Condition determination device, condition determination method, and condition determination program

Abstract

To reduce variations in judgment due to experience and individual differences, and to manufacture concrete products of consistent quality. [Solution] A condition determination device according to one aspect of this embodiment includes an acquisition unit and an estimation unit. It acquires first production data relating to the constituent materials of concrete to be processed, and first element information that fluctuates depending on the environment or period when manufacturing the concrete product. The estimation unit takes second production data relating to the constituent materials of concrete and second element information that fluctuates depending on the environment or period when manufacturing the concrete product as input data, and uses a trained model that has learned the mixing ratio or mixing conditions of the constituent materials when the concrete product was manufactured under the conditions based on the second production data and the second element information as ground truth data, to input the first production data and the first element information into the trained model and estimate the mixing ratio or mixing conditions of the constituent materials in the first production data.

Description

Technical Field

[0003] , ,

[0001] The present invention relates to a condition determination device, a condition determination method, and a condition determination program.

Background Art

[0002] Concrete products, particularly concrete with a slump value of about "3 to 5 cm" used for precast concrete piles, so-called stiff concrete, or so-called super-stiff concrete that does not easily deform under its own weight, are more easily affected by the degree of voids in the concrete than ordinary concrete. Therefore, it is difficult to evaluate the workability, including the consistency of the concrete or the compactability indicating the degree of ease of work in compacting the concrete, solely by the slump value. Therefore, there is a method of outputting blending information regarding the blending of constituent materials by a neural network learned using the physical property information, property information, and blending information of concrete (see, for example, Patent Document 1 and Patent Document 2).

Prior Art Documents

Patent Documents

[0003]

Patent Document 1

Patent Document 2

Summary of the Invention

Problems to be Solved by the Invention

[0004] However, in reality, due to factors such as the season and weather during manufacturing and the state of materials, the blending ratio of constituent materials and the mixing conditions change, so it will be determined based on the experience of experienced workers. Therefore, there is a problem that the quality of concrete varies due to empirical values and individual differences.

[0005] The present invention aims to reduce variability in judgment due to experience and individual differences, and to manufacture concrete products of consistent quality. [Means for solving the problem]

[0006] (1) A condition determination device according to one aspect of the present invention includes an acquisition unit and an estimation unit. It acquires first production data relating to the constituent materials of concrete to be processed and first element information that fluctuates depending on the environment or period when the concrete product is manufactured. The estimation unit takes second production data relating to the constituent materials of concrete and second element information that fluctuates depending on the environment or period when the concrete product is manufactured as input data, and uses a trained model that has learned the mixing ratio or mixing conditions of the constituent materials when the concrete product is manufactured under the conditions based on the second production data and the second element information as ground truth data, to input the first production data and the first element information into the trained model and estimate the mixing ratio or mixing conditions of the constituent materials in the first production data. According to the above configuration (1), by simply inputting production data and elemental information, the mixing ratio of the constituent materials of the concrete and the mixing conditions can be estimated using a trained model. This reduces variability in judgments due to experience and individual differences, and enables the production of concrete products of consistent quality.

[0007] (2) In some embodiments, in (1) above, The first element information and the second element information each include at least one of a basic element, a long-term element, a medium-term element, and a short-term element, wherein the basic element relates to inspection standards for the concrete product, the long-term element relates to at least one of the constituent materials of the concrete used to manufacture the concrete product and the characteristics of the factory, the medium-term element relates to at least one of the weather, season, and lot of the constituent materials used when manufacturing the concrete product, and the short-term element relates to the weather when manufacturing the concrete product. According to the above configuration (2), by classifying the elements as basic elements, long-term elements, medium-term elements, and short-term elements and reflecting the conditions in the parameters of the trained model, it is possible to generate a trained model that takes into account more detailed conditions, and to manufacture concrete products of appropriate quality according to the conditions.

