WO2022059720A1 - Structure diagnosis system, structure diagnosis method, and structure diagnosis program - Google Patents

Abstract

This structure diagnosis system includes a diagnosis unit which: for a feature amount indicating the state of a structure and generated from a time series of information relating to the structure, calculates an evaluation index, which is the ratio between a first periphery likelihood representing the feature amount probability distribution in a case where the structure is assumed to be in a sound state, and a second periphery likelihood representing the feature amount probability distribution in a case where the structure is assumed to not be in a sound state; and diagnoses the state of the structure on the basis of the evaluation index.

Description

Structure diagnostic system, structure diagnostic method, and structure diagnostic program

The present invention relates to a structure diagnosis system, a structure diagnosis method, and a structure diagnosis program.

Conventionally, based on the acceleration of structures such as bridges, the natural frequency of vibration, damping coefficient, feature quantities (vibration characteristics) such as mode shape are estimated, and the structure is based on the estimated changes in vibration characteristics. There is a technique for diagnosing abnormalities in.

For example, Patent Document 1 discloses the following technology. That is, independent component analysis (ICA) is performed on the sensor signals indicating vibrations acquired from accelerometers installed at multiple locations on the bridge, and spectral analysis is performed on the independent vibration components obtained by independent component analysis on the bridge. Find the natural frequency of. Then, based on the comparison between the natural frequency of the bridge in a healthy state acquired in advance and the natural frequency of the current bridge, the abnormality diagnosis of the entire bridge and the identification of the abnormality location are performed.

Japanese Unexamined Patent Publication No. 2008-25571

However, the sensitivity of change to damage varies depending on the location where vibration measurement is performed and the vibration mode. Furthermore, in the preceding technology, the estimation varies depending on the environmental conditions (external force, temperature, etc.) at the time of vibration measurement. However, it is difficult even for an expert to properly preset the installation location of the accelerometer and the vibration mode. For this reason, in, for example, in bridge abnormality diagnosis based on the natural frequency, if the vibration mode is set incorrectly, for example, changes in vibration characteristics due to bridge abnormalities and accidental changes in vibration characteristics due to other than bridge abnormalities are detected. There is a problem that quantitative abnormality diagnosis cannot be performed with high accuracy, such as being unable to discriminate.

It is an object of the present invention to be able to perform quantitative abnormality diagnosis of a structure with high accuracy in consideration of the above points.

In order to solve the above problems, the present invention has a structure diagnostic system of the present invention in which the structure is in a sound state with respect to a feature quantity indicating the state of the structure generated from a time series of information about the structure. A first marginal likelihood representing the probability distribution of the feature amount in the case of being present, and a second marginal likelihood representing the probability distribution of the feature amount in the case where the structure is assumed to be in the unhealthy state. It is characterized by having a diagnostic unit that calculates an evaluation index, which is a ratio of, and diagnoses the state of the structure based on the evaluation index.

According to the present invention, it is possible to perform quantitative abnormality diagnosis of a structure with high accuracy.

The figure which shows the schematic structure of the diagnostic system of Embodiment 1. FIG. The block diagram which shows the structure of the sensor node of Embodiment 1. FIG. The block diagram which shows the structure of the diagnostic apparatus of Embodiment 1. FIG. The sequence diagram which shows the reference time processing in the diagnostic system of Embodiment 1. FIG. The sequence diagram which shows the process at the time of diagnosis in the diagnosis system of Embodiment 1. FIG. The explanatory view of the process at the time of diagnosis in the diagnosis system of Embodiment 1. FIG. The figure which compares and shows the data amount of the coefficient matrix and the principal component matrix in Embodiment 1 and the prior art by the number of sensors. The flowchart which shows the diagnostic process in the diagnostic system of Embodiment 2. The figure for demonstrating the calculation method of the threshold value of abnormality determination of Embodiment 2. The figure for demonstrating another example of the abnormality determination method using a plurality of thresholds of Embodiment 2. FIG. The figure for demonstrating the experimental result of abnormality determination performed using the diagnostic system of Embodiment 2. The figure which shows the execution condition of the experiment performed using the diagnostic system of Embodiment 2. The figure which shows the experimental result which performed using the diagnostic system of Embodiment 2. The figure which shows the change of the natural frequency of a bridge which differs according to the vibration mode measured under the same conditions as the experiment which performed using the diagnostic system of Embodiment 2. FIG. The figure which shows the time course of the outside air temperature and the value of Bayes factor in a certain period of a certain girder of a bridge. The figure which shows the time-dependent change of the value of the deflection and the Bayes factor in a certain period of a bridge. The figure which shows the comparison of the time change of the value of the Bayes factor between two spans of a bridge. An explanatory diagram of sequential update of a coefficient matrix in the diagnostic system of the third embodiment. The figure which shows the convergence of the probability distribution of a coefficient matrix by sequential update of a coefficient matrix in the diagnostic system of Embodiment 3. The block diagram which shows the structure of the sensor node terminal used as the sensor node of an embodiment.

Hereinafter, embodiments of the present invention will be described with reference to the drawings. The following embodiments are examples for explaining the present invention in a necessary and sufficient manner, and are appropriately omitted and simplified. All of the configurations and processes described in the following embodiments are not essential in the practice of the present invention and may be omitted as appropriate. In addition to the following embodiments, the present invention can be carried out in various other embodiments capable of achieving the object of the present invention. Further, as long as it is consistent within the scope of the technical idea of the present invention, the embodiment of the present invention also includes a form in which some or all of the respective embodiments and modifications are combined.

In the following embodiments, the same reference numerals may be given to the later configurations and processes similar to the existing configurations and processes, the description may be omitted, and only the differences may be described.

In the following embodiment, a bridge will be described as an example of a structure to be diagnosed for the presence or absence of abnormalities such as damage or deterioration such as cracks. However, the present invention is applied not only to bridges but also to all civil engineering structures or building structures that support social infrastructure such as tunnels, road structures, river structures, harbor structures, water and sewage, and buildings, and whether or not there are any abnormalities thereof. Can be diagnosed.

In the description of the following embodiment, the expression in which the second symbol is continuously described in the first symbol, for example, “A ^”, is the expression in which the second symbol is described immediately above the first symbol. Is the same as.

<Mathematical background of the present invention>
First, the mathematical background of the present invention will be described prior to the description of the embodiments. In the present invention, the system matrix of the state equation in the state space model of the structure is represented by the matrix of the regression coefficients of the autoregressive model that expresses the observation signal at a certain time as a time series linear connection of the observation signals before a certain time. It uses what can be approximated. In the present invention, paying attention to the fact that the matrix of regression coefficients contains various information about the state of the structure included in the system matrix, the matrix of regression coefficients is used to represent the state of the structure, and the matrix of the regression coefficients is represented. The state of the structure is evaluated by analyzing.

In the following, the sensor information acquired by the sensor will be described as acceleration, but it is not limited to acceleration and may be another physical quantity.

Generally, the equation of motion of a structure is expressed as the following equation (1). It is assumed that n sensors (n is a natural number of 1 or more) are installed in the structure at each position, and the time series of the nth-order observation vector is acquired with the installation position as the observation point.

The equation of motion of equation (1) above can be converted as an equation of state as in equations (2-1) and (2-2) below.

Here, y in the above equation (2-2) is an observation vector, z in the above equation (2-3) is a state variable vector, and As in the above equation ( _2-5 ) is a system in a state space (diagnosis target). Represents each system matrix representing the state of the structure). C in the above equation (2-2) is a matrix that associates the observation vector y with the state of the system, and is an identity matrix when the measurement information at the observation point is regarded as the state of the system as it is.

