Ready-mixed concrete quality prediction system and ready-mixed concrete quality prediction method

Abstract

To provide a quality prediction system for ready mixed concrete capable of predicting a slump value of ready mixed concrete at the time of shipping and unloading with high accuracy, and a quality prediction method for ready mixed concrete.SOLUTION: A quality prediction system for ready mixed concrete comprises a storage unit that stores a first learned model generated by machine learning based on first teacher data, an input unit that receives input of input data for prediction, and an output unit that outputs a first output result derived by the first learned model. The first teacher data is input data for learning based on a first learning data group including mixing-related information of ready mixed concrete and slump information at the time of shipping, and data associated with output data for learning including slump information at the time of unloading ready mixed concrete when unloaded. The input data for prediction includes the mixing-related information of ready mixed concrete to be predicted and the slump information at the time of shipping. The first output result is the slump information at the time of unloading the ready mixed concrete to be predicted.SELECTED DRAWING: Figure 1

Description

[Technical Field]

[0001] The present invention relates to a system for predicting the quality of ready-mix concrete, and also to a method for predicting the quality of ready-mix concrete. [Background technology]

[0002] The required properties of ready-mixed concrete vary depending on the intended use and the environment of the site it is used in. For this reason, in the conventional manufacturing process of ready-mixed concrete, the mixing condition was checked visually based on intuition and experience to ensure that the ready-mixed concrete to be produced met the required properties, and then the concrete was shipped.

[0003] One of the important properties of ready-mixed concrete is the slump value, which is a value used as an index of the consistency of concrete before hardening. Generally, when ready-mixed concrete is shipped or unloaded, it is evaluated and inspected to see if its slump value is within the required range.

[0004] The evaluation and inspection results of the slump value of ready-mixed concrete are the basis for determining whether the manufactured ready-mixed concrete can be shipped or used at the unloading site. For this reason, it is expected that the slump value of ready-mixed concrete can be predicted with high accuracy, as much as possible, without relying on the worker's sense or experience, before it is shipped from the manufacturing plant.

[0005] Therefore, in recent years, in order to predict with greater accuracy the slump value at the time of shipment from the manufacturing plant, a method has been proposed for predicting the slump value of ready-mixed concrete using a trained model generated by machine learning that applies training data that associates image data of the ready-mixed concrete being mixed in the mixer with the slump value at the time of shipment (see Patent Document 1 below). [Prior art documents] [Patent documents]

[0006] [Patent Document 1] Japanese Patent Publication No. 2020-144132 Summary of the Invention [Problem to be solved by the invention]

[0007] After being produced at a manufacturing plant, ready-mixed concrete is transported to the unloading site by an agitator truck, which takes up to 90 minutes. As a result, the properties of ready-mixed concrete gradually change over time as it is transported from the manufacturing plant to the unloading site. Therefore, even if the slump value of ready-mixed concrete is within the required range at the time of shipment, it may fluctuate to the extent that it deviates from that range at the time of unloading due to the passage of time during transportation, etc.

[0008] Because situations like the one above can occur, there has long been a demand for methods and systems that can predict the slump value of ready-mixed concrete with high accuracy before it is unloaded and before it is shipped.

[0009] In view of the above problems, the present invention aims to provide a system for predicting the quality of ready-mixed concrete that can predict with high accuracy the slump value of ready-mixed concrete at the time of shipping and unloading. [Means for solving the problem]

[0010] The quality prediction system for ready-mixed concrete of the present invention comprises: a storage unit that stores a first trained model generated by machine learning based on first teacher data; an input unit that accepts input of prediction input data; an output unit that outputs a first output result derived by the first trained model; The first teacher data is data in which learning input data based on a first learning data group including mix-related information, which is information related to the mix of ready-mixed concrete, and shipping slump information, which is information related to the slump value or slump flow value of the ready-mixed concrete before shipping, is associated with learning output data including unloading slump information, which is information related to the slump value or slump flow value of the ready-mixed concrete at the time of unloading, The prediction input data includes the mix proportion information and the shipping slump information of the ready-mixed concrete to be predicted, The first output result is slump information at the time of unloading of the ready-mixed concrete to be predicted.

[0011] In this specification, unless there is a particular need to distinguish between them, the slump value and the slump flow value will be collectively referred to as the "slump value."

[0012] The mix-related information used in the quality prediction system of the present invention is used as a factor for predicting how the fluidity, etc. of ready-mixed concrete will change during transportation and how much slump loss (slump value at time of shipment - slump value at time of unloading) will occur over time.

[0013] The blending-related information can be used as a factor for predicting whether the slump will increase even when the minus (slump value at shipping) minus (slump value at unloading) is a negative value (when the slump will increase). Specific details of the blending-related information will be explained in the section entitled "Mode for Carrying Out the Invention" by listing examples.

[0014] The trained model used in the quality prediction system of the present invention is created by machine learning, and the learning method used in the machine learning, as well as the formulation-related information, are also described in detail in the section ``Forms for implementing the invention.''

[0015] In this specification, "slump information at time of shipment" refers to information related to the slump value and slump flow value of ready-mixed concrete immediately before shipment, such as the measured value of the slump value of ready-mixed concrete actually measured immediately before shipment, or the predicted value of the slump of ready-mixed concrete derived by a system that predicts the slump value at time of shipment.

[0016] Furthermore, in this specification, "at the time of unloading" refers to the time from when a transport vehicle loaded with ready-mixed concrete arrives at the pouring site until pouring is completed, and the slump value at the time of unloading refers to the slump value or slump flow value measured during the time from when a transport vehicle loaded with ready-mixed concrete arrives at the pouring site until pouring is completed. "At the time of production" refers to the time from when mix design begins until the start of adding the raw materials for ready-mixed concrete until the concrete is loaded onto the transport vehicle, and "at the time of shipment" refers to the time from when production begins until transportation begins.

[0017] The inventors applied the prediction method described in Patent Document 1 and conducted extensive research into a system that predicts the slump value of ready-mixed concrete at the time of unloading, based on image data obtained by photographing the inside of a mixer during mixing in the ready-mixed concrete manufacturing process.

[0018] However, with the prediction method based on image data obtained by photographing the inside of a mixer during mixing, as disclosed in Patent Document 1, it was difficult to predict the slump value at the time of unloading with the expected accuracy.

[0019] The reason for this is that the system described in Patent Document 1, which predicts slump values ​​using image data, contains a large amount of information in the image data, and this image data mainly functions as a factor for predicting the current slump, making it difficult to predict the slump at the time of unloading.

[0020] Furthermore, even if the transportation time is used to predict the slump value, the unloading slump cannot be predicted using the transportation time alone. Therefore, even if both the image data and the transportation time are used, the unloading slump value cannot be properly evaluated. Furthermore, the predicted value may vary for each pixel or each extracted region.

[0021] When predicting the slump value at the time of shipment, there is no need to predict the change in the slump value over the time required for transportation, so the variation in the values ​​derived for each pixel or each cut-out area is relatively small. Therefore, by using averaging processing, etc., a relatively stable predicted slump value can be obtained.

[0022] However, if one were to simply predict the slump value at the time of unloading using the same principle, the prediction would involve taking into account the time required for transportation in addition to the values ​​derived for each pixel or each extracted area. As a result, the prediction of the slump value at the time of unloading is significantly affected by the variation in the values ​​derived for each pixel or each extracted area compared to the prediction of the slump value at the time of shipment. Therefore, in a system that predicts slump value using image data, the predicted value of the slump at the time of unloading is less stable than the predicted value of the slump of ready-mixed concrete at the time of shipment, making it difficult to make predictions with the expected accuracy.

