JP7077514B2 - Building information processing device and building information processing model learning device - Google Patents

Description

The present invention relates to a building information processing device and a building information processing model learning device.

Conventionally, an earthquake-time building damage evaluation device for evaluating damage inside a building during an earthquake has been known (for example, Patent Document 1). This in-building damage assessment device during an earthquake calculates the seismic response based on building data and seismic motion data, and based on the seismic response, installation data and floor data, there is a risk of the installation falling over and a risk of moving the installation. , Evaluate the degree of damage to non-structural members, the possibility of human damage, and the degree of evacuation obstacles in the building.

Further, a method for displaying seismic performance in a building is known (for example, Patent Document 2). In this display method, the seismic hazard curve is selected based on the geographic information and the ground information, and the drawing data of the chart showing the damage evaluation curve according to the maximum acceleration of the ground surface and the inter-story deformation angle of the building is displayed.

Further, a method for determining the soundness of a concrete building is known (for example, Patent Document 3). This method for determining the soundness of a concrete building determines the soundness of concrete structural members based on the constant tremor (natural frequency) of the building and the temperature difference between the inside and outside of the building.

Further, an earthquake damage determination device for determining the state of earthquake damage to a building based on building data and an earthquake waveform is known (for example, Patent Document 4).

Further, there is known a method for confirming the soundness of a building that estimates the response of each floor of the building based on the earthquake response information of the building obtained by a sensor installed on a limited floor (for example, Patent Document 5).

A method for detecting damage to a structure by installing a large number of sensors on the structure, detecting the presence or absence and degree of damage to the structure, and estimating the position of damage is known (for example, Patent Document 6).

Japanese Unexamined Patent Publication No. 2007-109107 Japanese Patent No. 4087136 Japanese Patent No. 4859557 Japanese Patent No. 4914162 Japanese Patent No. 5967417 Japanese Patent No. 6032494

Various information can be obtained from existing buildings. For example, the actual information indicating the combination of the design information of the building and the defect information generated in the building is useful information when designing a new building. In addition, such performance information is also useful information when predicting what kind of trouble will occur in an existing building.

However, in the above-mentioned Patent Documents 1 to 6, the actual information of the buildings constructed in the past is not taken into consideration.

In view of the above facts, an object of the present invention is to obtain information on defects in a target building in consideration of actual information on buildings constructed in the past.

In order to achieve the above object, the building information processing apparatus of the present invention uses the acquisition unit for acquiring the design information of the target building and the design information acquired by the acquisition unit as the design information of the building for learning. A defect information acquisition unit that acquires the defect information of the target building by inputting into a pre-learned model based on the learning data associated with the defect information representing the defect of the building for learning. Is a building information processing device including. According to the building information processing apparatus of the present invention, it is possible to obtain defect information of the target building in consideration of the actual information of the building constructed in the past.

The acquisition unit of the present invention further acquires building state information indicating the state of each part of the target building, and the defect information acquisition unit further corresponds to the building state information of each part of the learning building. It is possible to acquire the defect information by inputting into the trained model trained in advance based on the attached learning data. As a result, it is possible to obtain defect information of the target building by further considering the state of each part of the building.

The target building of the present invention may be a design target building or an existing building. As a result, it is possible to obtain defect information of the building to be designed or an existing building in consideration of the past performance information.

The building information processing model learning device of the present invention is based on the learning data in which the design information of the building for learning and the defect information representing the defect of the building for learning are associated with each other, and the building is based on the design information of the building. It is a building information processing model learning device including a learning unit for machine-learning a model for outputting the defect information and obtaining a learned model for outputting the defect information from the design information. As a result, it is possible to acquire a trained model for obtaining defect information from the design information.

The learning data of the present invention is further associated with building state information representing the state of each part of the learning building, and the learning unit is based on the learning data of the building design information and the building. A model for outputting the defect information of the building can be machine-learned from the building state information, and a learned model for outputting the defect information can be obtained from the design information and the building condition information. As a result, it is possible to acquire a trained model for obtaining defect information from the design information and the building state information representing the state of the building.