By the way, it is known that the equations of state of the above equations (2-1) and (2-2) can be expressed by an autoregressive model which is a linear combination of the discretized observation signals as shown in the following equation (3). Has been done.

In the above equation (3), k is a time index, y (k) is a sensor information vector at time k, p is the order of the autoregressive model, A _i is a matrix of regression coefficients, and e (k) is autoregressive at time k. This is the error term of the model. The matrix Ai is a regression coefficient multiplied by each of the time series of the linearly combined sensor information vector _y (ki) in the autoregressive model.

Each element of the matrix _Ai , which is the regression coefficient in the self-return model of the above equation (3), is associated with the physical quantity of the structure included in the system _matrix As of the above equation (2-1), and a sensor is installed. It contains at least one of the information of displacement, velocity, acceleration, mass matrix m, damping coefficient matrix c, and rigidity matrix k at the observation point. Therefore, by using the matrix A _i , the state of the structure can be expressed so as to include various information regarding vibration. That is, by capturing the change in the matrix _Ai , it becomes possible to capture the change in the state of the structure.

However, when there is one sensor (n ₌ 1), the matrix Ai in the autoregressive model is a scalar.

Here, the matrix A _i is an nth-order square matrix as shown in the following equation (4). In the following equation (4), n is the number of sensors installed at each position of the structure.

Then, when the p-th order autoregressive model shown in the above equation (3) generated from the time series of the sensor information vector y (k) measured by n sensors is expressed by a determinant, it is as shown in the following equation (5). Become.

Here, in the above equation (5), Y _f on the left side is the predicted state, Y _p in the first term on the right side is the past acceleration of the pth order, and E in the second term on the right side is the uncertainty of state observation of the structure. Represents the error caused by. Further, the coefficient matrix A of the first term on the right side in the above equation (5) is an n × (n × p) matrix in which the matrices A _i for i from 1 to p are combined, and is as shown in the following equation (6). It is represented by.

The coefficient matrix A _inherits the information of the system matrix As. For example, the poles z _k , z _k of the system at time k expressed by using the natural frequency ω _k of the system at time k and the attenuation coefficient h _k . ^* Can be obtained from the above equation (5) as in the following equation (7).

The posterior distribution of the coefficient matrix A is converted into the posterior distribution of the poles, and the natural frequency ω _k and the damping coefficient h _k obtained as in the above equation (7) can also be used for the diagnosis of the structure. The natural frequency ω _k and the damping coefficient h _k obtained in this way have an advantage that the vibration characteristics such as the natural frequency and the damping coefficient can be identified in consideration of the variation as compared with the prior art. Further, by converting the posterior distribution of the coefficient matrix A into the posterior distribution of the poles and obtaining the vibration characteristics from the poles having a small variance, it is possible to automate the selection of the vibration characteristics that are physically meaningful. Further, the selection of the model order p can be automated based on the BIC (Bayesian information criterion).

By transmitting the coefficient matrix A obtained as described above to the diagnostic device as a feature quantity representing the state of the structure, the time series having the vibration characteristics of the structure is reproduced in the diagnostic device, and the state of the structure is reproduced. Diagnosis is possible.

However, the coefficient matrix A contains a component of inferior information that is not related to the diagnosis result. By removing the component of this inferior information and transmitting the reconstructed matrix A ^ to the diagnostic apparatus, the amount of data transmitted / received can be reduced. Since this matrix A ^ is obtained by the principal component analysis of the coefficient matrix A, it is called a principal component matrix.

For example, the coefficient matrix A estimated at the reference time (for example, sound time) of the structure is decomposed into singular values as shown in the following equation (8).

The matrix U and the matrix P ^T (where ^PT is the transposed matrix of the matrix P) in the above equation (8) are orthogonal matrices, and the matrix Λ is an eigenvalue of the coefficient matrix A. Then, on the right side of the above equation (8), as shown in the following equation (9), the matrix U is an n × N matrix U ₁ (N <n) and an n × (np−N) matrix U ₂ by principal component analysis. It is decomposed into.

The matrix U ₁ at the reference time (for example, at the healthy time) is a matrix that projects the coefficient matrix A in the probability space onto the principal component space. As shown in the following equation (10), by applying the transposed matrix U ₁ ^T of the matrix U ₁ to the coefficient matrix A estimated at each time after the reference time, the component of the inferior information is removed. The principal component matrix A ^, which is the N × np matrix, is generated.

As described above, the coefficient matrix A that _inherits the information of the system matrix As is estimated, the principal component matrix A ^ is generated from the coefficient matrix A, and the change in the principal component matrix A ^ is captured to detect the abnormality of the structure. can do. In the present invention, as shown in the following equation (11), the probability distribution of the elements of the principal component matrix A ^ is estimated by Bayesian estimation. Then, by using the estimated probability distribution for the state diagnosis of the structure, it is possible to detect the abnormality of the structure while considering the uncertainty included in the above equation (5).

In the present invention, since the probability distribution having the mean value and the variance is used as an evaluation index for evaluating the state of the structure, it is possible to capture a minute change in the mean value of the evaluation index at the time of damage to the sound state of the structure. Instead, the variance can show the probability of the evaluation index.

It should be noted that the state diagnosis of the structure can be performed by using the coefficient matrix A as the feature quantity instead of the principal component matrix A ^, but in that case, in the above equation (11), A ^ is replaced with A. It becomes an expression.

<Embodiment 1>
Hereinafter, the first embodiment will be described with reference to FIGS. 1 to 7.

(Configuration of Diagnostic System S of Embodiment 1)
FIG. 1 is a diagram showing a schematic configuration of the diagnostic system S of the first embodiment. The diagnostic system S has a sensor node 1 and a diagnostic device 2. The sensor node 1 generates a feature amount indicating the state of the bridge 4 based on the sensor information acquired from the sensor 3 installed on the bridge 4 via wireless communication (or wired communication). The diagnostic device 2 acquires a feature amount indicating the state of the bridge 4 generated in the sensor node 1 via wireless communication (or wired communication), and diagnoses the presence or absence of an abnormality in the bridge 4 using this feature amount.

The sensor 3 is, for example, an acceleration sensor, and is installed at each part of the bridge 4. In the present embodiment, the sensor 3 is a plurality of sensors installed at a plurality of parts of the bridge 4, but the sensor 3 is not limited to this and may be one sensor installed at one part. Further, in the present embodiment, the sensor 3 is assumed to be an acceleration sensor that detects acceleration, but the sensor 3 is not limited to this, and may be a sensor that can measure other physical quantities (for example, strain, displacement, velocity, etc.).

In this embodiment, for the sake of clarity of explanation, as shown in FIG. 1, a configuration in which the result of processing the sensor information acquired from the sensor 3 by the sensor node 1 is transmitted to the diagnostic device 2 will be described as an example. ing. However, the present invention is not limited to this, and a distributed coordination system may be configured in which the sensor 3 forms a sensor network and performs distributed coordination processing to realize a function equivalent to that of the sensor node 1. When there is one sensor 3, one sensor 3 realizes the same function as the sensor node 1.

(Configuration of Sensor Node 1 of Embodiment 1)
FIG. 2 is a block diagram showing the configuration of the sensor node 1 of the first embodiment. The sensor node 1 performs edge processing for generating a coefficient matrix A (or a principal component matrix A ^) as a feature amount representing the state of the bridge 4 from the sensor information acquired from the sensor 3. The sensor node 1 has a processor 11, a memory 12, a storage 13, and a communication I / F (Inter / Face) unit 14.