[0023] The quality prediction system of the present invention predicts the slump value at the time of unloading based on the uniquely determined slump value at the time of shipping. In addition, by combining information related to the blending and information on the time required from shipping to unloading, it is possible to predict the amount of slump loss that will occur up to the time of unloading.

[0024] Therefore, the quality prediction system of the present invention is able to predict the slump value of ready-mixed concrete at the time of unloading with higher accuracy than conventional systems, without the risk of the predicted value varying between pixels or regions of the image data.

[0025] Furthermore, according to the quality prediction system of the present invention, the slump value of ready-mixed concrete at the time of unloading can be predicted with high accuracy even during the production process. Therefore, by checking whether the slump value of the ready-mixed concrete at the time of unloading during production falls within the required numerical range, if a problem occurs with the quality of the ready-mixed concrete, workers, managers, transportation personnel, etc. can make decisions such as remaking the ready-mixed concrete or halting transportation as early as possible.

[0026] The above quality prediction system is the first teacher data includes at least one piece of information in which learning input data based on a second learning data group including at least one piece of information selected from production-related information, which is information related to the production of ready-mixed concrete, transportation-related information, which is information related to the transportation of ready-mixed concrete, and environment-related information, which is information related to the environment from the start of production of ready-mixed concrete to the completion of pouring, and the learning output data are associated with each other; The prediction input data may be configured to include information corresponding to the first learning data group and the second learning data group of the ready-mixed concrete to be predicted.

[0027] Furthermore, the quality prediction system At least one of the pieces of information in the first learning data group, the second learning data group, and the prediction input data may be configured to be acquired from an information management system at a ready-mixed concrete factory.

[0028] The manufacturing-related information is used as a factor for predicting what changes will occur in the fluidity, etc. of the ready-mixed concrete during transportation and how much slump loss will occur over time.

[0029] The transportation-related information is used as an element for identifying which transportation vehicle will be used to transport the ready-mixed concrete to the unloading site, and how long it will take.

[0030] The environmental information is used as an element for identifying the environment from the production of ready-mixed concrete, to the transportation of the produced ready-mixed concrete to the site, and to the completion of pouring.

[0031] The specific contents of the manufacturing-related information, transportation-related information, and environment-related information will be listed and explained as examples in the section "Form for carrying out the invention," similar to the formulation-related information described above.

[0032] The above-mentioned manufacturing-related information, transportation-related information, and environmental-related information are all used to predict the slump value at unloading as factors that affect the properties of ready-mixed concrete during transportation. By including these factors, it becomes possible to predict slump loss not simply over time, but also taking into account the effects of the temperature and humidity to which the ready-mixed concrete is subjected during transportation, the total amount transported, etc.

[0033] In other words, with the above configuration, the quality prediction system can predict the slump value of ready-mixed concrete at the time of unloading with higher accuracy.

[0034] In addition, the manufacturing-related information, transportation-related information, and environment-related information applied to the first trained model may each use only one of the listed information, or may use a combination of two or more types.

[0035] The above quality prediction system is The output unit may be configured to output the first output result for the fresh concrete to be predicted, and then, when first prediction update data, which is another prediction input data for the fresh concrete to be predicted, is input to the input unit, the first trained model may be configured to output an update output result derived based on the first prediction update data.

[0036] Furthermore, the quality prediction system The first trained model may be configured to output the updated output result during the production of the ready-mix concrete to be predicted, or during the time between shipment and unloading.

[0037] With the above configuration, the quality prediction system can input forecast update data that appropriately updates transportation-related information and environmental-related information in the event of changes in weather during transportation, and can check the predicted value of unloading slump in real time depending on the situation.

[0038] Furthermore, with the above configuration, for example, if a situation arises during transportation where ready-mixed concrete cannot be transported to the unloading site as planned, the quality prediction system can input the estimated time required to reach the unloading site as prediction update data, and use this to determine whether there are any problems in heading directly to the unloading site.

[0039] In particular, by configuring the system to output updated output results when the ready-mixed concrete to be predicted is being manufactured, the quality of the ready-mixed concrete can be confirmed before it is loaded onto a transport vehicle, making it possible to ship it with stable quality.

[0040] In addition, in the past, the slump value at the time of unloading could fluctuate due to various factors during transportation by a transport vehicle, and in order to determine whether or not this fluctuation occurred, it was necessary to check the measured slump value at the pouring site.In contrast, the present invention is configured to output updated output results between the time of shipment and the time of unloading, so that the presence or absence of fluctuation can be predicted early before actual testing is performed at the pouring site, allowing for appropriate quality maintenance.

[0041] The above quality prediction system is The first trained model may be configured to be updated by repeating machine learning using the prediction input data and the measured value of the slump value or slump flow value at the time of unloading of the fresh concrete to be predicted as training data.

[0042] By using the above configuration, the first trained model is updated to a model that can predict the slump value at the time of unloading with higher accuracy based on the prediction input data that is input sequentially and the measured value of the slump at the time of unloading.

[0043] In addition, the above quality prediction system The device may also be provided with an alarm means for notifying when the predicted value of the slump value or slump flow value at the time of unloading of the fresh concrete contained in the slump information at the time of unloading, which is the first output result, exceeds a predetermined threshold value.

[0044] The notification means may employ a mechanism or method that notifies by sound such as a buzzer or bell, or by displaying on an indicator light or display device in a form that can be visually recognized by a person.

[0045] With the above configuration, if a problem is detected in the quality of ready-mixed concrete during production or transportation, workers, managers, transporters, etc. can immediately and reliably recognize the occurrence of the problem. Therefore, if a problem occurs in the quality of ready-mixed concrete, workers, managers, transporters, etc. can make decisions such as remaking the ready-mixed concrete or halting transportation as early as possible.

[0046] In the above quality prediction system, the storage unit stores a second trained model generated by machine learning based on a plurality of second teacher data; The second teacher data is data in which learning input data based on at least one of the mix-related information, the manufacturing-related information, and the environment-related information is associated with learning output data including shipping slump information, which is information related to the slump value or slump flow value of the ready-mixed concrete at the time of shipping, the prediction input data in the second trained model includes at least one of the formulation-related information, the manufacturing-related information, and the environment-related information, The second output result derived by the second trained model includes information related to the shipment slump value or slump flow value of the ready-mixed concrete to be predicted derived by the second trained model, The second output result may be configured to be used as shipping slump information in the prediction input data when deriving the first output result using the first trained model.

[0047] In addition, the above quality prediction system It is equipped with a camera that photographs the ready-mixed concrete being mixed in the mixer and acquires image data, the storage unit stores a second trained model generated by machine learning based on second teacher data; the second teacher data is data in which learning input data based on the image data is associated with learning output data including shipping slump information which is information related to the slump value or slump flow value of the ready-mixed concrete at the time of shipment, The prediction input data in the second trained model includes the image data of the ready-mixed concrete to be predicted, The second output result derived by the second trained model, which includes information related to the shipment slump value or slump flow value of the fresh concrete to be predicted derived by the second trained model, may be configured to be used as shipment slump information in the prediction input data when deriving the first output result by the first trained model.

[0048] According to the above configuration, the quality prediction system can predict both the slump value of ready-mixed concrete at the time of shipment and the slump value at the time of unloading. In other words, it is possible to adjust the mixing time in the ready-mixed concrete production process or modify the mix ratio for the next batch or later so that the slump value at the time of unloading becomes the desired value.

