Estimation device, vibration sensor system, and method performed by the estimation device
Through the Kalman filter method at discrete time, combined with the data of external measurer and vibration sensor, the problem of dynamic range and frequency band expansion in seismometer sensor fusion is solved, and the practicality of seismic observation is improved. .
Patent Information
- Application Number
- CN202180015359.4
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Priority Date
- 2020-02-21
- Filing Date
- 2021-02-17
- Publication Date
- 2024-11-05
- Estimated Expiration
- 2041-02-17
AI Technical Summary
In existing seismic observations, the sensor fusion of multiple seismometers has not yet reached the practical level, and it is difficult to effectively expand the dynamic range and measure the frequency band.
The Kalman filter method at discrete time is adopted, and the data fusion of the external measurer and the vibration sensor is estimated, the state vector and the covariance matrix are adjusted, and the sensor fusion of multiple seismometers is realized.
The dynamic range of high-sensitivity seismometers and low-sensitivity seismometers has been expanded, and the practicality and accuracy of seismic observations have been improved.
Smart Images

Figure CN115210609B_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to an estimation device, a vibration sensor system, and a related method and program using the estimation device, the estimation device estimating the state of a vibration sensor such as a seismograph. In particular, the present invention relates to an estimation device, a vibration sensor system, and a related method and program estimating the state of a vibration sensor by integrating data from a plurality of sensors (sometimes referred to as detectors) through sensor fusion, a vibration sensor system, and a related method and program. Background Art
[0002] Sensor fusion refers to a technology that combines data from multiple sensors, takes advantage of their advantages, and compensates for their disadvantages to seek state estimation and control of the object. As a representative successful example, there is an integrated inertial navigation system (Chapter 9 of Non-Patent Document 1) that integrates position signals from a global navigation satellite system into an inertial navigation system composed of a gyroscope and an accelerometer. Recently, sensor fusion technology that is linked to AI (Artificial Intelligence) technology and integrates various sensors, such as millimeter wave radar, into the autonomous driving of automobiles has become a hot topic.
[0003] Among the techniques of sensor fusion, statistical inference such as Bayesian inference or maximum probability method is used (non-patent document 2), but one of the methods that is often used is Kalman filter (non-patent document 3). In fact, non-patent document 1 and others have discussed the integrated inertial navigation system as one of the application fields in the textbook of Kalman filter. Statistical inference discusses the data from the sensor, and how to integrate the data and make use of it is the purpose of sensor fusion. It is an outward-oriented technology.
[0004] In earthquake observation, high-sensitivity short-period velocimeters, wide-band velocimeters, and strong-motion seismometers have been used so far according to the purpose. Recently, for example, it has become a world trend to combine wide-band velocimeters and strong-motion seismometers to grasp the tendency of earthquakes in real time with a wider dynamic range. This integration is aimed at improving the performance of the sensor itself, so it is inward-oriented and has nothing to do with the splendor of global navigation satellite systems or automatic driving of cars. Having said that, sensor fusion, which integrates multiple seismometers, is one of the important technical issues in earthquake observation that seeks to expand the measurement frequency band and dynamic range. The basic sensor fusion in earthquake observation is the integration of two types of seismometers. For example, the sensor fusion of a wide-band velocimeter and a short-period velocimeter is used to expand the measurement frequency band. This can be easily achieved by complementary filtering (non-patent document 4). However, as far as the inventors know, sensor fusion that seeks to expand the dynamic range by fusing two types of seismometers has not yet reached a practical level at this stage.
[0005] [Prior art literature]
[0006] [Non-patent literature]
[0007] Non-patent literature 1: Grewal, MS and AP Andrews (2008). Kalman filtering, Wiley-IEEE Press.
[0008] Non-patent document 2: Shingo Kagami and Masatoshi Ishikawa (2005). Sensor fusion - Information processing structure of sensor network -, Journal of the Society of Electronics, Information and Communications Technology A, J88-A, No. 12, 1404-1412
[0009] Non-patent document 3: Kalman, RE (1960). A new approach to linear filtering and prediction problems, ASME. Journal of Basic Engineering, 82, 34-45
[0010] Non-patent document 4: Shigeo Kinoshita (2014), Characteristics compensation by state feedback type geophone and complementary filter, Earthquake, 66, 101-113
[0011] Non-patent document 5: Anderson, BDO and JB Moore (1979). Optimal filtering, Prentice-Hall, Inc. Summary of the invention
[0012] [Problems to be solved by the invention]
[0013] In view of the above situation, an object of the present invention is to provide a method for performing sensor fusion of multiple vibration sensors such as seismographs at a practical level.
[0014] [Methods used to solve the problem]
[0015] In order to solve the above problems, the present invention provides an estimation device, which estimates a state vector of a vibration sensor according to discrete moments, that is, a state vector that changes with time as an external input, and calculates a covariance matrix of the state vector according to the discrete moments. The estimation device comprises:
[0016] a measurement value acquisition unit that acquires, as an external input, a measurement value of a physical quantity related to vibration measured by an external measuring device;
[0017] A sensing value acquisition unit, which acquires a sensing value of a vibration sensor;
[0018] A data storage unit stores a state vector estimation value obtained by a previous estimation, a post-calculation covariance matrix obtained by a previous calculation, a measurement value corresponding to a time corresponding to a previous estimation and a previous calculation, and a covariance matrix of system noise (system interference);
[0019] a state vector prior estimation value calculation unit that calculates the state vector prior estimation value using the state vector estimated value obtained by the previous estimation and the measured value corresponding to the time corresponding to the previous estimation and the previous calculation;
[0020] The ex ante covariance matrix calculation unit calculates the ex ante covariance matrix using the ex post covariance matrix obtained by the previous calculation and the covariance matrix of the system noise;
[0021] A Kalman gain calculation unit calculates the Kalman gain using a prior covariance matrix;
[0022] a state vector estimated value calculation unit that calculates a state vector estimated value in current estimation using a state vector prior estimation value, a Kalman gain, and a sensed value; and
[0023] The post-covariance matrix calculation unit calculates the post-covariance matrix in this calculation using the pre-covariance matrix and the Kalman gain.
[0024] In addition, the present invention provides a vibration sensor system including a vibration sensor, an external measuring device, and the estimation device.
[0025] The external measuring device may be a device that can perform acceleration conversion by calculation and has a dynamic range that exceeds the upper limit of the dynamic range of the vibration sensor. The measurement value acquisition unit may be a device that acquires the measurement value of acceleration.
[0026] The vibration sensor may also have a pendulum, and the vibration sensor may measure one or more values of displacement, velocity, and acceleration. The sensing value acquisition unit may be a device that acquires one or more values of displacement, velocity, and acceleration.
[0027] The above-mentioned Kalman gain calculation unit can be a device that adjusts the Kalman gain by using a calculation using a Kalman gain adjustment term. The adjustment of the Kalman gain is performed in the following manner: when one or more of the displacement, velocity, and acceleration of the vibration sensor is greater than a predetermined value, the Kalman gain adjustment term can be increased compared to a situation where one or more of the values are smaller than the predetermined value, thereby reducing the Kalman gain.
[0028] The above-mentioned vibration sensor can also be a vibration sensor that has pendulum and measures one or more values of displacement, velocity, and acceleration; the sensor is provided with an analog sensor by the subsequent Kalman filter, and the analog sensor simulates the action of the vibration sensor with pendulum through calculation, and determines one or more values of displacement, velocity, and acceleration of the simulated pendulum through calculation; the state vector estimated value calculation unit uses one or more values of displacement, velocity, and acceleration measured by the vibration sensor and any one of the values obtained by multiplying one or more values of displacement, velocity, and acceleration of the simulated pendulum determined by the analog sensor by a coefficient as the sensing value according to the size of one or more values of displacement, velocity, and acceleration measured by the vibration sensor.
