JP2021193359A - Soundness evaluation system and soundness evaluation method - Google Patents

Abstract

To provide a soundness evaluation system capable of improving accuracy of an estimate indicating a response of a building based on a building model for estimating responses of buildings, and a soundness evaluation method.SOLUTION: A soundness evaluation system 1 for evaluating soundness of a building on the basis of a building model for estimating responses of buildings comprises a plurality of sensors Sn which detect an acceleration of a building B, and an arithmetic logical unit 14 which calculates a primary natural period of the building on the basis of detection values of the plurality of sensors, and updates parameters on distribution of masses and rigidity included in the building model on the basis of the primary natural period.SELECTED DRAWING: Figure 1

Description

The present invention relates to a soundness evaluation system and a soundness evaluation method for evaluating the soundness of a building.

Health monitoring is being studied to determine the position and degree of damage to a structure after an external force such as seismic motion is applied to the structure and to diagnose the soundness of the structure. In the health monitoring of a building, for example, whether or not the building is damaged, the position of the damage, and the degree of damage are evaluated based on the information of the sensor provided in the structure. It is not realistic to install a large number of sensors in the entire building for monitoring. Therefore, there is a structural health monitoring system that estimates the seismic response of the entire building at the time of an earthquake based on a limited number of sensor information.

Seismic shaking is detected by a seismograph with an accelerometer. Seismometers usually measure the weak accelerations that occur in buildings based on the detections of accelerometers. In the case of an earthquake, the building where the seismograph is installed will be monitored for shaking during the earthquake. By analyzing the waveform information of the acceleration recorded by the seismograph, it is possible to judge the soundness of the building structure and indoor damage (damage, safety).

Such a system for determining the soundness of a building during an earthquake has already been proposed in Patent Documents 1 and 2 and Non-Patent Document 1 as a "structural health monitoring system having a learning-type response estimation function" (see FIG. 4). The learning-type response estimation function updates the parameters of a building model that models a building based on Bayesian update, for example. Bayesian update is a method of updating the parameters of the building model every time information such as the detected value is obtained to improve the accuracy of the estimated value.

When constructing this system, a building model that models the dynamic response of the building is set as initial information. This building model is set based on the parameters of the mass point system response analysis model used for seismic response analysis. Generally, in a building model, the mass and stiffness distributions are used as parameters and are set by a high degree of expert design (see Figure 5).

A method for easily and automatically creating a mass and rigidity distribution by generalizing the initial information of such a building model has been proposed (see, for example, Patent Document 3). This method does not require a special dynamic analysis model, and an initial model can be easily and automatically created from the information on the number of building floors and the structure type (see Fig. 6). In the system shown above, the building model is given as initial information based on the measured values of the seismic response that actually occurred, and the response in all layers is estimated.

Japanese Unexamined Patent Publication No. 2013-195354 Japanese Unexamined Patent Publication No. 2015-001483 Japanese Unexamined Patent Publication No. 2017-125742

"Estimation of seismic response and probabilistic damage assessment using Bayesian update of building model" Tomoo Saito Architectural Institute of Japan Structural Papers 78 (683), 61-70, 2013

When the building model used for this response estimation is applied to an actual building as initial information, an error may occur compared to the response of the actual building, and the response estimation with high accuracy may not be performed. Therefore, it is necessary to set the parameters of the building model so that the estimated value by the building model is as close as possible to the response of the actual building. The learning type response estimation function (Bayesian update) also automatically corrects this building model, but since the model range that can be corrected is set and the convergence calculation is performed, it is corrected if the deviation of the parameter settings of the initial model is large. There is a problem that it cannot be done. Therefore, it is desirable that the building model to be set is as close to reality as possible.

It is an object of the present invention to provide a soundness evaluation system and a soundness evaluation method capable of improving the accuracy of an estimate indicating a building response based on a building model for estimating a building response.

In order to achieve the above object, the present invention is a soundness evaluation system that evaluates the soundness of the building based on a building model for estimating the response of the building, and is a plurality of soundness evaluation systems that detect the acceleration of the building. The primary natural period of the building is calculated based on the sensor and the detection values of the plurality of sensors, and the parameters related to the distribution of mass and rigidity included in the building model are updated based on the primary natural period. It is a soundness evaluation system characterized by having a calculation unit.

According to the present invention, in the calculation of seismic response estimation, the primary natural period of the building model can be calculated using the daily detection values of the sensor, and the calculated primary natural period can be used to calculate the initial natural period of the building model. It is possible to correct the initial stiffness distribution of the building model and improve the accuracy of the response estimation of the building model.

