Building information processing device and building information processing model learning device
The building information processing device automates the conversion of transaction information from construction estimates into standardized item information using a trained model, improving efficiency and error detection in facility management.
Patent Information
- Application Number
- JP2020178472
- Authority / Receiving Office
- JP · JP
- Patent Type
- Patents
- Current Assignee / Owner
- Filing Date
- 2020-10-23
- Publication Date
- 2024-11-19
- Estimated Expiration
- 2040-10-23
AI Technical Summary
Construction companies face the challenge of managing transaction information from multiple suppliers with varying formats in construction work estimates, requiring manual data entry into a predetermined format, which is time-consuming and inefficient.
A building information processing device that utilizes a trained model to automatically journalize transaction information from construction estimates into corresponding item information, incorporating BIM information for anomaly detection and reliability assessment.
Automatically converts and journals transaction information into standardized item information, facilitating efficient management and detection of errors in estimates, thereby enhancing facility management.
Smart Images

Figure 0007586422000001 
Figure 0007586422000002 
Figure 0007586422000003
Abstract
Description
[Technical field]
[0001] The present invention relates to a building information processing device and a building information processing model learning device. [Background technology]
[0002] Conventionally, a technique for classifying product names into categories has been known (for example, Patent Document 1). For example, this technique classifies a product name such as "Black GABRIE Shawl Collar Jacket" into the category "Jacket" (for example, [Figure 5] of Patent Document 1).
[0003] Also, an automatic accounting system that accounts for transaction information is known (for example, Patent Document 2). This automatic accounting system selects the presentation format of the accounting results to the user according to the classification conformity rate for the account items (for example, [Claim 1] of Patent Document 2). [Prior art documents] [Patent documents]
[0004] [Patent Document 1] JP 2014-92878 A [Patent Document 2] Patent No. 6161229 Summary of the Invention [Problem to be solved by the invention]
[0005] Incidentally, when maintenance work is carried out on an existing building, the construction company in charge of overseeing the entire work often receives quotations for each type of work from multiple suppliers (or "partner companies").
[0006] However, the formats of quotations submitted by multiple suppliers are different, and the transaction information indicating the quotation items written on the quotations is often in a different format. For this reason, for example, a construction company's staff must check the transaction information written on quotations in different formats received from multiple suppliers and unify the transaction information as appropriate. In this case, the construction company's staff must unify the transaction information written on the quotations and then manually input it into a specified format data, which is a time-consuming task.
[0007] On the other hand, transaction information such as maintenance work for existing buildings is important information from the viewpoint of facility management, and it is preferable to manage it appropriately. For example, by managing transaction information such as the breakdown of a certain maintenance work, the cost, and when the maintenance work occurred, it becomes possible to carry out facility management appropriately. For example, if the transaction information is "air conditioning equipment renovation work," from the viewpoint of facility management, "air conditioning equipment renovation work" is classified as an item called "repair costs."
[0008] However, as described above, the quotations submitted by the multiple suppliers are in various formats, and the transaction information described in the quotations is also various.
[0009] The technologies disclosed in the above Patent Documents 1 and 2 are technologies for classifying product names or transaction information, but are not related to construction work. Therefore, the technologies in the above Patent Documents 1 and 2 have the problem that they cannot classify the transaction information described in the estimate for construction work into the corresponding item information.
[0010] In consideration of the above, the present invention aims to categorize transaction information described in an estimate for construction work into corresponding item information. [Means for solving the problem]
[0011] A first aspect of the present invention is a building information processing device that includes an acquisition unit that acquires transaction information representing estimate items in an estimate for construction work, and an accounting processing unit that outputs item information corresponding to the transaction information acquired by the acquisition unit by inputting the transaction information acquired by the acquisition unit into a pre-trained accounting trained model for outputting item information representing an item corresponding to the transaction information from the transaction information.