According to the present invention, it is possible to obtain the defect information of the target building in consideration of the actual information of the building constructed in the past.

It is a block diagram which shows the schematic structure of the building information processing apparatus which concerns on 1st Embodiment. It is explanatory drawing for demonstrating an example of learning data of this Embodiment. It is a figure which shows an example of the sensor installed in each place of a building. It is a figure which shows an example of the trained model of 1st Embodiment. It is a figure which shows an example of the screen displayed by an output part. It is a figure which shows an example of the learning processing routine of this embodiment. It is a figure which shows an example of the building information processing routine of this embodiment. It is a figure which shows an example of the trained model of 2nd Embodiment. It is explanatory drawing for demonstrating an example of the method of acquiring the trouble information. It is explanatory drawing for demonstrating an example of the method of acquiring the trouble information.

Hereinafter, embodiments of the present invention will be described in detail.

<First Embodiment>

<System configuration of building information processing equipment>

FIG. 1 is a block diagram showing an example of the configuration of the building information processing apparatus 100 according to the first embodiment. Functionally, as shown in FIG. 1, the building information processing apparatus 100 can be represented by a configuration including a reception unit 10, a computer 20, and an output unit 30. The building information processing apparatus 100 according to the first embodiment predicts a defect in the building to be designed.

The reception unit 10 receives the design information of the building to be designed and the building state information indicating the state of each part of the building to be designed.

Further, the reception unit 10 is associated with the design information of the building for learning, the building state information indicating the state of each part of the building for learning, and the defect information indicating the defect of the building for learning. Accepts learning data. In addition, information indicating the type of the cause of the defect is added to the learning data. The details of the training data will be described later. The reception unit 10 is realized by, for example, a keyboard, a mouse, an external device, or the like. The reception unit 10 may be configured to be able to receive image information input from the outside (for example, a photograph of a defect that has occurred in each part of an existing building).

Examples of the design information of the building to be designed include BIM (Building Information Modeling) data. The BIM data, which is the design information, includes the structural design information of the building and the design (finishing material) design information. In addition, as building state information of each part of the building to be designed, time-series data of displacement of beam members, column members, etc. of the building to be designed obtained by structural analysis and thermal analysis for the building to be designed, and design. Time-series data of temperature changes of beam members and column members of the target building can be mentioned.

In the present embodiment, the design information of the design building represents the information of each member included in the BIM data, and the building state information of each part of the design building is the time series data of the displacement of the beam member and the pillar member and the time series of the temperature change. The case of representing data will be described as an example.

The computer 20 includes a CPU (Central Processing Unit), a ROM (Read Only Memory) that stores programs for realizing each processing routine, a RAM (Random Access Memory) that temporarily stores data, and a memory as a storage means. , Network interface, etc. are included. The computer 20 is configured to be able to process various data such as numerical information and image information. Functionally, as shown in FIG. 1, the computer 20 has an acquisition unit 22, a learning data storage unit 24, a selection unit 25, a learning unit 26, a learned model storage unit 28, and defect information acquisition. It is equipped with a part 29.

The acquisition unit 22 acquires the design information of the target building and the building state information of the target building. Further, the acquisition unit 22 acquires learning data in which the building state information of the building for learning, the design information of the building for learning, and the defect information indicating the defect of the building for learning are associated with each other. ..

The learning data storage unit 24 stores a plurality of learning data acquired by the acquisition unit 22. For example, as shown in FIG. 2, the building state information of each part of the building for learning, the design information of the building for learning, and the defect information indicating the defect of the building for learning are associated with each other. Stored.

The learning building is an existing building. The building state information of each part of the building for learning is, for example, as shown in FIG. 3, the data acquired by the sensor S representing the displacement sensor or the temperature sensor installed in each part of the existing building A. be. All the circles shown in FIG. 3 represent the sensor S.

