Moisture content prediction device, added water flow rate prediction device, moisture content prediction method, and added water flow rate prediction method

Abstract

To provide a moisture content prediction device capable of predicting a moisture content in wheat.SOLUTION: This moisture content prediction device comprises: a storage unit that stores a machine learning model machine-learned using teacher data in which a past actual value containing a pre-conditioned wheat moisture content included in wheat which is a wheat flour raw material yet to be selectively conditioned, a wheat flow rate during water addition, a pre-conditioned added water flow rate, and an air quantity before condition completion which is sprayed to the wheat before the condition completion is used as an input value and a past actual value of a wheat moisture content before milling contained in the wheat yet to be milled after the completion of the selective conditioning of the wheat is used as an output; an acquisition unit that acquires values of the pre-conditioned wheat moisture content, the wheat flow rate during the water addition, the pre-conditioned added water flow rate, and the air quantity before the condition completion; a prediction unit that predicts the wheat moisture content before milling by inputting, as input values, the values of the pre-conditioned wheat moisture content, the wheat flow rate during the water addition, the pre-conditioned added water flow rate, and the air quantity before the condition completion to the machine learning model, and acquiring the prediction value of the wheat moisture content before milling as an output value; and an output unit that outputs the prediction result of the prediction unit.SELECTED DRAWING: Figure 2

Description

[Technical Field]

[0001] This invention relates to a moisture content prediction device that predicts the moisture content contained in wheat flour or wheat, which is a raw material for wheat flour, a moisture content prediction method using said moisture content prediction device, a water addition flow rate prediction device that predicts the optimum water addition flow rate to add to wheat during conditioning, and a water addition flow rate prediction method using said water addition flow rate prediction device. [Background technology]

[0002] Because wheat will germinate if stored in the same condition as when it was harvested, it is distributed after being dried until the moisture content reaches around 10%. When producing flour from wheat in this dried state, impurities are removed from the wheat before it is ground, and the wheat is conditioned to have a moisture content suitable for grinding.

[0003] The edible parts of wheat are generally the endosperm and germ, and the bran, which is the outer layer, is removed during the grinding process. To prevent the bran from being mixed into the endosperm during this grinding process, it is necessary to prevent the bran from breaking into small pieces. To achieve this, it is desirable that the bran be removed from the endosperm during the grinding process while maintaining a size that allows easy separation from the endosperm. It is known that the toughness of bran increases when it contains a certain amount of moisture. Therefore, in order to produce wheat flour that is free of bran contamination, it is necessary to add an appropriate amount of moisture to the wheat before the grinding process.

[0004] Generally, before the milling process, a tempering process is carried out in which wheat is left to absorb water by adding water until the moisture content is suitable for milling. In this tempering process, the wheat absorbs the added moisture from the germ and distributes it to the endosperm. The tempering process is usually carried out in two stages, and the wheat is left to stand for a total of about 24 hours during the tempering process. [Prior art documents] [Non-patent literature]

[0005] [Non-Patent Document 1] Written by Japan Wheat Research Society, "Wheat flour - its raw materials and processed products", 4th edition, Japan Wheat Research Society, February 28, 2007, p.300-305 p.319-335 p.364-366 p.432-442 Summary of the Invention [Problem to be solved by the invention]

[0006] In the tempering process, the moisture added to wheat is distributed between the endosperm and bran, and the wheat is left to stand for a predetermined period of time. However, the rate at which moisture moves to the endosperm varies depending on factors such as temperature, variety, and crop yield, and is not constant. Furthermore, leaving wheat to stand until all moisture has migrated is not industrially practical. Furthermore, the surface area of ​​wheat increases during the grinding process, but moisture on that surface is prone to dispersal into the environment. Furthermore, moisture is present in various parts of the wheat, and flour particles of various particle sizes are produced during the grinding process, with each flour particle having a different moisture content and dispersal rate. Furthermore, moisture dispersed into the environment changes the environment, resulting in complex interactions between the moisture content and dispersal rate of each flour particle and environmental changes, resulting in complex behavior of the moisture content of wheat.