Further, the calculation unit of the present invention may be configured to calculate the damping constant of the building based on the calculated primary natural period and update the building model based on the damping constant.

According to the present invention, it is possible to modify the parameters of the initially set building model and update the building model by calculating the damping constant of the building using the first-order natural period calculated based on the detected value. can.

Further, the calculation unit of the present invention is configured to calculate the fine movement waveform of the building using the daily detection value and calculate the primary natural period based on the spectral analysis of the fine movement waveform. You may.

According to the present invention, the primary eigenperiod of a building can be calculated by spectrally analyzing the tremor waveform based on the detection value of the daily microvibration occurring in the building, and the calculated primary eigenperiod is used for the initial stage. The building model parameters set to can be modified to update the building model.

Further, the arithmetic unit of the present invention calculates a power spectrum from the fine movement waveform in the upper part of the building, cuts out a frequency range for one cycle near the peak of the lowest spectrum of the power spectrum, and obtains an autocorrelation function. It is configured to calculate the first-order natural period and calculate the attenuation constant so that the decay waveform calculated based on the autocorrelation function is non-linearly curve-fitted using the curve-fit method. You may.

According to the present invention, one peak including the peak of the lowest frequency spectrum is cut out from the frequency components obtained by spectral analysis of the microtremor waveform based on the detection value of the daily microvibration occurring in the building. The autocorrelation function can be calculated by regarding it as a period. The attenuation constant can be calculated by calculating the first-order natural period so that the error is minimized from the autocorrelation function calculated by using the curve fit method.

Further, the arithmetic unit of the present invention calculates the response waveform of the building at the time of an earthquake using the detected value detected at the time of an earthquake, and determines the primary natural period based on the mode analysis of the waveform at the time of the earthquake. It may be configured to calculate.

According to the present invention, the response waveform is calculated using the detection value detected at the time of an earthquake in which large vibration is observed, and the primary natural period of the building close to reality can be calculated by mode-analyzing the response waveform. can.

The arithmetic unit calculates the response waveform of the building at the time of an earthquake using the detected value detected at the time of an earthquake, and calculates the primary natural period based on the system identification method of the waveform at the time of the earthquake. It may be configured.

According to the present invention, by using a system identification method for estimating building model parameters from input / output data, it is possible to calculate the primary natural period of a building that is close to reality.

Further, the arithmetic unit of the present invention calculates a transfer function showing the relationship between the acceleration input to the building and the response acceleration generated in the building at the time of an earthquake by using the detected value detected at the time of an earthquake. It may be configured to calculate the first-order natural period based on the transfer function.

According to the present invention, the first-order natural period of a building close to reality can be calculated by calculating the transfer function using the detected value detected at the time of an earthquake.

It is a soundness evaluation method for evaluating the soundness of the building based on a building model for estimating the response of the building, and is a step of detecting the acceleration of the building by a plurality of sensors and a detection value of the plurality of sensors. It is characterized by including a step of calculating the primary natural period of the building based on the above, and a step of updating parameters related to the distribution of mass and rigidity included in the building model based on the primary natural period. This is a soundness evaluation method.

According to the present invention, in the calculation of seismic response estimation, the primary natural period of the building model can be calculated using the daily detection values of the sensor, and the calculated primary natural period can be used to calculate the initial natural period of the building model. It is possible to correct the initial stiffness distribution of the building model and improve the accuracy of the response estimation of the building model.

According to the present invention, it is possible to improve the accuracy of the estimated value indicating the response of the building based on the building model for estimating the response of the building.

It is a figure which shows the structure of the soundness evaluation system which concerns on embodiment of this invention. It is a figure which shows the method of calculating the primary natural period of a building based on the detected value. It is a figure which shows the other method of calculating the primary natural period of a building based on the detected value. It is a figure which shows the method of the full-thickness response estimation by a structural health monitoring system. It is a figure which shows the distribution of mass and rigidity of a building model, and the waveform example of the response estimation of all layers. It is a figure which shows the initial model information which is automatically created from the information of a building floor number and a structure type.

Hereinafter, embodiments of the soundness evaluation system and the soundness evaluation method according to the present invention will be described with reference to the drawings. The sanity assessment system corrects the parameters of the mass and stiffness distribution of the building model set as initial information so that the estimates of the building model used to estimate the response of the building are close to the actual response. Is.

As shown in FIG. 1, the soundness evaluation system 1 includes a detection unit 2 provided in the building, an evaluation device for evaluating the soundness of the building B based on the detection value detected by the detection unit 2, and an evaluation device. To prepare for. Building B is, for example, a high-rise building having a plurality of floors. The detection unit 2 includes, for example, a plurality of sensors S1, S2, ... Sn (n is a natural number). Hereinafter, a plurality of sensors are appropriately referred to as sensor Sp (p is a natural number from 1 to n).

