JP2020024184A - Structural inside deformation character detection device - Google Patents

Abstract

To solve the problem that a method for creating general teacher data cannot be applied because an inside deformation characteristic cannot clearly be regulated due to the difficulty of externally observing or measuring the inside deformation characteristic by actual measurement in applying a method for detecting the inside deformation characteristic existing in a structural inside by using an impulse applied to a structure and a response of the structure to the impulse.SOLUTION: A part or all of teacher data is set by numerical analysis and experiment about a structure model including an inside deformation state assumed to have a high possibility of existence in an object structure, a part considered to have a large relevance to a surface response corresponding to an inside deformation state assumed to have a high possibility of existence in the object structure about at least a part of the surface response obtained in actually measuring the object structure is extracted, and the measurement data of the parts is added to the teacher data.SELECTED DRAWING: Figure 4

Description

  The present invention generally uses not only the presence or absence of internal deformation existing inside a structure such as concrete (hereinafter referred to as a structure) but also the internal deformation characteristics such as its position, shape, range, type, and physical properties. The present invention relates to an apparatus for detecting based on a relationship between a stimulus acting on a structure and a response to the stimulus. More specifically, the present invention relates to an apparatus and a method having a feature in generating teacher data on a relationship between an internal deformation characteristic of a structure subjected to a specific stimulus and a response.

  Non-destructive detection of the presence, location, range, type, etc. of internal deformations that cannot be observed from the outside because they are inside the structure and there is no means to easily grasp their physical characteristics In order to do so, it is common practice to analyze the response obtained artificially by applying a physical stimulus from outside the structure, or obtained under the stimulus of natural physical action. The simplest stimulus is due to solar heating, and more advanced stimuli include natural tremor, impact, vibration, artificial heating, ultrasonic waves, and X-ray transmission.

  For example, if the target is a relatively small structure such as a human body or a compact article, the internal situation can be measured with high accuracy by measuring in a well-equipped enclosed space using X-rays or ultrasonic waves. Since the detailed observation can be performed, the detection of the internal deformed characteristic is relatively easy. However, at the other end of the spectrum, when targeting large structures that constitute social capital in the natural environment, it is usually not possible to directly observe the interior, and the structure surface is exposed to relatively simple stimuli. It is only practical to measure only the response at Although humans can recognize the degree of internal deformation only by looking at such surface response data, it is practically impossible to identify the characteristics of internal deformation.

  Therefore, a means for classifying and specifying the internal deformation characteristic from the obtained surface response through some logic is required. As one of such means, it is conceivable to obtain the internal deformation characteristics from the surface response by computer inverse analysis. At present, detailed surface response data and advanced and sophisticated analysis methods are required for the inverse analysis. Necessary, if the target is large and large, it is not entirely feasible from a cost-effective point of view.

  At present, technology such as autonomous driving, person authentication, fingerprint authentication, character reading, or voice authentication, which recognizes and / or identifies an object from obtained image data, wave data, etc. DL). It is conceivable to use such a DL as a means for associating the surface response data of the structure obtained for a specific stimulus with the characteristics of the internal deformation. If this detection method is used, it is not possible to visualize the internal deformation in detail using X-rays or the like, and to determine the internal deformation with specific numerical values by inverse analysis, but the maintenance target is large, and at present the cost is low. From the viewpoint of effectiveness, it is thought that it is possible to improve the detection of internal deformation of large structures of general social capital that can only be dealt with by visual inspection and tapping sound inspection. However, in order to use DL for recognizing and identifying internal deformation of such a large structure of social capital, it is necessary to solve some problems described below.

  FIG. 1 shows a comparison between a DL for general image recognition / identification and a DL for detecting a deformation characteristic inside a structure to which the present invention is applied, using image recognition / identification as an example. In the left half of the figure, and the latter in the right half. As shown in the left half of FIG. 1, in general image recognition / identification, the teacher data is usually composed of image data that enables a human to visually identify a target. In other words, the recognition / identification target included in the image of the teacher data, for example, a person, an animal, a car, a structure, a character, and the like, even if the individual forms are extremely diverse, each has a fixed form characteristic. And the feature of the object can be directly obtained visually. Moreover, such image data can be collected, stored, processed, etc. in a large amount simply and easily with the advance of related technology. Therefore, even if a part of the object is hidden in the image or the appearance changes variously due to different viewing angles, the morphological features and appearance of each object and the appearance and the object are paired. Basic information or method for associating in step 1 is stored. Even if it is not something that can be directly seen in this way, for example, if it is intended for organs inside the human body, it can be indirectly recognized by images obtained using X-rays and ultrasonic waves, In addition, since a large amount of anatomical information accumulated over many years is present in the background, the creation of teacher data is not significantly different from the case of general image recognition and identification. Even in large-scale structures of social capital, high-precision recognition and identification of phenomena on the surface itself, such as cracks on the surface of concrete structures and rust on the surface of steel structures, are required. It is considered that the technique can be relatively easily applied by applying an image obtained by a simple CCD camera or the like and a general image authentication / identification technique.

  Also, in the case of a DL that uses wave / vibration data such as voice, in a general DL, it is possible to associate the wave / vibration data with the characteristics of the source of the data. The same can be said.

  However, in the work of recognizing and identifying internal deformations from surface responses, such as when targeting large structures that constitute social capital, as described in the background section, humans view response data. Even if it is possible to recognize the degree of internal deformation by itself, it is practically impossible to identify even internal deformation characteristics. More specifically, as shown in the right half of FIG. 1, stimuli generally given to a large structure include natural heating, impact, vibration, artificial heating, ultrasonic waves, and the like. The wave distribution, displacement distribution, temperature distribution, etc. at a certain point in time or over time, which appeared at each point on the surface of the structure. Not only can the local presence of outliers related to internal transformations be detected in relatively stable data in relatively large temporal or spatial domains, but further inferences about the details of internal transformations Is virtually impossible.

  Therefore, when creating teacher data about the relationship between the response and the internal metamorphic property of a structure that has received a specific stimulus, if the response data only exists, the teacher data that associates the corresponding internal metamorphic property with that data Can not be created.

  It is also possible to infer from the teacher data of the similar structure in the past, if it is not the teacher data of the target structure itself, but the time is short due to the internal transformation of the large structure of social capital. Thus, there is virtually no measurement data relating the response to the internal deformation properties of a structure that has received a particular stimulus. Even if they exist, the physical and spatial characteristics of large structures built in the natural environment, as described above, are uneven due to the effects of materials used, construction methods, construction conditions, etc. Yes, furthermore, when such a structure is affected by the natural environment over time, causing mechanical, chemical or other causes to cause damage, damage, deformation, chemical change, corrosion, etc. over time, Since the diversity of deformations is multiplied by the loop, it is considered that at present it is almost impossible to divert between structures that are originally separate. In the future, in order to enrich the teacher data corresponding to such diversity, even if a large amount of measurement data on the response to some stimulus is obtained in a large structure, it may be in service, and in some cases even in the soil. Since it is practically impossible to cut out the location of the internal deformation from a large structure that may not be directly accessible because it is buried in the building, etc., it is practically impossible, teacher data on the characteristics of the internal deformation Is still incomplete. As a countermeasure, it is conceivable to collect teacher data by experiments using test specimens that can clearly set the characteristics of internal deformation and stimulus.However, due to cost, the number of characteristics of internal deformation that can be set is Compared to the diversity in structure, it is only a small thing, and it seems that even if it can be used supplementarily, it can not often be a fundamental solution.

  As described above, in the application of the DL to the recognition and identification of the internal deformation of a large structure targeted by the present invention, the generation of teacher data as in the recognition and identification of general images or wave / vibration data is performed. Is practically impossible, so that it is necessary to devise a method for creating new teacher data.

  At present, it is practically impossible to physically relate the relationship between the response when a structure having various internal deformations is stimulated and the internal deformations relatively physically to the above problem. Since it is easy to perform the analysis by, for example, finite element analysis, it is conceivable to create a part or all of the basic teacher data by analysis. That is, a physical model is set for (a) the structural characteristics of the structure to be detected, (b) the characteristics of the internal deformation of the target, (c) the stimulus given to the structure, and (d) the structure to the stimulus. Is analytically obtained. The analysis at this time is not a so-called inverse analysis of identifying the characteristics of internal deformation from the response as described above, but a forward analysis of obtaining a response from the physical model. It is widely used in research and design practice. Therefore, it is considered that the individual teacher data thus obtained almost guarantees practical physical consistency except for special cases. In the future, if the analysis technology advances further, it will be possible not only to improve the analysis accuracy but also to bring the teacher data closer to the one that reflects the actual diversity, such as by increasing the types of parameters and expanding the range of their fluctuation. It is possible.