[0029] In addition, the present invention provides a method executed by an estimation device, wherein the estimation device estimates a state vector of a vibration sensor according to discrete moments, that is, a state vector that changes with time as an external input, and calculates a covariance matrix of the state vector according to the discrete moments. The method executed by the estimation device comprises:
[0030] A step of obtaining a measurement value of a physical quantity related to vibration measured by an external measuring device as an external input;
[0031] The step of obtaining the sensing value of the vibration sensor;
[0032] A step of calculating a state vector prior estimation value using a state vector estimation value obtained by a previous estimation and a measurement value corresponding to a time corresponding to the previous estimation and the previous calculation;
[0033] A step of calculating an ex ante covariance matrix using the ex post covariance matrix obtained by the previous calculation and the covariance matrix of the system noise;
[0034] The step of calculating the Kalman gain using the prior covariance matrix;
[0035] A step of calculating a state vector estimated value in this estimation using a state vector prior estimation value, a Kalman gain, and a sensed value; and
[0036] The step of calculating the post-covariance matrix in this calculation using the pre-covariance matrix and the Kalman gain.
[0037] In addition, the present invention provides a program for causing a computer to execute a method, wherein the method estimates a state vector of a vibration sensor according to discrete moments, that is, a state vector that changes with time as an external input, and calculates a covariance matrix of the state vector according to the discrete moments, wherein the program causes the computer to execute the following steps:
[0038] A step of obtaining a measurement value of a physical quantity related to vibration measured by an external measuring device as an external input;
[0039] The step of obtaining the sensing value of the vibration sensor;
[0040] A step of calculating a state vector prior estimation value using a state vector estimation value obtained by a previous estimation and a measurement value corresponding to a time corresponding to the previous estimation and the previous calculation;
[0041] A step of calculating an ex ante covariance matrix using the ex post covariance matrix obtained by the previous calculation and the covariance matrix of the system noise;
[0042] The step of calculating the Kalman gain using the prior covariance matrix;
[0043] A step of calculating a state vector estimated value in this estimation using a state vector prior estimation value, a Kalman gain, and a sensed value; and
[0044] The step of calculating the post-covariance matrix in this calculation using the pre-covariance matrix and the Kalman gain.
[0045] [Effects of the invention]
[0046] According to the present invention, multiple vibration sensors such as seismometers can be integrated for use by sensor fusion of multiple vibration sensors. In one example, the dynamic range of a high-sensitivity geophone can be expanded by sensor fusion of a high-sensitivity geophone and a low-sensitivity geophone. BRIEF DESCRIPTION OF THE DRAWINGS
[0047] Figure 1 A diagram conceptually illustrating the process of sensor fusion.
[0048] Figure 2 A block diagram showing the configuration of sensor fusion using a Kalman filter.
[0049] Figure 3 FIG. 4 is a block diagram showing the structure of a vibration sensor system according to an embodiment of the present invention.
[0050] Figure 4 This is a block diagram showing an example of the device configuration of the estimation device.
[0051] Figure 5 This is a diagram showing an example of a vibration sensor (high-sensitivity speedometer).
[0052] Figure 6 A block diagram showing the construction of a simulated vibration sensor.
[0053] Figure 7 A block diagram showing the state-space representation of a continuous-time system of a high-sensitivity speedometer (FSF (full-state-feedback) sensor).
[0054] Figure 8 FIG. 1 is a diagram showing the relationship between the state vector of the restored acceleration of the FSF sensor and the feedback gain.
[0055] Fig. 9 Graph showing the VBB equivalent frequency response characteristics (model 1a2) with respect to the acceleration input of the FSF sensor.
[0056] Fig.10 This is a diagram showing the structure of an example of a speedometer (quoted from the instruction manual of the "VSE-15D-6" manufactured by Tokyo Shokin Co., Ltd.)
[0057] Fig.11 The flowchart shows the operation flow when designing a vibration sensor system.
[0058] Fig.12 The following is a flow chart showing the operation flow when the vibration sensor system is used.
[0059] Fig.13 To display Fig.11 , Fig.12 In the action flow, a flowchart of the action flow of the processing part using the Kalman filter is shown.
[0060] Fig.14 This is a diagram showing an example of a vibration sensor (high-sensitivity displacement meter).
[0061] Fig.15 A block diagram showing the configuration of sensor fusion using a Kalman filter using a high-sensitivity displacement meter and a strong motion sensor.
[0062] Fig.16 A block diagram showing the state space representation of a continuous-time system of a high-sensitivity displacement meter.
[0063] Fig.17 This is a diagram showing a high-sensitivity negative feedback type accelerometer (force parallel type) as an example of a vibration sensor (high-sensitivity accelerometer).
[0064] Fig.18 A block diagram showing the configuration of sensor fusion using a Kalman filter for high-sensitivity accelerometers and strong-motion seismometers.
[0065] Fig.19 A block diagram showing the state-space representation of a continuous-time system for a high-sensitivity accelerometer. DETAILED DESCRIPTION
[0066] The following describes the modes for implementing the present invention. However, it should be understood that the present invention is not limited to the specific modes described below, and can be appropriately changed within the scope of the claims of the present invention.
[0067] In the following embodiment, a method is proposed for expanding the dynamic range of a high-sensitivity seismic detector by fusing a high-sensitivity seismic detector (a high-sensitivity vibration sensor) with a low-sensitivity acceleration seismic detector (a low-sensitivity vibration sensor, an example of an "external measuring device") having a dynamic range exceeding the upper limit of the dynamic range of the high-sensitivity seismic detector. Although the method is initially developed using the sensor fusion of a wideband velocity meter (VBB) and a strong motion detector as a specific example, this method can also be applied to the integration of three types of high-sensitivity seismic instruments (displacement meter, velocity meter, and accelerometer) and a strong motion detector.
[0068] The concept of sensor fusion
[0069] Figure 1An example of the proposed sensor fusion process is conceptually shown in . This example fuses the observed acceleration record of the strong motion detector and the record from the high-sensitivity seismic detector observed at the same time to seek to expand the dynamic range. Therefore, software can be used to construct a virtual sensor (simulated sensor) equivalent to a high-sensitivity seismic detector (as described later, instead of using a virtual sensor, an actual high-sensitivity seismic detector can be used for sensor fusion, or a virtual sensor and an actual sensor can be combined and used separately). This virtual sensor is called a full-state-feedback (FSF) sensor (non-patent document 4). This FSF sensor is equivalent to a high-sensitivity seismic detector, so it can be represented in state space, the observed acceleration is provided as a determined signal to the input, and the record from the observed high-sensitivity seismic detector is applied to the output after adding normal interference. Here, the so-called "state" refers to the state of the oscillating motion inside the FSF sensor. Therefore, the output of the high-sensitivity geophone (that is, the output of the FSF sensor) can be estimated from the state of the oscillation obtained using the Kalman filter. The FSF sensor, which is a virtual sensor, is not a real machine, so it can avoid the limitation of the power supply voltage. Therefore, even if the output of the estimated FSF sensor is clipped from the record of the actually observed high-sensitivity geophone (even if it exceeds the predetermined value and is saturated), as long as the accelerometer operates normally, the observed interference (observation noise) added to the actual record of the high-sensitivity geophone will be increased and the Kalman gain will be reduced, thereby outputting it in an uncut state. In addition, as long as the observed interference is reduced and the Kalman gain is increased, the actually observed waveform will be correctly output from the Kalman filter. Hereinafter, when referred to as a "high-sensitivity geophone", it refers to a high-sensitivity geophone as a real machine.
[0070] Configuration of sensor fusion using virtual sensors (software sensors)
[0071] Figure 2 This is a block diagram showing the structure of sensor fusion using a Kalman filter. For simplicity of explanation, the FSF sensor is assumed to be an analog velocity meter here, but as described later, the FSF sensor may also be an analog displacement meter or accelerometer. In addition, by making the FSF sensor operate as both an analog displacement meter and a velocity meter, a low-sensitivity seismic detector (accelerometer) and a displacement meter and a velocity meter may be fused, and sensor fusion may be performed using a plurality of vibration sensors.