[0012] In addition, a building information processing device that is a second aspect of the present invention is a building information processing device, wherein the trained model for accounting further outputs tag information indicating the type of the transaction information, acquires BIM information indicating information of each part of the building from a BIM model of a building that is the target of construction work, and inputs the tag information and the BIM information into a trained model for anomaly detection that outputs a reliability of the transaction information from the tag information and the BIM information, thereby outputting a reliability of the transaction information acquired by the acquisition unit, and further includes an anomaly detection unit that outputs an alert if the reliability is below a threshold value.
[0013] In addition, a third aspect of the present invention is a building information processing model learning device that includes a learning unit that generates a learned accounting model that outputs, from transaction information representing estimate items, the subject information corresponding to the transaction information by learning an accounting model for outputting, from the transaction information representing estimate items, the subject information representing the subject corresponding to the transaction information, based on learning data in which learning transaction information representing estimate items for learning is associated with learning subject information representing the subject corresponding to the learning transaction information. Effect of the Invention
[0014] According to the present invention, it is possible to obtain the effect that transaction information described in an estimate for a construction work can be classified into corresponding item information. [Brief description of the drawings]
[0015] [Figure 1] FIG. 1 is a block diagram showing a construction journalization system according to a first embodiment. [Diagram 2] FIG. 13 is a diagram for explaining an example of a subject. [Diagram 3] FIG. 4 is a diagram for explaining learning data according to the first embodiment. [Figure 4] FIG. 2 is a diagram for explaining a model of the first embodiment. [Diagram 5] FIG. 13 is a diagram for explaining a life cycle cost. [Figure 6] 4 is a flowchart showing the contents of a learning processing routine of the construction journal entry system according to the first embodiment. [Figure 7] 4 is a flowchart showing the contents of a construction journal entry processing routine of the construction journal entry system according to the first embodiment. [Figure 8] FIG. 11 is a block diagram showing a construction journalization system according to a second embodiment. [Figure 9] FIG. 11 is a diagram for explaining learning data according to the second embodiment. [Figure 10] FIG. 11 is a diagram for explaining a model of the second embodiment. [Figure 11] FIG. 11 is a diagram for explaining a model of the second embodiment. [Figure 12] 13 is a flowchart showing the contents of a construction journal entry processing routine of the construction journal entry system according to the second embodiment. [Figure 13] FIG. 13 is a diagram showing an example of tag information incorporated into a life cycle cost form. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS
[0016] Hereinafter, an embodiment of the present invention will be described in detail with reference to the drawings.
[0017] [First embodiment]
[0018] <Configuration of the construction journal entry system according to this embodiment>
[0019] Fig. 1 is a block diagram showing an example of the configuration of a construction journalization system 10 according to the first embodiment. Functionally, the construction journalization system 10 can be represented as a configuration including an input unit 12, a scanner 13, a building information processing device 14, and a display unit 16, as shown in Fig. 1.
[0020] The input unit 12 receives data input by a user, etc. The input unit 12 is realized by, for example, a mouse, a keyboard, etc.
[0021] The scanner 13 scans the received paper to convert information appearing on the received paper into an image. Specifically, the scanner 13 scans a paper quotation received from a supplier to convert information appearing on the quotation into an image.
[0022] The building information processing device 14 is realized by a computer including 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, memory as a storage means, and a network interface, etc.
[0023] Functionally, as shown in Figure 1, the building information processing device 14 includes a learning data memory unit 20, a learning unit 22, a learned model memory unit 24, a character recognition unit 26, an acquisition unit 28, an accounting processing unit 30, a life cycle cost processing unit 32, and a life cycle cost memory unit 34.
[0024] When maintenance work is carried out on an existing building, the construction company that coordinates the entire work often receives quotations for various types of work from multiple suppliers. In this case, for example, while quotation A from one supplier describes maintenance work for an air conditioning system as "air conditioning system renovation work," quotation B from another supplier may describe the same maintenance work for the air conditioning system as "air conditioning equipment replacement work." However, from the perspective of facility management, these items correspond to the category of "repair costs."