However, as shown in the following embodiment, the number of parameters constituting the analysis model is large, and at present, there is a practical limit in setting each parameter value in consideration of actual variation. I do. Therefore, here, the following method is used as a method of creating teacher data that can cope with the diversity of the actual target structure and internal deformation while aiming for efficiency and rationalization by reducing the number of analysis cases as much as possible. It is also suggested to add more.
(a) From the measurement data on the response obtained when the target structure was actually measured, extract the parts that are considered to be highly relevant to the internal deformation that is likely to exist and measure them The part of the data is added to the teacher data in relation to the internal transformation. Thus, it is considered that data reflecting the relationship between the actual internal deformation occurring in the target structure and the response can be added to the teacher data.
(b) Perform at least a part of the internal deformation characteristic investigation on the target structure corresponding to the measurement data part extracted in (a) in more detail than other measurement points By doing so, the relationship between the characteristics of the actual internal deformation and its response is more accurately reflected on the teacher data, and the reliability of the teacher data is improved. It is conceivable to improve the accuracy of the DL by using the teacher data created in this way.

  Note that in a large structure that constitutes social capital, the measurable response is often limited to the response on the surface of the structure, and thus, in this specification, an example using only the surface response is shown. Is not limited to the surface response.

5 is a schematic diagram comparing a general DL with a DL for a structure targeted by the present invention. 1 is a schematic diagram illustrating a configuration of an apparatus according to an embodiment of the present invention. 5 is a schematic diagram schematically showing the outline of the growth over time of two internal deformations existing inside the structure and the relationship between the respective surface responses appearing on the measurement surface corresponding thereto. FIG. 3 is a schematic diagram showing data processing contents in each means constituting the apparatus shown in FIG. 2 and a flow of data between the means. 4 is a schematic diagram illustrating a flowchart of a method for detecting a characteristic of an internal metamorphosis from a response, according to an embodiment of the present invention. 1 shows an example of a crack analysis model assuming a case where a single crack exists in a structure. FIG. 7 is a perspective view showing a typical surface temperature distribution on the upper surface in the crack analysis model shown in FIG. 6. In the figure, (a) shows the case of front side heating in which the upper surface is heated, and (b) shows the case of back side heating in which the lower surface is heated. FIG. 6 shows respective temperature distributions on the vertical bisector of the crack opening in three cases where the inclination angle δ is 30 °, 45 °, and 60 ° in the crack analysis model shown in FIG. 6. In the figure, (a) shows the case of front side heating in which the upper surface is heated, and (b) shows the case of back side heating in which the lower surface is heated. 4 shows an example of a buoyancy analysis model in which a buoy exists inside a structure. FIG. 10 is a perspective view showing a typical surface temperature distribution on the upper surface in the floating analysis model shown in FIG. 9. In the figure, (a) shows the case of front side heating in which the upper surface is heated, and (b) shows the case of back side heating in which the lower surface is heated. FIG. 10 shows the temperature distribution on the center line of the float in three cases where the depth e is 10 mm, 30 mm, and 50 mm in the float analysis model shown in FIG. 9. In the figure, (a) shows the case of front side heating in which the upper surface is heated, and (b) shows the case of back side heating in which the lower surface is heated. 7 shows a surface temperature distribution to which noise is added when the inclination angle δ is 45 ° in the crack analysis model shown in FIG. 6. In the figure, (a) shows the case of front side heating in which the upper surface is heated, and (b) shows the case of back side heating in which the lower surface is heated. 10 shows a surface temperature distribution to which noise is added when the depth e is 30 mm in the floating analysis model shown in FIG. 9. In the figure, (a) shows the case of front side heating in which the upper surface is heated, and (b) shows the case of back side heating in which the lower surface is heated. The results of applying the present detection method to the surface temperature distribution in the case of front side heating at δ = 45 ° and no noise among the surface temperatures of the crack model are shown. For reference, the crack analysis model is shown on the left. The results of applying the present detection method to the surface temperature distribution when the backside heating is performed at δ = 45 ° and no noise is included among the surface temperatures of the crack model are shown. The results of applying the present detection method to the surface temperature distribution when the surface temperature is δ = 45 ° in the case of front side heating and noise is included among the surface temperatures of the crack model are shown. For reference, the crack analysis model is shown on the left. The results of applying the present detection method to the surface temperature distribution when noise is included in the case of backside heating at δ = 45 ° among the surface temperatures of the crack model are shown. 9 shows a positive sine wave-related component and a positive cosine wave-related component of the detection result of front side heating in the case of a multiple crack analysis model having four different inclination angles. 5 shows an orthogonal crack analysis model in which two cracks at δ = 45 ° are orthogonal. FIG. 20 shows a positive sine wave-related component and a positive cosine wave-related component of the detection result of front side heating in the case of the orthogonal crack analysis model shown in FIG. 19. FIG. 10 shows detection results when noise is included in the case of front side heating for the floating model shown in FIG. 9. FIG. 10 shows a detection result in the case of backside heating with respect to the floating model shown in FIG. 9 when noise is included. 1 shows a schematic view of a cracked test piece provided with a simulated crack. FIG. 24 shows a surface temperature distribution on the front side surface when heated by solar radiation on a sunny day, with the front side surface provided with the simulated crack of the crack specimen shown in FIG. In the figure, (a) is an RGB image of the surface temperature distribution itself, and (b) is an image showing a result obtained by dividing the average temperature from the surface temperature distribution by a maximum absolute value. 25 is a detection result of a surface temperature distribution forming an RGB image shown in FIG. FIG. 24B shows a detection result after noise removal has been performed on the surface temperature distribution forming the RGB image in advance. It is the RGB image and the situation photograph which displayed the surface temperature distribution under the floor slab when the lower surface side of the RC slab of the actual structure bridge was measured with an infrared imaging device on a sunny day in winter. 28 shows the result of applying the present detection method to the surface temperature distribution of the upper region R1 surrounded by a rectangular solid line in the RGB image shown in FIG. FIG. 28 illustrates a result of applying the present detection method to the surface temperature distribution of the intermediate region R2 surrounded by the rectangular solid line in the RGB image illustrated in FIG. 28 shows a result of applying the present detection method to the surface temperature distribution of the lower region R3 surrounded by a rectangular solid line in the RGB image shown in FIG. It is explanatory drawing of the method of hitting an edge part and detecting an internal deformation in the rod-shaped structure containing an internal deformation. 32 shows an example of acceleration data actually measured by an acceleration sensor at an end of a rod-shaped structure having a real structure similar to the rod-shaped structure of FIG. 31 when the end is hit. 31 shows a result obtained by analyzing acceleration data when vibration is propagated through the rod-shaped structure and returned to the upper end again when the upper end of the rod-shaped structure is hit with respect to the rod-shaped structure in FIG. 31. (A) is a case where the internal deformation is not included, and (b) is a case where the internal deformation is provided at the center of the rod-shaped structure. 33 shows a result obtained by applying noise removal by wavelet degeneration using multi-resolution analysis using orthogonal wavelets to the actually measured acceleration data shown in FIG. 32. FIG. 26 shows a schematic position of a crack occurrence point detected by using the DL method from the measured surface temperature distribution data of the test specimen shown in FIG. 24. FIG. 28 shows a schematic position of a crack occurrence point detected by using the DL method from surface temperature distribution data actually measured on the lower surface of the RC slab shown in FIG. 27.

1 Structural internal deformation detecting device according to the present invention
1.1 Configuration of Device FIG. 2 shows an example of a specific configuration of a device 100 for implementing the present invention, which detects internal deformation existing on the surface or inside of a structure from a response to a specific stimulus. The arrows in the figure generally represent the flow of data between the means.