The processor 11 is an arithmetic processing unit such as a CPU (Central Processing Unit), a PLD (Programmable Logic Device), or a microprocessor. The memory 12 is the main storage device. The storage 13 is an auxiliary storage device. The communication I / F unit 14 is a communication interface for the sensor node 1 to perform wireless communication (or wired communication) with the sensor 3 and the diagnostic device 2.

The processor 11 executes a program in cooperation with the memory 12, and is a functional unit of the sensor information acquisition unit 111, the autoregressive model generation unit 112, the feature amount generation unit 113, and the feature amount transmission unit 114. To realize.

The sensor information acquisition unit 111 acquires observation signals observed at each installation position of the bridge 4 by the sensor 3 in chronological order via the communication I / F unit 14, and stores them in the storage 13 as sensor information 131. In the present embodiment, the sensor information 131 acquired by the sensor information acquisition unit 111 is assumed to be continuous time time series information, but may be discrete time time series information. Further, the sensor information 131 may be constantly acquired or may be acquired for a certain period of time triggered by an acquisition instruction.

The autoregressive model generation unit 112 discretizes the sensor information 131 stored in the storage 13 in time. Then, as shown in the above equation (3), the autoregressive model generation unit 112 converts the sensor information 131 at a certain time k into p time ki (i = 1, ... , P) Generates an autoregressive model expressed by a linear combination of the sensor information 131 in time series. The details of the processing of the autoregressive model generation unit 112 will be described later with reference to the flowchart.

The feature quantity generation unit 113 is the state of the bridge 4 at a certain time k from the coefficient matrix A (the above equation (6)) in which the regression coefficients of the linear combination of the autoregressive model generated by the autoregressive model generation unit 112 are combined. Generates a feature quantity that represents. The feature amount may be the coefficient matrix A itself or the principal component matrix A ^ (the above equation (10)) obtained by projecting the coefficient matrix A onto the principal component space based on the principal component analysis. As will be described later, the state of the surface and the inside of the bridge 4 can be expressed as a probability distribution based on such features. The details of the processing of the feature amount generation unit 113 will be described later with reference to the flowchart.

The feature amount transmission unit 114 transmits the feature amount generated by the feature amount generation unit 113 to the diagnostic device 2 via the communication I / F unit 14.

(Configuration of Diagnostic Device 2 of Embodiment 1)
FIG. 3 is a block diagram showing the configuration of the diagnostic device 2 of the first embodiment. The diagnostic device 2 includes a processor 21, a memory 22, a storage 23, a communication I / F unit 24, and an output unit 25. The processor 21 realizes each functional unit of the feature amount receiving unit 211 and the diagnostic unit 212 by executing the program in cooperation with the memory 22.

The processor 21 is an arithmetic processing unit such as a CPU, PLD, or microprocessor. The memory 22 is the main storage device. The storage 23 is an auxiliary storage device. The communication I / F unit 24 is a communication interface for the diagnostic device 2 to perform wireless communication (or wired communication) with the sensor node 1. The output unit 25 is a monitor, a display, or the like, and outputs various information.

The feature amount receiving unit 211 receives the feature amount 231 indicating the state of the bridge 4 at each time k from the sensor node 1 via the communication I / F unit 24. The feature amount receiving unit 211 stores the received feature amount 231 in the storage 23.

The diagnostic unit 212 obtains the probability distribution of the feature amount at each time k (the above equation (11)) from the feature amount 231 stored in the storage 23 using Bayesian estimation as an evaluation index. Then, the diagnostic unit 212 calculates the Mahalanobis distance MD of the probability distribution obtained as an evaluation index at each time k, as shown in the following equation (12).

In the above equation (12), the matrix X is the coefficient matrix A (or the principal component matrix A ^), and the matrix S is the covariance matrix of the matrix X. The diagnostic unit 212 obtains the Mahalanobis distance MD of the coefficient matrix A (or the principal component matrix A ^) of the bridge 4 at the reference time (for example, when it is healthy) as the reference evaluation index 232 in advance and stores it in the storage 23. Further, the diagnosis unit 212 calculates the Mahalanobis distance MD of the coefficient matrix A (or the principal component matrix A ^) of the bridge 4 at the time of diagnosis when the presence or absence of damage and the degree of damage are unknown.

Then, the diagnosis unit 212 compares the Mahalanobis distance MD at the time of diagnosis with the reference evaluation index 232, and when the statistical distance such as the difference or ratio between them is a certain value or more, the bridge 4 is damaged as compared with the reference time. Diagnose that the abnormality is occurring or progressing. The diagnosis unit 212 outputs the diagnosis result to an output unit 25 such as a display.

The diagnostic unit 212 uses the Z value and other statistical indexes instead of the Mahalanobis distance MD to determine the statistical distance between the respective statistical indexes at the reference time and the diagnosis as a threshold value, whereby the abnormality of the bridge 4 is determined. Or the progress of deterioration may be diagnosed.

(Reference time processing in the diagnostic system S of the first embodiment)
FIG. 4 is a sequence diagram showing reference time processing in the diagnostic system S of the first embodiment. In the reference time processing in the diagnostic system S, the coefficient matrix A and the principal component matrix A ^ are generated from the time series of the sensor information acquired from the sensor 3 at the reference time (for example, when the bridge 4 is sound), and the principal component matrix A ^ is generated. Based on this, the process of calculating the reference evaluation index 232 is performed.

First, in step S101, the sensor information acquisition unit 111 of the sensor node 1 acquires the sensor information 131 from the sensor 3 and stores it in the storage 13. Next, in step S102, the autoregressive model generation unit 112 is the pth order of the sensor information vector y (k) at a certain time k in which the sensor information 131 acquired in step S101 is discretized based on the above equation (3). Generate an autoregressive model of. The order p of the autoregressive model is appropriately determined, but may be automatically determined from the above equation (5) based on the BIC, for example, as described above.

Next, in step S103, the feature amount generation unit 113 is a coefficient matrix A in which a matrix A _i representing the regression coefficient of the p-th order autoregressive model of the sensor information vector y (k) is combined as in the above equation (6). To generate. Next, in step S104, the feature amount generation unit 113 mainly projects the coefficient matrix A in the probability space representing the state of the bridge 4 at the reference time onto the main component space based on the above equations (8) to (9). Generate the component space matrix U ₁ ^T.

Next, in step S106, as shown in the above equation (10), the feature amount generation unit 113 causes the main component space matrix U ₁ ^T generated in step S105 to act on the coefficient matrix A, and causes the feature amount matrix as the feature amount. Generate A ^. Next, in step S106, the feature amount transmission unit 114 transmits the feature amount generated in step S105 to the diagnostic apparatus 2.

Next, in step S201, the feature amount receiving unit 211 of the diagnostic device 2 receives the feature amount 231 from the sensor node 1 and stores it in the storage 23. Next, in step S202, the diagnostic unit 212 obtains the probability distribution p (A ^ | Σ, Y) of the feature amount 231 as in the above equation (11), and the probability distribution p (A ^ | Σ, Y). The Mahalanobis distance MD is calculated as the reference evaluation index 232. Next, in step S203, the diagnostic unit 212 stores the reference evaluation index 232 calculated in step S202 in the storage 23.

Not limited to the illustration in FIG. 4, in the reference time processing, the diagnostic device 2 may execute the step executed by the sensor node 1, or the sensor node 1 may execute the step executed by the diagnostic device 2. .. That is, it is not limited whether the execution subject of each step shown in FIG. 4 is the sensor node 1 or the diagnostic device 2.

