JP6263455B2 - Structure deterioration diagnosis system - Google Patents

Description

  The present invention relates to a structure deterioration diagnosis system capable of monitoring deterioration of a mounting state of a structure such as a suspended structure in a tunnel or an overhang structure over a long period of time.

  As a method for detecting the deterioration state of the structure, it has been the mainstream that the inspection is carried out by visual inspection or some kind of instrument by a regular inspection by an inspector. In addition, a conventional technique is disclosed in which a quantitative inspection is easily and quickly performed on a crack that occurs over time in a structure such as a tunnel that is a degradation diagnosis target (see, for example, Patent Document 1).

  In Patent Document 1, a fluorescent dye that emits light by excitation light such as ultraviolet light or blue-based visible light is mixed in advance with a structure that is an object of deterioration diagnosis. Then, the structure is irradiated with a light source that emits ultraviolet light, blue-based visible light, or the like, and the occurrence of cracks is quantitatively determined by visual analysis or analysis processing of a captured image by a CCD camera or the like.

JP 2013-83493 A

  However, the prior art has the following problems.

  Although Patent Document 1 enables quantitative deterioration diagnosis, it is based on periodic inspection by an inspector to the last. Furthermore, it is necessary to mix a fluorescent dye in advance with respect to the structure to be deteriorated.

  On the other hand, in recent years, there has been a demand for a deterioration diagnosis system that can perform deterioration diagnosis of a structure with a shorter cycle than a periodic inspection and without an inspector. In addition, in tunnels, in addition to the deterioration of the walls and ceiling of the tunnel itself, it is necessary to quantitatively diagnose the deterioration of the mounting state of a suspended structure in the tunnel or a structure such as an overhang structure. There is sex. Furthermore, it is desired that not only a new structure but also an existing structure can be easily handled.

  The present invention has been made in order to solve the above-described problems, and is capable of quantitatively diagnosing the deterioration of the mounting state of the structure without requiring a periodic inspection by an inspector. The purpose is to obtain a deterioration diagnosis system.

  The structure deterioration diagnosis system according to the present invention is attached to a fixed surface, and the attachment state of the structure is normal based on the acceleration information of the structure output from the sensor head installed in the structure to be subjected to deterioration diagnosis. In a structure deterioration diagnosis system including a sensor controller for diagnosing whether or not the structure deterioration diagnosis system is executed by the sensor controller, the inclination of the structure is used as a feature amount based on acceleration information acquired from the sensor head. A feature amount extracting means for extracting at least one of information and natural frequency information, and a reference probability density distribution in a normal state and a diagnostic probability density at the time of deterioration diagnosis for the feature amount extracted by the feature amount extracting means The probability density distribution calculating means for calculating the distribution and the reference probability density distribution calculated by the probability density distribution calculating means When a determination area and a deterioration determination area are set and the distribution included in the deterioration determination area exists in the probability density distribution for diagnosis calculated by the probability density distribution calculation means at the time of deterioration diagnosis, the attachment state of the structure And determining means for determining that an abnormality has occurred.

  According to the present invention, the deterioration of the mounting state of the structure is periodically detected by the inspector by detecting the case where the feature amount cannot be obtained from the probability density distribution calculated from the feature amount based on the acceleration information. It is possible to obtain a structure deterioration diagnosis system that can be diagnosed at an early stage without requiring an inspection.

It is a block diagram of the structure deterioration diagnostic system in Embodiment 1 of this invention. It is a flowchart which shows the flow of the degradation diagnosis process by the 1st degradation diagnostic part in Embodiment 1 of this invention. It is the figure which showed the specific example regarding collection of the inclination information for 3 axes in Embodiment 1 of this invention. It is the figure which showed the specific example of probability density distribution regarding the inclination information for 3 axes in Embodiment 1 of this invention. It is the figure which showed the distance of probability distribution in Embodiment 1 of this invention. It is a flowchart which shows the flow of the degradation diagnostic process by the 2nd degradation diagnostic part in Embodiment 1 of this invention. It is the figure which showed the specific example regarding collection of the frequency information for 3 axes | shafts in Embodiment 1 of this invention. It is the figure which showed the calculation result of the power spectrum for 3 axes in Embodiment 1 of this invention. It is explanatory drawing of the degradation diagnostic method based on the degradation determination area | region in Embodiment 2 of this invention. It is a figure for demonstrating the comparison with the deterioration diagnosis in previous Embodiment 1, and the deterioration diagnosis in this Embodiment 2. FIG.

  Hereinafter, a preferred embodiment of a structure deterioration diagnosis system of the present invention will be described with reference to the drawings.