2, the apparatus 100 includes an actual measurement unit 101, a structure analysis unit 102, an experiment unit 103, an actual measurement data storage unit 104, a teacher data storage unit 105, a teacher data extraction unit 106, an estimation unit 107, Display means 108. The actual measurement unit 101 is a unit that measures at least the value of the response when the target structure receives a certain stimulus, and is, for example, an infrared imaging device that measures a surface temperature, an acceleration sensor that measures vibration, An acoustic sensor or the like for measuring sound is conceivable. As the structural analysis means 102, for example, general finite element analysis software capable of executing heat transfer analysis, mechanical analysis, acoustic analysis, and the like can be considered. The experimental means 103 may be, for example, one in which an internal deformation of a known specification is installed on a part or the whole of the target structure or on a full-scale or reduced simulation test specimen. As the actual measurement data storage unit 104, a HDD, an SDD, or the like of a computer that stores the actual measurement data acquired by the actual measurement unit 101, the target structure specifications, the measurement conditions, and the like can be considered. The teacher data extracting unit 106 extracts, from at least a part of the measured data stored in the measured data storage unit 104, a portion of the measured data that is evaluated to be strongly related to the response of the internal deformation of the detection target. This will be described in detail later. The teacher data storage unit 105 is a computer for storing the teacher data extracted from at least a part of the actually measured data via the teacher data extraction unit 106, and the teacher data created by the structure analysis unit 102 and the experiment unit 103. HDDs, SDDs, and the like are conceivable. The estimating unit 107 uses the teacher data stored in the teacher data storage unit 105 for each section constituting at least a part of the actually measured data stored in the storage unit 104, and This is a means for recognizing and identifying the association with the response caused by the existence. As the display means 108, a display or the like for displaying the result obtained by the estimating means 107 can be considered.
1.2 Procedure for detecting internal deformation Next, assuming a specific internal deformation and measurement method, the contents of data processing in each means constituting the device shown in Fig. 2 and the flow of data between the means are as follows. This will be described more specifically.

  In general, a point to be noted when assuming an internal deformation to be detected is that the problem becomes more complicated than necessary when various internal deformations are considered from the beginning. For this reason, it is important to start with the detection of internal deformations with a small number of basic characteristics that are likely to have occurred in the target structure based on engineering experience, especially in the early stages of deterioration progression. Efficiency is often high in detecting inspection levels. Another point to be noted is that the internal deformation does not remain in a constant state but changes with time, and in particular, scales up.

  More specifically, for example, FIG. 3A shows a case where two internal deformations 1 and 2 are present in one structural section of a flat plate structure, and a stimulus, for example, FIG. 3 schematically shows a surface response, for example, a surface temperature distribution, which appears on the measurement surface on the lower surface of the flat plate structure in response to each internal deformation when sunshine is given. The left side of FIG. 3A shows a cross section of the structure at an early stage of the progress of deterioration, and the surface response is also independent from each other in response to the presence of two internal deformations at an interval. On the other hand, the right side of FIG. 3A shows a stage where the deterioration has progressed, and the surface responses appearing on the measurement surface interfere with each other in response to the two internal deformations expanding and becoming closer to each other. The relationship between them is becoming more complicated. FIG. 3B is a plan view of a measurement surface having a flat plate structure, and schematically shows a surface response appearing on the measurement surface in response to two internal deformations. The cross section AA shown on the left side of FIG. 3B corresponds to the left side of FIG. 3A, and the cross section BB shown on the right side of FIG. 3B corresponds to the right side of FIG. Thus, in the initial stage when the growth of internal metamorphosis is not progressing much, it is considered that the temperature distribution associated with each internal metamorphosis is almost scattered independently, and in this case, the stimulation in each single internal metamorphosis is considered. It is highly possible that the characteristics such as the location of each internal deformation can be detected only by the relationship between the response and the response. However, as the growth of each internal deformation progresses, the interval between the response distributions related to the internal deformation narrows, and when a part where both interfere or overlap with each other appears, a plurality of internal deformations occur. It may be necessary to consider the effect of the response. However, in the detection of the inspection level at the early stage of the progress of deterioration, it is often sufficient to be able to detect a state in which each internal deformation exists independently.

  Therefore, in the following description, two internal deformations are independently present in one structural section, and a portion where the internal deformation has occurred in the initial stage where the growth of the internal deformation has not progressed much is described. The procedure for segmented detection will be described below. In the detection of internal deformation, the response may be measured at the upper and / or lower surface of the structural compartment when the upper and / or lower surface of the structural compartment is stimulated. Assume that a stimulus is applied and the response is measured on the lower surface. Also, the measured response may be a two-dimensional distribution and / or a temporal change at some point of the structure.

  When it is assumed that the target internal deformations are two independent internal deformations 1 and 2 as shown on the left side of FIG. 3A and the left side of FIG. For example, as shown in the upper left part of FIG. 4, the generation of the teacher data for the internal transformation is performed by using a model structure 301 having model internal transformations 501 and 502 that model the internal transformations 1 and 2, respectively. And 302, model responses 601 and 602 (for example, temperature distribution, vibration, etc.) when model stimuli 401 and 402 (for example, heating, impact, etc.) which model actual stimuli are obtained by the structural analysis means 102. It depends. In this case, although the type of the internal deformation is shown differently in the figure, the same type of internal deformation may be used.

  Data relating the model internal deformation 501 and the model response 601 and the model internal deformation 502 and the model response 602 are stored in the teacher data storage unit 105 together with analysis information such as analysis setting conditions. As described above, in the initial stage where the growth of the internal deformation has not progressed much, each internal deformation is often scattered almost independently. It seems that a relatively simple and independent specification that takes into account the target or average characteristics can be handled. However, if the actual phenomenon cannot be sufficiently dealt with by itself, or if the growth of the internal deformation further progresses, the types of the model internal deformation are sequentially increased, and / or each model internal deformation is configured. It is conceivable to take a stepwise approach of expanding the range of parameter values.

  On the other hand, as shown in the upper right part of FIG. 4, the actual response when the actual structure of the target is actually subjected to a specific actual stimulus is measured by using the actual measurement means 101, and the environment such as weather, temperature, and humidity at that time is measured. In addition, the actual response data is stored in the actual measurement data storage unit 104 together with the measurement information such as the structural dimensions of the target actual structure.

  In addition, optionally, from the actual response data stored in the actual measurement data storage unit 104, a range of a response considered to have a strong correlation with the response related to the assumed internal deformation 1 and the internal deformation 2 is instructed. Data extracted by the data extracting means 106 and data in which the response data in those ranges are associated with the internal abnormalities are additionally stored in the teacher data storage means 105. For example, when it is evaluated that there is a strong relationship between the actual response 1001 and the model response 601 of the model internal deformation 501, the actual internal deformation 901 that caused the actual response 1001 is similar to the model internal deformation 501. Considering that, the actual response 1001 is stored in the teacher data storage unit 105 as a part of the teacher data of the internal transformation 1. If the relationship between the actual response 1002 and the model response 602 is the same, the actual response 1002 is stored in the teacher data storage unit 105 as a part of the teacher data of the internal transformation 2. By this operation, it is considered that the relationship between the real internal deformation existing in the target real structure and the real response related thereto can be reflected in the teacher data. Further, optionally, when the data corresponding to the teacher data is created by the experiment means 103 assuming the actual structure, the actual internal deformation, the actual stimulus, etc., the data is additionally stored in the teacher data storage means 105. Remember.

  Teacher data created based on such structural analysis, actual measurement and experiments is not always obtained in a large amount and as easily as teacher data in general image recognition etc. There is. Therefore, by using a data expansion method such as that used in DL (processing data by rotation, scaling, parallel movement, shearing, blurring, adding noise, etc.), the data amount of the teacher data is increased and the teacher data is diversified. In some cases, it may be effective to add a property.

  Using the teacher data stored in the teacher data storage unit 105 and the measured data stored in the measured data storage unit 104 obtained as described above, learning, evaluation, and the like executed by the estimation unit 107 By recognizing / estimating a part of the actual response which is considered to be related to the assumed internal deformation from the actual response by the above operation, the actual response is displayed on the display means 108 in correspondence with the actual structure. At this time, it is also conceivable to display information about the type of internal deformation that is highly likely to be associated with the range of the actual response, the possibility of its existence, and the like.

  FIG. 5 shows an example of a flowchart describing the main procedure of the above procedure, and each step will be specifically described based on this flowchart. In step 1 (S1), for example, for a real structure whose characteristics such as the position, range, physical properties, and type of internal deformation are to be detected, the surface temperature with respect to a specific real stimulus such as heating, impact, vibration, or ultrasonic waves Are measured as actual response data by actual measurement means. In step 2 (S2), the actual response data is stored in the actual measurement data storage means together with the measurement information. In step 3 (S3), one or more internal deformations to be detected, which are considered to be present in the target structure, are assumed based on engineering experience and judgment, and a model corresponding to the assumed internal deformations An analysis is performed by the structural analysis means for a case where a model stimulus is applied to the model structure including the internal deformation to obtain model response data. In step 4 (S4), the model response data is stored in the teacher data storage means together with the analysis information.