[0025] An example of a subject is shown in Figure 2. As shown in Figure 2, subjects exist from major classifications to minor classifications. For example, an estimate item in one estimate may correspond to "repair costs (or repair and renewal costs)" and an estimate item in another estimate may correspond to "equipment (or maintenance costs)."
[0026] Therefore, the building information processing device 14 of this embodiment automatically journalizes transaction information (e.g., "air conditioning equipment renovation work") representing an estimate item in a construction work estimate to item information (e.g., "repair costs") representing an item. This makes it possible to journalize the transaction information described in a construction work estimate to the corresponding item information.
[0027] The specific details will be explained below.
[0028] A plurality of pieces of learning data are stored in the learning data storage unit 20. The learning data in this embodiment is data in which learning transaction information representing learning quotation items is associated with learning subject information representing subjects corresponding to the learning transaction information. The learning data is, for example, data based on past performance.
[0029] A diagram for explaining the learning data of this embodiment is shown in Fig. 3. As shown in Fig. 3, the learning data is stored such that learning transaction information and learning subject information are associated with each other for each data.
[0030] In the learning data with ID "00001" shown in FIG. 3, when a certain learning transaction information is "A1", the learning subject information corresponding to the learning transaction information is "Z1". The transaction information and the subject information are replaced with numerical data. For example, as an example of a method of replacing with numerical data, the transaction information "air conditioning equipment renovation work" is divided into "air conditioning", "equipment", "renovation", and "construction" by a known morphological analysis. Then, the occurrence frequency of each word obtained by the morphological analysis is expressed as a vector and treated as numerical data. This word occurrence frequency vector is a vector in a format such as one occurrence of "air conditioning", one occurrence of "equipment", one occurrence of "renovation", and one occurrence of "construction", and is in a format that can be input to the learned model for accounting, which will be described later. The same is true for the subject information. For example, a vector of [1,0,···,0] is set in advance to represent the subject "repair cost".
[0031] The learning unit 22 learns an accounting model for outputting item information from transaction information based on a plurality of learning data stored in the learning data storage unit 20. Then, the learning unit 22 generates a learned accounting model for outputting item information from transaction information.
[0032] Fig. 4 is a diagram for explaining the trained model of the first embodiment. As shown in Fig. 4, when transaction information to be journalized is input to the journalizing trained model M1, item information corresponding to the transaction information is output. From the journalizing trained model M1, for example, item information is output, such as the probability that it corresponds to "repair costs" is 0.8, and the probability that it corresponds to "administration", which represents the item of administrative labor costs, is 0.1. The journalizing trained model M1 is realized, for example, by a known neural network or the like.
[0033] When the trained model is realized by a known neural network, the learning unit 22 trains the neural network by known supervised machine learning based on a plurality of training data stored in the training data storage unit 20, and generates a trained neural network. Note that any method of supervised machine learning may be used.
[0034] Then, the learning unit 22 stores the generated trained model for journalization in the trained model memory unit 24.
[0035] The trained model memory unit 24 stores the trained model for journalization generated by the learning unit 22.
[0036] Character recognition unit 26 receives the image of the estimate obtained by scanner 13. Then, character recognition unit 26 recognizes character information appearing in the image of the estimate. For example, character recognition unit 26 is realized by existing OCR technology. Alternatively, character recognition unit 26 recognizes character information appearing in the image of the estimate using an existing trained model for character recognition. As a result, a character string such as "air conditioning equipment construction" is recognized as character information appearing in the image of the estimate. Then, character recognition unit 26 outputs the recognized character information to acquisition unit 28 as transaction information representing the estimate items in the estimate for construction work.
[0037] The acquisition unit 28 acquires the transaction information output from the character recognition unit 26. The acquisition unit 28 then performs an existing morphological analysis on the transaction information to obtain the morphological analysis result of the quotation item represented by the transaction information. For example, if the transaction information represents the quotation item "air conditioning equipment construction", the quotation item is divided into "air conditioning", "equipment", "renovation", and "construction" by the morphological analysis.