For example, the diagnostic device 2 may perform the processing after any one of steps S102 to S105 instead of the sensor node 1. Further, for example, the process of step S202 may be performed by the sensor node 1 instead of the diagnostic device 2, and the calculated reference evaluation index is transmitted to the diagnostic device 2 and used in the diagnostic device 2 for the state diagnosis of the bridge 4. ..

(Processing at the time of diagnosis in the diagnostic system S of the first embodiment)
FIG. 5 is a sequence diagram showing diagnostic processing in the diagnostic system S of the first embodiment. The diagnostic processing of FIG. 5 is different from the reference processing of FIG. 4 only in that it handles the sensor information and the feature amount at the time of diagnosis, not at the reference time, and steps S111 to S116 are the steps S101 of FIG. It is the same as S106.

When the principal component space matrix U ₁ ^T is generated in step S104 of the reference time processing of FIG. 4, in the diagnostic processing of FIG. 5, the step of generating the principal component space matrix U ₁ ^T of step S114 is performed. It can be omitted. When step S114 is omitted, in step S115 following step S113, the feature amount generation unit 113 generated the principal component space matrix _U1 ^T generated in step S104 of the reference time processing of FIG. 4 in step S113. By acting on the coefficient matrix A, the principal component matrix A ^ is generated as a feature quantity. Next, in step S116, the feature amount transmission unit 114 transmits the feature amount generated in step S115 to the diagnostic apparatus 2.

Next, in step S21, the feature amount receiving unit 211 of the diagnostic device 2 receives the feature amount from the sensor node 1. Next, in step S212, the diagnostic unit 212 uses Bayesian inference to determine the probability distribution p (A ^ | Σ, Y) of the feature amount based on the feature amount received in step S211, as in the above equation (11). Is calculated. Then, the diagnosis unit 212 calculates the Mahalanobis distance MD at the time of diagnosis as an evaluation index based on the above formula (12), compares the evaluation index with the reference evaluation index 232, and when the difference or ratio between them is a certain value or more. , It is diagnosed that an abnormality such as damage has occurred or progressed in the bridge 4 as compared with the reference time.

As in FIG. 4, FIG. 5 is not limited to whether the execution subject of each step is the sensor node 1 or the diagnostic device 2. For example, the calculation of the evaluation index may be performed by the sensor node 1 instead of the diagnostic device 2. The evaluation index calculated in this way is transmitted to the diagnostic device 2 and used in the diagnostic device 2 for diagnosing the state of the bridge.

Although the principal component matrix A ^ is used as the feature amount in the processes shown in FIGS. 4 and 5, the coefficient matrix A may be used as the feature amount. In this case, the probability distribution p (A | Σ, Y) is calculated by replacing A ^ with A in the above equation (11).

FIG. 6 is an explanatory diagram of diagnostic processing in the diagnostic system S of the first embodiment. The illustrated portion 601 of FIG. 6 outlines the reference time processing (FIG. 4) in which the feature amount is extracted from the observation data representing the vibration of the bridge 4 at the reference time and the statistical index based on the feature amount is calculated as the reference evaluation index. Shows. Further, the illustrated portion 602 shows an outline of a diagnostic processing (FIG. 5) in which a feature amount is extracted from observation data representing vibration at the time of diagnosis of the bridge 4 and a statistical index based on the feature amount is calculated as a diagnostic evaluation index. ing.

Then, in the diagnosis processing (FIG. 5), the presence or absence of an abnormality in the bridge 4 and its progress are determined based on the comparison result between the reference evaluation index and the diagnosis evaluation index. That is, as shown in the illustrated portion 603 of FIG. 6, in the probability space, the reference evaluation index based on the probability distribution of the reference feature amount and the diagnosis evaluation index based on the probability distribution of the diagnosis feature amount are compared, and the bridge 4 Abnormality is diagnosed. In the comparison between the standard evaluation index and the diagnostic evaluation index, the situation of how much the diagnostic evaluation index deviates from the standard evaluation index is evaluated using a statistical index such as the Mahalanobis distance. Judgment of the presence or absence of abnormality of the bridge 4 so that it is judged that the state of the bridge 4 at the time of diagnosis has changed from the reference time when the statistical distance between the evaluation index at the time of diagnosis and the standard evaluation index is a certain value or more. It can be performed.

In the present embodiment, the presence or absence of abnormality in the bridge 4 is determined based on the statistical distance of each evaluation index at the time of reference and at the time of diagnosis, but the present invention is not limited to this. For example, when the coefficient matrix A (or the principal component matrix A ^) at each time is clustered and the coefficient matrix A based on the new observation data is not classified into the existing cluster but is classified into the new cluster, the feature amount of the coefficient matrix A It may be determined that an abnormality has occurred in the bridge 4 on the assumption that the pattern of the above has changed.

(Effect of Embodiment 1)
In the above-described first embodiment, the state evaluation based on the feature quantity representing the state of the structure is a probability distribution including various physical quantities of the structure, and the sensitivity of the change of the measured value to the damage different depending on the measurement point. It is performed using an evaluation index based on a probability distribution that can consider the estimation error of the measured value while absorbing the difference. Therefore, according to the first embodiment, it is not necessary to set a damage-sensitive vibration mode that requires a high degree of expertise in setting, and a probability distribution having an average and a variance based on a feature amount sensitive to damage of a structure is obtained. By using it, the state of the structure can be evaluated with low cost and high accuracy without performing numerical analysis. In addition, the validity of the evaluation can be expressed by the variance.

Further, the evaluation index of the probability distribution used in the first embodiment captures minute changes that are easily buried due to noise during measurement or forced vibration such as live load, and slowly progress with damage to the structure. The state of the structure can be diagnosed with high accuracy.

Further, in the first embodiment, the coefficient matrix A representing the state of the structure is projected onto the main component space in order to remove the inferior information, and the main component matrix A ^ whose order is reduced is generated. Therefore, since the feature quantity representing the state of the structure is compressed into the feature quantity that can reproduce the state of the structure without degrading the quality of information, the edge processing is performed at a low cost such as a specific low power radio (Low Power Wide Area). It is possible to transfer features by a method using advanced technology, and it is expected that the sensor and transfer system will be configured at low cost and the management cost will be significantly reduced.

More specifically, in the past, it was necessary to use acceleration data for several minutes when evaluating such a vibration state, so a large amount of communication was required. However, in this method, since the necessary information is only the value of the principal component matrix A ^, the principal component matrix A ^ is obtained by edge processing in the measuring instrument, and by making it small data, a specific low power radio or the like can be obtained. It is possible to use inexpensive advanced technology, and as shown in FIG. 7, it is considered that there is an advantage in management cost. FIG. 7 is a diagram showing a comparison between the data amounts of the coefficient matrix A and the principal component matrix A ^ in the first embodiment and the prior art for each number of sensors. FIG. 7 shows a case where data measured for 60 s with a period of 5 ms is transmitted. According to FIG. 7, it can be seen that the amount of data transferred from the sensor node 1 can be significantly reduced by the principal component matrix A ^ regardless of the number of sensors of 8, 4, or 1.

<Embodiment 2>
In the first embodiment, the abnormality diagnosis of the bridge 4 is performed using a statistical index such as the Mahalanobis distance of the probability distribution based on the feature amount of the bridge 4 at the reference time and the diagnosis time, whereas in the second embodiment, the bridge 4 is diagnosed. An abnormality diagnosis of the bridge 4 is performed using the Bayes factor based on the feature amount as an evaluation index. The configuration of the diagnostic system S of the second embodiment is the same as the configuration of the diagnostic system of the first embodiment, except that a part of the processing of the diagnostic unit 212 of the diagnostic apparatus 2 is different.