[0008] (3) In some embodiments, in (2) above, When the estimation unit determines the blending ratio or mixing conditions on an annual or multi-year basis, it sets a higher weighting for the long-term factors so that they are reflected more than the medium-term or short-term factors. According to the above configuration (3), when considering the blending ratio or mixing conditions over a relatively long period, the optimal blending ratio or mixing conditions can be estimated by conditioned to give more consideration to long-term factors.

[0009] (4) In some embodiments, in the above configuration (2), When the estimation unit determines the blending ratio or mixing conditions on a weekly, weekly, monthly, or monthly basis, it sets a higher weighting for the medium-term factors so that they are reflected more than the long-term or short-term factors. According to the above configuration (4), when considering the blending ratio or mixing conditions over a shorter period than the long term, i.e., the medium term, the optimal blending ratio and mixing conditions can be estimated by conditioned to give more consideration to the medium-term factors.

[0010] (5) In some embodiments, in the above configuration (2), When the estimation unit determines the blending ratio or mixing conditions on a batch, hourly, or daily basis, it sets a higher weighting for the short-term factors so that they are reflected more than the long-term or medium-term factors. According to the above configuration (5), when considering the blending ratio or mixing conditions over a shorter period than the medium term, i.e., the short term, the optimal blending ratio and mixing conditions can be estimated by conditioned to give more consideration to the short-term factors.

[0011] (6) In one aspect of the present invention, the condition determination method is as follows: an acquisition means acquires first production data relating to the constituent materials of the concrete to be processed and first element information that varies depending on the environment or period when the concrete product is manufactured; an estimation means takes second production data relating to the constituent materials of the concrete and second element information that varies depending on the environment or period when the concrete product is manufactured as input data; and uses a trained model that has learned the mixing ratio or mixing conditions of the constituent materials when the concrete product is manufactured under the conditions based on the second production data and the second element information as ground truth data, inputs the first production data and the first element information into the trained model, and estimates the mixing ratio or mixing conditions of the constituent materials in the first production data. According to the above configuration (6), by simply inputting production data and element information, the mixing ratio of the constituent materials of the concrete or the mixing conditions can be estimated using a trained model. This reduces variability in judgments due to experience and individual differences, and enables the production of concrete products of consistent quality.

[0012] (7) A condition determination program according to one aspect of the present invention uses a computer as an acquisition means for acquiring first production data relating to the constituent materials of a concrete to be processed and first element information that varies depending on the environment or period when the concrete product is manufactured, and takes second production data relating to the constituent materials of a concrete and second element information that varies depending on the environment or period when the concrete product is manufactured as input data, and uses a trained model that has learned the mixing ratio or mixing conditions of the constituent materials when the concrete product is manufactured under the conditions based on the second production data and the second element information as ground truth data, to input the first production data and the first element information into the trained model and use it as an estimation means for estimating the mixing ratio or mixing conditions of the constituent materials in the first production data. According to the above configuration (7), by simply inputting production data and element information, it is possible to estimate the mixing ratio of the constituent materials of concrete or the mixing conditions using a learned model. Therefore, variations in judgment due to empirical values and individual differences can be reduced, and concrete products of a certain quality can be manufactured.

Effect of the Invention

[0013] According to the present invention, variations in judgment due to empirical values and individual differences can be reduced, and concrete products of a certain quality can be manufactured.

Brief Description of the Drawings

[0014] [Figure 1] FIG. 1 is a block diagram showing a condition determination device according to the present embodiment. [Figure 2] FIG. 2 is a conceptual diagram showing the learning and inference times of a learned model. [Figure 3] FIG. 3 is a flowchart showing an operation example of the condition determination device according to the present embodiment.

Mode for Carrying Out the Invention

[0015] Hereinafter, a condition determination device, a condition determination method, and a condition determination program according to an embodiment of the present invention will be described in detail with reference to the drawings. In the following embodiments, the same reference numerals are used for parts that perform the same operations, and repeated descriptions will be omitted.