  The present invention obtains a probability density distribution of a feature amount related to the inclination or natural frequency of a structure based on acceleration information obtained from a sensor attached to the structure to be diagnosed for deterioration, When there is a significant difference between the probability density distribution at the time of deterioration diagnosis, it is a technical feature that it is determined that some deterioration has occurred in the mounting state.

Embodiment 1 FIG.
FIG. 1 is a configuration diagram of a structure deterioration diagnosis system according to Embodiment 1 of the present invention. The structure deterioration diagnosis system according to the first embodiment includes a sensor controller 10 and N (N is an integer of 2 or more) sensor heads 20 (1) to 20 (N).

  In the structure deterioration diagnosis system according to the first embodiment, it is possible to perform the deterioration diagnosis if at least one sensor head 20 is provided. In addition, the functions of the N sensor heads when a plurality of sensor heads are used are all common. Therefore, in the following description, when it is not necessary to distinguish each sensor head, the sensor head 20 is simply described without using the subscripts (1) to (N).

  Each of the N sensor heads 20 includes a sensor unit 21 and an acceleration information output unit 22, and is installed at different positions of the structure that is the object of the deterioration diagnosis. Here, specific examples of the structure include a suspended structure in a tunnel or an overhang structure, and the structure deterioration diagnosis system according to the present invention performs deterioration diagnosis of the mounting state of the structure over a long period of time. It will be.

  Examples of such a structure include jet fans, information boards, signs, and the like installed in the tunnel, and heavy objects that are suspended or overhang from the concrete of the casing forming the tunnel. Then, preferably two sensor heads 20 are installed on the structure, and each of them is installed at a portion which becomes both ends with respect to the structure, so that one of the displacements of the structure is largely captured.

  Therefore, the N sensor heads 20 are installed at positions away from each other with respect to the direction of the installation surface in order to capture the displacement of the structure greatly. Furthermore, the sensor head 20 may be replaced with two or more in order to facilitate replacement, and the monitoring state can be continued by replacing each pair.

  The sensor unit 21 is, for example, a three-axis acceleration sensor that has excellent responsiveness and can measure acceleration with a measurement range of about DC to several tens of Hz. Of the acceleration, the DC component represents the direction of gravity acceleration, that is, the inclination, and the AC component represents vibration. Thus, by using a triaxial sensor as the acceleration sensor, leveling is not required, and sensor output can be performed regardless of the direction of inclination or vibration. Therefore, if the direction of inclination can be specified such as leveling, two or one axis may be used.

  The acceleration information output unit 22 converts an analog signal related to the triaxial acceleration of the structure at the installation location of the sensor head into a digital signal at a predetermined sampling rate (for example, a sampling rate of 50 Hz), and the sensor is used as acceleration information. Transmit to the controller 10. In addition, the acceleration information output part 22 can transmit the acceleration information which is a digital signal to the sensor controller 10 via communication I / F, such as CAN and RS232C, as an example.

  The sensor controller 10 is installed at a position where the wiring reaches the N sensor heads 20 in the vicinity of the structure, and the mounting state of the structure is normal based on the acceleration information acquired from each of the N sensor heads 20. Determine whether or not. Specifically, the sensor controller 10 extracts the inclination information as the first feature amount based on the acceleration information, and performs the deterioration diagnosis of the attachment state of the structure, and the acceleration information. Based on this, it has the second deterioration diagnosis unit 12 that extracts the natural frequency information as the second feature quantity and performs the deterioration diagnosis of the attachment state of the structure.

  In addition, the sensor controller 10 includes a large-capacity communication I / F such as Ethernet (registered trademark), for example, so that the diagnosis result or data used for the diagnosis can be transmitted to a control device (not shown). Can do.

  In the case of a tunnel facility provided with a jet fan or the like as a structure, information in the tunnel is collected in a control device installed in the vicinity of the tunnel, which is provided in the vicinity of the tunnel, and is not shown in detail. The information is transferred to the remote monitoring and control device where there is, and the information is shared by both parties. In accordance with the deterioration diagnosis process described later, the result of determining that the deterioration has occurred is notified to each of the control devices with a display, an alarm sound, or the like. Such a determination result may be sent separately to a center apparatus of an equipment contractor or an inspection contractor of the deterioration diagnosis system, and can promptly respond when an abnormality occurs.

Further, the level for determining deterioration may be subdivided as follows, and the correspondence may be changed according to the level. For example, the deterioration judgment level is
Level 1: “Necessary to conduct detailed inspections at regular inspections (confirmation level)”
Level 2: “Need to check the site and consider future improvement plans (plan level)”
Level 3: “There is a need for immediate improvement, and there is a need to urgently go to the site (improvement level)”
Level 4: Emergency situation including collapse (emergency level) ”
Levels 1 and 2 can be simply displayed on the panel of each control device, and at levels 3 and 4, detailed alarm display and alarm sounding may be performed. .

