an earthquake determination model according to the initial wave characteristic; andan earthquake determination module, for determining whether new earthquake data belong to an earthquake event or a non-earthquake event according to the earthquake determination model when the earthquake detecting module receives the new earthquake data.

11. The earthquake detecting system of claim 10, wherein the initial wave characteristic comprises a vertical component of a ground motion detected by the earthquake detecting module when an initial wave arrives.

12. The earthquake detecting system of claim 11, wherein the initial wave characteristic comprises a cumulative absolute velocity (CAV), an integral of absolute velocity (IAV) and an integral of absolute displacement (IAD) of the ground motion.

13. The earthquake detecting system of claim 10, wherein the initial wave characteristic related to the plurality of earthquake data corresponds to a plurality of vectors, each of which corresponds to one of the plurality of earthquake data and a target value.

14. The earthquake detecting system of claim 13, wherein the earthquake determination module determines whether the new earthquake data belong to the earthquake event or the non-earthquake event by utilizing a function:

15. The earthquake detecting system of claim 14, wherein the function outputs 1 when the new earthquake data belong to the earthquake event and outputs −1 when the new earthquake data belong to the non-earthquake event.

16. The earthquake detecting system of claim 14, wherein α1-αl and b are obtained by solving a primal problem:

17. The earthquake detecting system of claim 16, wherein the primal problem is converted into a dual form:

18. The earthquake detecting system of claim 17, wherein the kernel function is denoted by K(xi,xj)≡φ(xi)Tφ(xj).