Defect information is obtained from regular visual inspections of existing buildings. For example, information such as cracks in slabs and walls of existing buildings, cracks and peeling of finishing materials, and repair history can be obtained as defect information. In addition, the classification information that classifies the defect pattern may be added to the defect information. The classification information is, for example, information such as the presence or absence of regularity of defects and whether the defects are penetrating or non-penetrating. The defect information may include image information such as a photograph of a defect that has occurred in each part of the existing building.

In the learning data of the actual ID "00001" shown in FIG. 2, the displacement data is generated in the member of the building member ID "00001" of the building ID "00001" representing the identification information of the building for learning whose construction has already been completed. "X1" and temperature data are "Y1", the member cross section is "A1", the member rigidity is "B1", the member restraint degree is "C1", the material property of the member is "D1", and the member strength is "E1". , The finish material or specification was "F1", and when the connection between the frame and the finish material was "G1", the crack in the concrete was "H1" and the crack or peeling of the finish material was "I1". Is represented. If the defect information includes image information such as a photograph of a defect that has occurred in each part of the existing building, the image information may also be stored in the learning data storage unit 24.

In this case, each data shown in FIG. 2 is represented numerically. For example, the finishing material or specification is expressed in a label format such as "1" or "2" for each material.

Further, as shown in FIG. 2, factor type information indicating the cause of the defect is added to the defect information. Factor type information is manually given during regular visual inspections of existing buildings. Building defects are caused by a variety of factors. For example, defects may be caused by design or by construction. In this embodiment, since the defect that occurs in the target building at the design stage is predicted, only the defect of the "design factor" is used as the target data.

The selection unit 25 selects each of the learning data whose defect factor type information is the "design factor" from the plurality of learning data stored in the learning data storage unit 24.

The learning unit 26 machine-learns a model for outputting building defect information from building design information and building state information at each location of the building, based on a plurality of learning data selected by the selection unit 25.

Then, the learning unit 26 obtains a learned model that outputs defect information from the design information of the building to be designed and the building state information of the building to be designed. Since the training data is data obtained from an existing building, the design information of the existing building, the building state information of the existing building, and the defect information of the existing building are reflected in the trained model.

In this embodiment, a trained model as shown in FIG. 4 is generated. For example, as shown in FIG. 4, a neural network can be used as an example of a model, and deep learning can be used as an example of a learning algorithm.

The trained model storage unit 28 stores the trained model learned by the learning unit 26.

The defect information acquisition unit 29 inputs the design information and the building state information of the target building acquired by the acquisition unit 22 into the trained model stored in the trained model storage unit 28, and the defect information acquisition unit 29 inputs the defect of the target building. Get information. The building state information includes time-series data of displacements of building beam members and column members, etc. obtained by structural analysis and thermal analysis with respect to BIM data, which is design information of the building to be designed, and building beam members and. It is time-series data of temperature changes of pillar members and the like.

As shown in FIG. 4, when the design information of the target building and the building state information of the target building are input to the trained model, the defect information of the target building is output. As a result, defect information regarding what kind of defect occurs is output from the trained model, and it is possible to predict what kind of defect will occur in the building to be designed.

For example, the trained model outputs information in the form of defect information such that the probability of cracking of concrete is 80% and the probability of cracking or peeling of finishing material is 10%.

The output unit 30 outputs the defect information of the target building output by the defect information acquisition unit 29 as a result. For example, the output unit 30 is realized by a display. Therefore, the output unit 30 can display the image information.

For example, the output unit 30 displays a screen as shown in FIG. As shown in FIG. 5, BIM data B, which is design information of the target building, and “crack occurrence probability 30%”, which is defect information, are displayed.

The designer who is a user of the building information processing apparatus 100 confirms the defect information displayed on the output unit 30. Then, the user designs the target building with reference to the defect information. For example, when an alert as shown in FIG. 5 is output, the user considers a change in the structural design or a change in the finishing material.

<Action of building information processing equipment>

Next, the operation of the building information processing apparatus 100 will be described with reference to FIGS. 6 and 7. The building information processing apparatus 100 executes the learning processing routine of FIG. 6 and the building information processing routine of FIG. 7.