The sensor Sp is, for example, an acceleration sensor. The plurality of sensors Sp detect the acceleration of the building B. The sensor Sp is not attached to all floors of the building B, but is attached to a plurality of predetermined floors. The output value of the sensor Sp is output to the evaluation device 10.

The evaluation device 10 is an information processing terminal that evaluates the soundness of the building B based on the building model for estimating the response of the building. The evaluation device 10 is realized by, for example, a personal computer, a tablet terminal, a smartphone, or the like. The evaluation device 10 is electrically connected to the detection unit 2. The evaluation device 10 may be connected to the detection unit 2 through a network (not shown). The evaluation device 10 includes, for example, an acquisition unit 12 that acquires detection value data from the detection unit 2, an operation unit 14 that performs an operation related to the soundness of the building based on the detected value, and a storage unit 16 that stores the data related to the operation. And.

The acquisition unit 12 is a communication interface connected to a local network or a public network. The storage unit 16 has a storage medium such as a hard disk drive (HDD) and a flash memory. The storage unit 16 is not necessarily built in or externally connected to the evaluation device 10, and may be provided in a server (not shown) that provides data via a network.

The calculation unit 14 performs calculations related to the evaluation of the soundness of the building B, for example, based on the building model (response estimation model) showing the dynamic response of the building B. The calculation unit 14 performs a calculation related to learning-type response estimation based on the detection values of the plurality of sensors Sp, and corrects the parameters of the building model initially set based on the calculation result.

Specifically, in the processing of the arithmetic unit 14, first, the mass matrix M, the attenuation coefficient matrix C, and the rigidity matrix K of the building model are given, and the general eigenvalue problem shown in the equation (1) is solved to obtain the j-th order eigenvalue. Obtain the frequency w _j and the stimulus function φ _j.

Here, in the conventional structural health monitoring system having a learning type response estimation function, a function Δk (θ) for modifying the stiffness distribution k is introduced. As a result, the stiffness distribution is corrected to k'= k + Δk (θ), and the stiffness matrix K is corrected to K'(θ) correspondingly. At this time, the model parameters are represented by random variables as shown in the equation (2) ( _np is the number of parameters).

On the other hand, sensor placement story building response absolute acceleration y _p of the (observation layer) _(theta) is represented by the formula (3), the probability model of the building response absolute acceleration can be expressed by Equation (4).

n _s is the number of sensors installed in the building (excluding those for ground motion measurement), and y _p (^ (hat) above) (θ) is defined by M, C, K'(θ). It is the response absolute acceleration of the sensor installation floor at each time when the observed ground motion u is input to the modified design model, and it is shown that the value is independently normally distributed with the same _{variance σ y} ^{2 as the expected value.} There is.

Then, when the observation data D represented by the equation (5) is obtained at the time of an earthquake, the posterior distribution of θ is obtained by the equation (6) by Bayes' theorem.

Here, p (θ) is a prior distribution, which is a uniform distribution independent of each other and having an average of 0 as in Eq. (7). Further, p (D | θ) is a likelihood function and is obtained by the equation (8).

_{Let θ MAP} (^ (hat) above), which maximizes the posterior distribution p (θ | D) obtained in this way, be θ MAP (^ (hat) above), and then the stiffness matrix K'(θ _{MAP ) modified by θ MAP (^).} From ^)), the eigenvalue problem similar to Eq. (1) is solved, and the corresponding stimulus function φ _j'is obtained.

This means that the design model, which is prior information, has been updated to be closer to reality based on actual observation data. In addition, continuous learning may be performed by using this updated posterior distribution p (θ | D) as a prior distribution for the next earthquake.

Then, assuming that the mode that has a dominant influence on the response of the building is 1 to _nm order, the absolute response acceleration of the sensor installation floor can be approximated by the equation (9).

D is D = [1 ... 1] T ∈ Rns, Ф _p is a matrix obtained by extracting the row corresponding to the sensor installation floor from Ф = [φ1'... φnm'], and q is. , Q = [q1 (t) ... qnm (t)] It is a mode response relative acceleration vector of order 1 _{to nm} ^{represented by T.} Then, the observed response acceleration waveform y _p estimate of modal response relative acceleration from _(~ (tilde) above) q (^) is obtained by Equation (10).

Ф _p + is the generalized inverse matrix of _{Ф p.} As a result, the response y ∈ R ^{nf of} all layers (nf is the number of building layers) can be estimated by Eq. (11).