[0072] [Formula 1]
[0073]
[0074] exist Figure 2In the block diagram, [Formula 1] is set to the acceleration recorded by the low-sensitivity accelerometer (the unit is set to m (meter) / s 2 (seconds squared). The same applies below), y m (n) is the record of a high-sensitivity seismograph after adding normal interference. Figure 2 In the example, although m=2 is set, and the speed is recorded from a high-sensitivity speedometer (the unit is m (meter) / s (second). The same applies to the following), when m=3, the acceleration is recorded from a high-sensitivity accelerometer (the unit is m (meter) / s 2 (seconds squared). The same applies below), y1(n) when m=1 is set as the displacement record from the high-sensitivity displacement meter (the unit is set to m (meters). The same applies below). Here, n is an integer, representing a discrete moment. If the sampling time is set to △T, the actual moment corresponding to the discrete moment n becomes n△T (the unit is set to s. The same applies below), and it should be marked with this moment n△T, but △T is omitted for simplicity. In addition, Figure 2 In the example, Σ indicates that the signal values toward that point are summed and output in the direction from that point (the same applies to the following. In addition, when a negative sign is added, the signal value is subtracted rather than added.) z is a time shift operator that shifts discrete moments from n to n+1, z -1 It is an operator that performs the opposite operation (shifting discrete time from n+1 to n).
[0075] [Formula 2]
[0076] C′
[0077] [Formula 2] is an operator that converts the state vector into observation data (expressed as a matrix or vector corresponding to the observation object). When the state (state vector) at discrete time n is set to [Formula 3] (column vector), and the estimated value of the state vector at discrete time n made by the Kalman filter is set to [Formula 4] (column vector), the estimated value of the observation quantity can be given by the following observation equation [Formula 5].
[0078] [Formula 3]
[0079] X(n)
[0080] [Formula 4]
[0081]
[0082] [Formula 5]
[0083]
[0084] [Formula 6]
[0085] I3
[0086] [Formula 6] is a third-order unit matrix.
[0087] [Formula 7]
[0088] Φ
[0089] [Formula 7] is a transfer matrix that specifies the transition of the state vector.
[0090] [Formula 8]
[0091] B W
[0092] [Formula 8] is a column vector for selecting the input signal, and the value is specifically given according to the model.
[0093] [Formula 9]
[0094] K(n)
[0095] [Formula 9] is the Kalman gain, and in this example, it is determined in a filtering step as described later and is a vector (column vector).
[0096] like Figure 2 As shown, the Kalman filter input is the acceleration record of the low-sensitivity accelerometer [Formula 10] and the record of the high-sensitivity velocity meter [Formula 11] plus the observed disturbance r(n) m (n).
[0097] [Formula 10]
[0098]
[0099] [Formula 11]
[0100] v(n)
[0101] The observed interference r(n) is normal noise, its average value E[r(n)] is 0, and its variance E[r 2 (n)] is R(n) (E represents a statistical average. The same applies to the following). R(n) is a quantity set externally (by the user of the vibration sensor system, etc.), and the size of the Kalman gain can be adjusted by adjusting R(n).
[0102] [Formula 12]
[0103]
[0104] exist Figure 2 In the sensor fusion shown in the figure, [Equation 12] is regarded as the control input from the outside, and y m(n) is the observed value, and the above is used to perform the prediction step and filtering step of the linear Kalman filter to infer the state of the FSF sensor. m (n) as Figure 2 The y in m (n)( Figure 2 Although m=2 is set in the example, as mentioned above, m is arbitrary) input, the Kalman filter makes an estimation.
[0105] [Formula 13]
[0106]
[0107] [Formula 14]
[0108] y m (n) = C′x(n) + r(n)
[0109] (Formula 3)
[0110] Figure 2 The state estimation by the Kalman filter in is performed for the linear system described by [Formula 13] and [Formula 14]. Here, it is:
[0111] [Formula 15]
[0112] Q(n)=E[q(n)q′(n)]
[0113] (Formula 4)
[0114] [Formula 16]
[0115] R(n)=E[r 2 (n)]
[0116] (Formula 5)
[0117] In addition, [Formula 17] is a normal system interference (a column vector whose average is a zero vector), and [Formula 18] is a row vector given as a transposed matrix of [Formula 19].
[0118] [Formula 17]
[0119] q(n)
[0120] [Formula 18]
[0121] q′(n)
[0122] [Formula 19]
[0123] q(n)
[0124] In this specification, it is assumed that [*]' is a transposed matrix (transposed matrix) of a matrix (or vector) [*] or a vector given as a transposed matrix. The transposition is represented by the "'" symbol. In addition, [Formula 20] represented by (Formula 4) is a covariance matrix of system interference.
[0125] [Formula 20]
[0126] Q(n)
[0127] Specific composition of the vibration sensor system
[0128] Next, the specific structure of the vibration sensor system for performing sensor fusion according to the present invention is described. Figure 3 This is a block diagram showing the structure of a vibration sensor system according to an embodiment of the present invention. The vibration sensor system 1 includes an estimation device 2, an external measuring device 3, and a vibration sensor 4. The estimation device 2 includes a Kalman filter calculation unit 200 having parameters of a pseudo sensor.
[0129] The estimation device 2 is a device for performing the above-mentioned estimation by the Kalman filter, and includes a measurement value acquisition unit 201, a data storage unit 202, and a sensing value acquisition unit 203. In addition, the Kalman filter calculation unit 200 is a device for operating the Kalman filter using a pseudo sensor equivalent to the vibration sensor in the measurement band, and includes a state vector prior estimation value calculation unit 204, a prior covariance matrix calculation unit 205, a Kalman gain calculation unit 206, a state vector estimation value calculation unit 207, and a post covariance matrix calculation unit 208. The various functional units can also be realized by a computer for executing a state estimation program. An example of the device configuration of the estimation device as such an example is shown in Figure 4 However, the estimation device 2 does not necessarily need to be implemented by a computer, and the estimation device 2 may be constituted by integrated circuits such as ASIC (application specific integrated circuits) and FPGA (field-programmable gate array), or various calculation circuits or memory circuits may be combined to constitute various functional units of the estimation device 2.
[0130] Figure 4The estimation device 2 shown comprises: a processing unit 209 having a central processing unit (CPU) and a temporary memory (RAM); a memory unit 210 formed using memory elements such as a hard disk drive (hard disk drive) and an SSD (Solid State Drive); and an input / output unit 211 having a data input / output device, a display device, a keyboard, a mouse, etc. for inputting and outputting data in accordance with USB (Universal Serial Bus) standards. The memory unit 210 stores various programs such as a state estimation program, an OS (Operating system), data for the state estimation program, and other various data. By appropriately reading the programs and data into the RAM and executing the state estimation program and other programs by the CPU, the measurement value acquisition unit 201, the data storage unit 202, the sensing value acquisition unit 203, the state vector prior estimation value calculation unit 204, the prior covariance matrix calculation unit 205, the Kalman gain calculation unit 206, the state vector estimation value calculation unit 207, and the post covariance matrix calculation unit 208 are realized.
[0131] Specifically, first, the measured value acquisition unit 201 is a functional unit that acquires acceleration values from the external measuring device 3 (an accelerometer as a low-sensitivity seismic detector), and is implemented by a data input and output device of the input and output unit 211 controlled by a CPU that executes a state estimation program and various programs. The data storage unit 202 is implemented by a memory unit 210 controlled by a CPU that executes a state estimation program and various programs, and as described later, stores the state vector estimated value obtained by the previous estimation, the post-calculation matrix obtained by the previous calculation, the measured value corresponding to the time corresponding to the previous estimation and the previous calculation, and the system noise covariance matrix. The sensing value acquisition unit 203 is a functional unit that acquires the sensing value of the vibration sensor 4, and is implemented by a data input and output device of the input and output unit 211 controlled by a CPU that executes a state estimation program and various programs. The state vector prior estimation value calculation unit 204 is a functional unit realized by a CPU executing a state estimation program (especially a state vector prior estimation value calculation module) and various programs, and performs calculation of a state vector prior estimation value. The prior covariance matrix calculation unit 205 is a functional unit realized by a CPU executing a state estimation program (especially a prior covariance matrix calculation module) and various programs, and performs calculation of a prior covariance matrix. The Kalman gain calculation unit 206 is a functional unit realized by a CPU executing a state estimation program (especially a Kalman gain calculation module) and various programs, and performs calculation of a Kalman gain. The state vector estimated value calculation unit 207 is a functional unit realized by a CPU executing a state estimation program (especially a state vector estimated value calculation module) and various programs, and performs calculation of a state vector estimated value. The post hoc covariance matrix calculation unit 208 is a functional unit realized by a CPU executing a state estimation program (particularly, a post hoc covariance matrix calculation module) and various programs, and performs calculation of a post hoc covariance matrix.