[0049] The learning input data and prediction input data applied to the second trained model to derive the predicted slump value at time of shipment may be any data that can be associated with the predicted slump value at time of shipment and that can be applied as an element for deriving the predicted slump value at time of shipment, such as the power load value of the manufacturing equipment or image data inside the mixer when mixing fresh concrete.

[0050] In the above quality prediction system, The output unit may be configured to output the second output result for the fresh concrete to be predicted, and then, when second prediction update data, which is another prediction input data for the fresh concrete to be predicted, is input to the input unit, output an update output result derived by the second trained model based on the second prediction update data.

[0051] Furthermore, the quality prediction system The second trained model may be configured to output the updated output result during the production of the ready-mixed concrete to be predicted, or during the time between shipment and unloading.

[0052] Furthermore, the quality prediction system The device may also be provided with an alarm means for notifying when the predicted slump value or slump flow value of the ready-mixed concrete at the time of shipment, which is included in the shipping slump information, which is the second output result, exceeds a predetermined threshold value.

[0053] Furthermore, the quality prediction system The second trained model may be configured to be updated by repeating machine learning using the prediction input data and the measured value of the slump value or slump flow value of the ready-mixed concrete to be predicted at the time of shipment as second training data.

[0054] By using the above configuration, the second trained model is updated to a model that can predict the slump value at the time of shipment with higher accuracy based on the prediction input data that is input sequentially and the measured value of the slump at the time of shipment.

[0055] Furthermore, by providing an alarm means that issues an alarm when the slump value at the time of shipment or the predicted slump flow value exceeds a predetermined threshold, if it is detected that there may be a problem with the quality of the ready-mixed concrete being manufactured, workers, managers, transporters, etc. can be made aware of the problem immediately and reliably. Therefore, if a problem occurs in the quality of the ready-mixed concrete, workers, managers, etc. can make a decision to re-mix the ready-mixed concrete as early as possible. Note that the alarm means here may be, as in the case above, a method of alerting by sound such as a buzzer or bell, or a mechanism or method of alerting in a form that can be visually recognized by a person, such as by displaying on an indicator light or display device.

[0056] The above quality prediction system is A display device for displaying the first output result may be provided.

[0057] In addition, the above quality prediction system a display device for displaying the first output result; The display device may be configured to display at least one of the slump value at the time of unloading based on the first output result, the predicted value of the slump flow value at the time of unloading, the slump value at the time of shipment based on the second output result, and the predicted value of the slump flow value at the time of shipment.

[0058] Furthermore, the quality prediction system The first output result includes information related to the slump value or slump flow value of the fresh concrete at any time elapsed since production, The display device may be configured to display the slump value or slump flow value of the ready-mixed concrete at any time elapsed since production, based on the first output result.

[0059] With the above configuration, the person in charge of the manufacturing process, the person in charge of transport management, the person in charge of transportation, etc. can grasp the slump information at the time of unloading and the slump loss over time in a timely manner. Therefore, each person in charge can make appropriate decisions depending on the situation, such as adjusting the mixing time, continuing or stopping transportation, etc., based on the information on the slump value, and can ship high-quality ready-mixed concrete.

[0060] The method for predicting the quality of ready-mixed concrete of the present invention comprises: inputting prediction input data for ready-mixed concrete into a first trained model generated by machine learning based on first teacher data; obtaining a first output result derived by the first trained model; The first teacher data is data in which learning input data based on a first learning data group including mix-related information, which is information related to the mix of ready-mixed concrete, and shipping slump information, which is information related to the slump value or slump flow value of the ready-mixed concrete before shipping, and learning output data including unloading slump information, which is information related to the slump value or slump flow value of the ready-mixed concrete at the time of unloading, are associated; The prediction input data includes the mix proportion information and the shipping slump information of the ready-mixed concrete to be predicted, The first output result is slump information at the time of unloading of the ready-mixed concrete to be predicted.

[0061] The above quality prediction method is the first teacher data is data in which the learning output data is associated with learning input data based on a second learning data group including at least one of production-related information, which is information related to the production of ready-mixed concrete, transportation-related information, which is information related to the transportation of ready-mixed concrete, and environment-related information, which is information related to the environment from the start of production of ready-mixed concrete to the completion of pouring, The prediction input data may include information corresponding to the first learning data group and the second learning data group of the ready-mixed concrete to be predicted.

[0062] The above quality prediction method is Alternatively, after obtaining the first output result for the fresh concrete to be predicted, another prediction input data, which is first prediction update data for the fresh concrete to be predicted, may be input, and an update output result derived based on the first prediction update data by the first trained model may be obtained.

[0063] The above quality prediction method is The method may also be such that the updated output result is obtained using the first trained model during the production of the ready-mix concrete to be predicted, or during the time between shipment and unloading.

[0064] The above quality prediction method is The method may further include updating the first trained model by repeating machine learning using the prediction input data and the measured value of the slump value or slump flow value at the time of unloading of the fresh concrete to be predicted as training data.

[0065] The above quality prediction method is further comprising inputting prediction input data based on at least one of the mix proportion-related information, the production-related information, and the environment-related information of the ready-mixed concrete to be predicted to a second trained model generated by machine learning based on second teacher data, The second teacher data is data in which learning input data based on at least one of the mix-related information, the manufacturing-related information, and the environment-related information is associated with learning output data including shipping slump information, which is information related to the slump value or slump flow value of the ready-mixed concrete at the time of shipping, The second output result derived by the second trained model includes information related to the shipment slump value or slump flow value of the ready-mixed concrete to be predicted derived by the second trained model, The second output result may be used as shipping slump information in the prediction input data when deriving the first output result using the first trained model.

[0066] The above quality prediction method is The method further includes inputting prediction input data based on image data of the ready-mixed concrete to be predicted to a second trained model generated by machine learning based on the second teacher data, The second teacher data is data in which learning input data based on image data of ready-mixed concrete at the time of production is associated with learning output data including shipping slump information which is information related to the slump value or slump flow value of the ready-mixed concrete at the time of shipment, The method may also be such that the second output result, which includes information related to the shipment slump value or slump flow value of the fresh concrete to be predicted derived by the second trained model, is used as shipment slump information in the prediction input data when deriving the first output result by the first trained model.

[0067] The above quality prediction method is Alternatively, the second output result for the fresh concrete to be predicted may be output, and then second prediction update data, which is another prediction input data for the fresh concrete to be predicted, may be input, and an updated output result derived by the second trained model based on the second prediction update data may be obtained.

[0068] The above quality prediction method is The method may also be such that the updated output result is obtained using the second trained model during the production of the ready-mix concrete to be predicted, or during the time between shipment and unloading.

[0069] The above quality prediction method is The method may further include updating the second trained model by repeating machine learning using the prediction input data and the measured value of the slump value or slump flow value of the ready-mixed concrete to be predicted at the time of shipment as second training data. [Effects of the Invention]

[0070] According to the present invention, a system and method for predicting the quality of ready-mixed concrete are realized that can predict with high accuracy the slump value of ready-mixed concrete at the time of unloading at the time of shipment. [Brief explanation of the drawings]

[0071] [Figure 1] 1 is a schematic overall configuration diagram of one embodiment of a quality prediction system for ready-mixed concrete. FIG. [Figure 2] FIG. 2 is a block diagram illustrating a configuration of an embodiment of a main server. [Figure 3] FIG. 2 is a block diagram illustrating a configuration of an embodiment of a main server. [Figure 4] 2 is a diagram schematically illustrating a part of the inside of a mixer. [Figure 5] This diagram shows a display device connected to each terminal displaying images of the inside of the mixer captured by a camera and the predicted slump values ​​at the time of shipment and at the time of unloading of the ready-mixed concrete being mixed in the mixer. [Figure 6] 10 is a diagram showing a state in which a display device connected to each terminal displays a predicted curve of slump value over time from the time of shipment. DETAILED DESCRIPTION OF THE INVENTION

[0072] The quality prediction system for ready-mixed concrete of the present invention will be described below with reference to the drawings. Note that the drawings are all schematic illustrations, and the numbers on the drawings do not necessarily correspond to the actual numbers. Furthermore, the examples of the quality prediction method for ready-mixed concrete of the present invention correspond to the operations performed by the quality prediction system described below.