It should be noted that D'= [1 ... 1] T ∈ R ^nf . Further, by using the generalized inverse matrix in the equation (10), it is possible to arbitrarily set the number of the main modes used for the estimation.

In the present embodiment, the soundness confirmation method based on the Bayesian update of the building is used, and the parameters set as the initial information of the building model are corrected based on the following processing. In order to simplify the correction of the building model, the calculation unit 14 performs the processing in each process shown below. The calculation unit 14 calculates the first-order natural period of the building B based on the detected value of the actual response. The calculation unit 14 applies the calculated primary natural period to the primary natural period in the initial building model, and updates the parameters related to the distribution of mass and rigidity included in the building model based on the primary natural period. The calculation unit 14 corrects and updates the building model based on the updated parameters. By updating the building model, it is possible to bring the estimated value of the dynamic response of the building B closer to the actual response.

As shown in FIG. 2, the calculation unit 14 calculates the fine movement waveform of the building B using the daily detection values acquired from the plurality of sensors Sp. The calculation unit 14 calculates the first-order natural period of the building B based on the spectral analysis of the fine movement waveform. The calculation unit 14 calculates the power spectrum from, for example, the fine movement waveform in the upper part of the building B. The power spectrum represents the distribution of power on the frequency, and when the inverse discrete-time Fourier transform is performed, the power spectrum is converted into an autocorrelation function.

For example, the calculation unit 14 cuts out a frequency range for one cycle of a mountain near the peak of the lowest spectrum of the power spectrum, regards the cut out frequency range as a first-order natural period, and calculates an autocorrelation function. The calculation unit 14 calculates the first-order natural period by using the curve fit method. The curve fit method is a method of calculating parameters that match the detected values well. By the curve fitting method, function fitting is performed on the frequency axis or the time axis using the least squares method, and the optimum function is estimated to obtain an approximate curve.

The calculation unit 14 calculates the first-order natural period so that the decay waveform calculated based on the calculated autocorrelation function is non-linearly curve-fitted, and also calculates the attenuation constant. By using the first-order natural period obtained here for correcting the rigidity distribution of the initial building model used for response estimation, highly accurate response estimation can be performed.

The detection values detected from the sensor Sp include those detected at the time of an earthquake, in addition to the daily micro-vibrations. When a detected value is detected at the time of an earthquake, the primary natural period of the building B can be calculated using this detected value.

The calculation unit 14 calculates the response waveform of the building B at the time of the earthquake using the detected value detected at the time of the earthquake in the same manner as described above. The calculation unit 14 calculates the first-order natural period based on the spectral analysis of the response waveform at the time of an earthquake. The calculation unit 14 calculates the power spectrum from, for example, the response waveform at the time of an earthquake in the upper part of the building B. The calculation unit 14 cuts out a frequency range for one cycle near the peak of the lowest spectrum of the power spectrum and uses this as the first-order natural cycle to calculate the autocorrelation function.

The calculation unit 14 calculates the first-order natural period that minimizes the error so that the attenuation waveform calculated based on the autocorrelation function is non-linearly curve-fitted by using the curve fit method, and also calculates the attenuation constant. The arithmetic unit updates the parameters related to the mass and stiffness distribution included in the building model based on the calculated primary natural period and damping constant.

The calculation unit 14 may calculate the response waveform of the building B at the time of the earthquake using the detected value detected at the time of the earthquake, and calculate the primary natural period based on the mode analysis of the waveform at the time of the earthquake. The calculation unit 14 calculates the waveform of the building B at the time of the earthquake using the detected value detected at the time of the earthquake. The calculation unit 14 calculates the first-order natural period based on a system identification method such as an ARX (Auto Regressive with eXogenous) model of the waveform at the time of an earthquake. The system identification method estimates the parameters of the building model from the input / output data. The calculation unit 14 calculates the damping constant of the building B based on the calculated primary natural period, and updates the building model based on the damping constant.

As shown in FIG. 3, the arithmetic unit 14 may calculate the primary natural period by calculating the transfer function using the detected value detected at the time of the earthquake. The transfer function shows the relationship between the acceleration input to the building B at the time of an earthquake and the response acceleration generated in the building B. The calculation unit 14 calculates the transfer function and calculates the first-order natural period based on the transfer function. The calculation unit 14 calculates the damping constant of the building B based on the primary natural period calculated by each of the above processes, and updates the building model based on the damping constant. By identifying the first-order peak shown in the calculated transfer function, it can be used for correction of the initial building model as in the above processing.