[0132] Among them, the external measuring device 3 is an accelerometer as a low-sensitivity seismic detector (a device that can convert acceleration by calculation and has a dynamic range that exceeds the upper limit of the dynamic range of a high-sensitivity seismic detector), for example, a MEMS (Micro Electro Mechanical Systems) type acceleration sensor. The external measuring device 3 also has an input and output device for inputting a command signal from the outside and outputting a measurement record of acceleration to the outside.
[0133] (First Embodiment)
[0134] Model of a high-sensitivity speedometer
[0135] Next, an embodiment of a vibration sensor system using a model of a high-sensitivity speedometer will be specifically described as a first embodiment of the present invention. Figure 5 1 is a diagram showing an example of the vibration sensor 4 (VBB type seismograph as a high-sensitivity velocity meter).
[0136] The FSF sensor is assembled in the same way as a negative feedback seismograph designed using the classical PID control (Proportional Integral Differential Controller) theory and simulates Figure 5 The operation of the VBB seismograph is shown. Figure 5 The vibration sensor has a pendulum 4001 connected to a servo coil (for example, see the specific connection between the pendulum and the servo coil). Fig.10 In addition, the test coil and the drive coil are Fig.10 The cross-section shown on the paper is actually a ring shape. The magnet is also Fig.10 The cross section is shown on the paper, but it actually has a shape like a cylinder that is partially hollowed out. ), attenuation element 4002 (as a specific example, a damper that uses air resistance, which has no entity), rigid element 4003 (as a specific example, a spring that supports the oscillation), and displacement detector 4004 having three capacitor plates (electrodes), and the above elements are stored in a container 4005. Figure 5 As shown, the pendulum 4001 is connected to the central capacitor plate among the three capacitor plates included in the displacement detector 4004. In addition, the vibration sensor 4 has a displacement-voltage conversion amplifier A of the pendulum. D (For convenience, amplifier A D The conversion factor is A D [V / m] is used to express it. ), Differentiator G S (For convenience, the differentiator G S The coefficient of G S In the formula, the differentiator G S The effect of is expressed by [Formula 21]. ), impedance element R I (with resistance R I [Ω]), impedance element R D (with resistance R D [Ω]), impedance element R V (with resistance R V [Ω]), and also has an integrator represented by the integral symbol [Formula 22] (T(s) is set as a time constant. In the formula, the action of the integrator is represented by [Formula 23]).
[0137] [Formula 21]
[0138]
[0139] [Formula 22]
[0140] ∫
[0141] [Formula 23]
[0142]
[0143] in addition, Figure 5 Although not explicitly shown, magnets are arranged around the servo coil to which the pendulum 4001 is connected (see Fig.10 ), the servo coil moves in the magnetic field.
[0144] If the ground on which the vibration sensor 4 is installed vibrates, and an input displacement w(t) is given to the vibration sensor 4 (the unit is set to m. In this specification, quantities without explicit units are set to quantities expressed in the MKSA unit system.), the pendulum 4001 vibrates (the relative displacement of the pendulum in the storage container 4005 relative to the container is set to x(t) (the unit is set to m)). Due to the displacement of the pendulum 4001, the intervals between the three capacitor plates contained in the displacement detector 4004 change. Specifically, in a stationary state, if the interval between the upper capacitor plate and the central capacitor plate, and the interval between the central capacitor plate and the lower capacitor plate are set to d[m] respectively, the pendulum 4001 will vibrate. Figure 6 The displacement x in the direction shown in the figure causes the capacitor plate connected to the center of the pendulum 4001 to be displaced in the same manner. Accordingly, the distance between the upper capacitor plate and the central capacitor plate becomes dx, and the distance between the central capacitor plate and the lower capacitor plate becomes d+x. Therefore, the capacitance between the upper capacitor plate and the central capacitor plate, and the capacitance between the central capacitor plate and the lower capacitor plate change respectively, and the displacement x of the pendulum 4001 is reflected by the electrical circuit A. D The voltage is detected as a voltage proportional to the displacement. The voltage is integrated by the above circuit elements, and the current i(t) is made to flow through the servo coil, and the oscillation 4001 is controlled to be stationary. That is, the current i(t) is made to flow (feed back) to the servo coil driven by the magnet (refer to Fig.10) generates a force in the servo coil in the magnetic field formed by the servo coil, and the force is used to control the pendulum to be stationary. Specifically, [Formula 24] is generated relative to the input displacement w(t) as the restoring acceleration (G[N / A] is the motor generator constant of the servo coil, and m[kg] is the mass of the pendulum 4001.), and the restoring acceleration [Formula 25] is realized (the reason why "˙˙" is added to w is to indicate the second derivative with respect to time.
[0145] [Formula 24]
[0146]
[0147] [Formula 25]
[0148]
[0149] When "˙" is added to the top, it means that the time differential is performed once. For quantities other than w that change over time, the time differential is also appropriately marked. ). The acceleration, velocity, and displacement are obtained from the current required for this control as the sensing value. Specifically, from the differentiator G S The output end of the pendulum 4001 obtains the acceleration output, and the displacement-voltage conversion amplifier A D The velocity output of the pendulum 4001 is obtained from the output end of the integrator, and the displacement output of the pendulum 4001 is obtained from the output end of the integrator.
[0150] Next, according to non-patent document 4, a method for simulating a FSF sensor that is a virtual sensor is described. Figure 5 The model of the operation of the VBB type seismograph shown in FIG. First, the parameters of the oscillation system of the mechanism and the design target of the FSF sensor are provided as specifications as shown below.
[0151] [Oscillating system of the mechanism]
[0152] Mass of pendulum = m [kg], attenuation constant = h [dimensionless quantity], natural circular vibration frequency = ω0 [rad], servo coil motor generator constant = G [N / A]
[0153] [FSF sensor characteristics]
[0154] Speed output sensitivity S V [V / (m / s)], low frequency cutoff circular vibration number = ω C [rad], and [Equation 26]
[0155] [Formula 26]
[0156] Decay constant = ζ(zeta) = 0.6321
[0157] According to the above specifications, the FSF sensor is implemented in a four-stage process. First, in step 1, the feedback gain vector (column vector) [Equation 27] is determined as follows.
[0158] [Formula 27]
[0159] k=[k1,k2,k3]′
[0160] (Formula 8)
[0161] Here, the high-frequency cutoff circular vibration number ω3 for oscillation prevention is provided as a control parameter for trial calculation. Usually, it is set to 10 for safety. 3 [Hz] approximately.
[0162] [Step 1: Feedback gain vector]
[0163] [Formula 28]
[0164] ω3≈2π·10 3 (Control parameter) (Equation 9)
[0165] [Formula 29]
[0166]
[0167] Next, in step 2, the sensitivity, S, of the acceleration and displacement signals are determined. A and S D At this time, given the time constant T of the integrator and the coefficient G of the differentiator S As a control parameter. On the contrary, given S A and S D To find A D , T, G S No problem.
[0168] [Step 2: Sensitivity]
[0169] [Formula 30]
[0170]
[0171] (T, G S : Control parameters)
[0172] (Formula 11)
[0173] Finally, in step 3, decide Figure 5 Feedback impedance R V , R D , R I .
[0174] [Step 3: Feedback Impedance]
[0175] [Formula 31]
[0176]
[0177] Step 4 is the confirmation operation and is the calculation of the transfer function of the obtained FSF sensor. Here, H A (s), H V (s), H D (s) becomes the transfer function of the acceleration, velocity, and displacement output of the FSF sensor relative to the acceleration, velocity, and displacement input. The polar structure of the transfer function can be obtained from the common characteristic equation Δ(s)=0.
[0178] [Step 4: Transfer function]
[0179] [Formula 32]
[0180]
[0181] The block diagram of the FSF sensor constructed in the above steps becomes, if the state vector [Equation 33] is used, Figure 7 As shown. Figure 7 In the output terminals D, V, and A of the FSF sensor, [Formula 34] is obtained as the output (it should be noted that no observation noise is introduced at this point).