[0073] The quality prediction system for ready-mixed concrete of the present invention is a system that predicts the slump value of ready-mixed concrete at the time of unloading using a first trained model generated by machine learning based on a plurality of first teacher data that associates learning input data including information related to the mix of ready-mixed concrete, manufacturing information, transportation information, and slump information at the time of shipping, as described below, with learning output data including slump information at the time of unloading.

[0074] First, we will explain the details of the mix-related information, manufacturing-related information, transportation-related information, and environmental-related information used as learning input data to generate the first trained model, and the relationship between this information and the slump value prediction at the time of unloading fresh concrete.

[0075] [Composition information] Specific examples of the mix-related information include the water-to-cement ratio (W / C), cement type, admixture type, admixture amount, and unit water volume. The mix-related information applied to the first trained model used in the present invention may be one of the above mix-related information or a combination of two or more of them.

[0076] The mix proportion information is information that contributes to the rate at which the hydration reaction in the ready-mixed concrete to be produced progresses, and is an element for predicting what changes will occur in the fluidity, etc. of the ready-mixed concrete during transportation and how much slump loss will occur over time. Therefore, the unloading slump information, which relates to the slump value at the time of unloading the ready-mixed concrete, can be associated with the mix proportion information and shipping slump information, which is information related to the slump value at the time of shipping.

[0077] Therefore, by applying a first trained model for predicting the slump value at unloading of ready-mixed concrete generated by machine learning using learning input data including mix-related information and shipping slump information and learning output data including unloading slump information, it is possible to predict the slump value at unloading of the ready-mixed concrete to be produced.

[0078] Other examples of mix-related information include, but are not limited to, data on the raw materials of cement clinker, data on the burning conditions of cement clinker, and data on the grinding conditions of cement. Mix-related information may include mix-related information for cement production, as well as mix-related information for aggregates, water, and other components mixed into ready-mix concrete. For example, concrete design strength, unit water content, unit cement content, cement density, cement specific surface area, coarse aggregate type, unit coarse aggregate volume, fine aggregate type, unit fine aggregate volume, aggregate density, coarse particle ratio, fine aggregate ratio, particle size distribution, fine and coarse aggregate density, water absorption, water content, surface water content, maximum dimensions, cement chemical composition, mineral composition, particle size distribution, sieve test residue, color, mineralogical and crystalline properties of each mineral contained in cement, and data related to the entire cement, such as the hemihydration rate of gypsum contained in cement, may also be used. In particular, the design slump value may be used preferably. Since information on the design slump value is highly correlated with the slump value at the time of unloading, the slump value at the time of unloading can be predicted more accurately by using the design slump value as the composition-related information, especially when there is little other composition-related information.

[0079] Furthermore, in order to keep slump loss within a certain range of predictions, a standard time is set for the time required for the transportation of ready-mixed concrete from the manufacturing plant to the unloading site (JIS A 5308). As long as the concrete is transported within this standard time, the slump loss of ready-mixed concrete that occurs due to time differences is small compared to the slump loss based on mix-related information.

[0080] Therefore, the quality prediction system 1, which applies a first trained model that has undergone machine learning using a learning input model that includes mix-related information and shipping slump information, can predict the slump value of ready-mixed concrete at the time of unloading before shipping with higher accuracy than a human prediction based on experience or intuition.

[0081] [Manufacturing related information] Specific examples of the production-related information include the mixer type, mixer model, mixing time, mixing noise, power load value, ready-mixed concrete temperature during mixing, temperature and humidity inside the mixer, etc. The production-related information applied to the first trained model may be one of the above production-related information or a combination of two or more of the above production-related information.

[0082] The manufacturing-related information is an element that can confirm the hardness and fluidity of fresh concrete during mixing, and is an element for predicting the slump value at the time of shipment. In addition, when combined with the mix proportion-related information, it can be used to understand the progress of the hydration reaction at the time of shipment, and can therefore be an element for predicting the slump value at the time of unloading.

[0083] Therefore, a quality prediction system that applies a first trained model generated by machine learning using mix-related information, shipping slump information, as well as training input data including manufacturing-related information and training output data including unloading slump information, can predict the slump value of fresh concrete at unloading with higher accuracy.

[0084] Other production-related information that can be used include the amount of air, the temperature of each material, the temperature and humidity of the storage location (container), batcher data (type of heating / cooling equipment, stored water temperature, stored aggregate temperature, stored cement temperature, measuring bottle type, discharge volume, maximum discharge pressure, etc.), chloride content, crack resistance, dynamic modulus of elasticity, dynamic shear modulus of elasticity, dynamic Poisson's ratio, hardened body void volume and void size distribution, durability, color, separation status of each material in the fresh concrete, fluidity, rheological values ​​(plastic viscosity, yield value, etc.), chloride content in the fresh concrete, etc.

[0085] [Transportation related information] The transportation-related information is information related to the environment from the start of production of ready-mixed concrete to the completion of pouring, and specifically includes information related to the transport vehicle, such as the agitator truck or dump truck, the agitator truck number, the transport speed, mixing performance, the agitator truck mixing sound, the mixer rotation speed, the load capacity, the power load value inside the agitator truck drum, and the volume of ready-mixed concrete; information related to the transport vehicle, such as the transport time, which is the actual time taken for transport, the normally expected transport time to the destination, information on the transport situation, such as traffic congestion information and transport distance; and information on the site environment of the transport destination, such as the thickness of the components to be placed. The transportation-related information applied to the first trained model may use only one of the above transportation-related information, or a combination of two or more types.

[0086] The transportation-related information is an element for identifying which transportation vehicle will be used, how long it will take for the ready-mixed concrete to be transported to the unloading site, etc. The slump value at the time of unloading does not vary greatly if the ready-mixed concrete is transported and unloaded within a specified time range from the time of shipment, but by taking into account a more accurate transportation time, the slump value at the time of unloading can be predicted with higher accuracy.

[0087] Therefore, a quality prediction system that applies a first trained model generated by machine learning using training input data that includes mixing-related information and shipping slump information as well as transportation-related information, and training output data that includes unloading slump information, can predict the slump value of fresh concrete at the time of unloading with higher accuracy.

[0088] [Environment-related information] Specific examples of the environment-related information include the temperature inside the agitator vehicle, the amount of solar radiation on the agitator vehicle, wind speed, wind direction, or the weather, humidity, altitude, and outside temperature along the route the agitator vehicle travels. The weather, humidity, altitude, and outside temperature at the production site of the ready-mixed concrete can also be used as the environment-related information, with the outside temperature at the production site being particularly suitable. While special means are required to acquire environment-related information during transportation, such as the temperature inside the agitator vehicle or the weather along the route the agitator vehicle travels, using the outside temperature at the production site allows the environment-related information to be acquired in a relatively simple manner. The amount of solar radiation, wind speed, wind direction, weather, humidity, altitude, and outside temperature at the pouring site of the ready-mixed concrete can also be used as the environment-related information. The environment-related information applied to the first trained model may be one of the above-mentioned environment-related information, or a combination of two or more of them. In particular, when the temperature inside the mixer of the agitator car rises due to the effect of the amount of solar radiation on the agitator car, the slump loss increases, so using information on the amount of solar radiation allows for more accurate prediction.