[0007] Meanwhile, wheat flour as a product undergoes secondary processing at the purchaser's facility. At this time, the purchaser adds water to the wheat flour according to a recipe for each secondary processed product. However, if the moisture content of the flour before adding water is not consistent, the moisture content after adding water will not be stable, resulting in variations in the finished secondary processed product. For this reason, the moisture content of wheat flour is specified as a standard value. However, as mentioned above, not all of the water added in the tempering process is used to increase the moisture content of the wheat flour, and the moisture content of wheat is not constant due to various factors, making it difficult to predict the moisture content of wheat flour as a product.

[0008] In order to achieve the target moisture content in flour, flour mill operators intuitively understand various environmental factors such as the properties of the wheat used as the raw material, temperature and humidity during milling, and the conditions during grinding, and set the flow rate of water to add to the wheat.However, due to issues such as technical transfer and differences in years of experience, it is often impossible to achieve the target moisture content in flour.

[0009] An object of the present invention is to provide a moisture content prediction device that can predict the moisture content of wheat or wheat flour, a moisture content prediction method using said moisture content prediction device, a hydration flow rate prediction device that can predict the optimal hydration flow rate to be added to wheat during conditioning, and a hydration flow rate prediction method using said hydration flow rate prediction device. [Means for solving the problem]

[0010] The moisture content prediction device of the present invention is a moisture content prediction device that predicts a pre-grinding wheat moisture content, which is the moisture content contained in wheat before grinding after the completion of refinement of wheat, which is a raw material for wheat flour. The moisture content prediction device stores a first machine learning model that uses as input values ​​past actual values ​​including at least the pre-conditioning wheat moisture content, which is the moisture content contained in the wheat before refinement, the wheat flow rate at hydration, which is the flow rate of the wheat when hydration is added before tempering, the pre-conditioning hydration flow rate, which is the flow rate of hydration added to the wheat before tempering, and the pre-conditioning air volume, which is the air volume when air is blown onto the wheat before tempering is finished, and uses as output values ​​the past actual values ​​of the pre-grinding wheat moisture content. a first acquisition unit that acquires at least the value of the pre-conditioning wheat moisture content, the value of the wheat flow rate when hydration is added, the value of the pre-conditioning hydration flow rate, and the value of the air volume before the end of conditioning, which are used when predicting the pre-grinding wheat moisture content; and a first prediction unit that predicts the pre-grinding wheat moisture content by inputting at least the value of the pre-conditioning wheat moisture content, the value of the wheat flow rate when hydration is added, the value of the pre-conditioning hydration flow rate, and the value of the air volume before the end of conditioning acquired by the first acquisition unit as input values ​​into the first machine learning model stored in the first storage unit, and acquiring the predicted value of the pre-grinding wheat moisture content as an output value.

[0011] The moisture content prediction device of the present invention is a moisture content prediction device that predicts the moisture content contained in wheat flour, and is a second moisture content prediction device that has been machine-learned using training data in which input values ​​include past actual values ​​including at least the pre-milling wheat moisture content, which is the moisture content contained in wheat before milling after the completion of screening and conditioning of wheat, which is the raw material of the wheat flour, and the milling temperature, which is the temperature of the ground wheat obtained by milling the wheat and the air that comes into contact with the wheat flour when producing the wheat flour from the wheat, and the milling relative humidity, which is the relative humidity, and output values ​​are past actual values ​​of the post-milling flour moisture content, which is the moisture content contained in the flour after milling. The system is characterized by comprising a second memory unit that stores two machine learning models; a second acquisition unit that acquires at least the value of the wheat moisture content before grinding, the value of the temperature during milling, and the value of the relative humidity during milling, which are used when predicting the wheat moisture content after milling; and a second prediction unit that predicts the wheat moisture content after milling by inputting at least the value of the wheat moisture content before grinding, the value of the temperature during milling, and the value of the relative humidity during milling acquired by the second acquisition unit as input values ​​into the second machine learning model stored in the second memory unit, and acquiring the predicted value of the wheat moisture content after milling as an output value.