[0182] [Formula 33]
[0183]
[0184] [Formula 34]
[0185] y(t)=[d(t) v(t) a(t)]′
[0186] (Formula 15)
[0187] [Formula 35]
[0188]
[0189] [Formula 36]
[0190] k=[k1,k2,k3]′
[0191] (Formula 17)
[0192] [Formula 37]
[0193] b(t)=k′·x(t)
[0194] (Formula 18)
[0195] In addition, if the state feedback gain [Formula 36] is used to restore the acceleration [Formula 35], it becomes [Formula 37] (inner product calculation) (this means full state feedback.) The details are as follows Figure 8 shown. Figure 8 To extract only Figure 7 , which is a graph of the restoration acceleration b(t) and the state vector x(t), and is a graph that visualizes Equation 18.
[0196] If you will Figure 7 and Figure 8 The block line diagram of is expressed numerically, that is, in state space, as shown below.
[0197] [Formula 38]
[0198]
[0199] Among these are:
[0200]
[0201] and
[0202]
[0203] [Formula 39]
[0204]
[0205] If only the speed output is observed in the observation vector, (Equation 20) becomes the following equation.
[0206] [Formula 40]
[0207] v(t)=C′x(t), C=[0 A D 0]′
[0208] (Formula 21)
[0209] As for the design example of FSF sensor, if VBB type is used as the object, it will be as follows. First, consider the seismic detector elements with the mass of the pendulum m = 0.09 [kg], the natural vibration frequency f0 = 1.5 [Hz], the attenuation constant of 0.4, and the motor generator constant G = 56 [N / A]. Next, as the characteristics of VBB, let S V =750[V / (m / s)] and [Formula 41].
[0210] [Formula 41]
[0211]
[0212] If f3=1.062[kHz] is given under this condition, the feedback gain becomes [Formula 42], and [Formula 43] is obtained.
[0213] [Formula 42]
[0214] k1=18.2975[s -3 ],k2=352.96[s -2 ],k3=6,667[s -1 ]
[0215] (Formula 23)
[0216] [Formula 43]
[0217] A d =5×10 6 [V / m]
[0218] (Formula 24)
[0219] In addition, if [Formula 44] is assumed, the sensitivity of the acceleration output [Formula 45] and the sensitivity of the displacement output [Formula 46] can be obtained.
[0220] [Formula 44]
[0221] T=80[s],G S =0.003[s -1 ]
[0222] (Formula 25)
[0223] [Formula 45]
[0224] S A =2.2523[V / (m / s 2 )]
[0225] (Formula 26)
[0226] [Formula 46]
[0227] S D =9.375[V / m]
[0228] (Formula 27)
[0229] The frequency response characteristics of the acceleration, velocity and displacement outputs of the designed FSF sensor relative to the acceleration input are Fig. 9 This is the characteristic of a standard VBB. As with a real VBB, the output from the displacement (D) end has a characteristic that has DC sensitivity to the acceleration input. In the dB display of the gain characteristic in the figure, assume that [Formula 47] is 0 [dB].
[0230] [Formula 47]
[0231] 1[V / (m / s 2 )],1[V / (m / s)],1[V / m]
[0232] The model of the high-sensitivity speedometer described above is summarized as follows.
[0233] When the state vector is set to [Formula 48], it becomes [Formula 49],
[0234] [Formula 48]
[0235]
[0236] [Formula 49]
[0237]
[0238] Among these are:
[0239]
[0240] and
[0241]
[0242] [Formula 50]
[0243] v(t)=C′x(t), C=[0 A D 0]′
[0244] (Formula 30)
[0245] Here, equation 51 is a quantity that does not include observation noise.
[0246] [Formula 51]
[0247] v(n)
[0248] Although (Equation 28) to (Equation 30) are expressions of continuous-time systems, if the sampling time ΔT is used as a parameter to convert them into discrete systems, through [Equation 52], under the condition of [Equation 53], the following equation [Equation 54] becomes available.
[0249] [Formula 52]
[0250] nΔT≤t<(n+1)ΔT
[0251] (Formula 31)
[0252] [Formula 53]
[0253]
[0254] [Formula 54]
[0255]
[0256] Here:
[0257] [Formula 55]
[0258] Φ≡e AΔT
[0259] (Formula 34)
[0260] and
[0261] [Formula 56]
[0262]
[0263] [Formula 57]
[0264] v(n)=C′x(n)
[0265] (Formula 36)
[0266] The linear model for (Equation 33) to (Equation 36) (system noise and observation noise are not introduced at this stage. The block diagram of the system is shown in Figure 6 . ), the system noise [Formula 58] explained in (Formula 2) is introduced (assuming that the average is 0 and the covariance matrix Q(n) is a normal interference.) In addition, the quantity after adding the observation noise r(n) to [Formula 59] is redefined as y2(n) as [Formula 60].
[0267] [Formula 58]
[0268] q(n)
[0269] [Formula 59]
[0270] v(n)
[0271] [Formula 60]
[0272] y2(n)=v(n)+r(n)
[0273] (Formula 37)
[0274] The observed noise r(n) is assumed to be a normal disturbance with an average of 0 and a variance of R(n). The above (Equation 37) is a signal model of a high-sensitivity speedometer with observed noise.
[0275] In summary, Figure 2 The configuration of sensor fusion is described by the following linear system including system noise and observation noise.
[0276] [Formula 61]
[0277]
[0278] y2(n)=C′x(n)+r(n)
[0279] (Formula 38)
[0280] In this model, since [Formula 63] is obtained when (Formula 38) is expressed as [Formula 62], in the range where VBB is not cut off, assuming that [Formula 64] is obtained with sufficient accuracy, it becomes [Formula 65], so it becomes [Formula 66].
[0281] [Formula 62]
[0282] y2(n)=v(n)+r(n)
[0283] (Formula 39)
[0284] [Formula 63]
[0285] r(n)=y2(n)-v(n)
[0286] (Formula 40)
[0287] [Formula 64]
[0288]
[0289] [Formula 65]
[0290]
[0291] [Formula 66]
[0292] r(n)≈0
[0293] (Formula 42)
[0294] In the range where VBB is cut off, if [Formula 67] is obtained, then [Formula 68] is obtained. However, it does not indicate [Formula 69] but indicates [Formula 70].
[0295] [Formula 67]
[0296] |y2(n)|>>|v(n)|
[0297] (Formula 43)
[0298] [Formula 68]
[0299]
[0300] [Formula 69]
[0301] v(n)≈0
[0302] (Formula 45)
[0303] [Formula 70]
[0304] |v(n)|<<|r(n)|
[0305] (Formula 46)
[0306] The signal observed by VBB is modeled as an observation disturbance r(n) of this nature. In (Equation 38), the interpretation is expanded and r(n) is modeled as a time-varying normal disturbance with a mean of zero and a variance of R(n).
[0307] State estimation by Kalman filter
[0308] In the system of (Equation 38), a linear Kalman filter with control input can be applied. The prediction step and filtering step of the Kalman filter in this case are as follows.
[0309] [Formula 71]
[0310] Prediction Steps:
[0311]
[0312] Filtering steps:
[0313]
[0314] In (Formula 47), [Formula 72] is the prior state estimation value of the state vector [Formula 73], and [Formula 74] is the prior covariance matrix (prior error covariance matrix) of the prior value of the covariance matrix (error covariance matrix) of the state vector [Formula 75].
[0315] [Formula 72]
[0316]
[0317] [Formula 73]
[0318] x(n)
[0319] [Formula 74]
[0320]
[0321] [Formula 75]
[0322] x(n)
[0323] [Formula 76]
[0324] P(n)
[0325] [Formula 77] is the estimated value of the state vector [Formula 78] at discrete time n-1, and [Formula 79] is the value of the covariance matrix [Formula 80] at discrete time n-1 calculated by the Kalman filter, and is the post-covariance matrix.
[0326] [Formula 77]
[0327]
[0328] [Formula 78]
[0329] X(n-1)
[0330] [Formula 79]
[0331]
[0332] [Formula 80]
[0333] P(n-1)
[0334] [Formula 81] is the estimated value of the state vector [Formula 82] at discrete time n, [Formula 83] is the value of the covariance matrix [Formula 84] at discrete time n calculated by the Kalman filter, and is the post-covariance matrix.