[0038] The journalizing processing unit 30 inputs transaction information to be journalized into a trained model for journalizing stored in the trained model memory unit 24, and outputs item information corresponding to the transaction information.
[0039] Specifically, the journalizing processing unit 30 converts the transaction information into data that can be input to the trained model. First, the journalizing processing unit 30 generates a vector that represents the transaction information based on the morphological analysis result of the transaction information obtained by the acquisition unit 28. For example, the journalizing processing unit 30 converts the transaction information into numerical data by expressing the occurrence frequency of each word obtained by the morphological analysis result by the acquisition unit 28 as a vector. The vector of the occurrence frequency of the words expresses, for example, one occurrence of "air conditioning", one occurrence of "equipment", one occurrence of "renovation", and one occurrence of "construction". As a result, the transaction information becomes data that can be input to the trained model.
[0040] Then, the accounting processing unit 30 inputs the transaction information converted into numerical data into the accounting trained model stored in the trained model memory unit 24, thereby outputting item information corresponding to the transaction information.
[0041] The journalization trained model outputs, for example, each probability as shown in FIG. 4 above. For example, the journalization processing unit 30 identifies the item information with the highest probability as the item information corresponding to the input transaction information. For example, when the transaction information "air conditioning equipment construction" is input, if the highest probability among the probabilities output from the trained model is 0.9 and that part corresponds to "repair costs," the journalization processing unit 30 journalizes the transaction information "air conditioning equipment construction" to "repair costs." Then, the journalization processing unit 30 outputs the journalization result.
[0042] In this way, the building information processing device 14 of the first embodiment automatically categorizes transaction information (e.g., "air conditioning equipment renovation work") representing an estimate item in an estimate for construction work into item information (e.g., "repair costs") representing an item. This makes it possible to categorize the transaction information described in an estimate for construction work into the corresponding item information.
[0043] The life cycle cost processing unit 32 associates the journalization results obtained by the journalization processing unit 30 with the price information and date information of the transaction information corresponding to the journalization results, and stores them in the life cycle cost storage unit 34.
[0044] Data on the life cycle cost of each building is stored in the life cycle cost storage unit 34. The life cycle cost processing unit 32 processes the data stored in the life cycle cost storage unit 34 to visualize the life cycle cost of the target building.
[0045] Fig. 5 shows a graph of the life cycle cost (LCC) of a building. In the example shown in Fig. 5, rent is recorded as income for a building, and various expenses such as repair and renewal costs are recorded as expenditures. The above-mentioned journalization processing by the journalization processing unit 30 journalizes the transaction information written on the estimate into appropriate item information, so that the life cycle cost as shown in Fig. 5 can be automatically generated.
[0046] The display unit 16 displays the information output from the building information processing device 14. The display unit 16 is realized by, for example, a display or the like. For example, the display unit 16 displays the result of the life cycle cost obtained by the life cycle cost processing unit 32. The user checks the result displayed on the display unit 16 and uses it for, for example, facility management of the building.
[0047] The journalizing unit 30 may output, for example, the subject information together with the probability output from the trained model, instead of outputting the journalized information. In this case, the user refers to each of the presented probabilities and each of the subject information, and selects the subject information that seems appropriate from among them.
[0048] <Function of the construction journal entry system>
[0049] Next, the operation of the construction journalization system 10 of the first embodiment will be described with reference to the drawings. When the building information processing device 14 of the construction journalization system 10 receives learning data via the input unit 12, it stores the learning data in the learning data storage unit 20. Then, when the building information processing device 14 of the construction journalization system 10 receives an instruction signal to start the learning process, it executes the learning process routine shown in FIG.
[0050] <Learning processing routine>
[0051] In step S100 , the learning unit 22 acquires a plurality of pieces of learning data stored in the learning data storage unit 20 .
[0052] In step S102, the learning unit 22 generates a learned accounting model for outputting item information from transaction information by training the accounting model based on the multiple learning data acquired in step S100 above.