[0089] Environmental information is an element for identifying the environment in which ready-mixed concrete is transported. In other words, it is an element for understanding whether the progress of the hydration reaction of ready-mixed concrete is speeding up or slowing down due to the influence of the outside temperature, etc.

[0090] Therefore, a quality prediction system that applies a first trained model generated by machine learning using training input data that includes mix-related information and shipping slump information as well as environmental information, and training output data that includes unloading slump information, can predict the slump value of fresh concrete at unloading with higher accuracy.

[0091] From the above, the quality prediction system, which applies the first trained model that has undergone machine learning using learning input data including mix-related information and shipping slump information, can predict the slump value of ready-mixed concrete at the time of unloading before shipping with higher accuracy than a human prediction based on experience or intuition.

[0092] Furthermore, a quality prediction system that applies a first trained model for predicting the slump value of fresh concrete at the time of unloading, which is generated by performing machine learning using learning input data that includes not only mix-related information and shipping slump information but also manufacturing-related information, transportation-related information, and environmental information, and learning output data that includes unloading slump information, can predict the slump value of fresh concrete at the time of unloading with higher accuracy.

[0093] The above-mentioned mix-related information, production-related information, and transportation-related information may also be configured to be obtained from an existing information management system at the ready-mixed concrete plant. Specifically, in an existing ready-mixed concrete plant, the above information can be obtained from a system that manages mix-related information for specified materials normally used when shipping ready-mixed concrete, a system that sets the measurement values ​​for storing materials and controls the production equipment, and a system that starts transporting ready-mixed concrete and manages information during transport. For example, the transportation-related information may be information obtained from a dispatch system such as Sky One II (manufactured by Pacific Systems Co., Ltd.), a ready-mixed concrete dispatch system that utilizes GPS.

[0094] The above information management system is merely an example, and there are no particular limitations on the method of acquiring the blend-related information, production-related information, and transportation-related information.

[0095] The number of samples of first training data used for training the first trained model is preferably 100 or more, more preferably 1,000 or more, even more preferably 10,000 or more, even more preferably 50,000 or more, and particularly preferably 100,000 or more, from the viewpoint of extracting features necessary for deriving unloading slump information from training input data and further improving prediction accuracy. Furthermore, the number of training times for the first trained model is preferably 1,000 or more, more preferably 8,000 or more, and particularly preferably 10,000 or more, from the viewpoint of improving prediction accuracy, but is not particularly limited. For example, only one high-quality sample may be used as training data.

[0096] Next, a machine learning method for creating the first trained model of the present invention will be described. Examples of machine learning methods for creating the first trained model used in the present invention include neural networks, linear regression, decision trees, support vector regression, ensemble methods, support vector machines, discriminant analysis, naive Bayes methods, and nearest neighbor methods. These methods may be used alone or in combination of two or more.

[0097] Among these methods, machine learning using a neural network is preferably selected from the viewpoint of being able to predict quality with higher accuracy. As the neural network, a hierarchical neural network having one or more intermediate layers between an input layer and an output layer is preferable from the viewpoint of being able to predict quality with higher accuracy.

[0098] Examples of neural networks include convolutional neural networks (CNNs) such as 3D convolutional neural networks (3DCNNs), deep neural networks (DNNs), recurrent neural networks (RNNs), and long short-term memory (LSTM) neural networks (recurrent neural networks improved using LSTM).

[0099] Among these, convolutional neural networks (neural networks having convolutional layers, pooling layers, etc. as intermediate layers) are more suitable, as they have excellent performance in the field of image recognition. Convolutional neural networks can detect features from image data and create prediction models capable of classification or regression using the features. The number of layers in a convolutional neural network, each consisting of a combination of convolutional layers and pooling layers, is preferably two or more, more preferably three or more, from the viewpoint of enabling predictions with higher accuracy.

[0100] In addition, tools that can be used for machine learning include, for example, "TensorFlow (registered trademark)," a software library developed by Google, and "IBM Watson (registered trademark)," a system developed by IBM.

[0101] Next, a specific configuration of each embodiment of the quality prediction system 1 will be described.

[0102] [First embodiment] The configuration of a first embodiment of the ready-mixed concrete quality prediction system 1 of the present invention will be described. Fig. 1 is a schematic overall configuration diagram of one embodiment of the ready-mixed concrete quality prediction system 1. As shown in Fig. 1, the quality prediction system 1 is composed of a main server 10, and a data server 20, an operation terminal 30, and a plurality of reference terminals 40, all of which are connected to the main server 10 via a network.

[0103] The number of data servers 20, operation terminals 30, and reference terminals 40 connected to the main server 10 via a network is arbitrary, and they are all connected to each other so as to perform data communication via wire or wirelessly.

[0104] The main server 10 in the first embodiment starts a prediction operation when an operator operates the operation terminal 30 to start the operation. Then, when prediction input data including information about the ready-mixed concrete to be predicted to be applied to the first trained model is input to the main server 10 from the data server 20 or the operation terminal 30, a predetermined process is performed within the main server 10, and unloading slump information related to the predicted value of the slump at unloading is output from the main server 10.

[0105] The output data including the unloading slump information output from the main server 10 is transmitted to each reference terminal 40. Then, each reference terminal 40 displays the unloading slump value obtained from the output data output from the main server 10 on its respective display device 50.

[0106] 1, each reference terminal 40 in the first embodiment is a desktop PC separate from the display device 50, but it may also be a terminal in which the display device 50 is integrated, such as a notebook PC, a smartphone, or a tablet. Also, each reference terminal 40 may also function as the operation terminal 30.

[0107] Fig. 2 is a block diagram schematically showing the configuration of a first embodiment of the main server 10. As shown in Fig. 2, the main server 10 in the first embodiment includes a data input unit 11, a data output unit 12, and a first calculation processing unit 13. The first calculation processing unit 13 includes a first storage unit 13a in which a first trained model M1 is recorded.

[0108] The data input unit 11 in the first embodiment accepts input of prediction input data d1 (in the first embodiment, information related to the mix, manufacturing, and transportation of the ready-mix concrete for prediction) sent from the data server 20 or the operation terminal 30 to be applied to the first trained model M1.

[0109] When the data input unit 11 has received all of the input data for prediction d1, it generates input data for calculation d2 and outputs the input data for calculation d2 to the first calculation processing unit 13.

[0110] The calculation input data d2 may include at least the blending-related information and the shipping slump information (first learning data group). Alternatively, the calculation input data d2 may include only a part of the manufacturing-related information, the transportation-related information, and the environment-related information (second learning data group) in addition to the blending-related information and the shipping slump information.

[0111] When the data output unit 12 receives the calculation output data d3 output from the first calculation processing unit 13, it generates transmission data d4 that can be received by the reference terminal 40 based on the calculation output data d3 and transmits the transmission data d4 to each reference terminal 40.

[0112] When the first calculation processing unit 13 receives the calculation input data d2 output from the data input unit 11, it reads out the first trained model M1 recorded in the first memory unit 13a and applies the calculation input data d2 to the first trained model M1.