  At this time, an optional step 10 (S10) is additionally performed to extract, from the actual response data, portions that can be evaluated as having a strong relationship with the response corresponding to the assumed internal deformation, and update them. May be additionally stored in the teacher data storage means as teacher data. Further, an optional step 11 (S11) is additionally performed, and the experiment data obtained from the experiment assuming the actual structure, the actual internal deformation, the actual stimulus, and the like as the object are further used as teacher data storage means. May be additionally stored. Further, the optional step 12 (S12) is additionally performed, and the teacher data stored in the teacher data storage unit is processed by the data extension method or other methods, so that the teacher data can be improved in quality and quantity. Good.

  Next, in step 5 (S5), using the teacher data stored in the teacher data storage means obtained as described above and the actual response data stored in the measurement data storage means, by a method such as DL. Evaluate and infer properties such as the position, range, physical properties, and type of internal deformation existing in the structure.

  Finally, in step 6 (S6), the detected characteristic of the internal deformation is displayed on a display means such as a display. At this time, related information about the internal deformation, for example, the position in the actual structure where the internal deformation occurs, the type of the internal deformation that is likely to be associated with the range of the actual response, the possibility of its existence, etc. Information may be displayed.

  What remains as an issue here is a specific extraction method that constitutes the teacher data extraction means 106 and extracts a portion strongly related to the response corresponding to the assumed internal deformation from the actual response of the actually measured data. Is necessary. As described above, the response of a structure having an internal deformation that has undergone some stimulus includes an internal deformation in a relatively small and steady-state response that reflects most healthy general parts. The reflected, non-stationary response part with relatively large fluctuation appears locally. In a conventional signal processing technique, there is a scanning method using a filter effective for extracting an unsteady response from such a steady response. For example, by applying the scanning method using this filter to the measured data of the response of the structure, an unsteady response part similar to the one considered to be related to the assumed internal deformation is extracted, and this is Can be added to The specific contents regarding this will be described later in detail.

  In the above description, the internal deformation is defined as the internal deformation 1 and the internal deformation 2. However, by changing the filter characteristics variously, a new characteristic from among the responses not corresponding to the internal deformation 1 and the internal deformation 2 is obtained. If the part of the response can be extracted, the part can be set as the internal deformation 3. In this case, even if the characteristics of the internal deformation 3 cannot be specifically determined by a clear specification such as that set by an analysis model, the teacher data is regarded as a new characteristic internal deformation of the target structure. It seems that the accuracy of classification and identification can be improved step by step by adding.

As described above, the present invention reflects the various characteristics of internal deformation in the actual structure by using measurement data in the actual actual structure, expanding the data, etc. even when starting from teacher data based on relatively few analysis data. The teacher data can be propagated.
2 Example of how to create teacher data
2.1 Setting of the overall situation Next, in order to specifically explain the method of creating the teacher data, an example in which the internal deformation 1 and the internal deformation 2 exist in the structural section shown in FIG. More specifically, cracks and floats, which are typical internal deformations existing in the concrete plate structure, are selected, and the side of the plate structure on the side where the cracks and floats are present and the opposite side of the plate structure are stimulated by heating. Consider a case in which the surface temperature of the surface on the side where cracks and floating are present, corresponding to heating (hereinafter referred to as front side heating and back side heating), is measured as a response.

Here, a crack is a thin, wide, planar space that exists locally in the structure, and a part of the space intersects with the surface of the structure. It is assumed that the space is a thin and spread planar space that exists locally in the space, and that the space does not intersect the structural surface.
2.2 Creating teacher data by analysis
2.2.1 Surface temperature distribution related to cracks Fig. 6 shows a crack analysis model in which cracks exist in the structure. When a steady-state heat transfer analysis is specifically performed on this model, for example, a vertical a (500 mm), a horizontal b (500 mm), and a thickness b made of a uniform material having a thermal conductivity κ (1.6 W / (m · K)) are used. A flat plate having a height of h (100 mm) has a crack opening having a width of 0.1 mm and a length of c (150 mm) near the center of the upper surface, and the inclination angle formed from the crack opening with the upper surface of the flat plate is δ (45 °). Many conditions need to be set such that cracks reaching the depth e (50 mm) from the upper surface of the flat plate have penetrated. Furthermore, the heat transfer coefficient of the heat transfer boundary of the lower surface or the upper surface is 6 W / m ² K, the ambient temperature T is 22 °, the four side surfaces are the adiabatic boundary surfaces, and the heat flux φ flowing from the upper surface or the lower surface is 100 W / m. ^It is necessary to further set thermal conditions such as ² . FIG. 7 is a perspective view showing a typical surface temperature distribution on the upper surface in each of the front side heating and the back side heating in which the upper surface is the observation surface and the upper surface or the lower surface is heated under such conditions. (a) and 7 (b). By varying the parameters that make up these conditions, various model cases can be considered. For example, for three cases where the inclination angle δ was 30 °, 45 ° and 60 °, those showing the respective temperature distributions on the vertical bisector of the opening of the crack at the time of front side heating or back side heating are shown. Figures 8a and 8b. The shapes of both the front side heating and the back side heating have substantially similar temperature distributions when turned upside down. The shape of these temperature distributions is roughly similar to a sine wave whose origin is the crack opening. The shape of this partial surface temperature distribution is considered to be one of the clues for detecting the location where the crack has occurred.
2.2.2 Surface temperature distribution related to floating Next, the effect of floating on the surface temperature distribution when the flat concrete structure has floating was determined by analysis. FIG. 9 shows the analysis model. In this model, the depth e (from the top surface of a flat plate of vertical a (500 mm), horizontal b (500 mm), and thickness h (100 mm) made of a uniform material having a thermal conductivity κ (1.6 W / (m · K)) At 30 mm), there is a flaky space of 0.1 mm thick with a square of side length c (150 mm) and d (1500 mm). FIGS. 10 (a) and 10 (b) show typical surface temperature distributions on the entire upper surface at the time of heating on the front side and the heating on the back side when the heating conditions are the same as those of the crack model. Also in this case, various model cases can be examined by changing these parameters. For example, FIGS. 11A and 11B show the temperature distribution on the center line of the float in three cases where the depth e is 10 mm, 30 mm, and 50 mm. As for the floating, the shapes of both the front side heating and the back side heating have substantially similar temperature distributions when the top and bottom are reversed. In addition, the shape of these temperature distributions is roughly similar to a sine wave whose origin is at the center of the float. This shape is considered to be one of the clues for detecting the place where the floating occurs. Although FIG. 11A and FIG. 11B have different temperature distributions, the temperature distribution in the same range on the upper surface of the analysis model is represented by 250 × 250 pixels in the former and pixels 125 × 125 in the latter. It can be seen that the difference in the number of pixels has little effect on the roughness of the image.
2.2.3 Diversification of teacher data As described above, it is possible to create teacher data on the surface temperature distribution related to cracks and floats, which are internal deformations of the flat concrete structure, using the analysis model. Further, if necessary, regarding many parameters set above, for example, changing individual dimensions, changing thermal conditions of concrete, changing heating conditions, intersecting cracks, connecting cracks and floats. The teacher data may be reinforced by increasing the diversity of the model by changing the flaky space from a single one to a combination of multiple ones as in, etc., and considering the unsteady state. . Further, in order to more appropriately match the conditions of the actual structure, it is also possible to incorporate the effects of reinforcing bars, aggregates, and the like existing inside the general concrete structure into the analysis model.

In addition, as described above, in addition to changing the configuration, parameters, and the like of the analysis model, as described above, data expansion methods used in the DL method, that is, rotation, scaling, translation, shearing, blurring, and noise addition Etc. may be applied to the surface temperature distribution.
2.3 How to create teacher data from measured surface temperature distribution
2.3.1 Analysis method for extracting the teacher data part As a filter to detect local temperature change parts related to internal deformation contained in the surface temperature distribution of the structure, for example, There is a filter based on a two-dimensional wavelet transform that is effective for detecting a local change of an unsteady signal. A method using this filter will be described below.

  Assuming that the coordinate axes are x in the horizontal direction and y in the vertical direction, the basic expression of the two-dimensional wavelet transform is as follows.

Here, t (x, y) is a function related to the surface temperature at the coordinates (x, y), and not only the temperature distribution but also the stimulus values constituting the brightness image data or the color image reflecting the temperature distribution can be substituted. Although it is possible, hereinafter, these are referred to as a temperature distribution for simplicity. ψ (x, y) is a function called mother wavelet, which is a localized function. (x ₀ , y ₀ ) is a shift parameter of ψ (x, y), and a is a scaling parameter.