[0016] The condition determination device according to the present embodiment will be described with reference to the conceptual diagram of FIG. 1. The condition determination device 1 shown in FIG. 1 includes a processing circuit 11, a storage unit 12, and a communication interface 13, which are connected via a bus, respectively.

[0017] The processing circuit 11 is a processor such as a CPU (Central Processing Unit), GPU (Graphics Processing Unit), NPU (Neural network Processing Unit), ASIC (Application Specific Integrated Circuit), or FPGA (Field Programmable Gate Array), and is realized by any one or a combination of these. The processing circuit 11 includes an acquisition unit 111, a learning unit 112, an estimation unit 113, and an output unit 114.

[0018] The acquisition unit 111 acquires first production data regarding the constituent materials of the concrete to be processed, that is, the concrete to be kneaded from now on, and first element information that varies depending on the environment or period when manufacturing the concrete product. Note that when manufacturing the concrete product, for example, it refers to the time of kneading, placing, or curing the concrete used in the manufacturing. The learning unit 112 uses, as input data, second production data regarding the constituent materials of the concrete and second element information that varies depending on the environment and period when manufacturing the concrete product, and uses learning data in which at least one of the mixing ratio of the constituent materials and the kneading conditions when manufacturing the concrete product under the conditions based on the second production data and the second element information is used as correct data to learn a machine learning model and generate a learned model. The second production data is data regarding the constituent materials of the concrete kneaded in the past. It is assumed that the first production data and the second production data, and the first element information and the second element information are each the same type of data. For example, the first production data and the second production data are information regarding the material temperature and the mixing ratio of the materials.

[0019] The estimation unit 113 uses the learned model to input the first production data and the first element information into the learned model, and estimates the mixing ratio of the constituent materials and the kneading conditions in the first production data. The output unit 114 outputs the mixing ratio of the constituent materials and the kneading conditions in the first production data to the outside via the communication interface 13.

[0020] The storage unit 12 consists of an HDD (Hard Disk Drive), an SSD (Solid State Drive), and the like. The storage unit 12 stores sensor data, captured images, inspection data, and trained models that can be measured in the equipment and environment related to mixing materials.

[0021] The communication interface 13 is an interface for sending and receiving data with the device shown in Figure 1, which conforms to a predetermined communication standard. The communication standard may be a wireless network such as a Wi-Fi® compliant wireless LAN (Local Area Network) or Bluetooth®, or a wired network using signal lines, LAN cables, etc.

[0022] In this embodiment, the condition determination device 1 is shown to include a learning unit 112, but it is not limited to this. The learning unit 112 may be provided on an external device (e.g., an external server), and the external device may perform the machine learning model training process and store the generated trained model. In this case, the condition determination device 1 can use the trained model stored on the external device, and the estimation unit 113 can estimate the mixing ratio of the constituent materials and the mixing conditions in the first production data.

[0023] Next, the training and inference processes of the trained model used in the condition determination device 1 according to this embodiment will be explained with reference to the conceptual diagram in Figure 2. During training of the trained model, the learning unit 112 inputs production data 21 (also called second production data) and element information 22 (also called second element information) as input data to the machine learning model 20, and inputs the mixing ratio 23 of the constituent materials of the concrete and the mixing conditions 24 as ground truth data, and trains the machine learning model 20 through supervised learning.

[0024] The machine learning model 20 in this embodiment is assumed to be a convolutional neural network including one or more convolutional layers, but other network models commonly used in machine learning, such as decision trees, random forests, and support vector machines, may also be used. Furthermore, the training method for the machine learning model 20 using supervised learning can be a general training method, such as training to minimize the loss between the output from the machine learning model 20 and the ground truth data for the input data, so a detailed explanation is omitted here. As a result of the training being completed, a trained model 25 is generated.