  Furthermore, data used for continuous determination at the center device of the supplier may be collected and analyzed separately from multiple aspects, for example, to predict the time until reaching level 3. Tunnels are often monitored by a remote monitoring and control device at all times. However, in some cases, and in cases where daily management facilities are not sufficient, the primary emergency in the event of an emergency through the contractor's center device. It is also possible to make a staff member of the dealer take a response to confirm the situation and report the situation as a response.

The first deterioration diagnosis unit 11 and the second deterioration diagnosis unit 12 both compare the probability density distribution at the normal time of the structure to be monitored with the probability density distribution at the time of deterioration diagnosis, and are significant. This is common in that the deterioration of the structure is detected due to the difference. However, a specific signal processing method is used in the first deterioration diagnosis unit 11 that performs the deterioration diagnosis using the inclination information as the feature amount and the second deterioration diagnosis unit 12 that performs the deterioration diagnosis using the natural frequency information as the feature amount. These are different and will be described in detail below.
(1) Flow of deterioration diagnosis processing based on inclination information by first deterioration diagnosis unit 11 FIG. 2 is a flowchart showing a flow of deterioration diagnosis processing by the first deterioration diagnosis unit 11 according to Embodiment 1 of the present invention. It is. A specific deterioration diagnosis process by the first deterioration diagnosis unit 11 based on the inclination information will be described based on the flowchart of FIG.

  First, in step S <b> 201, the first deterioration diagnosis unit 11 extracts inclination information based on the acceleration information acquired from the acceleration information output unit 22 in the sensor head 20. Specifically, the first deterioration diagnosis unit 11 acquires X, Y, and Z-axis acceleration information at a sampling rate of 50 Hz, for example. And the 1st degradation diagnostic part 11 performs the low-pass filter process by a digital filter for every axis | shaft using this acceleration information. The cut-off frequency of the low-pass filter is, for example, about 0.1 Hz.

  Here, the low-frequency component of the triaxial acceleration including the DC component is equivalent to the tilt information. Then, the low-pass filtered slope information does not include a rapidly changing high frequency component. For this reason, information is not lost even if the sampling rate of the inclination data for detecting the deterioration is slowed. Specifically, if there is a sampling rate that exceeds twice the cutoff frequency, the sampling theorem can be satisfied, and for cutoff frequencies below 0.1 Hz, a sampling frequency of 0.2 Hz, ie, Sampling at intervals of 5 seconds is sufficient.

  FIG. 3 is a diagram showing a specific example relating to collection of inclination information for three axes in the first embodiment of the present invention. As shown in FIG. 3, the first deterioration diagnosis unit 11 collects inclination information for three axes at each sampling interval ΔT. The “tilt information” in the present invention means a tilt in a range from 180 degrees to −180 degrees when a certain reference direction is defined as 0 degrees, and is shown on the vertical axis in FIG. “Inclination information” is obtained by normalizing the inclination from 180 degrees to −180 degrees with respect to the reference direction from 1 to −1.

  Next, the first deterioration diagnosis unit 11 corrects the mounting direction of the sensor head 20 in step S202. Specifically, the first deterioration diagnosis unit 11 performs a matrix operation on the calibration data (affine transformation matrix elements) in the mounting direction of the sensor head 20 and the sampling data obtained in the previous step S201. The error in the mounting direction is corrected so that the coordinate axes of the respective sensor heads 20 coincide.

  The purpose of this step S202 is to align the coordinate axes regardless of the mounting direction of the sensor head 20, and since the sensor head 20 can detect tilt and vibration regardless of the direction with three axes, the sensor output is used as it is. Is not necessary.

  Next, in step S203, the first deterioration diagnosis unit 11 performs logarithmic processing (decibelization) on the inclination information after correcting the error in the mounting direction of the sensor head 20. For measurement of minute inclination information, for example, if an attempt is made to obtain a probability density distribution, which will be described later, based on a resolution of 1/1 million [G] units, the storage area required for this is enormous. Become. Therefore, the first deterioration diagnosis unit 11 of the first embodiment performs efficient quantization by logarithmizing the acceleration (decibelization) and obtaining the probability density distribution by classification based on the decibel data. Yes.

  Next, in step S <b> 204, the first deterioration diagnosis unit 11 performs a probability density distribution generation process in order to perform deterioration diagnosis based on inclination information. For example, the first deterioration diagnosis unit 11 creates a cumulative frequency distribution in units of 0.5 decibels using the slope information decibeled in the previous step S203, and performs a local smoothing process on the cumulative frequency distribution. Generate a probability density distribution. As the smoothing process, for example, a Gaussian function can be applied.