[0012] In addition, the moisture content prediction device of the present invention is a moisture content prediction device that predicts the moisture content contained in wheat flour, and is equipped with the moisture content prediction device of the present invention described above and an output unit that outputs the prediction result by the second prediction unit, and is characterized in that the second acquisition unit acquires the predicted value of the moisture content of the wheat before grinding predicted by the first prediction unit.

[0013] The moisture content prediction device of the present invention is a moisture content prediction device that predicts the moisture content contained in wheat flour, and is machine-learned using training data in which input values ​​are past actual values ​​including at least the pre-conditioning wheat moisture content, which is the moisture content contained in wheat, which is the raw material for the wheat flour before refinement and refinement, the wheat flow rate during hydration, which is the flow rate of the wheat when hydration is added before hydration, the pre-conditioning hydration flow rate, which is the flow rate of hydration added to the wheat before hydration, the pre-conditioning air volume, which is the air volume when air is blown onto the wheat before hydration is completed, and the milling temperature, which is the temperature of the air that comes into contact with the ground wheat obtained by grinding the wheat and the flour when producing wheat flour from the wheat, and the milling relative humidity, which is the relative humidity, and output values ​​are past actual values ​​of the post-milling flour moisture content, which is the moisture content contained in the flour after milling. an acquisition unit that acquires at least the value of the wheat moisture content before tempering, the value of the wheat flow rate when hydration is added, the value of the hydration flow rate before tempering, the value of the air volume before tempering is completed, the temperature during milling, and the value of the relative humidity during milling, which are used when predicting the flour moisture content after milling; a prediction unit that inputs at least the value of the wheat moisture content before tempering, the value of the wheat flow rate when hydration is added, the value of the hydration flow rate before tempering, the value of the air volume before tempering is completed, the temperature during milling, and the values ​​of the relative humidity during milling acquired by the acquisition unit as input values ​​into the machine learning model stored in the storage unit, and acquires the predicted value of the flour moisture content after milling as an output value, thereby predicting the flour moisture content after milling; and an output unit that outputs the prediction result by the prediction unit.

[0014] Furthermore, the moisture content prediction device of the present invention is characterized in that, when water is added to the wheat multiple times during conditioning, the value of the wheat flow rate during hydration and the value of the hydration flow rate before conditioning are values ​​each time water is added to the wheat.

[0015] The moisture content predicting device of the present invention is also characterized in that the pre-conditioning air volume is determined by whether or not air is blown onto the wheat before the end of conditioning.

[0016] The moisture content prediction device of the present invention is a moisture content prediction device that predicts a pre-milling wheat moisture content, which is the moisture content contained in wheat before milling after the wheat, which is a raw material for wheat flour, has been refined and refined, and stores a first machine learning model that uses as input values ​​past actual values ​​including at least the milling temperature, which is the temperature of the ground wheat obtained by milling the wheat, and the air that comes into contact with the wheat flour when producing the wheat flour from the wheat, and the milling relative humidity, which is the relative humidity, and the post-milling flour moisture content, which is the moisture content contained in the flour after milling, and stores as output values ​​past actual values ​​of the pre-milling wheat moisture content. a first acquisition unit that acquires at least the value of the milling temperature, the value of the milling relative humidity, and the desired value of the after-milling wheat moisture content to be used when predicting the pre-milling wheat moisture content so as to set the post-milling wheat moisture content to the desired moisture content; and a first prediction unit that predicts the pre-milling wheat moisture content by inputting at least the value of the milling temperature, the value of the milling relative humidity, and the desired value of the after-milling wheat moisture content acquired by the first acquisition unit as input values ​​into the first machine learning model stored in the first storage unit, and acquiring the predicted value of the pre-milling wheat moisture content as an output value.