[0113] The first calculation processing unit 13 performs calculation processing using the first trained model M1 and outputs to the data output unit 12 calculation output data d3 including unloading slump information, which is the first output result derived based on the calculation input data d2.

[0114] The information adopted as the learning input data and the prediction input data d1 in the first embodiment is as follows:

[0115] [Table 1]

[0116] Here, we will explain the results of a verification experiment to confirm how accurately the quality prediction system 1, in which the first trained model M1 that applies the above information is recorded, can predict the unloading slump value and slump flow value.

[0117] The output data consisted of two values: the slump value at unloading of the produced concrete, and the slump flow value at unloading. The number of output data for each value was 779 for the slump value at unloading and 1,163 for the slump flow value at unloading.

[0118] The first trained model M1 was generated by applying a deep neural network (DNN), using Tensor Flow (registered trademark) as the training library, and training was performed 200,000 times.

[0119] The accuracy rate for the slump value was confirmed with tolerances of ±1.0 cm, ±1.5 cm, ±2.0 cm, and ±2.5 cm. The accuracy rate for the slump flow value was confirmed with tolerances of ±2.5 cm, ±5.0 cm, ±7.5 cm, and ±10.0 cm.

[0120] The results were as follows:

[0121] [Table 2]

[0122] The above results confirm that the quality prediction system 1 of the present invention can predict the slump value at the time of unloading with an accuracy of 90% or more within a tolerance of ±2.0 cm, and the slump flow value at the time of unloading with an accuracy of 90% or more within a tolerance of ±5.0 cm.

[0123] As described above, with the above configuration, the quality prediction system 1 of the present invention predicts the slump value at unloading based on the uniquely determined slump value at shipping, so there is no risk of the predicted value varying for each pixel or region of the image data. In other words, the slump value of ready-mixed concrete at unloading can be predicted with higher accuracy than conventional methods.

[0124] Furthermore, according to the quality prediction system 1 configured as described above, the slump value at the time of unloading of ready-mixed concrete can be predicted with high accuracy even during the production process, so that it is possible to fine-tune the mixing time in the mixer while checking whether the slump value at the time of unloading of the ready-mixed concrete being produced falls within the required numerical range. In other words, production losses of ready-mixed concrete are suppressed.

[0125] [Second embodiment] The configuration of the second embodiment of the quality prediction system 1 of the present invention will be described, focusing on the differences from the first embodiment.

[0126] In the quality prediction system 1 of the second embodiment, the shipping slump information applied to the first trained model M1 is a predicted slump value of the fresh concrete before shipment obtained based on image data taken inside the mixer while the fresh concrete is being mixed.

[0127] Fig. 3 is a block diagram showing a schematic configuration of the main server 10 according to a second embodiment, and Fig. 4 is a diagram showing a schematic configuration of a part of the mixer 2. As shown in Fig. 3, the main server 10 of the quality prediction system 1 according to the second embodiment includes a second calculation processing unit 14. The quality prediction system 1 according to the second embodiment also includes a camera 60.

[0128] For ease of explanation, the main server 10 shown in FIG. 3 is shown configured with two sets of processing units and memory units, but the main server 10 may be configured with each processing unit (13, 14) being a single processing unit such as a CPU or MPU, and each memory unit (13a, 14a) being a single memory such as a flash memory or hard disk, with the whole server constituting a single processing unit.

[0129] As shown in Fig. 4, the camera 60 captures the inside of the mixer 2 while mixing the ready-mixed concrete, and acquires image data d5. When the image data d5 acquired by the camera 60 is input to the second arithmetic processing unit 14, the second arithmetic processing unit 14 derives a predicted slump value at the time of shipment and outputs the predicted slump value as prediction input data d1 to the data input unit 11. Note that, although the camera 60 and the main server 10 in the second embodiment are configured to transmit and receive the image data d5 via wireless communication, they may also be configured to transmit and receive via wired communication.

[0130] The second calculation processing unit 14 has a second memory unit 14a in which the second learned model M2 is recorded, and when prediction input data including applicable information is input, it outputs a second output result including information related to the predicted value of the shipping slump.

[0131] The second trained model M2 is a predictive model created by machine learning using multiple pieces of second training data consisting of a combination of learning input image data obtained by photographing the inside of mixer 2 during the ready-mixed concrete manufacturing process and measured values ​​of the slump of the ready-mixed concrete at the time of shipment. Note that the machine learning method and learning library of the second trained model can similarly adopt the various methods and libraries used to generate the first trained model described above.

[0132] Examples of image data that can be used as learning input image data include image data related to the production of ready-mixed concrete, image data photographed of the mixing of the ingredients in a mixer for mixing the ingredients of ready-mixed concrete, image data photographed of the pouring of ready-mixed concrete from the mixer 2 into a hopper, and image data photographed of a monitor displaying the history of the power load value of the mixer 2 while mixing the ingredients of ready-mixed concrete (a display that visually shows the change over time in the power load value using a graph or the like). These may be used alone or in combination of two or more.

[0133] From the viewpoint of being able to predict quality with higher accuracy, the number of image data used for training the second trained model M2 is preferably 100 or more, more preferably 1,000 or more, even more preferably 10,000 or more, even more preferably 50,000 or more, and particularly preferably 100,000 or more. Furthermore, from the viewpoint of improving prediction accuracy, the number of times of training in the second trained model M2 is preferably 1,000 or more, more preferably 8,000 or more, and particularly preferably 10,000 or more.

[0134] The image data may be captured from video data. The image data used as the learning input image data may be a grayscale image or a color image.

[0135] Furthermore, in order to predict quality with higher accuracy, the machine learning for generating the second trained model M2 may use, in addition to the image data captured by the camera 60, other data (e.g., formulation-related information, manufacturing-related information, etc.) as input data for learning.

[0136] Furthermore, from the viewpoint of being able to predict quality with higher accuracy, it is preferable that the prediction image data d5 applied to the second trained model M2 is acquired by photographing the stirring blades (mixer blades) rotating in the mixer 2 when they are positioned at a specific, arbitrarily determined location within the mixer 2. The specific, arbitrarily determined location may be one location or two or more locations.

[0137] Furthermore, although the second trained model M2 in the second embodiment derives a predicted slump value at the time of shipment based on image data d5, the data for deriving the predicted slump value at the time of shipment is not limited to image data. The data applied to the second trained model M2 may be, for example, the power load value of the manufacturing equipment. Any data may be used as appropriate as long as it can be associated with the predicted slump value at the time of shipment and can be used as an element for deriving the predicted slump value at the time of shipment. Furthermore, by using such a predicted slump value at the time of shipment derived based on image data d5 or other applicable data as the calculation input data d2 applied to the first trained model M1, the slump value at the time of unloading can be predicted without the need to measure the slump at the time of shipment, thereby enabling efficient and stable quality evaluation.

[0138] 5 is a diagram showing a state in which a display device 50 connected to each reference terminal 40 displays video of the inside of the mixer 2 captured by a camera 60, as well as the predicted shipping slump and unloading slump of the ready-mixed concrete being mixed in the mixer 2. As shown in FIG. 5, the display device 50 connected to the reference terminal 40 may be configured to display video of the inside of the mixer 2 captured by the camera 60, as well as simultaneously display the predicted shipping slump and unloading slump. By simultaneously displaying the predicted shipping slump and unloading slump, the mixing status, current slump, and unloading slump can be evaluated, and slump loss can be predicted, allowing ready-mixed concrete to be shipped while maintaining high quality.