[0335] [Formula 81]
[0336]
[0337] [Formula 82]
[0338] x(n)
[0339] [Formula 83]
[0340]
[0341] [Formula 84]
[0342] P(n)
[0343] In addition, [Formula 85] is a unit matrix (a matrix of 3 rows and 3 columns in the current embodiment).
[0344] [Formula 85]
[0345] I
[0346] In summary, by introducing system noise and observation noise, sensor fusion using a Kalman filter (estimation of the state of the vibration of the FSF sensor or high-sensitivity geophone) can be performed.
[0347] Operation flow of the vibration sensor system
[0348] Next, the overall operation flow of the vibration sensor system 1 will be described by dividing it into a design phase and an operation phase.
[0349] Design time
[0350] Fig.11 The flowchart shows the action flow when designing a vibration sensor system. In step S1101, the processing unit 209 of the estimation device 2 executes the pseudo sensor creation program stored in the memory unit 210, and accordingly performs parameter calculations of a (full) state feedback type seismic detector having equivalent characteristics to a high-sensitivity seismic detector as described in (Equation 6) to (Equation 36) above to create a simulated sensor. Based on this, a discretized model of the state feedback type seismic detector described by (Equation 33) to (Equation 36) is generated (step S1102). The various parameters and other data used in this model are used in the processing using the Kalman filter performed by the estimation device 2, so the various parameters and other data are stored in the memory unit 210 of the estimation device 2 (refer to Figure 4 ).
[0351] Next, the input / output unit 211 of the estimation device 2 obtains the low-sensitivity geophone discrete signal (acceleration record) from the external measuring device 3 and the high-sensitivity geophone discrete signal (velocity record) from the vibration sensor 4 at all times (step S1103), and stores the obtained recorded data in the storage unit 210. The simulated sensor (processing unit 209) adds the observation signal from the observation interference model to the high-sensitivity geophone discrete signal as described in (Equation 37) (step S1104), and generates a linear system of (Equation 38) including system noise. In the following, it is assumed that the simulated sensor obtains the acceleration record from the external measuring device 3 at all times, and generates system noise and observation noise from each interference model (assumed to be pre-defined and stored in the storage unit 210. As described later, the interference model can be adjusted afterwards.), and continuously generates a high-sensitivity velocity record y2(n) for each discrete moment n according to the model of (Equation 38). In addition, it is assumed that the initial value of the state vector [Formula 86] in (Formula 38) at n=0 is given in advance and stored as data for the state estimation program in the memory unit 210 (the same applies to other variables that require initial values, such as system noise).
[0352] [Formula 86]
[0353] X(n)
[0354] Next, the processing unit 209 of the estimation device 2 executes a state estimation program by using the acquired recorded data and various data stored in the memory unit 210, and estimates the oscillation motion of the high-sensitivity seismic detector by a Kalman filter with exogenous input according to (Equation 47), and calculates the estimated value [Equation 88] of the state vector [Equation 87] and the value [Equation 90] of the post-covariance matrix [Equation 89] of the estimation error by the Kalman filter at all times (step S1105).
[0355] [Formula 87]
[0356] x(n)
[0357] [Formula 88]
[0358]
[0359] [Formula 89]
[0360] P(n)
[0361] [Formula 90]
[0362]
[0363] [Formula 91]
[0364] R(nn o )
[0365] In addition, [Formula 91] in the Kalman gain in (Formula 47) is the variance of the observed noise and is a quantity ("Kalman gain adjustment term") that can be set externally (in one example, it is input by the user through the input / output unit 211 and stored in the memory unit 210 as data for the state estimation program by the processing unit 209 that executes the state estimation program.). When the speed recorded from the vibration sensor 4 is larger than a predetermined value (in one example, it is input by the user through the input / output unit 211 and stored in the memory unit 210 as data for the state estimation program by the processing unit 209 that executes the state estimation program.) (when the speed record of the high-sensitivity speedometer is saturated, etc.), the processing unit 209 that executes the state estimation program can increase the Kalman gain adjustment term compared to a case where the speed recorded is smaller than the predetermined value, thereby reducing the Kalman gain.
[0366] The output of the state feedback type seismic detector calculated from the oscillation motion of the high-sensitivity seismic detector estimated by the Kalman filter can be obtained by the processing unit 209 that executes the state estimation program (S1106). This output is a fusion signal of high sensitivity (vibration sensor 4) and low sensitivity (external measuring device 3) (S1107). The user can know the estimated state of the oscillation by displaying each component of the estimated state vector through the display device of the input / output unit 211.
[0367] In addition, when designing, the processing unit 209 executing the state estimation program performs a comparison judgment between the recorded values of acceleration, velocity, etc. indicated by the signals from the external measuring device 3 and the vibration sensor 4 and the values of acceleration, velocity, etc. calculated from the state vector estimated using the Kalman filter (step S1108). In one example, at a certain discrete moment, when the difference between the recorded values indicated by the signals from the external measuring device 3 and the vibration sensor 4 and the values calculated from the state vector estimated using the Kalman filter exceeds a predetermined value, the processing unit 209 executing the state estimation program determines that the consistency between the recorded value and the calculated value is low ("NO" in the conditional branch of step S1108), and adjusts the observation disturbance model by changing the size of the variance of the observation noise (step S1109). Using the observation disturbance from the adjusted observation disturbance model, the processing from step S1104 to step S1108 is performed again. The above processing is repeated until it is determined that the consistency is high in step S1108. When the consistency is judged to be high in step S1108 (in one example, when the difference between the recorded value indicated by the signal from the external measuring device 3, the vibration sensor 4, and the value calculated from the state vector estimated using the Kalman filter does not exceed the predetermined value. "YES" in the conditional branch of step S1108), the design is completed.
[0368] When using
[0369] By using Fig.11 The design process described includes various calculation parameters or interference models to determine the linear system of (Equation 38) and the Kalman filter of (Equation 47). When in operation, the state vector of the vibration sensor 4 is estimated at discrete moments (the estimation is repeated while increasing by n=1, 2, etc.). The flowchart when in operation is as follows Fig.12 Steps S1201, S1202, S1203, S1204, and S1205 may be respectively the same as steps S1103, S1104, S1105, S1106, and S1107 during design.
[0370] Details of the calculations performed by the Kalman filter
[0371] Will be in Fig.11 Step S1105 in the flowchart of Fig.12 The detailed process of the calculation performed by the Kalman filter in step S1203 of the flowchart is shown in Fig.13 Flowchart of the process.
[0372] First, the state vector prior estimation value calculation unit 204 reads out the previous state vector estimation value data, the previous post-covariance matrix data, the previous measured value data, and the calculation parameters from the storage unit 210 (step S1301). The read data is temporarily stored in the RAM of the processing unit 209 (the temporary storage of various data in the RAM is appropriately omitted in the subsequent description and the description so far. The reading of various data from the storage unit is also appropriately omitted.). In the initial estimation, that is, when n=1 is set in (Formula 47) for processing, the previous state vector estimation value data, the previous post-covariance matrix data, and the previous measured value data (the data of each in n=0) are each initial value data that is previously stored in the storage unit 210 by the user through the input / output unit 211 and provided from the outside.
[0373] Next, the state vector advance estimation value calculation unit 204 calculates the advance estimation value of the state vector [Formula 93] in (Formula 47) according to [Formula 92] (Formula 48) of the prediction step (step S1302), and temporarily stores it in the RAM of the processing unit 209.
[0374] [Formula 92]
[0375]
[0376] [Formula 93]
[0377]
[0378] Next, the prior covariance matrix calculation unit 205 calculates the prior covariance matrix [Formula 95] according to [Formula 94] of the prediction step in (Formula 47) (step S1303), and temporarily stores it in the RAM of the processing unit 209.
[0379] [Formula 94]
[0380]
[0381] [Formula 95]
[0382]
[0383] Next, the Kalman gain calculation unit 206 calculates the Kalman gain in (Formula 47) according to [Formula 96] of the filtering step (step S1304), and temporarily stores it in the RAM of the processing unit 209.