Next, it is necessary to select a mother wavelet ψ (x, y) suitable for characterizing the temperature distribution near the internal deformation of the concrete structure. Since such a temperature distribution has a feature that the directionality, gradient, magnitude, and the like change depending on a place, the present invention is effective for detecting, for example, local directionality, periodicity, and the like of a signal. A two-dimensional Gabor wavelet using a Gabor function was used as an example.
2.3.2 Characteristics of two-dimensional Gabor wavelet suitable for detection The mother wavelet ψ (x, y) is redefined as a two-dimensional Gabor wavelet (hereinafter abbreviated as Gabor wavelet) shown in equation (2). This Gabor wavelet is obtained by applying a two-dimensional Gaussian window to a complex vibration e ^{i (2πfx + φ)} traveling in the x-axis direction. By applying a two-dimensional Gaussian window, the complex oscillation e ^{i (2πfx + φ)} becomes localized around the origin, reflecting the local periodicity and directionality of the temperature distribution near it. The result is obtained.

Here, f represents the center frequency, σ _x and σ _y represent the window widths in the x-axis direction and the y-axis direction, and φ represents the phase.

  By using such a Gabor wavelet, a Gabor wavelet transform Ψ (x, y) of the temperature distribution t (x, y) is given by the following equation.

However, the wavelet in the equation (2) is a wave traveling in only one direction on the xy plane, whereas the direction of the temperature change due to the deformation is arbitrary, and therefore, the temperature distribution t (x, y In order to detect the characteristic of the temperature change at each point in any direction, it is necessary to rotate the wavelet around each point. Therefore, ψ _r (x, y, θ _r ) obtained by rotating ψ (x, y) counterclockwise at each point by an angle θ _r is defined by the following equation, and this is calculated over a necessary range in the analysis. Rotate.

Here, the configuration of the Gabor wavelet of the formula (2) suitable for detecting the shape of the surface temperature distribution related to cracks or floating is considered. As described with reference to FIGS. 4 to 6, the surface temperature distribution related to the crack is such that in the direction perpendicular to the straight line of the opening, the surface temperature changes abruptly at the opening, and the peak including the highest hot spot and the lowest temperature. Since valleys including points are present on both sides of the opening, it may be considered to be similar to a sine wave having an origin at the opening. Therefore, it is considered that there is a possibility that the temperature distribution related to the crack can be detected by the sine wave sin (2πfx) of the complex vibration e ^{i (2πfx + φ)} .

On the other hand, as described with reference to FIGS. 9 to 11, the surface temperature distribution related to the float is such that the shape of the surface temperature distribution near the float in the vertical cross section is the highest temperature point and the lowest temperature for both the front side heating and the back side heating. It may be considered to be similar to the characteristic of a cosine wave having an origin at a temperature point. Therefore, it is considered that there is a possibility that the surface temperature distribution related to the floating can be detected by the cosine wave cos (2πfx) of the complex vibration e ^{i (2πfx + φ)} .

Further, in order to analyze the characteristics of the surface temperature distribution related to the internal deformation in more detail, the surface temperature distribution of the detection range is changed to the surface temperature of the sound part which has little or no influence of the internal deformation (reference value). It is conceivable to obtain a temperature distribution composed of positive and negative temperatures expressed as a difference from the temperature distribution, and apply a wavelet transform to this temperature distribution. The sine wave component and the cosine wave component obtained as described above evaluate not only the magnitude of the value but also the characteristics of the internal deformation from various angles depending on whether the value is positive or negative. be able to. The surface temperature of the healthy part is, for example, about 31 ° C. for front side heating and about 29 ° C. for back side heating in the cases shown in FIGS. If it is not easy to set the surface temperature of the healthy part, the average temperature of the surface temperature distribution in the detection target range may be simply used as the reference temperature.
2.3.3 Procedure for detecting temperature distribution related to internal deformation Using Gabor wavelet ψ (x, y) as described above, it is affected by the presence of internal deformation existing in surface temperature distribution t (x, y). The procedure for detecting the location of the received temperature distribution, that is, the location of the characteristic temperature distribution related to the internal deformation is as follows.

For example, when detecting a temperature distribution portion related to a deformation similar to a crack, a temperature distribution having a high relevance to the sine wave sin (2πfx) in the complex vibration e ^{i (2πfx + φ)} of the equation (2). You will find a place. Therefore, in order to give the sine wave sin (2πfx) a range of characteristics, a set of values of the frequency f and the window width σ _x , σ _y within an appropriate range is sequentially selected, and the Gabor wavelet ψ (x , y). Angles provided at appropriate intervals in a counterclockwise direction

The Gabor wavelet expansion coefficient y _ri of the surface temperature distribution t (x, y) of the target area is obtained by ψ _r (x, y, θ _r ) rotated by only. Then, the value theta _r0 angle theta _ri the square sum of the absolute values of the expansion coefficients [psi _ri is maximum determined, the set of values of the orientation of this value theta _r0 the frequency f and window width sigma _x and sigma _y Is the traveling direction of the target sine wave.

  Next, when the absolute values of the Gabor wavelet expansion series of the entire target area are arranged in the descending order, and only the expansion coefficient of the upper p% (expansion coefficient ratio) is left to be zero and the others are corrected, as a result, a minute temperature change occurs. Since the associated value is removed, relatively large temperature changes that tend to be characteristic of the deformation will be emphasized. When the distribution of the squared value of the corrected expansion coefficient is displayed in the figure, the position of the location where the surface temperature distribution related to the internal deformation exists is shown. Even in this case, the influence of noise is large, and when detection is not sufficiently performed only by the method of the present invention, it is effective to apply the noise removal method to the surface temperature distribution of the target area in advance. As an example of the noise removal method, a method based on wavelet degeneration using multi-resolution analysis using orthogonal wavelets can be considered.

In addition, when detecting a temperature distribution occurrence location related to deformation similar to a float, the relationship with the cosine wave cos (2πfx) of the complex vibration e ^{i (2πfx + φ)} of Expression (2) is large. The temperature distribution point is found, and the procedure is the same as above.

In the above, the complex vibration e ^{i (2πfx + φ)} of the filter is configured by a sine wave or a cosine wave of a single frequency, but a filter is configured by combining a sine wave of a plurality of frequencies and a plurality of cosine waves. Is also good. Further, the filter may be expanded to an arbitrary function. As a result, if it is possible to extract a part of a new characteristic surface response, it is not necessary to clearly specify the characteristic of the internal deformation related to the part, and the internal deformation may be set as a new internal deformation. .
2.3.4 Verification of detectability of analysis data
(1) Target surface temperature data and noise to be considered For the crack analysis or the floating analysis model shown in FIG. 6 or the floating analysis model shown in FIG. The validity of the detection method for detecting the location of the surface temperature distribution was verified.

  A point to be considered here is that the actual surface temperature distribution is such that noises due to various factors are added to the temperature distributions as shown in FIGS. Under the influence of noise, the temperature distribution characteristic of the internal deformation may be significantly disturbed. Therefore, when examining the detectability, it is necessary to show that the detection can be sufficiently performed even for the surface temperature distribution to which noise is added.

  Considering the surface temperature distribution noise, the noise characteristics include the unevenness of local thermal properties due to the fact that concrete is composed of composite materials such as cement, fine aggregate, coarse aggregate, unevenness and dirt on the concrete surface, Because it is affected by non-uniform heating conditions and environmental conditions, it cannot be specified unconditionally, but here, Gaussian white noise whose average is 0 based on the standard deviation σ of the surface temperature distribution of concrete with internal deformation As given.

For example, in the crack analysis model shown in FIG. 6, when δ = 45 °, the standard deviation σ of the temperature distribution on the observation surface at the time of front side heating and back side heating is 0.157, respectively. In the floating analysis model shown in FIG. 9, the standard deviation σ of the temperature distribution on the observation surface in the case of d = 30 mm in the case of front side heating and the case of back side heating is 0.261, respectively. These cracks analysis model and float analysis model, respectively the temperature distribution in the case of adding the respective double 2σ standard deviation sigma as noise Gaussian white noise with standard deviation sigma _w to the surface temperature distribution diagram 12 and As shown in FIG. Since the surface temperature of the location of the internal deformation is higher or lower than the surrounding area, it may be possible to visually recognize the presence of the internal deformation by displaying the color. It is difficult to see the temperature change.
(2) Verification of the detectability of cracks Regarding the surface temperature distribution obtained by the crack analysis model in the case of front side heating and back side heating, when they do not contain noise, and when the above Gaussian white noise is added Verification was performed on the surface temperature distribution.