[0139] [Another embodiment] Another embodiment will be described below.

[0140] <1> After input data for calculation d2 is input to the data input unit 11 and a first output result is output from the data output unit 12, when another input data for calculation d2, that is, first prediction update data, is input to the data input unit 11, the first trained model M1 may be configured to output an update output result derived based on the first prediction update data to the data output unit 12.

[0141] With the above configuration, the quality prediction system 1 can predict the slump value at the time when unloading is expected to occur in real time if there is a change in weather or a change in transportation time during transportation.

[0142] The forecasting update data may be data that includes only a portion of the information included in the forecasting input data d1 (mixture-related information, manufacturing-related information, transportation-related information, environment-related information, and shipping slump information). The mixture-related information, manufacturing-related information, and shipping slump information are not values ​​that change after shipping. For this reason, the forecasting update data may be configured to include, for example, only transportation-related information and environment-related information.

[0143] Similarly, with regard to the second trained model M2, when image data d5 is input to the second calculation processing unit 14 and the second calculation processing unit 14 outputs prediction input data d1 derived based on the image data d5, and then when second prediction update data, which is another image data d5, is input to the second calculation processing unit 14, the second trained model M2 may be configured to output prediction input data d1 derived based on the second prediction update data to the data input unit 11.

[0144] 6 is a diagram showing a state in which a display device 50 connected to each reference terminal 40 displays a predicted curve of the slump value over time from the time of shipment. The first trained model M1 may be created to output, as the unloading slump information, calculation output data d3 relating to the fluctuation of the slump value over time from the time of shipment to the time of unloading.

[0145] The display device 50 may be configured to display a predicted curve of the slump value over time from the time of shipment, as shown in FIG. 6, based on the calculation output data d3.

[0146] With the above configuration, even if the transporter finds that he or she cannot transport the ready-mixed concrete to the unloading site within the originally scheduled transport time before shipping, the transporter can estimate the time until unloading from the current situation and determine whether there is any problem in transporting the ready-mixed concrete to the unloading site as is.

[0147] In addition, ready-mix concrete manufacturers can adjust the selection of raw materials and mixing ratios by anticipating slump loss due to the time required for transportation, allowing them to respond to more specific requirements.

[0148] Furthermore, the first trained model M1 may be created to output calculation output data d3 relating to the predicted slump value of fresh concrete at any time as the unloading slump information. Note that the predicted slump value of fresh concrete at any time may be configured to be obtained by extracting the value at a specified time from the above-mentioned prediction curve.

[0149] The display device 50 may be configured to display the slump value of the ready-mixed concrete at any time based on the calculation output data d3.

[0150] <3> The quality prediction system 1 of the present invention may be configured to update the first trained model by repeating machine learning using the prediction input data d1 and the measured value of the slump value at the time of unloading of the ready-mixed concrete to be predicted as first training data.

[0151] By using the above configuration, the first trained model is updated to a model that can predict the slump value at the time of unloading with higher accuracy based on the prediction input data that is input sequentially and the measured value of the slump at the time of unloading.

[0152] In addition, the quality prediction system 1 of the present invention may be configured to update the second learned model by repeating machine learning using the prediction input data d1 and the measured value of the slump value at the time of shipment of the ready-mixed concrete to be predicted as the second training data.

[0153] By using the above configuration, the second trained model is updated to a model that can predict the slump value at the time of unloading with higher accuracy based on the prediction input data that is input sequentially and the measured value of the slump at the time of shipment.

[0154] <4> The quality prediction system 1 may include a notification means for providing notification when the predicted slump value or slump flow value of the ready-mixed concrete at the time of unloading, which is included in the unloading slump information (the first output result), or the predicted slump value or slump flow value of the ready-mixed concrete at the time of shipment, which is included in the shipping slump information (the second output result), exceeds a predetermined threshold. The notification means may be, for example, a buzzer or bell, or a visually perceptible notification mechanism or method, such as an indicator light or a display device. The notification means may include separate notification means for providing notification when the predicted slump value or slump flow value at unloading exceeds the predetermined threshold and for providing notification when the predicted slump value or slump flow value at shipment exceeds the predetermined threshold. Alternatively, a single notification means may be provided and configured to respond to both notifications.

[0155] With the above configuration, if a problem is detected in the quality of ready-mixed concrete during production or transportation, workers, managers, transporters, etc. can immediately and reliably recognize the occurrence of the problem. Therefore, if a problem occurs in the quality of ready-mixed concrete, workers, managers, transporters, etc. can make decisions such as remaking the ready-mixed concrete or halting transportation as early as possible.

[0156] <5> The configuration of the quality prediction system 1 described above is merely an example, and the present invention is not limited to the illustrated configurations. [Explanation of symbols]

[0157] 1: Quality prediction system 2: Mixer 10: Main server 11: Data input section 12: Data output section 13: First processing unit 13a: First memory section 14: First processing unit 14a: First memory section 20: Data Server 30: Operation terminal 40: Reference terminal 50: Display device 60: Camera M1: First trained model M2: Second trained model

Claims

1. a storage unit that stores a first trained model generated by machine learning based on first teacher data; an input unit that accepts input of prediction input data; an output unit that outputs a first output result derived by the first trained model; The first teacher data is data in which learning input data based on a first learning data group including mix-related information, which is information related to the mix of ready-mixed concrete, and shipping slump information, which is information related to the slump value or slump flow value of the ready-mixed concrete before shipping, and a second learning data group including at least one of production-related information, which is information related to the production of ready-mixed concrete, transportation-related information, which is information related to the transportation of ready-mixed concrete, and environment-related information, which is information related to the environment from the start of production of ready-mixed concrete to the completion of pouring, is associated with learning output data including unloading slump information, which is information related to the slump value or slump flow value of the ready-mixed concrete at the time of unloading, the input data for prediction includes the mix proportion information and the shipping slump information of the ready-mix concrete to be predicted, and information corresponding to the first learning data group and the second learning data group of the ready-mix concrete to be predicted, The first output result is slump information at the time of unloading of the ready-mixed concrete to be predicted, A system for predicting the quality of ready-mixed concrete, characterized in that at least one piece of information included in the first learning data group, the second learning data group, and the prediction input data is acquired from an information management system at a ready-mixed concrete factory.

2. The ready-mixed concrete quality prediction system described in claim 1, characterized in that after the output unit outputs the first output result for the ready-mixed concrete to be predicted, when first prediction update data, which is another prediction input data for the ready-mixed concrete to be predicted, is input to the input unit, the output unit outputs an update output result derived by the first trained model based on the first prediction update data.

3. The ready-mixed concrete quality prediction system described in claim 2, characterized in that the first trained model outputs the updated output result during the production of the ready-mixed concrete to be predicted or during the time between shipment and unloading.

4. The quality prediction system for ready-mixed concrete described in any one of claims 1 to 3, characterized in that the first trained model is updated by repeating machine learning using the prediction input data and the measured value of the slump value or slump flow value at the time of unloading of the ready-mixed concrete to be predicted as training data.

5. A quality prediction system for ready-mixed concrete as described in any one of claims 1 to 4, characterized in that it is provided with an alarm means for notifying when the predicted value of the slump value or slump flow value at the time of unloading of the ready-mixed concrete contained in the slump information at the time of unloading, which is the first output result, exceeds a predetermined threshold value.