[0384] [Formula 96]
[0385]
[0386] [Formula 97]
[0387] R(nn o )
[0388] Here, the Kalman gain calculation unit 206 adjusts the Kalman gain by using the Kalman gain adjustment term [Formula 97]. Specifically, when Fig.11 Step S1103 in the process of Fig.12 In the case where the value (v(n)) of the high-sensitivity seismic detector discrete signal obtained in step S1201 of the process is larger than the preset value, the Kalman gain adjustment term is increased compared to the case where the value is smaller than the preset value, thereby reducing the Kalman gain, and the Kalman gain is adjusted accordingly. In addition to the velocity record, when the sensing value from the vibration sensor 4 is a displacement record or an acceleration record, the Kalman gain is also adjusted by the Kalman gain adjustment term according to the comparison result between each preset value and the sensing value through the Kalman gain calculation unit 206 (when one or more of the displacement, velocity, and acceleration of the oscillation as the sensing value from the vibration sensor 4 is larger than the preset value, the Kalman gain adjustment term is increased compared to the case where one or more of the values is smaller than the preset value).
[0389] Next, the state vector estimated value calculation unit 207 calculates the state vector estimated value [Formula 99] according to [Formula 98] of the filtering step in (Formula 47) (step S1305), and temporarily stores it in the RAM of the processing unit 209.
[0390] [Formula 98]
[0391]
[0392] [Formula 99]
[0393]
[0394] Next, the post-covariance matrix calculation unit 208 calculates the post-covariance matrix [Formula 101] according to [Formula 100] of the filtering step in (Formula 47) (step S1306), and temporarily stores it in the RAM of the processing unit 209.
[0395] [Formula 100]
[0396]
[0397] [Formula 101]
[0398]
[0399] Next, in step S1307, the state vector estimated value calculation unit 207 stores the state vector estimated value estimated in step S1305 in the memory unit 210 (the state vector estimated value at discrete time n is used as the “previous state vector estimated value data” in the Kalman filter processing at discrete time n+1). In addition, the post-covariance matrix calculation unit 208 stores the post-covariance matrix calculated in step S1306 in the memory unit 210 (the post-covariance matrix at discrete time n is used as the “previous post-covariance matrix data” in the Kalman filter processing at discrete time n+1). In addition, the processing unit 209 that executes the state estimation program stores various data such as the measurement value obtained from the external measuring device 3 (the measurement value at discrete time n is used as the “previous measurement value data” in the Kalman filter processing at discrete time n+1). The processing of steps S1301 to S1307 is repeatedly performed all the time along with the data signal acquisition of steps S1103 and S1201 (the data acquisition can be performed uniformly first, and then the Kalman filter processing can be performed uniformly, or the data acquisition and Kalman filter processing can be synchronized by the following process: after the data is acquired at a discrete moment, the Kalman filter processing is immediately performed at the discrete moment, and after adding 1 to the discrete moment (incrementing), data is acquired at the next discrete moment to perform Kalman filter processing.), when the low-sensitivity seismic detector discrete signal and the high-sensitivity seismic detector discrete signal are no longer acquired, or the discrete moment reaches the predetermined end value and meets the predetermined end conditions, the processing by the Kalman filter is terminated.
[0400] Include Fig.13 The Kalman filter processing of the process shown in the figure is repeated all the time based on Fig.12 The process of the operation flow shown in the figure estimates the state vector (the state vector of the swing) of the vibration sensor 4, which changes moment by moment. The estimated value of the state vector can be displayed on the display device of the input / output unit 211, etc., by the processing unit 209 executing the state estimation program, so that the user can confirm the estimated value of the state vector.
[0401] (Second Embodiment)
[0402] Model of high sensitivity displacement meter
[0403] Next, an embodiment in which a high-sensitivity displacement meter is used as the vibration sensor 4 will be described as a second embodiment of the present invention. However, description of the same parts as the first embodiment will be omitted.
[0404] Fig.14 This is a diagram showing an example of a vibration sensor (VBB type seismograph which is a high-sensitivity displacement meter). Fig.15 This is a block diagram showing the structure of sensor fusion using a high-sensitivity displacement meter and a Kalman filter of a strong earthquake meter. Figure 5 The same configuration is implemented as for the speedometer.
[0405] exist Fig.14 In the vibration sensor, the displacement signal can be obtained from Fig.14 The position #1 of the measurement band can be taken out by an external integrator with attenuation characteristics outside the measurement band, such as the position #2. The state space of the continuous-time system in the measurement band is represented as Fig.16 As shown, the speed signal [Formula 103] is also output simultaneously with the displacement signal [Formula 102], so the above can be combined and utilized.
[0406] [Formula 102]
[0407] d(t)
[0408] [Formula 103]
[0409] v(t)
[0410] At this time, the state space shows that if the state vector of the vibration sensor 4 is set to [Formula 104], it becomes [Formula 105],
[0411] [Formula 104]
[0412]
[0413] [Formula 105]
[0414]
[0415] In this, become:
[0416]
[0417] and
[0418]
[0419] [Formula 106]
[0420]
[0421] When the representation of this time continuous system is converted into a discrete system by taking the sampling time ΔT as a parameter, through [Formula 107], under the condition of [Formula 108], it becomes the following formula [Formula 109].
[0422] [Formula 107]
[0423] nΔT≤t<(n+1)ΔT
[0424] (Formula 56)
[0425] [Formula 108]
[0426]
[0427] [Formula 109]
[0428]
[0429] In this, become:
[0430]
[0431] [Formula 110]
[0432]
[0433] Therefore, if the normal interference [Formula 112] in which the average vector is a zero vector (two-dimensional column vector) and the covariance matrix is [Formula 111] is considered, the signal model of the high-sensitivity displacement meter becomes [Formula 113], and sensor fusion can be performed by the following Kalman filter [Formula 114].
[0434] [Formula 111]
[0435] R(n)
[0436] [Formula 112]
[0437]
[0438] [Formula 113]
[0439]
[0440] [Formula 114]
[0441] Prediction Steps:
[0442]
[0443] Filtering steps:
[0444]
[0445] The device configuration, design, and operation flow of the vibration sensor system 1 can be implemented in the same manner as in the first embodiment, after the system noise [Formula 115] is introduced in the same manner as in the first embodiment, except that [Formula 116] is used as the equation of the linear system and (Formula 62) is used as the equation of the Kalman filter.
[0446] [Formula 115]
[0447] q(n)
[0448] [Formula 116]
[0449]
[0450] y m (n) = C′x(n) + r(n)
[0451] (Formula 63)
[0452] (Third Embodiment)
[0453] Model of a high-sensitivity accelerometer
[0454] Next, an embodiment in which a high-sensitivity accelerometer is used as the vibration sensor 4 will be described as a third embodiment of the present invention. However, description of the same parts as the first embodiment will be omitted.
[0455] Fig.17 A diagram showing an example of a vibration sensor (high-sensitivity accelerometer). Fig.18 A block diagram showing the construction of sensor fusion using a high-sensitivity accelerometer and a Kalman filter for a strong-seismometer. The high-sensitivity accelerometer can be used with Figure 5 The same configuration as the speedometer is implemented.
[0456] Figure 5 In the case of a VBB type accelerometer, the acceleration signal is Figure 7 The A terminal of is output as [Formula 117].
[0457] [Formula 117]
[0458] a(t)
[0459] [Formula 118]
[0460] v(t)
[0461] [Formula 119]
[0462] a(t)
[0463] If [Formula 118] of the high-sensitivity speedometer is replaced by [Formula 119], sensor fusion can be performed. However, Figure 5The sensitivity of the seismograph with the structure of is also zero at a frequency of 0, which is disadvantageous compared to the commonly used force balance type negative feedback type accelerometer. Fig.17 The conventional negative feedback type accelerometer shown is treated as a high sensitivity accelerometer. Fig.17 The state space display of the continuous-time system of the accelerometer becomes as follows Fig.19 shown.
[0464] Fig.19 The state space representation of is [Formula 121] when the state vector of the vibration sensor 4 is set to [Formula 120],
[0465] [Formula 120]
[0466]
[0467] [Formula 121]
[0468]
[0469] In this, become:
[0470]
[0471] [Formula 122]
[0472] a(t)=C′x(t), C=[k1 k2]′
[0473] (Formula 66)
[0474] The representation of this time continuous system, if converted into a discrete system by taking the sampling time ΔT as a parameter, becomes the following equation [Equation 125] through [Equation 123] under the condition of [Equation 124].