  The probability density distribution as the initial learning data at the time of installation is generated using, for example, data for about one month. On the other hand, the number of data of the probability density distribution created as current data at the time of deterioration diagnosis is generated based on data of the past 3 hours to 24 hours, for example.

  Further, it is not always necessary to continuously use the data acquired at the time of installation as the learning data at the normal time serving as the reference data. Since deterioration diagnosis is performed over a long period of time, it is possible to continuously update learning data in the process of performing deterioration diagnosis. And the probability density distribution adopted as continuous learning data at the time of deterioration monitoring is generated based on data of the past several hours to several weeks, for example.

  FIG. 4 is a diagram showing a specific example of the probability density distribution regarding the inclination information for three axes in the first embodiment of the present invention. The one-dimensional probability density distribution of each axis is expressed as a distribution that spreads from the midpoint to both sides as the positive slope and the negative slope increase with respect to the midpoint without slope.

  Next, in step S205, the first deterioration diagnosis unit 11 calculates the probability density distribution of each axis obtained during learning (corresponding to the reference data) from the probability density distribution of each axis obtained during deterioration diagnosis (corresponding to the data during diagnosis). And the probability distribution distance data is generated for each axis. As reference data, learning data updated by continuous learning data can be adopted in addition to the data at the time of initial setting.

  FIG. 5 is a diagram showing the distance of the probability distribution in the first embodiment of the present invention. More specifically, it is an explanatory diagram showing the distance between probability density distributions related to a certain uniaxial inclination information, the probability density distribution obtained at the time of deterioration diagnosis (corresponding to the current probability density distribution), and the initial setting or learning An example of the distance between the two distributions of the probability density distribution (corresponding to the probability density distribution in the normal state) sometimes obtained is shown.

  As shown in FIG. 5, the first deterioration diagnosis unit 11 obtains a distance between probability density distributions related to the inclination calculated for each axis, and uses it as an evaluation value for deterioration diagnosis. As typical examples of the calculation of the inter-distribution distance, there are the following three examples.

(1-1) Cullback Library Distance KL (p‖p ') when the current probability density distribution is p (x) and the probability density distribution in the normal state is p ′ (x) is It can be calculated using equation (1).

(1-2) Pearson distance Pearson distance PE (p‖p ′) when p (x) is a current probability density distribution and p ′ (x) is a probability density distribution in a normal state is expressed by the following equation (2) Can be used to calculate.

(1-3) L ² distance L ² distance L ² (p‖p ′) when p (x) is the current probability density distribution and p ′ (x) is the probability density distribution in the normal state, It can be calculated using (3).

  In addition, for calculating the distance of the probability distribution, a calculation formula such as f distance or relative Pearson distance can be used.

  Finally, in step S206, the first deterioration diagnosis unit 11 determines that the probability distribution is normal if the probability distribution distance is less than a predetermined deterioration determination reference value set in advance for all three axes. However, if there is an inter-distribution distance that exceeds the reference value, it is determined that degradation has occurred.

  In the deterioration diagnosis, the first deterioration diagnosis unit 11 detects, for example, a case where a distance between distributions exceeding a predetermined reference value set in advance is detected continuously for a predetermined time, or a distance between distributions exceeding the reference value. It can also be determined that deterioration has occurred when it is detected a predetermined number of times or more.

When the above contents are arranged, the following processing is performed when the deterioration diagnosis of the attachment state of the structure is performed using the probability density distribution regarding the inclination information.
[Feature 1] The inclination information is acquired by extracting the low frequency component of the acceleration information.
[Characteristic 2] The sampling rate of the inclination information can be slowed so that the information is not lost. For example, data can be collected at intervals of 5 seconds.
[Characteristic 3] The tilt data whose attachment direction has been corrected is converted to decibels to reduce the storage capacity and perform efficient quantization.
[Feature 4] Based on a predetermined decibel unit, the probability density distribution at the time of learning and at the time of deterioration diagnosis is calculated, and the distance of the probability distribution at the time of three-axis learning and at the time of deterioration diagnosis (inter-distribution distance) is obtained respectively. The evaluation value is used for diagnosis.
[Feature 5] If there is an evaluation value exceeding a predetermined reference value set in advance, it is determined that the inclination of the structure has reached an unacceptable level at the portion where the sensor head 20 is installed, and the deterioration state is determined. Judge quantitatively.

(2) Flow of deterioration diagnosis processing based on natural frequency information by the second deterioration diagnosis unit 12 Next, specific processing contents by the second deterioration diagnosis unit 12 will be described.

  FIG. 6 is a flowchart showing a flow of the deterioration diagnosis process by the second deterioration diagnosis unit 12 in the first embodiment of the present invention. A specific deterioration diagnosis process by the second deterioration diagnosis unit 12 based on the natural frequency information will be described based on the flowchart of FIG.