Among the surface temperatures of the crack model shown in FIG. 6, the results of applying the present detection method to the surface temperature distribution when δ = 45 ° and front side heating and back side heating and no noise are included are shown. 14 (a) to (d) and FIG. 15 (a) to (d). In the case of the front side heating, FIG. 14A shows a positive portion (hereinafter, a positive sine wave related component ⁾ of a portion related to the sine wave constituting the complex vibration e ^{i (2πfx + φ} ) of the Gabor wavelet. FIG. 14B shows a negative portion (hereinafter referred to as a negative sine wave related component) of the portion related to the sine wave, and FIG. 14C shows a portion related to the cosine wave. FIG. 14 (d) shows a positive portion (hereinafter, referred to as a positive cosine wave-related component) of the portion corresponding to the negative cosine wave (hereinafter referred to as a negative cosine wave-related component). Component). In this study, since the main problem is whether or not there are a sine wave-related component and a cosine wave-related component in a very narrow range of the expansion coefficient ratio p = 4%, the value of each component has no particular meaning. .

  The position where the positive sine wave related component shown in FIG. 14 (a) is present almost coincides with the position where the surface temperature at the crack opening is higher than the reference temperature, and FIG. 14 (b) The position where the negative sine wave-related component is present substantially coincides with the position where the surface temperature at the crack opening is lower than the reference temperature. Therefore, it can be seen that the position of the crack opening can be detected by the positive and negative sine wave related components.

  Further, the range where the positive cosine wave related component shown in FIG. 14 (c) exists, as shown in the graph of FIG. 8 (a), on the side opposite to the side where the cracked surface exists with the crack opening as a boundary. The temperature almost falls in the range where the temperature gradually drops from the highest temperature point in FIG. Further, the negative cosine wave function component shown in FIG. 14 (d) has a reference temperature at which the temperature rises gently from the lowest hot spot on the side where the cracked surface exists as shown in the graph of FIG. 8 (b). The temperature almost coincides with the range where the temperature is lower than the temperature. Therefore, it can be seen that the side where the cracked surface exists can be detected by the positive and negative cosine wave related components.

  As described above, information on cracks from the sine wave-related components and cosine wave-related components in FIGS. 14A to 14D, that is, the position of the crack opening and the side where the crack surface exists with respect to the crack opening Can be detected.

  On the other hand, in the case of the backside heating, similarly to the case of the frontside heating, in FIGS. 15A to 15D, the position of the crack opening is indicated by positive and negative sine wave related components. In the case of the backside heating, as shown in FIG. 8 (b), there is a maximum hot spot on the side where no cracked surface exists and the temperature gradually decreases from that point. On the opposite side of the opening, there is a range in which the temperature gradually rises from the lowest temperature point and is lower than the reference temperature. The former temperature range is detected by the positive cosine wave-related component shown in FIG. 15C, and the latter temperature range is detected by the negative cosine wave-related component shown in FIG. 15D. Therefore, also in the case of the backside heating, the information regarding the crack from the sine wave-related component and the cosine wave-related component in FIGS. 15A to 15D, that is, the position of the crack opening and the cracked surface with respect to the crack opening The existence side can be detected.

  Next, the detection results when the δ shown in FIG. 9 is 45 ° and the front-side heating and the back-side heating include noise (see FIG. 13) are shown in FIGS. 16 (a) to (d) and FIG. a) to (d). In the case of the front side heating, even if noise is included, the influence of the noise is almost completely removed, and as clearly as the case without noise shown in FIGS. 14 (a) to (d), FIGS. The position of the crack opening can be determined from the positive and negative sine wave related components in (b), and the side where the cracked surface exists can be determined from the positive and negative cosine wave related components in FIGS. 16 (c) and 16 (d). On the other hand, in the case of backside heating, as shown in FIGS. 17A to 17D, the temperature change on the observation surface of the observation target is smaller than in the case of front side heating, so that the influence of noise cannot be completely removed. However, the position of the crack opening and the side where the cracked surface exists can be sufficiently determined.

  In the previous studies, the case where one image includes a crack with one crack inclination angle (δ = 45 °) was targeted. Next, a plurality of images having different inclination angles in one image were examined. Consider the case where cracks are included. A total of four crack models obtained by adding a crack model having a crack with a δ of 75 ° to three crack models having a crack with a tilt angle δ of 30 °, 45 ° and 60 ° shown in FIG. Consider a crack model with a total size of 1000 mm x 1000 mm x 100 mm, connected two by two in each direction. FIG. 18 shows a positive sine wave-related component and a positive cosine wave-related component among the results of applying the present detection method to the surface temperature distribution obtained by heating this crack model under the same conditions as the front side heating conditions. .

  In FIG. 18 as well, the sine wave-related component indicates the position of the crack opening and the cosine wave-related component indicates the side where the crack surface exists, as in the past. The position of the crack opening and the side where the crack surface exists are detected as signals in a wider range as the inclination angle δ is smaller. That is, the range represented by each component is wide in the order of δ = 30 ° and δ = 45 °. For a crack with δ of 60 °, no signal appears on the side where the cracked surface exists, and a crack with δ of 75 ° Does not show any signal about the position of the crack opening and the side where the crack surface exists.

  The reason is that, as can be understood from FIG. 8, in both the front side heating and the back side heating, the smaller the inclination angle δ, the higher the temperature at the highest hot spot, so that the sine wave related component appears more strongly. In this case, since the crack depth is the same, the smaller the inclination angle δ, the larger the area of the crack surface extending from the crack opening to the inside of the structure, and accordingly, the range detected as the cosine wave-related component is wider. It is considered that Thus, it can be seen from the difference in the characteristics of the surface temperature distribution that the difference in the inclination angle of the crack can be detected.

Further, as a more complicated state, as shown in FIG. 19, two cracks obtained by adding another δ = 45 ° crack perpendicular to the δ = 45 ° crack shown in FIG. Consider the model. 20 (a) to (d) show the results of applying the present detection method to the surface temperature distribution when noise is included in the case of front side heating. Also in this case, it can be seen that the positions of two orthogonal crack openings can be detected from the positive and negative sine wave related components in FIGS. 20 (a) and 20 (b). Also, the side where the two cracked surfaces exist can be determined from the positive and negative cosine wave related components in FIGS. 20 (c) and 20 (d).
(3) Verification of Float Detectability The detection results of the floating model shown in FIG. 9 in the case of front side heating and back side heating and including noise are shown in FIGS. 21 (a) to 21 (d) and FIG. a) to (d). In the case of the floating model, the positive and negative sine wave-related components shown in FIGS. 21 (a) and 21 (b) and FIGS. 22 (a) and 22 (b) do not appear. The reason is that the surface temperature does not suddenly change at the crack opening as in the case of the crack model. Then, a portion higher in temperature than the dome-shaped reference temperature is generated on the observation surface by the front side heating, so that a positive cosine wave related component appears. On the other hand, since a portion lower than the dome-shaped reference temperature is generated on the observation surface due to the backside heating, a negative cosine wave-related component appears.

Thus, by comprehensively considering the appearance of the heating surface and the sine wave-related component and the positive / negative cosine wave-related component, the presence of the floating can be detected, and therefore, the surface temperature distribution related to the floating can be detected. Can be specified as teacher data.
2.4 Verification of experimental data on crack detection
2.4.1 Outline of Experiment After verification by analysis data, the detection performance of the method of the present invention for a crack exhibiting a surface temperature distribution that is relatively more complex than that of a float was verified by an experiment.

  FIG. 23 shows a concrete specimen of 450 mm × 600 mm × 200 mm (a1 = 200 mm, a2 = 250 mm, a3 = 150 mm, b1 = 100 mm, b2 = 250 mm, b3 = 100 mm), and the inclination angle δ with the surface is 30. An acrylic plate simulating a cracked space having a thickness of 1 mm (width of 350 mm, length of 200 mm and 115 mm) was inserted so that the angle was 60 ° and 60 °.

  Since the thermal conductivity of the acrylic plate and the air is about 0.21 W / (mK) and about 0.025 W / (mK), respectively, an acrylic plate with a thickness of 1 mm will be cracked with a width of about 0.12 mm. It is considered to have a corresponding thermal insulation performance.

FIG. 24 (a) is a surface temperature image on the front side surface when the front side surface provided with the simulated crack is heated by sunlight on a sunny day with the front side right above. At this time, the maximum temperature of the surface temperature distribution is 19.4 ° C, the minimum temperature is 13.8 ° C, the average temperature is 16.3 ° C, and the standard deviation is 1.1 ° C. FIG. 24 (b) shows a result obtained by dividing the average temperature from the surface temperature distribution and normalized by their maximum absolute values. From these images, the position of the crack opening with the inclination angle δ of 30 ° is visible, but the position of the crack opening with the inclination angle δ of 60 ° is not visible. As shown in the previous section, the larger the inclination angle δ, the more difficult it is for a crack to have a characteristic temperature distribution. Also, due to the positional relationship between the two acrylic plates, the amount of heat to be stored at the location where the highest hot spot appears for a crack with an inclination angle δ of 60 ° is determined by the low It is also thought that it was further promoted by the flow to the country.
2.4.2 Detection Results FIGS. 25 (a) to 25 (d) are detection results for the image data shown in FIG. 24 (b), and (a) a positive sine wave related component, and (b) a negative , A (c) positive cosine-wave-related component, and (d) a negative cosine-wave component.