[0053] In step S104, the learning unit 22 stores the trained model for journalization generated in step S102 above in the trained model storage unit 24, and ends the learning processing routine.
[0054] Next, when the building information processing device 14 of the construction journalization system 10 receives an image of the estimate to be journalized input from the scanner 13, it executes a construction journalization processing routine shown in FIG.
[0055] <Construction journal entry processing routine>
[0056] In step S200, the character recognition unit 26 acquires an image of the estimate obtained by the scanner 13.
[0057] In step S202, the character recognition unit 26 acquires character information from the image of the estimate acquired in step S200, and regards the character information as transaction information. The character information is acquired based on, for example, a known OCR technique.
[0058] In step S204, the acquisition unit 28 acquires the transaction information acquired in step S202. Then, the acquisition unit 28 performs an existing morphological analysis on the transaction information to obtain a morphological analysis result of the quotation item represented by the transaction information.
[0059] In step S206, the journalizing processing unit 30 converts the morphological analysis result obtained in step S204 into numerical data that can be input to the journalizing trained model.
[0060] In step S208, the journalizing processor 30 reads out the journalizing trained model stored in the trained model storage unit 24. Then, the journalizing processor 30 inputs the numerical data obtained in step S206 above into the journalizing trained model, thereby outputting item information corresponding to the transaction information.
[0061] In step S210, the accounting processor 30 outputs the item information obtained in step S208 as a result.
[0062] The life cycle cost processing unit 32 associates the journalization results obtained by the journalization processing unit 30 with the price information and date information of the transaction information corresponding to the journalization results, and stores them in the life cycle cost storage unit 34.
[0063] Then, the life cycle cost processing unit 32 processes the data stored in the life cycle cost storage unit 34 to visualize the life cycle cost of the target building.
[0064] The display unit 16 displays the result of the life cycle cost obtained by the life cycle cost processing unit 32. The user checks the result displayed on the display unit 16 and uses it for, for example, facility management of the building.
[0065] As described above in detail, the building information processing device of the first embodiment outputs item information corresponding to the transaction information by inputting the transaction information representing the estimate items in the estimate for construction work to a pre-trained journalization model for outputting item information representing the item corresponding to the transaction information from the transaction information. This makes it possible to journalize the transaction information described in the estimate for construction work into the corresponding item information.
[0066] [Second embodiment]
[0067] Next, a second embodiment will be described. The second embodiment is different from the first embodiment in that a trained model for anomaly detection is used to determine whether details of estimate items in an estimate for construction work on a target building are consistent with BIM information of the target building.
[0068] Fig. 8 is a block diagram showing an example of the configuration of the construction journalization system 210 according to the second embodiment. The construction journalization system 210 can be functionally represented as a configuration including an input unit 12, a scanner 13, a building information processing device 214, and a display unit 16, as shown in Fig. 8.
[0069] Functionally, as shown in Figure 8, the building information processing device 214 includes a learning data memory unit 220, a learning unit 222, a learned model memory unit 224, a BIM information memory unit 225, a character recognition unit 26, an acquisition unit 28, an accounting processing unit 230, and an anomaly detection unit 231.
[0070] Two types of learning data are stored in the learning data storage unit 220. Specifically, journalization learning data for generating a journalization trained model described later, and anomaly detection learning data for generating an anomaly detection trained model described later are stored in the learning data storage unit 220.
[0071] FIG. 9 shows an example of journalization learning data and an example of anomaly detection learning data.
[0072] As shown in FIG. 9, the journalization learning data S is stored such that learning transaction information, learning subject information, and learning tag information are associated with each other for each data.
[0073] The tag information is information that indicates the type of transaction information. For example, if the transaction information is an estimate item such as "air conditioning equipment construction X building 2nd floor men's restroom electrical equipment repair fluorescent light replacement", the transaction information is classified under the heading "repair cost". Furthermore, in the second embodiment, tag information such as "fluorescent light", "2F", "restroom", and "electricity" is added to information such as "2nd floor men's restroom electrical equipment repair fluorescent light replacement". This allows the estimate item in the estimate to identify which part of the building it corresponds to for what repair. The tag information is also information that indicates the details of the transaction information, and the building information processing device 214 of the second embodiment uses this tag information to determine whether the estimate submitted by the supplier is appropriate. Details will be described later.