[0025] Production data 21 includes information on the types of constituent materials of concrete, the amount produced, and information on concrete products. The types of constituent materials of concrete include cement, coarse aggregate, fine aggregate, water, admixtures, and additives (e.g., fly ash). Admixtures may include those that improve the workability and freeze-thaw resistance of concrete, such as air entraining (AE) agents; those that increase strength, such as water-reducing agents; those that adjust the hardening or setting time, such as hardening accelerators and setting retarders; and those that provide waterproofing effects, such as waterproofing agents. The amount produced may include, for example, the weight or volume of concrete produced per day. Information on concrete products includes, for example, information on products using concrete, such as concrete piles, Hume pipes, concrete poles, and precast members.

[0026] Element information 22 includes information on basic elements 221, long-term elements 222, medium-term elements 223, and short-term elements 224.

[0027] Basic element 221 is elemental information related to inspection standards for concrete products. Specifically, it includes information on inspection standards for compressive strength tests and workability tests for mixed concrete, or information on inspection standards for visual inspection and bending strength tests for manufactured concrete products. The compressive strength test is a standard value for the maximum load indicated by the testing machine before the specimen breaks, when a load is applied at a uniform rate without impacting the specimen. The workability test evaluates inspection items such as slump value and air content. Workability is influenced by the resistance of mixed concrete to deformation and flow, and resistance to material segregation. Visual inspection evaluates inspection items such as surface smoothness, gloss, color tone, and unevenness. The bending strength test is a standard value for the maximum load indicated by the testing machine before the specimen breaks, when a load is applied at a uniform rate without impacting the specimen using a so-called three-point loading device or a central point loading device, while adjusting the rate of increase in edge stress.

[0028] Long-term element 222 refers to conditions related to the characteristics of the factory that manufactures the constituent materials and concrete products (location, equipment, surrounding environment, etc.). Specifically, this includes information about the factory's manufacturing equipment, information about the climate at the factory's location (whether it is warm or cold, etc.), and information about the type of constituent materials and their source (such as information about water hardness). Furthermore, long-term element 222 assumes conditions that are fixed (specified) on an annual or multi-year basis.

[0029] Medium-term element 223 includes conditions related to at least one of the following: weather, season, and lot of constituent material when manufacturing concrete products, as well as the wear and tear of equipment parts and the date of replacement of parts. Specifically, this includes information on weather conditions on a monthly basis, information on seasons (spring, summer, autumn, winter), and information on the lot of the concrete's constituent material. Lot information includes test results specified in JIS regarding material quality (such as particle size, moisture content, and compressive strength), the manufacturing date, or the slump value and compressive strength of concrete using that material. Equipment parts include mixing blades of mixers and protective materials (liners) for mixing containers. Furthermore, medium-term element 223 assumes conditions that are fixed (specified) on a weekly, weekly, monthly, or monthly basis.

[0030] Short-term element 224 is information about the weather during the manufacturing of concrete products. For example, this includes atmospheric information such as whether the day of manufacture is sunny, cloudy, rainy, or snowy, as well as temperature, humidity, moisture content in the constituent materials (moisture content, surface moisture content, etc.), material temperature, and water temperature. Furthermore, short-term element 224 assumes conditions that are fixed (specified) on a batch basis, an hourly basis, or a daily basis.

[0031] The mixing ratio 23 is the mixing ratio of the types of concrete materials, such as the cement, coarse aggregate, fine aggregate, water, admixtures, and other admixtures mentioned above. The mixing conditions 24 include information about the mixing time, such as the time from the start of mixing until the various ingredients are added.

[0032] The training data, which consists of pairs of input data and correct answer data, can be, for example, stored as elemental information from when concrete products that actually passed inspection were manufactured, the mixing ratio of the constituent materials of the concrete, and the mixing conditions, and used during training.

[0033] On the other hand, when inference is performed using the trained model 25, the production data 26 (also called the first production data) and element information 27 (also called the first element information) to be processed are input to the trained model 25. The element information 27 includes various elements (basic elements 271, long-term elements 272, medium-term elements 273, and short-term elements 274) at the time of processing. The trained model 25 outputs the optimal blending ratio 28 and mixing conditions 29 according to the production data 26 as the inference result.