[0475] [Formula 123]
[0476] nΔT≤t<(n+1)ΔT
[0477] (Formula 67)
[0478] [Formula 124]
[0479]
[0480] [Formula 125]
[0481]
[0482] Among these are:
[0483]
[0484] [Formula 126]
[0485] a(n)=C′x(n)
[0486] (Formula 70)
[0487] Therefore, if the normal disturbance with mean 0 and variance [Formula 127] is considered [Formula 128], the signal model of the high-sensitivity accelerometer becomes [Formula 129], which can be obtained by Fig.18 Sensor fusion performed by the Kalman filter [Equation 130] is shown.
[0488] [Formula 127]
[0489] R(n)
[0490] [Formula 128]
[0491] r(n)
[0492] [Formula 129]
[0493] y3(n)=a(n)+r(n)
[0494] (Formula 71)
[0495] [Formula 130]
[0496] Prediction steps:
[0497]
[0498] Filtering steps:
[0499]
[0500] The device configuration, design, and operation flow of the vibration sensor system 1 can be implemented in the same manner as in the first embodiment, after the system noise [Formula 131] is introduced in the same manner as in the first embodiment, except that [Formula 132] is used as the equation of the linear system and (Formula 72) is used as the equation of the Kalman filter.
[0501] [Formula 131]
[0502] q(n)
[0503] [Formula 132]
[0504]
[0505] y3(n)=C′x(n)+r(n)
[0506] (Formula 73)
[0507] [Industrial Applicability]
[0508] The present invention can be used for state estimation of all vibration sensors including seismographs.
[0509] Description of Reference Numerals
[0510] 1 Vibration sensor system
[0511] 2 Estimation device
[0512] 3 External measuring device (low sensitivity seismic detector)
[0513] 4 Vibration sensors (high-sensitivity seismic detectors)
[0514] 200 Kalman filter calculation part
[0515] 201 Measurement value acquisition unit
[0516] 202 Data Storage Department
[0517] 203 Sensing value acquisition unit
[0518] 204 State vector pre-estimated value calculation unit
[0519] 205 Prior covariance matrix calculation part
[0520] 206 Kalman gain calculation unit
[0521] 207 State vector estimated value calculation unit
[0522] 208 Post hoc covariance matrix calculation part
[0523] 209 Processing Department
[0524] 210 Memory
[0525] 211 Input and Output
[0526] 4001 connected to the oscillator of the servo coil
[0527] 4002 Attenuation Factor
[0528] 4003 Rigidity Elements
[0529] 4004 Displacement Detector (Three Capacitor Plates)
[0530] 4005 container.
Claims
1. An estimation device for estimating a state vector of a vibration sensor according to discrete moments, that is, a state vector that changes with time as an external input, and calculating a covariance matrix of the state vector according to the discrete moments, the estimation device comprising: a measurement value acquisition unit that acquires a measurement value of a physical quantity related to vibration measured by an external measuring device as the external input; A sensing value acquisition unit, which acquires the sensing value of the vibration sensor; A data storage unit stores a state vector estimation value obtained by a previous estimation, a post hoc covariance matrix obtained by a previous calculation, the aforementioned measured values corresponding to the time corresponding to the previous estimation and the previous calculation, and a covariance matrix of system noise; a state vector prior estimation value calculation unit that calculates a state vector prior estimation value using the state vector estimated value obtained by the previous estimation and the measurement value corresponding to the time corresponding to the previous estimation and the previous calculation; The ex ante covariance matrix calculation unit calculates the ex ante covariance matrix using the ex post covariance matrix obtained by the previous calculation and the covariance matrix of the system noise; A Kalman gain calculation unit that calculates the Kalman gain using the prior covariance matrix; a state vector estimated value calculation unit that calculates a state vector estimated value in this estimation using the state vector prior estimation value, the Kalman gain, and the sensed value; and The post-covariance matrix calculation unit calculates the post-covariance matrix in this calculation using the pre-covariance matrix and the Kalman gain.
2. A vibration sensor system comprising: the vibration sensor; the external measuring device; and the estimation device according to claim 1.
3. The vibration sensor system according to claim 2, wherein the external measuring device is a device that can perform acceleration conversion by calculation, and has a dynamic range that exceeds an upper limit of a dynamic range of the vibration sensor, and the measurement value acquisition unit acquires the measurement value of the acceleration.
4. The vibration sensor system according to claim 2 or claim 3, wherein the vibration sensor has a vibration and measures one or more values of displacement, velocity, and acceleration, and the sensing value acquisition unit acquires one or more values of the displacement, velocity, and acceleration.
5. The vibration sensor system according to claim 4, wherein the Kalman gain calculation unit adjusts the Kalman gain by using a calculation of a Kalman gain adjustment term, and the Kalman gain is adjusted in the following manner: when one or more of the displacement, velocity, and acceleration of the vibration sensor is greater than a predetermined value, the Kalman gain adjustment term is increased compared to a case where the one or more values are smaller than the predetermined value, thereby reducing the Kalman gain.
6. The system according to claim 2 or claim 3, wherein the vibration sensor has a pendulum and measures one or more values of displacement, velocity, and acceleration; The Kalman filter at the subsequent stage is provided with a simulated sensor, the simulated sensor simulates the action of a vibration sensor having a pendulum by calculation, and determines one or more values of displacement, velocity, and acceleration of the simulated pendulum by calculation; The state vector estimated value calculation unit uses, as the sensing value, one of the values of the displacement, velocity, and acceleration measured by the vibration sensor and one of the values of the displacement, velocity, and acceleration of the simulated oscillation determined by the simulated sensor multiplied by a coefficient, based on the size of the one or more values of the displacement, velocity, and acceleration measured by the vibration sensor.
7. A method performed by an estimation device, wherein the estimation device estimates a state vector of a vibration sensor according to discrete moments, that is, a state vector that changes with time with external input, and calculates a covariance matrix of the state vector according to the discrete moments, the method performed by the estimation device include: A step of obtaining a measurement value of a physical quantity related to vibration measured by an external measuring device as the external input; The step of obtaining the sensing value of the vibration sensor; A step of calculating a state vector prior estimation value using a state vector estimation value obtained by a previous estimation and a measurement value corresponding to a time corresponding to the previous estimation and the previous calculation; A step of calculating an ex ante covariance matrix using the ex post covariance matrix obtained by the previous calculation and the covariance matrix of the system noise; The step of calculating the Kalman gain using the prior covariance matrix; The step of calculating the estimated value of the state vector in this estimation using the previously estimated value of the state vector, the Kalman gain, and the sensed value; and The step of calculating the post-covariance matrix in this calculation using the previous covariance matrix and the previous Kalman gain.
8. A memory element having a program stored therein, the program being used to cause a computer to execute a method for estimating a state vector of a vibration sensor according to discrete moments, i.e., a state vector that changes with time as an external input, and calculating a covariance matrix of the state vector according to the discrete moments, the program causing the computer to execute the following steps: A step of obtaining a measurement value of a physical quantity related to vibration measured by an external measuring device as the external input; The step of obtaining the sensing value of the vibration sensor; A step of calculating a state vector prior estimation value using a state vector estimation value obtained by a previous estimation and a measurement value corresponding to a time corresponding to the previous estimation and the previous calculation; A step of calculating an ex ante covariance matrix using the ex post covariance matrix obtained by the previous calculation and the covariance matrix of the system noise; The step of calculating the Kalman gain using the prior covariance matrix; The step of calculating the estimated value of the state vector in this estimation using the previously estimated value of the state vector, the Kalman gain, and the sensed value; and The step of calculating the post-covariance matrix in this calculation using the previous covariance matrix and the previous Kalman gain.
Citation Information
Patent Citations
Velocity and acceleration two-parameter digital geophone
CN102628960A
Method for automatically picking up seismic reflection event based on Kalman filtering
CN103149591A
Integrated system of GNSS receiver and seismometer
CN103760594A
PP-PS joint inversion system based on Kalman filtering
CN109521468A
Processing method and configuration method for configuring seismometer from plurality of acceleration sensors
JP2009031032A