[0074] For this reason, the journalization trained model of the second embodiment outputs the probability of the subject information corresponding to the transaction information, as well as the probability of the tag information corresponding to the transaction information. Figure 10 shows an example of the journalization trained model of the second embodiment. As shown in Figure 10, the journalization trained model M2-1 of the second embodiment outputs the probability of the subject information corresponding to the transaction information and the probability of the tag information corresponding to the transaction information.
[0075] 9, the anomaly detection learning data I is stored with the learning BIM information, the learning tag information, and the learning reliability associated with each other for each data. The BIM information is information that represents the BIM (Building Information Modeling) of the target building.
[0076] When the BIM information of the target building and the tag information corresponding to the transaction information are input, the trained model for anomaly detection of the second embodiment outputs the reliability of the transaction information corresponding to the tag information. FIG. 11 shows an example of the trained model for anomaly detection of the second embodiment. As shown in FIG. 11, the trained model for anomaly detection M2-2 of the second embodiment outputs the reliability of the transaction information corresponding to the tag information. This reliability represents the degree to which the estimated items of the estimate submitted by the supplier are reliable, by matching the tag information obtained from the transaction information of the estimated items of the estimate submitted by the supplier with the BIM information of the building that the estimate targets. Details will be described later.
[0077] The learning unit 222 learns a model for outputting item information and tag information from transaction information based on a plurality of journalization learning data stored in the learning data storage unit 220. Then, the learning unit 222 generates a journalization trained model for outputting item information and tag information from transaction information.
[0078] Furthermore, the learning unit 222 learns a model for outputting the reliability of transaction information corresponding to the tag information from the BIM information and tag information, based on a plurality of learning data for anomaly detection stored in the learning data storage unit 220. Then, the learning unit 222 generates a trained model for anomaly detection for outputting the reliability of the transaction information of the tag information from the BIM information and tag information.
[0079] The learning unit 222 stores the trained model for journalization and the trained model for anomaly detection in the trained model memory unit 224.
[0080] The journalizing processing unit 230 acquires the probability of the item information and the probability of the tag information by inputting the numerical data of the transaction information acquired by the acquisition unit 28 into the journalizing trained model stored in the trained model storage unit 224. Then, the journalizing processing unit 230 journalizes the item information corresponding to the highest probability among the item information probabilities as the item information corresponding to the input transaction information. In addition, the journalizing processing unit 230 identifies the tag information corresponding to the highest probability among the tag information probabilities as the tag information corresponding to the input transaction information.
[0081] The anomaly detection unit 231 reads out the BIM model of the building that is the target of the construction work from the BIM information storage unit 225. Then, the anomaly detection unit 231 acquires BIM information representing information on each part of the building from the read BIM model. Attribute information is assigned to each part of the BIM model. For example, attribute information such as "2F," "men's toilet," and "3 fluorescent lights" is assigned in advance to the toilet room on the second floor of the building represented by the BIM model.
[0082] Therefore, the anomaly detection unit 231 acquires BIM information that indicates attribute information of each location of the BIM model based on the tag information identified by the accounting processing unit 230. For example, when the tag information is information such as "2F", "men's restroom", and "3 fluorescent lights", the anomaly detection unit 231 acquires the BIM information of "2F" and "men's restroom" from the BIM model.
[0083] Next, the anomaly detection unit 231 uses a known method to convert the subject information identified by the journalization processing unit 230 into numerical data that can be input to the trained model for anomaly detection. Also, the anomaly detection unit 231 uses a known method to convert the tag information identified by the journalization processing unit 230 into numerical data that can be input to the trained model for anomaly detection.