(Diagnosis processing in the diagnostic system S of the second embodiment)
FIG. 8 is a flowchart showing a diagnostic process in the diagnostic system S of the second embodiment. The flowchart shown in FIG. 8 is another example of the process of step S212 of the process at the time of diagnosis of the first embodiment shown in FIG. That is, in the second embodiment, the diagnostic apparatus 2 executes the diagnostic process shown in FIG. 8 following step S211. In the second embodiment, the diagnostic apparatus 2 can execute the diagnostic time processing of FIG. 5 without performing the reference time processing of FIG.

First, in step S2121, the diagnostic unit 212 of the diagnostic apparatus 2 is defined by the following equation (13) as an evaluation index based on the coefficient matrix A (or the principal component matrix A ^) which is the feature amount received in step S211. Bayes factor B is calculated. When distinguishing from the local Bayes factor B ^j described later, the Bayes factor B is called a global Bayes factor.

In the above equation (13), H ₀ is the null hypothesis (healthy state of the bridge 4 in this embodiment), H ₁ is the alternative hypothesis (abnormal / damaged state of the bridge 4 in this embodiment), and Y _t is the hypothesis test. Target feature amount (in this embodiment, coefficient matrix A or principal component matrix A ^), _Φt is a parameter under the hypothesis of H ₀ or H ₁ (for example, the mean value or standard deviation of the probability distribution of the feature amount). ).

In the Bayes factor B shown in the above equation (13), the denominator p (Y _t | Φ _t , H ₀ ) has the same feature amount Y _t at the time of diagnosis as the state (healthy state) at the reference time of the bridge 4. The marginal likelihood is shown, and the molecule p (Y _t | Φ _t , H ₁ ) shows the marginal likelihood in which the feature amount Y _t at the time of diagnosis is different from the state of the bridge 4 at the reference time. Therefore, when Bayes factor B, which is the ratio of p (Y _t | Φ _t , H ₁ ) to p (Y _t | Φ _t , H ₀ ), exceeds the threshold value, the bridge 4 is abnormally damaged or damaged based on a certain feature amount. Can be determined to have occurred.

Next, in step S2122, the diagnostic unit 212 determines whether or not the Bayes factor B calculated in step S2121 is larger than the threshold value. This threshold value is set as a classification determined by the inspection engineer of the bridge 4 to be abnormal from the past inspection data and the like. The diagnostic unit 212 determines that there is an abnormality in the bridge 4 when the Bayes factor B is larger than the threshold value (step S2122Yes) (step S2123), and when the Bayes factor B is equal to or less than the threshold value (step S2122No), the bridge 4 Is normal (step S2124).

The diagnostic process shown in FIG. 8 may be performed by the sensor node 1 instead of the diagnostic device 2.

(Calculation method of threshold value for abnormality judgment)
Here, a method of calculating the threshold value for abnormality determination used in step S2122 of the diagnostic process (FIG. 8) will be described. FIG. 9A is a diagram for explaining a method of calculating a threshold value for determining an abnormality according to the second embodiment. In the graph (a) on the left side of FIG. 9A, the horizontal axis is time and the vertical axis is the logarithm of Bayes factor B, and the value of Bayes factor B at each time is plotted. Further, in the graph (b) on the right side of FIG. 9A, the horizontal axis is the frequency and the vertical axis is the logarithm of the Bayes factor B, and the frequency distribution of each value of the Bayes factor B is shown. The calculation of the threshold value for abnormality determination may be performed by the diagnostic device 2 or another information processing device.

In an actual bridge, various situations that affect the abnormality judgment occur, such as a truck with a large load passing through or braking on the bridge. Bayes factor B is a variable value including specific data under various circumstances. When calculating the threshold value of Bayes factor B, by including it in the basis of threshold value calculation instead of excluding specific data of multiple measurement data in the past predetermined period and data variation, in an actual bridge It is possible to calculate the threshold value considering various situations that occur.

Of the past measurement data of the bridge 4, the measurement data (see graph (a) of FIG. 9A) for a predetermined period (for example, one year) as a reference is applied to the above equation (13) to obtain the distribution of the value of Bayes factor B. Create (see graph (b) in FIG. 9A). Based on the mean value μ and the variance σ of the distribution of the values of Bayes factor B, for example, any value such as μ, μ + σ, μ + 2σ can be determined as a threshold value. In the graph (b) of FIG. 9A, since the value of Bayes factor B is normalized, the normal distribution has an average μ = 0, and the case where the threshold value is the average μ is shown. The threshold value of Bayes factor B can be quantitatively determined using the average μ and the variance σ based on the characteristics of the bridge 4 and index values such as usage conditions, geographical conditions, and meteorological conditions. In this way, it is possible to improve the accuracy of abnormality determination by using the threshold value considering various situations that occur in an actual bridge.

(Other examples of abnormal diagnosis threshold)
Another example of the threshold value for abnormality diagnosis will be described. FIG. 9B is a diagram for explaining another example of the abnormality determination method using the plurality of threshold values of the second embodiment. In FIG. 9B (graph a), the horizontal axis is the frequency distribution and the vertical axis is the logarithm of the Bayes factor B, and the frequency distribution of each value of the Bayes factor B is shown. In (graph b) of FIG. 9B, the horizontal axis is the day and the vertical axis is the logarithm of the Bayes factor B, and each value of the Bayes factor B on each day is plotted.

The value of Bayes factor B is determined to be abnormal when it exceeds the threshold value. From this, in the distribution of the values of Bayes factor B based on the above-mentioned measurement data with reference to FIG. 9A, for values larger than the average μ, a plurality of thresholds are set using the average μ and the standard deviation σ, and an abnormality is obtained. Can be determined step by step. Based on a plurality of thresholds, a range indicating the diagnostic level may be defined, for example, a range of "normal", a range of "need attention", and a range of "abnormality".

For example, a plurality of values based on the mean value μ of the distribution of the values of Bayes factor B and the variance σ (μ, μ + σ, μ + 2σ, etc. in (graph a) of FIG. 9B) are used for the abnormality diagnosis of the bridge 4 based on the Bayes factor B. Determine with multiple gradual thresholds. As shown in FIG. 14 (graph a), if the Bayes factor B is B ≦ μ + σ, it is diagnosed as “normal”, if μ + σ <B ≦ μ + 2σ, it is diagnosed as “need attention”, and if μ + 2σ <B. If so, it can be diagnosed as "abnormal". In the example of FIG. 9B, the number of the plurality of thresholds for diagnosing abnormalities is set to “2”, but the number is not limited to this. The stepwise threshold value of Bayes factor B and its number shall be quantitatively determined using the average μ and the variance σ based on the characteristics of the bridge 4 and the index values such as usage conditions, geographical conditions, and meteorological conditions. Can be done.

In this way, it is possible to improve the accuracy of abnormality determination by using multiple threshold values that take into consideration various situations that occur in actual bridges. Further, by determining the threshold value of the abnormality diagnosis of the bridge 4 based on the Bayes factor B step by step, the manager of the bridge 4 can grasp the abnormality of the bridge 4 gradually progressing with the passage of time, and plans. The bridge 4 can be maintained.

(Local Bayes Factor B)
By the way, the Bayes factor B based on the above equation (13) is an evaluation index for detecting an abnormality in the bridge 4 as a whole. On the other hand, it is also possible to calculate an individual Bayes factor based on the feature amount obtained from the time series of the sensor information of each sensor 3 installed on the bridge 4. Individual Bayes factors are called local Bayes factors. For example, the local Bayes factor Bj based on the time-series features of the sensor information of the jth ( ^j = 1, ..., N) sensor 3 among the plurality of sensors 3 installed on the bridge 4 is as follows. It is defined by the equation (14).