The above-mentioned arithmetic unit 14 is realized by executing a program (software) by a processor such as a CPU (Central Processing Unit) or a GPU (Graphics Processing Unit). Part or all of these functional parts may be realized by hardware such as LSI (Large Scale Integration), ASIC (Application Specific Integrated Circuit), FPGA (Field-Programmable Gate Array), or software. It may be realized by the cooperation of hardware. The program may be stored in advance in a storage device such as an HDD (Hard Disk Drive) or a flash memory, or may be stored in a removable storage medium such as a DVD or a CD-ROM, and the storage medium is stored in the drive device. It may be installed in the storage device by being attached.

As described above, according to the soundness evaluation system 1, in the calculation of the seismic response estimation, the primary natural period of the building model can be calculated by using the daily detection value of the sensor Sp, and the calculated primary period can be calculated. The natural period can be used to correct the initial stiffness distribution of the early building model and improve the accuracy of the response estimation of the building model.

Furthermore, according to the soundness evaluation system 1, the response estimation of the updated building model is performed by giving the damping constant calculated as information close to reality to the initial building model in the learning type response estimation function (Bayes update). The accuracy can be improved. According to the soundness evaluation system 1, by correcting the initial building model, the model assumed in the design can be made closer to reality, and as a result, the accuracy of response estimation can be improved.

Although one embodiment of the present invention has been described above, the present invention is not limited to the above-mentioned one embodiment and can be appropriately modified without departing from the spirit of the present invention. For example, the sanity assessment system may be applied not only to buildings but also to response estimation of other structures.

1 Soundness evaluation system 2 Detection unit 10 Evaluation device 12 Acquisition unit 14 Calculation unit 16 Storage unit B Building S1-Sn Sensor

Claims (10)

It is a soundness evaluation system that evaluates the soundness of the building based on the building model for estimating the response of the building.
A plurality of sensors that detect the acceleration of the building,
With a calculation unit that calculates the primary natural period of the building based on the detection values of the plurality of sensors and updates the parameters related to the distribution of mass and rigidity included in the building model based on the primary natural period. , Featuring,
Health assessment system. The calculation unit calculates the damping constant of the building based on the calculated primary natural period, and updates the building model based on the damping constant.
The soundness evaluation system according to claim 1. The calculation unit calculates the fine movement waveform of the building using the daily detection value, and calculates the primary natural period based on the spectral analysis of the fine movement waveform.
The soundness evaluation system according to claim 2. The calculation unit calculates a power spectrum from the fine movement waveform in the upper part of the building, cuts out a frequency range for one cycle near the peak of the lowest spectrum of the power spectrum, calculates an autocorrelation function, and calculates an autocorrelation function. Using the method, the first-order natural period is calculated so that the decay waveform calculated based on the autocorrelation function is non-linearly curve-fitted, and the decay constant is calculated.
The soundness evaluation system according to claim 3. The calculation unit calculates the response waveform of the building at the time of an earthquake using the detected value detected at the time of an earthquake, and calculates the primary natural period based on the spectral analysis of the response waveform at the time of the earthquake.
The soundness evaluation system according to claim 2. The calculation unit calculates the power spectrum from the response waveform at the time of the earthquake in the upper part of the building, cuts out the frequency range for one cycle near the peak of the lowest spectrum of the power spectrum, and calculates the autocorrelation function. , The first-order natural period is calculated so that the decay waveform calculated based on the autocorrelation function is non-linearly curve-fitted using the curve fit method, and the decay constant is calculated.
The soundness assessment system according to claim 5. The calculation unit calculates the response waveform of the building at the time of an earthquake using the detected value detected at the time of an earthquake, and calculates the primary natural period based on the mode analysis of the waveform at the time of the earthquake.
The soundness evaluation system according to claim 2. The calculation unit calculates the response waveform of the building at the time of an earthquake using the detected value detected at the time of an earthquake, and calculates the primary natural period based on the system identification method of the waveform at the time of the earthquake.
The soundness assessment system according to claim 7. The arithmetic unit calculates a transfer function indicating the relationship between the acceleration input to the building and the response acceleration generated in the building at the time of an earthquake using the detected value detected at the time of an earthquake, and is based on the transfer function. Calculate the primary natural period,
The soundness assessment system according to claim 7 or 8. It is a soundness evaluation method for evaluating the soundness of the building based on a building model for estimating the response of the building.
The process of detecting the acceleration of the building with a plurality of sensors,
A step of calculating the primary natural period of the building based on the detection values of the plurality of sensors, and
It is characterized by comprising a step of updating parameters relating to mass and rigidity distribution included in the building model based on the primary natural period.
Soundness assessment method.

Images