[0084] Then, the anomaly detection unit 231 outputs the reliability of the transaction information acquired by the acquisition unit 28 by inputting the numerical data of the tag information identified by the accounting processing unit 230 and the numerical data of the acquired BIM information into the trained model for anomaly detection stored in the trained model memory unit 224.
[0085] As shown in Figure 11, the trained model for anomaly detection M2-2 outputs the reliability of the transaction information. This reliability is the reliability of the quotation items in the quotation submitted by the supplier, so if there is an error in the quotation items, the reliability will be low. For example, if the tag information is information such as "2F", "men's restroom", and "4 fluorescent lights", while the BIM information represents "2F", "men's restroom", and "3 fluorescent lights", there is an error in the quotation items, so the reliability will be low.
[0086] The anomaly detection unit 231 outputs an alert when the reliability output from the trained model for anomaly detection is equal to or less than a threshold. As described above, when the reliability is equal to or less than a threshold, there is a high possibility that an error exists in the estimated item. Therefore, the anomaly detection unit 231 outputs an alert when the reliability is equal to or less than a threshold.
[0087] The display unit 16 displays the alert output from the anomaly detection unit 231. The user checks the alert displayed on the display unit 16 and confirms whether or not there are any errors in the quotation items submitted by the supplier.
[0088] <Function of the construction journal entry system>
[0089] Next, the operation of the construction journalization system 210 of the second embodiment will be described with reference to the drawings. When the building information processing device 214 of the construction journalization system 210 receives journalization learning data and anomaly detection learning data via the input unit 12, it stores the data in the learning data storage unit 220. Then, when the building information processing device 214 of the construction journalization system 210 receives an instruction signal to start the learning process, it executes a learning process routine similar to the learning process routine shown in FIG. 6 above.
[0090] Specifically, the learning unit 222 generates a trained model for journalization for outputting item information and tag information from transaction information, based on a plurality of journalization learning data stored in the learning data storage unit 220. The learning unit 222 also generates a trained model for anomaly detection for outputting the reliability of the transaction information of the tag information from the BIM information and tag information, based on a plurality of anomaly detection learning data stored in the learning data storage unit 220. Then, the learning unit 222 stores the trained model for journalization and the trained model for anomaly detection in the trained model storage unit 224.
[0091] Next, when the building information processing device 214 of the construction journalization system 210 receives an image of the estimate to be journalized input from the scanner 13, it executes a construction journalization processing routine shown in FIG.
[0092] The processes from step S200 to step S206 are executed in the same manner as in the first embodiment.
[0093] In step S308, the accounting processing unit 230 acquires item information and tag information corresponding to the transaction information by inputting the numerical data of the transaction information acquired by the acquisition unit 28 into the learned model for accounting stored in the learned model memory unit 224.
[0094] In step S310, the abnormality detection unit 231 acquires the BIM information stored in the BIM information storage unit 225 based on the tag information acquired in step S308.
[0095] In step S311, the anomaly detection unit 231 uses a known method to convert the subject information acquired in step S310 into numerical data that can be input to the trained model for anomaly detection. Also, the anomaly detection unit 231 uses a known method to convert the tag information acquired in step S310 into numerical data that can be input to the trained model for anomaly detection.
[0096] In step S312, the anomaly detection unit 231 obtains the reliability of the transaction information by inputting the numerical data of the tag information and the numerical data of the BIM information obtained in step S311 above into the trained model for anomaly detection stored in the trained model memory unit 224.
[0097] In step S314, the anomaly detection unit 231 determines whether the reliability obtained in step S312 is equal to or less than a threshold. If the reliability is equal to or less than the threshold, the process proceeds to step S316. On the other hand, if the reliability is greater than the threshold, the process ends.
[0098] In step S316, the abnormality detection unit 231 outputs an alert.
[0099] The display unit 16 displays the alert output from the anomaly detection unit 231. The user checks the alert displayed on the display unit 16 and confirms whether or not there are any errors in the quotation items in the quotation submitted by the supplier.