By substituting the local Bayes factor B ^j for the Bayes factor B and performing the diagnostic processing shown in FIG. 8, the abnormality of the bridge 4 at the location where the sensor 3 is installed at the jth position from the index j of the local Bayes factor B ^j indicating an abnormality. And can be identified as having damage.

The local Bayes factor ^Bj can also be used as follows. The type of sensor information that can sensitively detect an abnormality using the Bayes factor may differ depending on the type of the structure such as the bridge 4 and the abnormality or damage that occurs in the structure. Therefore, sensors 3 of a plurality of sensor types are installed in a structure such as a bridge 4, and diagnosis is performed based on the local Bayes factor ^Bj based on the feature amount generated for each sensor type.

That is, sensors 3 having different physical quantities to be detected (different sensor types) are installed on the bridge 4, and the local Bayes factor ^Bj is calculated based on the time-series feature quantities of the sensor information for each sensor type. When an abnormality is determined based on the local Bayes factor B ^j of the jth sensor type, an abnormality or damage occurs based on the sensor information of the sensor 3 of the jth sensor type from the index j of the local Bayes factor B ^j . It can be identified as being. By using the local Bayes factor B ^j in this way, any one of the plurality of physical quantities can be made to detect an abnormality, so that the sensitivity of abnormality detection can be increased.

(Abnormality determination in the bridge diagnosis system S of the second embodiment)
FIG. 10 is a diagram for explaining the experimental results of abnormality determination performed using the diagnostic system S of the second embodiment. FIG. 10 shows the time transition of the binary logarithm of 10 types of local Bayes factor B ^j of A1 to A10, with the horizontal axis representing time and the vertical axis representing the binary logarithm axis of local Bayes factor B ^j . In FIG. 10, A1 to A10 show the installation position of the sensor 3 as an example. INT (initial), DMG1, and DMG2 in FIG. 10 represent each period.

In INT, since the bridge 4 is not damaged, the local Bayes factor B ^j based on the sensor information of the sensor 3 at any installation position is also approximately 0 and does not exceed the threshold value. On the contrary, if the local Bayes factor B ^j has a value as shown in FIG. 10, it can be estimated that the bridge 4 is in a healthy state without any abnormality or damage.

In DMG1 where the part near A6 of the bridge 4 was damaged following INT, the value of the local Bayes factor based on the sensor information of sensor 3 of A6 in particular among A1 to A10 exceeded the threshold value compared with INT. It's getting bigger. On the contrary, if the local Bayes factor B ^j is a value like DMG1 as compared with INT, it can be estimated that damage has occurred in the portion of the bridge 4 near A6.

Further, in the DMG2 in which the portion of the bridge 4 near A6 following the DMG1 is further damaged, the local Bayes factor Bj based on the sensor information of the sensor 3 of the A6 is further larger than that of the ^DMG1 . On the contrary, if the local Bayes factor B ^j has a value similar to that of DMG2 as compared with DMG1, it can be estimated that the damage generated in the vicinity of A6 of the bridge 4 has further progressed. Further, by comparing the sum or average of the local Bayes factor B ^j of ^DMG1 with the sum or average of the local Bayes factor Bj of DMG2, the degree of progress of damage from DMG1 to DMG2 can be evaluated.

Even if the Global Bayes factor is used, the same evaluation as in FIG. 10 can be performed.

By using ^Bayes factor B or local Bayes factor Bj in this way, it is possible to detect that the bridge 4 has changed from a healthy state to an abnormal state, and to evaluate the progress of damage that has already been found by inspection or the like. can.

(Experimental result of Embodiment 2)
FIG. 11 is a diagram showing execution conditions of an experiment performed using the diagnostic system S of the second embodiment. FIG. 12 is a diagram showing the results of experiments performed using the diagnostic system S of the second embodiment.

FIG. 11 shows the pattern of the load applied to the bridge over time, with the horizontal axis representing the time and the vertical axis representing the load. As shown in FIG. 11, times t0 to t1 (Stage1), t2 to t3 (Stage2), t4 to t5 (Stage3), t6 to t7 (Stage4), t7 to t8 (Stage5), t9 to t10 (Stage6), At t10 to t11 (Stage7) and t12 to (Stage8), vibration was applied to the bridge. At times t1 to t2 (Loading 1), a crack load was applied to the bridge. At times t3 to t4 (Loading 2), a yield load was applied to the bridge. Times t5 to t6 (Loading 3) are load continuation periods for the bridge. At times t8 to t9 (Loading 4), a design load was applied to the bridge. At times t11 to t12 (Loading 5), the maximum load was applied to the bridge.

As a result of conducting an experiment under such execution conditions, as shown in FIG. 12, it can be said that the Bayes factor tends to increase with the progress of Stage accompanied by loading. Therefore, since a clear interpretation can be given to the observation result of the change of Bayes factor, it is possible to easily detect the abnormality or damage generated in the bridge.

As described above, the advantage of this embodiment that the abnormality or damage generated in the bridge can be easily detected regardless of the increase or decrease of the natural frequency of the bridge can be seen from FIG. As a comparative example, FIG. 13 is a diagram showing changes in the natural frequencies of bridges that differ depending on the vibration mode measured under the same conditions as the experiment conducted using the diagnostic system S of the second embodiment shown in FIG. ..

Graph 1101 in FIG. 13 shows changes in the natural frequency of the primary bending mode in Stages 1 to 8. Further, graph 1102 in FIG. 13 shows changes in the natural frequency of the secondary bending mode in Stages 1 to 8. As shown in graphs 1101 and 1102, the natural frequency increases or decreases due to a change in the state of the bridge. Theoretically, the natural frequency decreases due to the decrease in rigidity due to damage, but the natural frequency may increase, for example, when the state of the bearing of the bridge changes. Further, as shown in the graphs 1101 and 1102, the change in the natural frequency differs depending on the mode.

In this way, in the abnormality diagnosis based on the natural frequency, it is difficult to set an appropriate vibration mode and determine the presence or absence of damage and the part because the mode of change of the natural frequency differs depending on the damaged part and the vibration mode. .. In this respect, the abnormality diagnosis using the Bayes factor according to the present embodiment does not require the setting of the vibration mode, and the abnormality or damage generated in the bridge can be quantitatively detected.

(Changes in temperature and Bayes factor values over time)
FIG. 14 is a diagram showing changes over time in the values of the outside air temperature and the Bayes factor in a certain period of a certain girder of the bridge 4. FIG. 14 (graph a) is a graph showing the time course of the outside air temperature for a certain period, and (graph b) is a graph showing the time course of the Bayes factor B of the outside air temperature. The solid line in FIG. 14 (graph b) indicates the threshold value for determining the abnormality of Bayes factor B, and when this threshold value is exceeded, it is determined to be abnormal.

FIG. 14 exemplifies the data of the change in the outside air temperature of the girder of the bridge 4 in which it is confirmed that there is no abnormality in the structure itself. Although the outside air temperature fluctuates in a one-year cycle, it is determined that the change in the outside air temperature during the target period for determining the abnormality is not an abnormal change in the abnormality determination when the present embodiment is applied to the outside air temperature. That is, from FIG. 14, it can be seen that in the present embodiment, it is possible to determine the abnormality of the structure in consideration of the presence or absence of the influence of periodic fluctuations occurring in the natural world such as temperature changes.