  First, in step S <b> 601, the second deterioration diagnosis unit 12 extracts frequency information based on acceleration information acquired from the acceleration information output unit 22 in the sensor head 20. Any structure has a natural frequency depending on the material, weight, shape, and support structure. For example, X, Y, and Z axis acceleration information is output from the sensor head 20 at a sampling rate of 50 Hz. However, large vibrations are not always generated. On the other hand, when performing frequency analysis, a high sampling rate is required, and a huge storage area is required to obtain data at all times.

Therefore, the second deterioration diagnosis unit 12 according to the first embodiment, when acceleration exceeding a predetermined reference level set in advance is obtained in at least one of the acquired three-axis acceleration information. , A constant of ²ⁿ (for example, 512) is acquired from the past to the future.

  FIG. 7 is a diagram showing a specific example related to collection of frequency information for three axes in the first embodiment of the present invention. In FIG. 7, when acceleration data measured on the X-axis exceeds a reference level, a certain number of data consisting of past data and future data is stored for each of the three axes, using that time as an event trigger. Is illustrated.

  Next, the second degradation diagnosis unit 12 corrects the mounting direction of the sensor head 20 in step S602. Specifically, the second deterioration diagnosis unit 12 performs a matrix of calibration data (affine transformation matrix elements) in the mounting direction of the sensor head 20 and a certain number of data obtained in the previous step S601. By calculating, the error in the mounting direction is corrected so that the coordinate axes of the respective sensor heads 20 coincide with each other.

Next, in step S603, the second deterioration diagnosis unit 12 calculates a power spectrum by FFT processing for the error-corrected three-axis data, and obtains a natural frequency. FIG. 8 is a diagram showing calculation results of power spectra for three axes in the first embodiment of the present invention. The acceleration data includes a direct current component (slope component). Therefore, the second deterioration diagnosis unit 12 removes the low-frequency component from the power spectrum, the maximum maximum value (corresponding to f _1x , f _1y , and f _1z in FIG. 8) from the remaining spectrum, and the second (Corresponding to f _2x , f _2y , and f _2z in FIG. 8) are obtained, and these are set as the first natural frequency and the second natural frequency of each axis, respectively.

  Next, in step S604, the second deterioration diagnosis unit 12 creates a cumulative frequency distribution for each of the first natural frequency and the second natural frequency, and performs a local smoothing process on the cumulative frequency distribution. Individual probability density distributions for the first natural frequency and the second natural frequency are generated. Specifically, the second deterioration diagnosis unit 12 is defined by, for example, the first natural frequency and the second natural frequency obtained in the previous step S603 for each event trigger in the previous step S601. Each cumulative frequency distribution is created, and a local smoothing process is performed on each of the cumulative frequency distributions, thereby generating individual probability density distributions related to the first natural frequency and the second natural frequency. As the smoothing process, for example, a Gaussian function can be applied.

  The probability density distribution as the initial learning data at the time of installation is generated using, for example, data for about one month. On the other hand, the number of data of the probability density distribution created as current data at the time of deterioration diagnosis is generated based on data of the past 3 hours to 24 hours, for example.

  Further, it is not always necessary to continuously use the data acquired at the time of installation as the learning data at the normal time serving as the reference data. Since deterioration diagnosis is performed over a long period of time, it is possible to continuously update learning data in the process of performing deterioration diagnosis. The probability density distribution adopted as continuous learning data during deterioration monitoring is generated based on data of the past several hours to several weeks, for example.

  Next, in step S605, the second deterioration diagnosis unit 12 obtains individual probability density distributions (corresponding to diagnosis time data) regarding the first natural frequency and the second natural frequency of each axis obtained at the time of deterioration diagnosis, The distance between the probability distributions of the individual probability density distributions (corresponding to the reference data) relating to the first natural frequency and the second natural frequency of each axis obtained during learning is calculated, and the first natural frequency is calculated for each axis. And inter-distribution distance data relating to the second natural frequency. As reference data, learning data updated by continuous learning data can be adopted in addition to the data at the time of initial setting.

  The distance between the probability density distributions related to the natural frequency information is the same concept as the distance between the probability density distributions related to the slope information shown in FIG. Therefore, more specifically, the second deterioration diagnosis unit 12 uses the above equations (1) to (3) in the same manner as the first deterioration diagnosis unit 11, so that the first natural frequency and the second Individual inter-distribution distance data relating to the natural frequency can be calculated.

  Finally, in step S606, the second deterioration diagnosis unit 12 determines that the distance between distributions is normal if the distance between distributions is less than a predetermined deterioration determination reference value set in advance for all three axes. If there is an inter-distribution distance that exceeds the reference value, it is determined that deterioration has occurred.