6. the storage unit stores a second trained model generated by machine learning based on second teacher data; The second teacher data is data in which learning input data based on at least one of the mix-related information, production-related information which is information related to the production of ready-mixed concrete, and environment-related information which is information related to the environment from the start of production of ready-mixed concrete to the completion of pouring is associated with learning output data including shipping slump information which is information related to the slump value or slump flow value of the ready-mixed concrete at the time of shipment, the prediction input data in the second trained model includes at least one of the formulation-related information, the manufacturing-related information, and the environment-related information, The second output result derived by the second trained model includes information related to the shipment slump value or slump flow value of the ready-mixed concrete to be predicted derived by the second trained model, A quality prediction system for ready-mixed concrete as described in any one of claims 1 to 5, characterized in that the second output result is used as shipping slump information in the prediction input data when deriving the first output result using the first trained model.

7. It is equipped with a camera that photographs the ready-mixed concrete being mixed in the mixer and acquires image data, the storage unit stores a second trained model generated by machine learning based on second teacher data; the second teacher data is data in which learning input data based on the image data is associated with learning output data including shipping slump information which is information related to the slump value or slump flow value of the ready-mixed concrete at the time of shipment, The prediction input data in the second trained model includes the image data of the ready-mixed concrete to be predicted, A quality prediction system for ready-mixed concrete as described in any one of claims 1 to 5, characterized in that a second output result derived by the second trained model, which includes information related to the shipping slump value or slump flow value of the ready-mixed concrete to be predicted, is used as shipping slump information in the prediction input data when deriving the first output result by the first trained model.

8. The ready-mixed concrete quality prediction system described in claim 6 or 7, characterized in that after outputting the second output result for the ready-mixed concrete to be predicted, when second prediction update data, which is another prediction input data for the ready-mixed concrete to be predicted, is input to the input unit, the output unit outputs an update output result derived by the second trained model based on the second prediction update data.

9. The ready-mixed concrete quality prediction system described in claim 8, characterized in that the second trained model outputs the updated output result during the production of the ready-mixed concrete to be predicted or during the time between shipment and unloading.

10. The quality prediction system for ready-mixed concrete according to any one of claims 6 to 9, characterized in that it is provided with an alarm means for notifying when the predicted value of the slump value or slump flow value of the ready-mixed concrete at the time of shipment contained in the slump information at the time of shipment, which is the second output result, exceeds a predetermined threshold value.

11. A ready-mixed concrete quality prediction system as described in any one of claims 6 to 10, characterized in that the second trained model is updated by repeating machine learning using the prediction input data and the measured value of the slump value or slump flow value at the time of shipment of the ready-mixed concrete to be predicted as second training data.

12. The system for predicting the quality of ready-mixed concrete according to any one of claims 1 to 11, further comprising a display device for displaying the first output result.

13. a display device for displaying the first output result; The quality prediction system for ready-mixed concrete according to any one of claims 6 to 11, characterized in that the display device displays at least one of the slump value at the time of unloading based on the first output result, the predicted value of the slump flow value at the time of unloading, the slump value at the time of shipment based on the second output result, and the predicted value of the slump flow value at the time of shipment.

14. The first output result includes information related to the slump value or slump flow value of the fresh concrete at any time elapsed since production, The ready-mixed concrete quality prediction system according to claim 12 or 13, characterized in that the display device displays the slump value or slump flow value of the ready-mixed concrete at any elapsed time from production based on the first output result.

15. inputting prediction input data for ready-mixed concrete into a first trained model generated by machine learning based on first teacher data; obtaining a first output result derived by the first trained model; The first teacher data is data in which learning input data based on a first learning data group including mix-related information, which is information related to the mix of ready-mixed concrete, and shipping slump information, which is information related to the slump value or slump flow value of the ready-mixed concrete before shipping, and a second learning data group including at least one of production-related information, which is information related to the production of ready-mixed concrete, transportation-related information, which is information related to the transportation of ready-mixed concrete, and environment-related information, which is information related to the environment from the start of production of ready-mixed concrete to the completion of pouring, and learning output data including unloading slump information, which is information related to the slump value or slump flow value of the ready-mixed concrete at the time of unloading, are associated; the input data for prediction includes the mix proportion information and the shipping slump information of the ready-mix concrete to be predicted, and information corresponding to the first learning data group and the second learning data group of the ready-mix concrete to be predicted, The first output result is slump information at the time of unloading of the ready-mixed concrete to be predicted, a method for predicting the quality of ready-mixed concrete, the method comprising: acquiring at least one piece of information from an information management system at a ready-mixed concrete factory, the piece of information being included in the first learning data group, the piece of information being included in the second learning data group, and the piece of information being included in the prediction input data.

16. The method for predicting the quality of ready-mixed concrete described in claim 15, characterized in that after obtaining the first output result for the ready-mixed concrete to be predicted, first prediction update data, which is another prediction input data for the ready-mixed concrete to be predicted, is input, and an update output result derived based on the first prediction update data by the first trained model is obtained.

17. The method for predicting the quality of ready-mixed concrete described in claim 16, characterized in that the updated output result is obtained using the first trained model during the production of the ready-mixed concrete to be predicted or during the time between the time of shipment and the time of unloading.

18. The method for predicting the quality of ready-mixed concrete according to any one of claims 15 to 17, further comprising updating the first trained model by repeating machine learning using the prediction input data and the measured value of the slump value or slump flow value at the time of unloading of the ready-mixed concrete to be predicted as training data.

19. further comprising inputting prediction input data based on at least one of the mix proportion-related information of the ready-mixed concrete to be predicted, production-related information which is information related to the production of the ready-mixed concrete, and environment-related information which is information related to the environment from the start of production of the ready-mixed concrete to the completion of pouring, to a second trained model generated by machine learning based on second teacher data; The second teacher data is data in which learning input data based on at least one of the mix-related information, the manufacturing-related information, and the environment-related information is associated with learning output data including shipping slump information, which is information related to the slump value or slump flow value of the ready-mixed concrete at the time of shipping, The second output result derived by the second trained model includes information related to the shipment slump value or slump flow value of the ready-mixed concrete to be predicted derived by the second trained model, A method for predicting the quality of ready-mixed concrete according to any one of claims 15 to 18, characterized in that the second output result is used as shipping slump information in the prediction input data when deriving the first output result using the first trained model.

20. The method further includes inputting prediction input data based on image data of the ready-mixed concrete to be predicted to a second trained model generated by machine learning based on the second teacher data, The second teacher data is data in which learning input data based on image data of ready-mixed concrete at the time of production is associated with learning output data including shipping slump information which is information related to the slump value or slump flow value of the ready-mixed concrete at the time of shipment, A method for predicting the quality of ready-mixed concrete described in any one of claims 15 to 18, characterized in that a second output result including information related to the shipping slump value or slump flow value of the ready-mixed concrete to be predicted derived by the second trained model is used as shipping slump information in the prediction input data when deriving the first output result by the first trained model.

21. The method for predicting the quality of ready-mixed concrete described in claim 19 or 20, characterized in that after outputting the second output result for the ready-mixed concrete to be predicted, second prediction update data, which is another prediction input data for the ready-mixed concrete to be predicted, is input, and an update output result derived by the second trained model based on the second prediction update data is obtained.

22. The method for predicting the quality of ready-mixed concrete described in claim 21, characterized in that the updated output result is obtained using the second trained model during the production of the ready-mixed concrete to be predicted or during the time between shipment and unloading.

23. The method for predicting the quality of ready-mixed concrete according to any one of claims 19 to 22, further comprising updating a second trained model by repeating machine learning using the prediction input data and the measured value of the slump value or slump flow value of the ready-mixed concrete to be predicted at the time of shipment as second training data.

Images