<Learning processing routine>

When a plurality of learning data are input to the reception unit 10 of the building information processing apparatus 100, the acquisition unit 22 of the computer 20 acquires the plurality of learning data received by the reception unit 10. Then, the acquisition unit 22 stores the learning data in the learning data storage unit 24. When the building information processing apparatus 100 receives the instruction signal for starting the learning process, the building information processing apparatus 100 executes the learning process routine shown in FIG.

In step S100, the selection unit 25 selects each of the learning data whose defect factor type information is the "design factor" from the plurality of learning data stored in the learning data storage unit 24.

In step S102, the learning unit 26 provides a model for outputting building defect information from the building design information and the building state information of each part of the building based on the plurality of learning data selected in step S100. Let them learn by machine. Then, the learning unit 26 obtains a learned model that outputs defect information from the building design information and the building state information of the building.

In step S104, the learning unit 26 stores the learned model obtained in step S102 in the learned model storage unit 28, and ends the learning processing routine.

<Building information processing routine>

When the design information of the building to be designed and the building state information of each part of the building to be designed are input to the reception unit 10 of the building information processing apparatus 100, the computer 20 of the building information processing apparatus 100 is shown in FIG. Execute the building information processing routine shown in.

In step S200, the acquisition unit 22 acquires the design information and the building state information of the building to be designed.

In step S202, the defect information acquisition unit 29 reads out the trained model stored in the trained model storage unit 28.

In step S204, the defect information acquisition unit 29 inputs the design information and the building state information of the building to be designed acquired in step S200 to the trained model read out in step S202, and is to be designed. Get building defect information.

In step S206, the defect information acquisition unit 29 outputs the defect information acquired in step S204 as a result.

The output unit 30 outputs the information output from the defect information acquisition unit 29 as a result.

As described in detail above, the building information processing apparatus of the present embodiment inputs the design information of the building to be designed and the building state information of each part of the building to be designed into the trained model, and is the target of the design. Get building defect information. As a result, it is possible to obtain defect information of the building to be designed in consideration of the actual information of the building constructed in the past.

Further, the building information processing apparatus of the present embodiment is for learning in which the design information of the building for learning, the building state information of each part of the building for learning, and the defect information indicating the defect of the building for learning are associated with each other. Based on the data, the model for outputting defect information from the design information of the building and the building condition information of each part of the building is machine-learned, and the defect information is output from the design information and the building condition information of each part of the building. Get a trained model. As a result, it is possible to acquire a trained model for obtaining defect information of the building to be designed from the design information of the building to be designed and the building state information of each part of the building.

Further, the building information processing apparatus of the present embodiment can generate objective and versatile defect information and present it to a user who is a designer. In addition, it is possible to reduce the occurrence of defects in the frame and finish of the building to be designed.

<Second Embodiment>

Next, the second embodiment will be described. The second embodiment is different from the first embodiment in that a defect of an existing building is predicted. Since the configuration of the building information processing apparatus according to the second embodiment has the same configuration as that of the first embodiment, the same reference numerals are given and the description thereof will be omitted.

In the first embodiment, a trained model that outputs defect information of the building to be designed is used, but this trained model can also be used when acquiring defect information of an existing building.

Therefore, the reception unit 10 of the second embodiment receives the design information and the building state information of the existing building of the defect prediction target instead of the data of the building to be designed.

As shown in FIG. 3, the building state information of each part of the existing building targeted for defect prediction is the sensor S (displacement sensor and temperature sensor) installed at each part of the existing building targeted for defect prediction. The data obtained by.

The acquisition unit 22 of the second embodiment acquires the design information and the building state information of the existing building targeted for defect prediction.

The defect information acquisition unit 29 of the second embodiment inputs the design information and the building state information acquired by the acquisition unit 22 into the trained model stored in the trained model storage unit 28, and inputs the existing model of the defect prediction target. Get the defect information of the building.