  In the deterioration diagnosis, for example, the second deterioration diagnosis unit 12 detects, for example, a case where a distance between distributions exceeding a predetermined reference value set in advance is detected continuously for a predetermined time or a distance between distributions exceeding the reference value. It can also be determined that deterioration has occurred when it is detected a predetermined number of times or more. In addition, the probability density distribution used in the deterioration diagnosis may use both the first natural frequency and the second natural frequency, or may use one of the distributions.

When the above contents are arranged, the following processing is performed when the deterioration diagnosis of the mounting state of the structure is performed using the probability density distribution regarding the natural frequency information.
[Feature 1] With respect to acceleration information, natural frequency information is acquired by extracting a certain number of acceleration data including the point in time when acceleration data exceeding a reference level is obtained on any axis.
[Feature 2] Since the natural frequency information is used for subsequent frequency analysis, it is not appropriate to downsample the output rate from the sensor head 20. Therefore, when the acceleration data that exceeds the reference level in any axis is obtained as an event trigger, a certain number of acceleration data consisting of past data and future data is extracted and the natural frequency information is stored. The storage capacity is reduced.
[Characteristic 3] The power spectrum is calculated by performing frequency analysis on the natural frequency information in which the mounting direction is corrected, and the maximum maximum value and the second value are calculated from the remaining spectrum obtained by removing the low frequency component from the power spectrum. Is obtained, the first natural frequency and the second natural frequency of each axis are calculated.
[Feature 4] From the cumulative frequency distribution of each axis created for each of the first natural frequency and the second natural frequency calculated for each event trigger, the first natural frequency and the second natural frequency at the time of learning and deterioration diagnosis are calculated. An individual probability density distribution relating to the natural frequency is calculated, and further, a distance between the distributions at the time of learning and deterioration diagnosis for each of the three axes is obtained and used as an evaluation value for deterioration diagnosis.
[Feature 5] If there is an evaluation value exceeding a predetermined reference value set in advance, it is determined that the vibration of the structure has reached an unacceptable level at the portion where the sensor head 20 is installed, and the deterioration state is determined. Judge quantitatively.

  As described above, according to the first embodiment, the first feature value related to the inclination and the second feature value related to the natural frequency are extracted based on the acceleration information obtained from the structure to be diagnosed for deterioration. Yes. For each feature amount, the distance between the distributions is calculated by comparing the probability density distribution during learning corresponding to the reference data at normal time and the probability density distribution based on the measurement results during the deterioration diagnosis, and a significant difference is detected. If it is, it is determined that deterioration has occurred. As a result, the deterioration diagnosis of the mounting state of the structure can be performed quantitatively over a long period of time.

  In the above-described first embodiment, the deterioration diagnosis based on the first feature value related to the inclination and the deterioration diagnosis based on the second feature value related to the natural frequency have been described. It is possible to carry out the deterioration diagnosis quantitatively over a long period of time by performing only one.

  The deterioration diagnosis according to the first embodiment described above can be performed even when one sensor head is used as a minimum configuration. However, even if no deterioration has occurred at the location where the sensor head is installed, there is a possibility that the deterioration has occurred at other locations. Therefore, it is possible to expect improvement in detection accuracy by installing sensor heads at a plurality of locations and determining that the attachment deterioration of the structure has occurred when a deterioration state is detected by any of the sensor heads.

  Further, when the data of the two sensor heads can be used, in addition to the individual diagnosis results, the inclination is the difference between the inclinations of the two sensor heads, and the vibration frequency is the phase difference between the vibrations of the two sensor heads. As described above, deterioration diagnosis based on the probability density distribution can be performed, and further improvement in detection accuracy can be expected. In addition, regarding the vibration frequency, not only the phase difference but also the deterioration diagnosis can be performed similarly as the amplitude difference.

  Moreover, in Embodiment 1 mentioned above, the suspended structure in a tunnel or the overhang | projection structure was mentioned as an example of a degradation diagnosis object, However, This invention is not limited to this. If the structure is attached to the fixed surface and the attachment state may change over time, the deterioration diagnosis can be performed quantitatively over a long period of time. Further, by attaching a sensor head to an existing structure, it is possible to quantitatively diagnose a secular change in the mounting state of the structure after the installation of the sensor head.

  In the above-described first embodiment, the case of learning at the initial stage and the case of updating by learning at the time of deterioration diagnosis have been described as normal reference data. However, the present invention is not limited to this. Probability density distribution for judging normality does not need to be unique. For example, individual reference data is provided for each time zone, or separate reference data is provided for each event occurring in a structure. You can also. Furthermore, considering the secular change, it is also possible to use the probability density distribution at the initial stage and a new probability density distribution obtained by learning at the time of deterioration diagnosis.