  The positive and negative sine wave related components clearly indicate the location of the crack opening with an inclination angle δ of 30 °. However, cracks with an inclination angle δ of 60 ° were not detected at all. This is presumably because, for the above-described reason, the characteristic temperature distribution targeted by the present detection method did not occur at a crack having an inclination angle δ of 60 °.

  The components related to the positive and negative cosine waves are located slightly away from the crack opening having the inclination angle δ of 30 °, but correctly capture the temperature change on the side where the crack surface exists and on the opposite side. Therefore, the cosine wave-related component, by being superimposed with the sine wave-related component, not only becomes useful supplementary information for detecting the crack occurrence location, but also alone, the surface temperature changes relatively slowly, It is considered that there is a possibility that a region having a higher temperature than a peripheral portion which is not affected by the deformation can be detected.

  25 (c) to 25 (d), there are fine point-like components. This is because the concrete is composed of composite materials such as cement, fine aggregate, coarse aggregate, etc. as described above, and the local thermal properties are uneven, and the noise due to unevenness and dirt on the concrete surface is caused. Is considered to be the effect of The effect of preprocessing for removing such noise components was studied. The noise removal method used is based on wavelet degeneration using multi-resolution analysis using the orthogonal wavelets described above. The image data shown in FIGS. 23A and 23B are degenerated using only the upper α = 1% when the scaling coefficients of the wavelet level j = 5 of Daubechies (N = 10) and the absolute values of the wavelet expansion integers are arranged in descending order. 26 (a) to 26 (d) show the result of performing a detection process with the expansion coefficient ratio p = 4% again after performing noise removal in advance. The minute temperature change is removed, and the presence of a crack opening having an inclination angle δ of 30 ° can be more clearly detected.

It is also assumed that the temperature distribution of the actual structure includes a noise component equal to or greater than this test piece. In such a case, the noise removal method described here is considered to be effective.
2.5 Examination of applicability to actual measurement data in actual structure
2.5.1 Measured data in actual structure As an example of application to surface temperature distribution in an actual structure, application to a slab concrete structure generally called a floor slab will be described. Such a structure is widely and generally used for a bridge portion of a road or a railway or an elevated structure portion, and is exposed to severe deterioration factors such as traffic load and rainwater. Therefore, the application of the present invention is one of the desirable application examples. The stimulus in this case can be applied similarly to the case of artificial heating, but here, the simplest and least costly, the upper side of the structure is directly or indirectly heated under the influence of sunlight. Think about it.

  In this case, as the internal deformation to be detected, it is assumed that cracks and lifts which are particularly typically seen in the floor slab of the concrete structure, and the surface response related to the cracks and lifts from the surface temperature distribution of the lower surface of the structure which is a surface response. Consider detection of the location where the surface temperature distribution exists.

The thermal state in which such a structure is placed is almost the same as that shown in the schematic diagram of FIG. When this is compared with the examination by the above-described analysis model, the observation surface and the heating surface are different from each other in a structural aspect as shown in FIG. 6B for cracks and FIG. 9B for floating. This corresponds to the case where the heating condition of the back surface heating works.
2.5.2 Detection results Next, from the surface temperature distribution obtained when the target structure was actually measured, a study was conducted to extract a range that is highly relevant to the surface temperature distribution corresponding to cracks and floats assumed in the analysis. The results are shown.

  FIG. 27 is a photograph of the situation when the lower surface side of the RC slab of the bridge is measured by an infrared imaging device on a sunny day in winter, and an RGB image displaying the surface temperature distribution on the lower surface of the slab. FIGS. 28 to 30 show the results of applying the present detection method to the surface temperature distributions of the regions R1 to R3 surrounded by the rectangular solid lines in the RGB image, respectively. As before, in each figure, (a) positive sine wave related component, (b) negative sine wave related component, (c) positive cosine wave related component, and (d) negative cosine wave related component The components are shown.

Under these structural conditions and the heating conditions of backside heating, as shown in the analysis data analysis, the range related to floating in this case is lower in temperature than the surroundings, so only negative cosine wave related components appear and cracks The positive and negative sine wave related components appear as one set, and the positive sine wave related component appears accompanying the periphery of the crack. In this way, the range of the surface temperature distribution that is highly relevant to the internal deformation of the target is extracted from the measured data, and the surface temperature distribution in that range is converted into teacher data related to cracking and lifting. Can be.
3 Creating teacher data when response is oscillatory data
3.1 Description of the overall situation to be considered As one embodiment according to the present invention, when the stimulus given to the structure having the internal deformation is a blow and the obtained response is oscillatory data, the internal deformation and the vibration data A method of creating teacher data on the relationship with the teacher will be described below.

  As an example of the target structure and the internal deformation, it is assumed that an internal deformation exists in a part of the intermediate portion of the rod-shaped structure as shown in FIG. By striking the end of such a rod-shaped structure and observing the vibration that propagates through the structure and returns to the end again, it is determined whether or not internal deformation has occurred, and at what position and to what extent. In some cases, detection may be performed as to how many internal deformations exist in the image. This is because a part of the vibration propagating in the rod-shaped structure is reflected and a part is transmitted at the location where the internal deformation exists, so that the shape of the vibration is related to the existence of the internal deformation.

  For example, when this rod-shaped structure is underground, the details such as the length of the structure are unknown and the relationship between the given vibration data and internal deformation cannot be directly related. It is difficult to create teacher data on a regular basis. As the vibration data, acceleration data detected by an acceleration sensor installed near the end of the structure, or acoustic data detected by an acoustic sensor installed in the air around the end where the vibration propagated in the air can be considered. . FIG. 32 shows an example of acceleration data measured by the acceleration sensor for the actual structure.

The example of the rod-like structure shown here is generated by stimulus such as impact on the structure surface when there is internal deformation such as cracks and floating in the flat plate structure as shown in FIGS. The same can be applied to the generation of teacher data when vibration, sound, or the like is used as a response.
3.2 Creation of teacher data by analysis When the rod-shaped structure as shown in FIG. 31 does not include the internal deformation, and when the internal deformation is in the center of the rod-shaped structure, when the upper end of the rod-shaped structure is hit, FIGS. 33 (a) and 33 (b) show the results obtained by analyzing the acceleration data among the wave data which propagate through the rod-shaped structure and return to the upper end portion again. When the internal deformation shown in FIG. 33 (a) is not included, the wave propagating in the rod-like structure is periodically attenuated by repeating the reflection at the tip and the upper end of the rod-like structure. On the other hand, if there is an internal deformation in the middle part of the rod-like structure, as shown in FIG. 33 (b), the propagating wave is reflected and transmitted in the internal deformation, so that the wave observed at the upper end of the rod-like structure Shows a very complicated shape. Even if the internal deformation exists only in one place, it exhibits such a complicated shape. Therefore, when the number of the internal deformations increases, the position of the internal deformation can be identified simply by observing the shape of the wave. It is difficult.
3.3 Diversification of teacher data By setting the above analysis model, it is possible to create teacher data of acceleration data at the end of the rod-shaped structure with internal deformation by hitting the end by analysis. . Further, if necessary, regarding many parameters set in the rod-shaped structure shown in the figure, for example, changing individual dimensions, changing the strength conditions of the constituent materials, adding constraints from the periphery, the position of internal deformation The teacher data may be reinforced by increasing the diversity of the model structure by an operation such as changing the range, the number, or the like. Further, in order to better match the conditions of the actual structure, it is also possible to incorporate the influence of reinforcing bars, aggregates, and the like existing inside the general concrete structure into the analysis model.

  Even in the case of this rod-shaped structure, since the measured data obtained in the actual structure often includes a lot of noise, it is sometimes desirable to remove the noise to make the shape of the wave more clear. As an example of the noise removal method applied to the actually measured data shown in FIG. 32, a method based on wavelet degeneration using multi-resolution analysis using orthogonal wavelets is considered. FIG. 34 shows the result of applying this noise removal to the measured data of FIG.

  In addition to the target vibration data, not only the measured data itself, but also the decomposition of the measured data by Fourier transform, multi-resolution analysis using orthogonal wavelets, etc. In some cases the impact can be considered in more detail.