As described in detail above, the building information processing apparatus of the second embodiment transfers the design information of the existing building targeted for defect prediction and the building state information of each part of the existing building targeted for defect prediction to the trained model. Enter to get the defect information of the existing building targeted for defect prediction. As a result, it is possible to obtain defect information of an existing building in consideration of the actual information of the building constructed in the past.

In addition, defects in the frame and finish of existing buildings can be detected at an early stage. In addition, when an earthquake occurs, the degree of damage to each member can be evaluated by inputting the data obtained from each sensor into the trained model.

<Third Embodiment>

Next, the third embodiment will be described. The third embodiment is different from the first and second embodiments in that the defect information of the target building is predicted without using the building state information of each part of the target building. Since the configuration of the third embodiment is the same as that of the first embodiment, the same reference numerals are given and the description thereof will be omitted.

The building information processing apparatus of the third embodiment acquires defect information of the target building from the design information of the target building. In the third embodiment, the target building may be either a design target building or an existing building.

The reception unit 10 of the third embodiment receives the design information of the target building. Further, the reception unit 10 of the third embodiment receives learning data in which the design information of the building for learning and the defect information indicating the defect of the building for learning are associated with each other. In the third embodiment, defect information is acquired from the design information. Therefore, the learning data of the third embodiment is data in which the design information of the building for learning and the defect information for learning are associated with each other.

The acquisition unit 22 of the third embodiment acquires the design information of the target building. Further, the acquisition unit 22 of the third embodiment acquires learning data in which the design information of the building for learning and the defect information of the building for learning are associated with each other.

The learning data storage unit 24 of the third embodiment stores a plurality of learning data acquired by the acquisition unit 22. The learning data of the third embodiment does not include the building state information as shown in FIG.

The learning unit 26 of the third embodiment machine-learns a model for outputting building defect information from the building design information based on a plurality of learning data selected by the selection unit 25.

Then, the learning unit 26 obtains a learned model that outputs defect information from the design information of the target building. Since the training data is data obtained from an existing building, the design information of the existing building and the defect information of the existing building are reflected in the trained model.

In the third embodiment, for example, a trained model as shown in FIG. 8 is generated. For example, as shown in FIG. 8, a neural network can be used as an example of a model, and deep learning can be used as an example of a learning algorithm.

The defect information acquisition unit 29 of the third embodiment inputs the design information acquired by the acquisition unit 22 into the trained model stored in the trained model storage unit 28, and acquires the defect information of the target building. ..

As shown in FIG. 8, when the design information of the target building is input to the trained model, the defect information of the target building is output. As a result, it is possible to predict what kind of defect will occur in the target building by using only the design information.

As described in detail above, the building information processing apparatus of the third embodiment inputs the design information of the target building into the trained model and acquires the defect information of the target building. As a result, it is possible to obtain defect information of the target building to be designed by using only the design information.

Further, the building information processing apparatus of the present embodiment has defect information from the building design information based on the learning data in which the design information of the building for learning and the defect information indicating the defect of the building for learning are associated with each other. The model for outputting is trained by machine, and the trained model that outputs the defect information from the design information is obtained. As a result, it is possible to acquire a trained model for obtaining defect information of the target building from the design information of the target building.

The present invention is not limited to the above-described embodiment, and various modifications and applications can be made without departing from the gist of the present invention.

For example, in the above embodiment, the defect information in the learning data has been described as an example of the case where it is obtained by periodic inspection of an existing building, but the present invention is not limited to this. For example, as shown in FIG. 9, it is assumed that the sensor S1 (displacement sensor and temperature sensor) is installed in the pillar P1 and the sensor S2 (displacement sensor and temperature sensor) is installed in the pillar P2. In this case, for example, the length L (expansion / contraction) of the beam is sequentially calculated from the data obtained by each displacement sensor installed on the column.