Embodiment 2. FIG.
In the first embodiment, the case where the deterioration diagnosis is performed focusing on the distance between the probability density distributions has been described. On the other hand, in the second embodiment, as a method different from the abnormality determination based on the distance between the probability density distributions, degradation determination is performed on a region where no distribution exists based on the probability density distribution of each axis obtained during learning. A method for determining that deterioration has occurred when there is data included in the deterioration determination area in the probability density distribution of each axis determined at the time of deterioration diagnosis will be described.

  This method is common to the probability density distribution related to the slope information and the probability density distribution related to the natural frequency information, and will be described as a deterioration diagnosis based on the “probability density distribution” common to both.

  FIG. 9 is an explanatory diagram of a deterioration diagnosis method based on the deterioration determination region in the second embodiment of the present invention. Here, the probability density distribution shown in FIG. 9 shows the probability density distribution in the normal state, and the slope obtained by steps S201 to S204 in FIG. 2 in the first embodiment at the time of initial setting or learning. This corresponds to the probability density distribution related to information or the probability density distribution related to the natural frequency information obtained in steps S601 to S604 in FIG.

  As shown in FIG. 9, in the second embodiment, the “degradation determination region” does not occur in the normal state based on the probability density distribution in the normal state (or occurs but does not easily occur). ) It is set as an area of probability density distribution.

  FIG. 10 is a diagram for explaining a comparison between the deterioration diagnosis based on the distance between the established density distributions in the first embodiment and the deterioration diagnosis based on the area where the probability density distribution does not exist in the second embodiment. FIG. 10A illustrates a case where the probability density distribution at the time of initial setting or learning and the probability density distribution at the time of deterioration diagnosis are different from each other, but the distribution ranges are the same. On the other hand, FIG. 10B shows a probability density distribution at the time of deterioration diagnosis that is specific to the left end (range 0 to 5 in FIG. 10B) as compared to the probability density distribution at the time of initial setting or learning. The case where distribution is included is illustrated.

In the case of the distribution as shown in FIG. 10A, the deterioration cannot be diagnosed by the “deterioration determination region” as in the second embodiment. However, as in the first embodiment, it is possible to diagnose deterioration based on the “inter-distribution distance”. Indeed, FIG. 10 L ² distance determining the 0.0284 next relative distribution (a), the by appropriately setting the threshold value, it is possible deterioration diagnosis based on "inter-distribution distance."

On the other hand, in the case of the distribution as shown in FIG. 10B, since the two distributions are almost the same, it becomes difficult to diagnose the degradation by the “inter-distribution distance” as in the first embodiment. . Indeed, finding the 0.00002 next the L ² distance to the distribution of FIG. 10 (b), the degradation diagnosis based on the "inter-distribution distance" is found to be difficult. However, the deterioration can be diagnosed by the “deterioration determination region” as in the second embodiment.

  In the deterioration diagnosis, the first deterioration diagnosis unit 11 or the second deterioration diagnosis unit 12 may, for example, detect a distribution existing in the “deterioration determination region” continuously for a predetermined time, It can also be determined that the degradation has occurred when the distribution existing in the “region” is detected a predetermined number of times or more.

  As described above, according to the second embodiment, the deterioration diagnosis by the “degradation determination region” is performed even for the probability density distribution that is determined not to be deteriorated by the “inter-distribution distance”. It is possible to accurately detect deterioration when a random probability distribution occurs.

  That is, according to the second embodiment, based on the acceleration information obtained from the deterioration diagnosis target structure, the first feature value related to the inclination and the second feature value related to the natural frequency are extracted, and the respective features are extracted. With respect to the quantity, it is determined that deterioration has occurred when the feature quantity is in an area where it cannot be taken from the probability density distribution during learning corresponding to the reference data at normal time. As a result, the deterioration diagnosis of the attachment state of the structure can be diagnosed at an early stage.

  In other words, it takes time to determine the probability density distribution, but as shown in FIG. 9, the region showing a high probability density has a high probability of inclination information or natural frequency from the probability density distribution during learning. The value which information can take is shown, and the area | region which shows a low probability density becomes a value which is hard to take. Accordingly, when the probability density distribution in the normal state (during learning) is set as a normal determination region as a normal value range, a value that cannot be obtained if normal is set as a deterioration determination region, and the slope information is in the deterioration determination region. It is determined that deterioration has occurred. This determination method is simpler than the determination based on the calculation of the inter-distribution distance according to the first embodiment, from the inclination information acquisition or the natural frequency information acquisition to the determination, and detects the deterioration immediately after the occurrence. Is possible.