(Bridge 4 deflection and Bayes factor value change over time)
FIG. 15 is a diagram showing changes over time in the values of deflection and Bayes factor of a bridge over a certain period of time. Deflection is detected using a communication pipe. In FIG. 15, one year from April to March of the following year is set as a yearly unit, (graph a) shows the change over time of the deflection for a certain period, and (graph b) shows the change over time of the Bayes factor B of the deflection. show. The solid line in FIG. 15 (graph b) indicates the threshold value for determining the abnormality of Bayes factor B, and when this threshold value is exceeded, it is determined to be abnormal. The threshold value for abnormality judgment is set based on the measurement data for one year of the base year.

The temperature fluctuates throughout the year, but as shown in (Graph a) of FIG. 15, the determination result by this embodiment shows that it is abnormal at the time when the amount of deflection becomes large (for example, from September of the fourth year to January of the following year). It can be seen that it has been judged. In addition, it can be seen that after the 5th year, it is clearly determined to be abnormal.

(Comparison of changes over time in Bayes factor values between two spans of a bridge)
FIG. 16 is a diagram showing a comparison of changes over time in the values of Bayes factor between two spans of a bridge. In FIG. 16, one year from May to April of the following year is set as the annual unit. The bridge 4 whose data is shown in FIG. 16 is the same as the bridge 4 whose data is shown in FIG.

In (Graph a) and (Graph b) of FIG. 16, abnormality determination is made based on the measurement results of acceleration at two different points (damaged first span and second span with relatively little damage) of the same bridge 4. The result of is calculated. In FIG. 16, the threshold value is set based on the measurement data for one year of the base year. However, since the base year, the first year, the second year, and the third year of FIG. 16 are different from those of FIG. 15, they are distinguished by notation such as the first year of FIG. 15 and the first year of FIG. ..

In the abnormality determination according to the present embodiment, the damaged abnormality between the first spans was detected in September of the first year of FIG. 16, and the deterioration over time can be grasped thereafter. On the other hand, in the second span where the damage is relatively small, even in the year when the abnormality is detected in the first span in FIG. 16, it is judged to be normal except for the temporary abnormal value. Since the result of this first span is the same as the time when the fluctuation of the deflection amount shown in FIG. 15 becomes large (the fourth and fifth years and thereafter in FIG. 15), an appropriate abnormal value can be determined. it is conceivable that.

(Modified Example of Embodiment 2)
In the present embodiment, the diagnostic apparatus 2 uses the coefficient matrix A or the principal component matrix A ^ of the bridge 4 as a feature quantity representing the state of the bridge 4, and uses the Bayes factor B or the local Bayes factor B ^j based on the feature quantity to bridge the bridge. It was decided to make an abnormality diagnosis. However, not limited to this, feature quantities are extracted from time-series data obtained by continuously measuring one or more physical quantities of each part of the structure, and an abnormality diagnosis of the structure is performed using the Bayes factor based on these feature quantities. You may. Physical quantities are displacement, velocity, acceleration, external force, strain, temperature, and the like. For example, an abnormality diagnosis of a structure may be performed using a Bayes factor based on a feature amount extracted from data for a predetermined period in time-series data obtained by observing the temperature of the bridge 4 at a fixed point.

Further, the diagnostic unit may correct the Bayes factor using the result of the on-site inspection by the inspection worker for the bridge 4, and diagnose the state of the bridge 4 based on the corrected Bayes factor. For example, if an abnormality is determined based on the Bayes factor of a certain feature, but there is no abnormality as a result of on-site inspection, the Bayes factor is corrected so that the same feature is not determined to be abnormal thereafter. To. Alternatively, if there is an abnormality as a result of the on-site inspection, the threshold value may be set or corrected by referring to the Bayes factor based on the sensor installed in the vicinity of the site where the abnormality is detected.

(Effect of Embodiment 2)
In the second embodiment, the Bayes factor, which is the ratio of the probability of the healthy state and the abnormal state calculated from the time series of the observation data of the structure in each time domain, and the inspection engineer with a high degree of expertise are abnormal in the past. The state of the structure is diagnosed based on the threshold set from the damaged state. Therefore, even a person who does not have a high degree of specialized knowledge does not require high-cost processing such as numerical analysis, and uses a quantitative evaluation index to form a structure equivalent to that of an inspection engineer who has a high degree of specialized knowledge. It is possible to evaluate the soundness of an object with high accuracy.

<Embodiment 3>
In the third embodiment, in the first or second embodiment, the coefficient matrix A, which is a feature quantity representing the state of the bridge 4, is sequentially learned by using Bayesian estimation to improve the abnormality detection accuracy of the bridge 4. The sequential learning of the coefficient matrix A may be performed by the sensor node 1 (for example, the feature amount generation unit 113) or the diagnostic device 2 (for example, the diagnostic unit 212). The same effect can be obtained by performing the same processing on the principal component matrix A ^ instead of the coefficient matrix A.

FIG. 17 is an explanatory diagram of sequential update of the coefficient matrix A in the diagnostic system S of the third embodiment. In FIG. 17, the acceleration detected by the sensor 3 is shown on the vertical axis, and the time is shown on the horizontal axis. The posterior probability p (A, Σ | Y) _i of the coefficient matrix A at the time i shown in FIG. 17 is obtained by using Bayesian estimation as shown in the following equation (15-1). However, the prior probability p (A, Σ) _i at time i is the posterior probability p (A, Σ | Y) _i-1 at time (i-1) as shown in the following equation (15-2). .. By replacing i in the following equations (15-1) and (15-2) with i + 1, the posterior probabilities p (A, Σ | Y) _{i + 1} of the coefficient matrix A in the time (i + 1) can be obtained in the same manner. ..

(Effect of Embodiment 3)
FIG. 18 is a diagram showing the convergence of the probability distribution of the coefficient matrix by sequentially updating the coefficient matrix in the diagnostic system of the third embodiment. In FIG. 18, the horizontal axis is a value that can be taken by an element of a feature amount (for example, a coefficient matrix A), and the vertical axis is a posterior probability at each value of the element. In the third embodiment, by sequentially learning the probability distribution of the coefficient matrix A, as shown in FIG. 18, the average value of the probability distribution approaches the true value and converges in the direction in which the variance becomes smaller (the probability distribution shown in FIG. 18 is). As it converges from a thin line to a broken line and then to a thick line), the coefficient matrix A can represent the state of the structure with higher accuracy. Then, the diagnosis unit 212 diagnoses the state of the structure using the coefficient matrix A or the principal component matrix A ^ after the sequential learning as the feature quantity. Therefore, it is possible to improve the accuracy of the abnormality estimation of the state of the structure. Since the coefficient matrix A is a feature quantity that includes various features of the vibration data of the structure, it captures finer state changes of the structure than the conventional technique for determining an abnormality using the natural frequency or the like. It becomes possible to perform abnormality judgment.

(Other embodiments)
In the above-described embodiment, the diagnostic system S has been described as having the sensor node 1 and the diagnostic device 2, but the embodiment is not limited thereto. For example, the diagnostic system S may be configured without the diagnostic device 2. In such a case, the diagnostic system S includes a sensor node 1 having an acquisition unit, an autoregressive model generation unit, a feature amount generation unit, and a diagnostic unit. That is, the sensor node 1 has an acquisition unit that acquires sensor information from a sensor attached to the structure in time series, and a sensor that acquires sensor information acquired by the acquisition unit at a certain time before a certain time. A feature quantity that generates a feature quantity that indicates the state of a structure at a certain time based on a self-regression model generator that generates a self-regression model expressed by a linear combination of information in a time series and a regression coefficient of the self-regression model. With respect to the generation part and the feature quantity indicating the state of the structure generated from the time series of information about the structure, the first periphery representing the probability distribution of the feature quantity when the structure is assumed to be in a healthy state. An evaluation index, which is a ratio of the likelihood and the second peripheral likelihood representing the probability distribution of the feature amount when the structure is not in a healthy state, is calculated, and the state of the structure is calculated based on the evaluation index. It has a diagnostic unit for diagnosing.