In addition, as described above, in addition to changing the configuration, parameters, and the like of the model structure used in the analysis, the data expansion method used in the DL method for response data obtained by the analysis, that is, rotation, scaling, translation, Shearing, blurring, noise addition, etc. may be applied.
4 Example of detecting the location of crack occurrence using DL
4.1 Detection Procedure The detection procedure mainly follows the flowchart shown in FIG. However, since only the teacher data obtained by the analysis is used, optional steps S10 and S11 are excluded.

  According to step 1, the surface temperature distribution in the structural section from which the internal deformation is to be detected is measured, and according to step 2, this is stored as surface temperature distribution data.

  Next, according to step 3, one or more internal deformations to be detected existing in the target structural section are assumed based on engineering experience and judgment. Here, the lower surface of the RC floor slab structure is assumed. Only the cracks present in the Then, the surface temperature distribution at the place where the cracks occurred was determined from the analysis using the crack analysis model shown in FIG. As described in 2.2.1, the parameters constituting the analysis model are diversified, and the numerical values of each parameter can be set freely within a certain range. However, in this case, as the most basic case, the value of each parameter shown in 2.2.1 was used, and one crack analysis model having a crack with a crack inclination angle δ of 45 ° was targeted.

  Next, according to an optional step 12, processing is applied to the surface temperature distribution at the crack occurrence location determined from the one crack analysis model. Here, in order to generalize the obtained surface temperature distribution, the minimum temperature is normalized to be 0 and the maximum temperature is set to 1, and the normalized temperature distribution is further reduced by -90 of the data expansion method described above. Rotation in the range of ° to + 90 °, enlargement in the range of 1.0 to 1.5 times, and change of the aspect ratio in the range of 1: 1.5 to 1.5: 1 By diversifying, a total of 50 teacher data were created.

  Next, in step 5, the actually measured surface temperature distribution data of the target was read, and DL detection was performed using the above teacher data.

Finally, according to step 6, the location where the occurrence of the detected crack is assumed is shown, for example, superimposed on the image of the original surface temperature distribution.
4.2 Examination with experimental data
4.2.1 Target experimental data
The surface temperature distribution data shown in FIG. 24B obtained by an experiment using the crack specimen shown in 2.4 is targeted.
4.1.2 Detection results The DL method was applied to the surface temperature distribution described above, and detection was performed using a location where a local surface temperature distribution related to a crack, which was the source of the teacher data, exists. In FIG. 35, the approximate position of such a portion is indicated by a square frame drawn over the image of the surface temperature distribution. 25 and 26, the positions of the openings of the cracks having the inclination angle δ of 30 ° can be clearly detected, as with the results shown in FIGS. 25 and 26. , Only the upper right part of the figure has been detected.
4.2 Examination with measured data on actual structures
4.2.1 Actual Measurement Data of Target The surface temperature distribution data actually measured on the lower surface of the RC slab shown in FIG. 27 was used.
4.2.2 Detection results The DL method was applied to the above surface temperature distribution, and detection was performed using a location where a local surface temperature distribution related to a crack as the base of the teacher data exists. In FIG. 36, the approximate position of such a portion is indicated by a square frame drawn over the image of the surface temperature distribution.

  In these two examples, the result of executing DL using teacher data created from only one basic crack analysis model is shown. However, the analysis model is further diversified, for example, a crack inclination angle and a crack size. Changing the depth or depth, increasing the number of cracks, making the relationship of multiple cracks parallel, intersecting, etc., and / or adding a float which is a typical internal deformation to the internal deformation of the object It is thought that this can contribute to improvement in detection accuracy.

REFERENCE SIGNS LIST 100 structural internal deformation detecting device 101 actual measuring means 102 structural analyzing means 103 experimental means 104 actual measured data storing means 105 teacher data storing means 106 teacher data extracting means 107 estimating means 108 display means

Claims (20)

A structural internal deformation detection device for detecting characteristics of internal deformation existing on the surface or inside of a real structure given a physical stimulus,
Structural analysis means for analyzing the response of the model structure when a physical stimulus is applied to the model structure including internal deformation having typical characteristics,
Teacher data recording means for recording the response of the model structure obtained by the structure analysis means as teacher data,
Actual measurement data recording means for recording the response obtained by giving the physical stimulus to the actual structure as actual measurement data,
Estimating means for estimating characteristics of internal deformation existing on the surface or inside of the actual structure using the teacher data and the actual measurement data,
Display means for displaying on the screen the characteristics of the estimated internal deformation,
A structural internal deformation detection device comprising:   Further comprising teacher data extracting means for extracting a part of the response of the real structure having a strong correlation with the response of the model structure from the response of the real structure, wherein the teacher data recording means is provided by the teacher data extracting means The structural internal deformation detection device according to claim 1, further comprising, as the teacher data, a portion of the extracted response of the real structure having a strong correlation with the response of the model structure.   3. The structural internal deformation detection according to claim 1, wherein the teacher data recording unit further includes, as the teacher data, experimental data obtained from a test body simulating the actual structure and the internal deformation. 4. apparatus.   The structural internal deformation detecting device according to any one of claims 1 to 3, wherein the estimating means includes software for executing a deep learning technique.   The structural internal deformation detecting device according to any one of claims 1 to 4, wherein the teacher data recording unit or the measurement data recording unit increases a data amount of the teacher data or the measurement data by a data extension method.   The structural internal deformation detecting device according to any one of claims 1 to 5, wherein the stimulus includes any of heating, impact, vibration, and ultrasonic waves.   The structural internal deformation detection device according to claim 1, wherein the characteristics of the internal deformation include an existing position, an existing range, a physical property, and a type of the internal deformation.   The teacher data extracting means uses, when the response of the real structure is data distributed in two dimensions, a characteristic obtained by processing using a two-dimensional filter for each specific region of the response of the real structure. The structural internal deformation detecting device according to any one of claims 1 to 7.   The apparatus for detecting a deformed portion in a structure according to claim 8, wherein the two-dimensional filter is a two-dimensional Gabor wavelet.   The said teacher data extraction means uses the characteristic obtained by processing using the one-dimensional filter for every specific area | region of the said sound or wave, when the response of the said real structure is a sound or wave. The structural internal deformation detecting device according to any one of claims 7 to 7. A structural internal deformation detection method for detecting characteristics of internal deformation existing on the surface or inside of a real structure given a physical stimulus,
A structural analysis step of analyzing a response of the model structure when a physical stimulus is applied to the model structure including the internal deformation having typical characteristics,
A teacher data recording step of recording the response of the model structure obtained by the structural analysis step as teacher data,
An actual measurement data recording step of recording a response obtained by giving the physical stimulus to the actual structure as actual measurement data,
An estimating step of estimating a characteristic of an internal deformation existing on the surface or inside of the actual structure using the teacher data and the actual measurement data;
A display step of displaying the property of the estimated internal deformation on a screen,
A method for detecting deformation inside a structure, comprising:   Further comprising a teacher data extracting step of extracting a response portion of the real structure having a strong correlation with the response of the model structure from the responses of the real structure, wherein the teacher data recording step is performed by the teacher data extracting step. 12. The structural internal deformation detection method according to claim 11, further comprising adding, as the teacher data, a part of the extracted response of the real structure having a strong correlation with the response of the model structure.   The inside of the structure according to any one of claims 11 to 12, wherein the teacher data recording step further includes adding, as the teacher data, experimental data obtained from a test body simulating the actual structure and the internal deformation. Deformation detection device.   14. The method according to claim 11, wherein the inferring step further includes performing a deep learning method. 15.   15. The structural internal deformation detection method according to claim 11, wherein the teacher data recording step or the actually measured data recording step further includes increasing a data amount of the teacher data by a data extension method.   16. The method according to any one of claims 11 to 15, wherein the physical stimulus includes heating, impact, vibration, and ultrasonic waves.   The structural internal deformation detection method according to claim 11, wherein the characteristic of the internal deformation includes an existing position, an existing range, a physical property, and a type of the internal deformation.   The teacher data extracting step includes, when the real response of the real structure is data distributed two-dimensionally, a feature obtained by processing using a two-dimensional filter for each specific region of the real response of the real structure. The structural internal deformation detection method according to any one of claims 11 to 17, wherein the method is used.   The method according to claim 18, wherein the two-dimensional filter is a two-dimensional Gabor wavelet. 12. The teacher data extracting step uses a feature obtained by processing using a one-dimensional filter for each specific section of the sound or the wave when the response of the real structure is a sound or a wave. 18. The method for detecting internal deformation of a structure according to any one of claims 17 to 17.