  10A and 10B, the deterioration diagnosis based on the distance between the established density distributions in the first embodiment and the deterioration diagnosis based on the area where the probability density distribution does not exist in the second embodiment are used in combination. For any probability density distribution of (b), the abnormal state can be determined.

  Further, as a method other than Embodiments 1 and 2, deterioration diagnosis can be performed by obtaining a probability density ratio that is a ratio of a probability density distribution during learning and a probability density distribution during deterioration diagnosis. More specifically, the vector length is obtained as the Euclidean distance of the three-axis probability density ratio, the vector length obtained from the data at the time of learning is used as a reference value, and the vector length obtained from the data at the time of deterioration diagnosis exceeds the reference value. Therefore, it can be determined that deterioration has occurred.

In the method according to the first and second embodiments, the first deterioration diagnosis unit 11 performs logarithmic processing (decibelization) on the inclination information after correcting the error in the mounting direction of the sensor head 20 in step S203. Although it was performed, it is not always necessary to logarithmize.
In addition, the natural frequency information is obtained by extracting a certain number of acceleration data including the point in time when acceleration data exceeding the reference level is obtained on any axis. May be extracted to obtain the natural frequency information.

Embodiment 3 FIG.
The first and second embodiments described above are apparatuses for diagnosing long-term deterioration of a structure using the sensor unit 21 that is a three-axis acceleration sensor. The apparatus measures the inclination from the static acceleration and the natural vibration from the dynamic acceleration. The deterioration of the structure is diagnosed by measuring the number and detecting the change over time. An advantage of using a three-axis acceleration sensor is that a change in inclination can be detected regardless of the posture at the time of installation. The direction of gravity can be known by measuring the static acceleration (DC component) applied to the acceleration sensor. The acceleration information obtained from the triaxial acceleration sensor has three components of three axes, and by appropriately setting α of the so-called affine transformation matrix, the sensor output with the gravitational direction as the Z axis is always set regardless of the posture. Can be obtained.

  If long-term deterioration diagnosis is continued using such a triaxial acceleration sensor, the life of the device will eventually reach its end, but if the triaxial acceleration sensor has a shorter lifetime than the structure, the old triaxial acceleration It is necessary to match the measurement data of the sensor (old sensor) with the measurement data of the three-axis acceleration sensor (new sensor) newly installed by replacement.

  At this time, when the old sensor and the new sensor are mounted adjacent to each other in a random posture, even if the Z-axis direction can be matched only by applying gravitational acceleration, the other axis directions can be matched. Therefore, it is necessary to apply acceleration from other directions. That is, the positioning of the two sensors can be accurately performed under conditions in which the application of gravitational acceleration and the application of periodic vibration are measured, for example, when the jet fan is attached.

  However, when the three-axis acceleration sensor is installed in a structure to which a clear periodic vibration is not applied, the two sensor postures cannot be matched only by external acceleration information. Calibration of the postures of the two sensors in such an environment can be performed by providing a constraint that the old sensor and the new sensor are located on the same plane.

  Therefore, in the third embodiment, by using a base on which the old sensor and the new sensor can be installed on the same plane, the old sensor and the new sensor are attached to only one at the time of monitoring and both at the time of calibration. The gravitational acceleration direction can be measured and the error can be corrected.

  DESCRIPTION OF SYMBOLS 10 Sensor controller, 11 1st degradation diagnostic part, 12 2nd degradation diagnostic part, 20 Sensor head, 21 Sensor part, 22 Acceleration information output part

Claims (2)

A sensor controller for diagnosing whether or not the mounting state of the structure is normal based on acceleration information of the structure output from a sensor head installed on a structure that is attached to a fixed surface and is subject to deterioration diagnosis A structural deterioration diagnosis system equipped with
Feature amount extraction means for extracting at least one of the tilt information and the natural frequency information of the structure as the feature amount based on the acceleration information acquired from the sensor head;
A probability density distribution calculating means for calculating a reference probability density distribution in a normal state and a probability density distribution for diagnosis at the time of deterioration diagnosis with respect to the feature quantity extracted by the feature quantity extracting means;
A normal determination region and a deterioration determination region are set from the reference probability density distribution calculated by the probability density distribution calculating unit, and the diagnosis probability density distribution calculated by the probability density distribution calculating unit at the time of deterioration diagnosis When there is a distribution included in the deterioration determination area, the structure deterioration diagnosis system includes determination means for determining that an abnormality has occurred in the attachment state of the structure. In the structure deterioration diagnosis system according to claim 1,
From the sensor head, triaxial acceleration information is output,
The feature amount extraction unit extracts at least one information of inclination information for three axes and natural frequency information for three axes.

Images