In the above-described embodiment, a case where a feature amount indicating the state of the structure at a certain time is generated based on the regression coefficient of the autoregressive model has been described, but the embodiment is not limited to this. For example, a feature quantity indicating the state of the structure may be generated from other than the regression coefficient of the autoregressive model.

(Implementation example of sensor node 1)
FIG. 19 is a block diagram showing a configuration of a sensor node terminal 1B used as the sensor node 1 of the embodiment. The sensor node terminal 1B mounted as the sensor node 1 further includes an acceleration sensor 15 as compared with the configuration of the sensor node 1 (FIG. 2). Such a sensor node terminal 1B exhibits the same functions as the sensor node 1 and the sensor 3 by executing a predetermined application. That is, the sensor node terminal 1B is installed on the bridge 4 in the same manner as the sensor 3, generates a feature amount of the bridge 4 from the acquired acceleration data, and transmits it to the diagnostic device 2.

The diagnostic device 2 is built on a cloud server, for example, and provides diagnostic results based on feature quantities as a cloud service. The diagnostic device 2 transmits the diagnostic result to the sensor node terminal 1B or another terminal device. The user confirms the diagnosis result by looking at the screen output of the sensor node terminal 1B or other terminal device.

Further, a plurality of sensor node terminals 1B may be mounted as a plurality of sensor nodes 1 and installed at a plurality of locations on the bridge 4 in the same manner as the sensor 3. In this case, the bridge 4 is obtained from the acceleration data acquired by all of the plurality of sensor node terminals 1B by the independent processing of one sensor node terminal 1B representing the plurality of sensor node terminals 1B or the cooperative processing of a predetermined number of sensor nodes 1. The feature amount of is generated and transmitted to the diagnostic apparatus 2.

The present invention is not limited to the above-described embodiment, and the configuration of each embodiment can be added, deleted, replaced, integrated, or dispersed. Further, the configurations and processes shown in the embodiments can be appropriately distributed, integrated, or replaced based on the efficiency of the processes or implementations. The program that executes each process of the diagnostic system described in the above-described embodiment is installed in one or more computers via a recording medium or a transmission medium, or is provided as an embedded program.

S: Diagnostic system, 1: Sensor node, 2: Diagnostic device, 3: Sensor, 4: Bridge, 111: Sensor information acquisition unit, 112: Self-return model generation unit, 113: Feature amount generation unit, 114: Feature amount transmission Unit, 131: Sensor information, 211: Feature amount receiving part: 212: Diagnosis part, 231: Feature amount, 232: Reference evaluation index

Claims (14)

1. With respect to the feature quantity indicating the state of the structure generated from the time series of information about the structure, the first marginal likelihood representing the probability distribution of the feature quantity when the structure is assumed to be in a healthy state. , Calculate an evaluation index which is a ratio of the second marginal likelihood representing the probability distribution of the feature amount when it is assumed that the structure is not in the healthy state, and based on the evaluation index, the structure. A structure diagnostic system characterized by having a diagnostic unit for diagnosing a condition.
2. The time series of the information is acquired for each position in the structure.
   The diagnostic unit calculates the evaluation index for each position based on the time series for each position in the structure, and diagnoses the state for each position of the structure based on the evaluation index for each position. The structure diagnostic system according to claim 1.
3. The time series of the information is acquired for each of a plurality of types of the information.
   The first aspect of claim 1 is characterized in that the diagnostic unit calculates the evaluation index for each type based on the time series for each type and diagnoses the state of the structure based on the evaluation index for each type. The described structure diagnostic system.
4. An acquisition unit that acquires sensor information in chronological order from sensors attached to the structure, and
   An autoregressive model generation unit that generates an autoregressive model that expresses the sensor information acquired at a certain time by the acquisition unit by a linear combination of the sensor information acquired before the certain time.
   Any of claims 1 to 3, further comprising a feature amount generation unit that generates the feature amount indicating the state of the structure at a certain time based on the regression coefficient of the autoregressive model. Or the structure diagnostic system according to item 1.
5. The structure diagnostic system is
   A sensor including the acquisition unit, the autoregressive model generation unit, the feature amount generation unit, and a transmission unit that transmits the feature amount generated by the feature amount generation unit to the diagnosis unit. The structure diagnostic system according to claim 4, wherein the structure diagnostic system has a node.
6. The number of the sensors is plural,
   The sensor information acquired from the plurality of sensors at each time in the time series is a sensor information vector at each time having the number of the sensors as the order.
   In the autoregressive model, the sensor information vector acquired by the acquisition unit at a certain time is represented by a time-series linear combination of the sensor information vectors acquired before the certain time.
   The regression coefficient is a matrix multiplied by each of the time series of the sensor information vectors linearly combined in the autoregressive model.
   The structure diagnostic system according to claim 4, wherein the feature amount is information based on a coefficient matrix obtained by combining the matrices.
7. The structure diagnosis system according to claim 6, wherein the feature amount generation unit generates a principal component matrix of the coefficient matrix based on the singular value decomposition of the coefficient matrix as the feature amount.
8. The feature amount generation unit is
   A claim characterized in that the feature amount is generated by sequential learning that repeats the process of making the posterior distribution obtained by Bayesian estimation using the probability distribution of the feature amount generated previously as the prior distribution the probability distribution of the feature amount generated this time. Item 6. The structure diagnosis system according to any one of Items 4 to 7.
9. Claims 1 to 8 include the feature amount including at least one of information on displacement, velocity, acceleration, mass, rigidity, and damping coefficient of the structure at the observation point of the feature amount. The structure diagnostic system according to any one of the above items.
10. The diagnostic unit is characterized in that the evaluation index is corrected by using the inspection result of the on-site inspection of the structure, and the state of the structure is diagnosed based on the corrected evaluation index. 9. The structure diagnostic system according to any one of 9.
11. The diagnostic unit diagnoses the state of the structure based on the comparison between the evaluation index and the threshold value.
    The structure according to any one of claims 1 to 10, wherein the threshold value is determined based on the average and variance of a plurality of measurement data of the structure in the past predetermined period. Diagnostic system.
12. A plurality of the threshold values are determined based on the mean and the variance.
    The structure diagnostic system according to claim 11, wherein the diagnostic unit diagnoses the state of the structure based on a comparison between the evaluation index and a range based on the plurality of threshold values.
13. It is a structure diagnosis method performed by a structure diagnosis system that diagnoses the state of a structure.
    With respect to the feature quantity indicating the state of the structure generated from the time series of the information about the structure, the first marginal likelihood representing the probability distribution of the feature quantity when the structure is assumed to be in a healthy state. And a second marginal likelihood representing the probability distribution of the feature amount when it is assumed that the structure is not in the healthy state, an evaluation index which is a ratio is calculated.
    A structure diagnosis method comprising each process for diagnosing the state of the structure based on the evaluation index.
14. Computer,
    With respect to the feature quantity indicating the state of the structure generated from the time series of information about the structure, the first marginal likelihood representing the probability distribution of the feature quantity when the structure is assumed to be in a healthy state. , Calculate an evaluation index which is a ratio of the second marginal likelihood representing the probability distribution of the feature amount when it is assumed that the structure is not in the healthy state, and based on the evaluation index, the structure. A structure diagnosis program to function as a diagnostic unit for diagnosing a condition.
    

Images