[0034] Next, an example of the processing of the condition determination device 1 according to this embodiment will be described with reference to the flowchart in Figure 3.

[0035] In step SA1, the acquisition unit 111 acquires the production data and element information to be processed. For example, the production data should include the constituent materials for manufacturing the desired concrete product, which have been input by the user into a control device (not shown) that performs material mixing, and the production quantity, if necessary. The element information is assumed to have been pre-input by the user into the control device, and this information can be used. For short-term elements among the element information, weather information can be acquired, for example, from the internet, at the time the condition determination device 1 performs processing. For temperature (air temperature, material temperature, water temperature, etc.) and humidity, values ​​measured in the factory by temperature sensors and humidity sensors may be used.

[0036] In step SA2, the estimation unit 113 inputs the production data and element information acquired in step SA1 into the trained model, and the blending ratio of the constituent materials and the mixing conditions are generated as inference results from the trained model.

[0037] In step SA3, the output unit 114 outputs the mixing ratio and mixing conditions to an external device. For example, the mixing ratio and mixing conditions may be output to a display for the purpose of presenting them to the user, while simultaneously being output to a control device to perform mixing of the materials under the estimated conditions.

[0038] Furthermore, when inputting production data and element information to be processed into a trained model, the types of element information may be weighted according to the period over which the blending ratios and mixing conditions are determined. For example, when determining the blending ratios and mixing conditions of constituent materials on a weekly or monthly basis, the long-term elements among the element information may be weighted more highly so as to reflect at least one of the other elements. Alternatively, when determining the blending ratios and mixing conditions of constituent materials on a batch basis, the medium-term elements among the element information may be weighted more highly so as to reflect at least one of the other elements. Alternatively, when determining the blending ratios and mixing conditions of constituent materials on a daily basis, the short-term elements may be weighted more highly so as to reflect at least one of the other elements.

[0039] Furthermore, the above embodiment shows an example in which the mixing ratios and mixing conditions of the constituent materials of concrete are input as ground truth data, and a trained model is generated that estimates the mixing ratios and mixing conditions of the constituent materials, but it is not limited to this. For example, a trained model may be generated that estimates either the mixing ratios or the mixing conditions of the constituent materials. Alternatively, a trained model may be generated that estimates the mixing ratios of the constituent materials of concrete by inputting only the mixing ratios of the constituent materials of concrete as ground truth data. A trained model may be generated that estimates the mixing conditions of the constituent materials by inputting only the mixing conditions of concrete as ground truth data.

[0040] According to the embodiment described above, a machine learning model is trained using second production data relating to the constituent materials of concrete and second element information that fluctuates depending on the environment or period during the manufacture of concrete products as input data, and training data that uses the mixing ratio or mixing conditions of the constituent materials when concrete products are manufactured under conditions based on the second production data and second element information as ground truth data, thereby generating a trained model. Using this trained model, the first production data and first element information are input into the trained model to estimate the mixing ratio or mixing conditions of the constituent materials of concrete in the first production data. This allows the system to estimate the mixing ratio or mixing conditions of the concrete's constituent materials using a trained model simply by inputting production data and elemental information. Therefore, it reduces variability in judgments due to experience and individual differences, and enables the production of concrete products of consistent quality that have the desired properties—that is, properties that can produce products that meet product inspection requirements, properties that are easy to manufacture, or properties that result in a low defect rate.

[0041] In this embodiment, we assume the mixing of concrete components, but it may also be used for mixing mortar materials or rubber materials. In other words, the condition determination device according to this embodiment can be similarly applied to any mixing process in which adjusting the mixing ratio of components and mixing conditions is effective.