[0100] For example, when the reliability is greater than a threshold, the anomaly detection unit 231 may reflect the tag information in the data of the life cycle cost described in the first embodiment. For example, when the reliability is greater than a threshold, the anomaly detection unit 231 reflects the tag information TAG obtained by the accounting processing unit 230 in the data of the life cycle cost as shown in Fig. 13. This allows the data of the life cycle cost to be automatically updated.
[0101] Other configurations and operations of the construction journalization system according to the second embodiment are similar to those of the first embodiment, and therefore will not be described.
[0102] As described above in detail, the building information processing device of the construction journalization system of the second embodiment acquires tag information indicating the type of transaction information, and acquires BIM information indicating information on each part of the building from the BIM model of the building that is the target of construction work.The building information processing device then inputs the tag information and BIM information into a trained model for anomaly detection that outputs the reliability of the transaction information from the tag information and BIM information, thereby outputting the reliability of the transaction information indicating the estimated items of the estimate, and outputs an alert if the reliability is equal to or lower than a threshold.This makes it easier to find errors in the estimated items of the estimate.
[0103] The present invention is not limited to the above-described embodiment, and various modifications and applications are possible without departing from the spirit and scope of the present invention.
[0104] For example, in each of the above embodiments, a neural network or the like is used as an example of a trained model, but the present invention is not limited to this. Any model obtained by machine learning may be used.
[0105] In addition, a trained model for predicting the life cycle cost in the first embodiment may be further generated.
[0106] In addition, although the above describes an embodiment in which the program according to the present invention is pre-stored (installed) in a memory unit (not shown), the program according to the present invention can also be provided in a form in which it is recorded on a recording medium such as a CD-ROM, a DVD-ROM, or a micro SD card. [Explanation of symbols]
[0107] 10,210 Construction accounting system 12 Input section 13. Scanner 14,214 Building information processing equipment 16 Display 20,220 Learning data storage unit 22,222 Learning Department 24,224 trained model memory 28 Acquisition Department 30 Journal entry processing section 32 Life Cycle Cost Processing Division 34 Life cycle cost memory section 225 BIM Information Storage Unit 231 Anomaly detection unit
Claims
1. An acquisition unit for acquiring transaction information representing estimate items in an estimate for a construction work; a journalization processing unit that outputs the item information corresponding to the transaction information acquired by the acquisition unit by inputting the transaction information acquired by the acquisition unit into a pre-trained journalization model for outputting item information representing an item corresponding to the transaction information from the transaction information; Including, The journalization trained model further outputs tag information indicating the type of the transaction information, Obtaining BIM information representing information on each part of a building from a BIM model of a building to be constructed; The tag information and the BIM information are input to a trained model for anomaly detection for outputting a reliability of the transaction information from the tag information and the BIM information, thereby outputting a reliability of the transaction information acquired by the acquisition unit; and an anomaly detection unit that outputs an alert when the reliability is equal to or lower than a threshold. Building information processing device.
2. An acquisition unit for acquiring transaction information representing estimate items in an estimate for a construction work; a journalization processing unit that outputs the item information corresponding to the transaction information acquired by the acquisition unit by inputting the transaction information acquired by the acquisition unit into a pre-trained journalization model for outputting item information representing an item corresponding to the transaction information from the transaction information; a life cycle cost processing unit that associates the item information, which is the journalization result obtained by the journalization processing unit, with price information of the transaction information corresponding to the journalization result, and date information, and stores the associated information in a life cycle cost storage unit; The life cycle cost processing unit processes various information stored in the life cycle cost storage unit to visualize the life cycle cost of the building that is the target of the construction work, and outputs the visualized life cycle cost. Building information processing device.
Citation Information
Patent Citations
Magnetic recording medium
JP1986061229A
Building life-cycle analysis system
JP1998046832A
Construction work management system, construction work management method, and construction work management program
JP2006072547A
Classification model update support device, method, and program
JP2014092878A
Accounting input system, terminal equipment, server device, method, and program
JP2016194802A