FIG. 10 shows a graph in which the beam length L calculated sequentially and the temperature data T sequentially measured by the temperature sensor are plotted. The graph shown in FIG. 10 is a biaxial graph, the horizontal axis represents the time when the measurement was performed, and the vertical axis represents the length (mm) and the temperature (° C.) of the beam. For example, "1/1" on the horizontal axis indicates January 1. Further, in the graph shown in FIG. 10, in order to compare the relative relationship between the temperature data T and the beam length L, the beam length L is superimposed on the graph representing the temperature data T at each time period. It was done.

As shown on the left side of FIG. 10, when there is no defect in the building, the relationship between the temperature data T at each period and the beam length L at each period is almost constant. On the other hand, as shown on the right side of FIG. 10, when a defect occurs in the building, the relationship between the temperature data T at each period and the beam length L at each period is not a constant relationship. Specifically, as shown in the elliptical region X of the graph on the right side of FIG. 10, the relative relationship between the temperature data T and the beam length L due to the occurrence of the crack H as shown in FIG. Changes. Therefore, for example, these relationships may be set as training data and reflected in the trained model. Specifically, the data on the left side of FIG. 10 and the data other than the elliptical region X on the right side of FIG. 10 are set as learning data in which no problem has occurred, and the data in the elliptical region X on the right side of FIG. 10 is set. Set as learning data where a problem has occurred.

Further, in the above embodiment, the case where the neural network model as an example of the model is trained by deep learning has been described as an example, but the present invention is not limited to this. For example, another model different from the neural network model may be trained by another learning method different from deep learning.

Further, in the above embodiment, the building information processing apparatus has been described as an example in which the learning processing routine shown in FIG. 6 and the building information processing routine shown in FIG. 7 are performed, but the present invention is not limited thereto. For example, the system may be configured by the building information processing model learning device that performs the learning processing routine shown in FIG. 6 and the building information processing device that performs the building information processing routine shown in FIG. 7.

Further, in the above embodiment, the case where temperature data and displacement data are used as an example of building state information has been described as an example, but the present invention is not limited to this, and any information indicating the state of each part of the building is used. , Any information may be used. For example, the data (acceleration, angle of rotation, etc.) obtained by the 3-axis acceleration sensor and the 3-axis gyro sensor, expansion / contraction, end bending, and degree of deformation at each location may be used. Similarly, the design information may be any information, not limited to BIM data, and the defect information may be any information as long as it represents a defect in the building.

Further, in the above description, the mode in which the program according to the present invention is stored (installed) in advance in the storage unit (not shown) has been described, but the program according to the present invention includes a CD-ROM, a DVD-ROM, a micro SD card, and the like. It is also possible to provide the form recorded on the recording medium of the above.

10 Reception unit 20 Computer 22 Acquisition unit 24 Learning data storage unit 25 Selection unit 26 Learning unit 28 Learned model storage unit 29 Defect information acquisition unit 30 Output unit 100 Building information processing device

Claims (4)

The acquisition department that acquires the design information of the target building,
The design information acquired by the acquisition unit is transferred to a pre-learned model based on learning data in which the design information of the building for learning and the defect information representing the defect of the building for learning are associated with each other. A defect information acquisition unit that inputs and acquires the defect information of the target building,
Including
Information on the cause type of the defect of the building is given in advance to each of the plurality of learning data.
The trained model is a trained model that has been trained in advance based on learning data in which the factor type information of the defect of the building is a design factor.
Building information processing equipment. The acquisition unit further acquires building state information indicating the state of each part of the target building.
The defect information acquisition unit inputs the defect information to the trained model that has been learned in advance based on the learning data to which the building state information of each part of the learning building is further associated, and acquires the defect information. The building information processing apparatus according to claim 1. The target building is a design target building or an existing building.
The building information processing apparatus according to claim 1 or 2. A plurality of learning data in which the design information of the building for learning and the defect information representing the defect of the building for learning are associated with each other, and a plurality of learning to which the factor type information of the defect of the building is given in advance. A selection unit that selects learning data for which the factor type information is a design factor from the data for
Based on the learning data selected by the selection unit, a trained model that machine-learns a model for outputting the defect information of the building from the design information of the building and outputs the defect information from the design information. With the learning department to get
Building information processing model learning device including.

Images