[0042] The instructions shown in the processing procedure described in the above-described embodiment can be executed based on a software program. A general-purpose computer system can also obtain the same effect as the identification device described above by pre-storing this program on a recording medium and reading the stored program. Furthermore, the recording medium in this embodiment is not limited to a medium independent of the computer or embedded system, but also includes a recording medium on which a program transmitted via a LAN, the Internet, etc., has been downloaded and stored or temporarily stored. [Explanation of Symbols]

[0043] 1...Condition determination device, 11...Processing circuit, 12...Storage unit, 13...Communication interface, 20...Machine learning model, 21,26...Production data, 22,27...Element information, 23,28...Blending ratio, 24,29...Mixing conditions, 111...Acquisition unit, 112...Learning unit, 113...Estimation unit, 114...Output unit, 221,271...Basic elements, 222,272...Long-term elements, 223,273...Medium-term elements, 224,274...Short-term elements.

Claims

1. An acquisition unit that acquires first production data relating to the constituent materials of the concrete to be processed, and first element information that fluctuates depending on the environment or period during the manufacture of the concrete product, An estimation unit takes second production data relating to the constituent materials of concrete and second element information that fluctuates depending on the environment or period during the manufacture of the concrete product as input data, and uses a trained model that has learned the mixing ratio or mixing conditions of the constituent materials when the concrete product was manufactured under conditions based on the second production data and the second element information as ground truth data, to input the first production data and the first element information into the trained model and estimate the mixing ratio or mixing conditions of the constituent materials in the first production data. A condition determination device equipped with the following:

2. The condition determination device according to claim 1, wherein the first element information and the second element information include at least one of basic elements, long-term elements, medium-term elements, and short-term elements, the basic element relating to inspection standards for the concrete product, the long-term element relating to at least one of the constituent materials of the concrete used to manufacture the concrete product and the characteristics of the factory, the medium-term element relating to at least one of the weather, season, and lot of the constituent materials used when manufacturing the concrete product, and the short-term element relating to the weather when manufacturing the concrete product.

3. The condition determination device according to claim 2, wherein when the estimation unit determines the blending ratio or mixing conditions on an annual or multi-year basis, it sets the weighting higher so that the long-term elements are reflected more than the medium-term elements or the short-term elements.

4. The condition determination device according to claim 2, wherein when the estimation unit determines the blending ratio or mixing conditions on a weekly, weekly, monthly, or monthly basis, it sets a higher weighting for the medium-term elements so that they are reflected more than the long-term elements or the short-term elements.

5. The condition determination device according to claim 2, wherein when the estimation unit determines the blending ratio or mixing conditions on a batch, hourly, or daily basis, it sets the weighting higher so that the short-term elements are reflected more than the long-term or medium-term elements.

6. The acquisition means acquires first production data relating to the constituent materials of the concrete to be processed, and first element information that fluctuates depending on the environment or period during the manufacture of the concrete product. Condition determination method wherein the estimation means takes second production data relating to the constituent materials of concrete and second element information that fluctuates depending on the environment or period during the manufacture of the concrete product as input data, and uses a trained model that has learned the mixing ratio or mixing conditions of the constituent materials when the concrete product was manufactured under conditions based on the second production data and the second element information as ground truth data, inputs the first production data and the first element information into the trained model, and estimates the mixing ratio or mixing conditions of the constituent materials in the first production data.

7. Computers, An acquisition means for acquiring first production data relating to the constituent materials of the concrete to be processed, and first element information that fluctuates depending on the environment or period during the manufacture of the concrete product, A condition determination program that takes second production data relating to the constituent materials of concrete and second element information that fluctuates depending on the environment or period during the manufacture of concrete products as input data, and uses a trained model that has learned the mixing ratio or mixing conditions of the constituent materials when the concrete product is manufactured under conditions based on the second production data and the second element information as ground truth data, inputs the first production data and the first element information into the trained model, and functions as an estimation means for estimating the mixing ratio or mixing conditions of the constituent materials in the first production data.

Images