[0017] The water addition flow rate prediction device of the present invention is a water addition flow rate prediction device that predicts an optimal water addition flow rate to be added to wheat during conditioning of wheat, which is a raw material for wheat flour, and is a second machine learning model trained by machine learning using training data in which past actual values ​​including at least a pre-conditioning wheat moisture content, which is the moisture content contained in the wheat, which is a raw material for wheat flour, before refinement and refinement, a water addition wheat flow rate, which is the flow rate of the wheat when hydration is added before conditioning, a pre-conditioning air volume, which is the air volume when air is blown onto the wheat before the end of conditioning, and a pre-grinding wheat moisture content, which is the moisture content contained in the wheat before grinding after the end of refinement and refinement of the wheat, are used as input values, and the past actual values ​​of the pre-conditioning hydration flow rate, which is the water addition flow rate to be added to the wheat before conditioning, are used as output values. a second acquisition unit that acquires at least the value of the wheat moisture content before tempering, the value of the wheat flow rate during hydration, the value of the air volume before tempering, and the desired value of the wheat moisture content before grinding, which are used when predicting the optimal hydration flow rate for adjusting the post-milling wheat moisture content to the desired moisture content; and a second prediction unit that predicts the optimal hydration flow rate by inputting at least the value of the wheat moisture content before tempering, the value of the wheat flow rate during hydration, the value of the air volume before tempering, and the desired value of the wheat moisture content before grinding, which are acquired by the second acquisition unit, as input values ​​to the second machine learning model stored in the second storage unit, and acquiring the predicted value of the optimal hydration flow rate as an output value.

[0018] In addition, the water addition flow rate prediction device of the present invention is a water addition flow rate prediction device that predicts the optimal water addition flow rate to be added to wheat, which is a raw material for wheat flour, when conditioning the wheat, and is equipped with the moisture content prediction device of the present invention described above and the water addition flow rate prediction device of the present invention described above, and an output unit that outputs the prediction result by the second prediction unit, and is characterized in that the second acquisition unit acquires the predicted value of the pre-grinding wheat moisture content predicted by the first prediction unit as the desired pre-grinding wheat moisture content value.

[0019] The water addition flow rate prediction device of the present invention is a water addition flow rate prediction device that predicts an optimal water addition flow rate to be added to wheat during conditioning of wheat, which is a raw material for wheat flour, and is configured to use a machine learning model trained by machine learning using training data in which, as input values, past actual values ​​including at least the pre-conditioning wheat moisture content, which is the moisture content contained in the wheat, which is a raw material for wheat flour before refinement, the hydration wheat flow rate, which is the flow rate of the wheat when hydrating before conditioning, the air volume before end of conditioning, which is the air volume when air is blown on the wheat before end of conditioning, the milling temperature, which is the temperature and milling relative humidity of the air that comes into contact with the ground wheat obtained by grinding the wheat and the flour when producing wheat flour from the wheat, and the post-milling flour moisture content, which is the moisture content contained in the flour after milling, are used, and the output value is the past actual value of the pre-conditioning hydration flow rate, which is the water volume to be added to the wheat before conditioning. an acquisition unit that acquires at least the value of the wheat moisture content before tempering, the value of the wheat flow rate during hydration, the value of the air volume before tempering, the value of the temperature during milling, the value of the relative humidity during milling, and the desired value of the flour moisture content after milling, which are used when predicting the optimal water addition flow rate to make the flour moisture content after milling the desired moisture content; a prediction unit that predicts the optimal water addition flow rate by inputting at least the value of the wheat moisture content before tempering, the value of the wheat flow rate during hydration, the value of the air volume before tempering, the value of the temperature during milling, the value of the relative humidity during milling, and the desired flour moisture content after milling, which are acquired by the acquisition unit, as input values ​​into the machine learning model stored in the storage unit, and acquiring the predicted value of the optimal water addition flow rate as an output value; and an output unit that outputs the prediction result by the prediction unit.

[0020] Furthermore, the water addition flow rate prediction device of the present invention is characterized in that, when water is added to the wheat multiple times during conditioning, the value of the wheat flow rate during hydration is the value for each time water is added to the wheat, and the value of the water addition flow rate before conditioning, which is the flow rate of water added to the wheat for each time water is added to the wheat other than the optimal water addition flow rate, is included in the input value.

[0021] The water addition flow rate predicting device of the present invention is also characterized in that the air volume before the end of conditioning is determined by whether or not air is blown onto the wheat before the end of conditioning.

[0022] The moisture content prediction method of the present invention is a moisture content prediction method that uses a moisture content prediction device to predict the pre-grinding wheat moisture content, which is the moisture content contained in wheat, a raw material for wheat flour, before grinding after the completion of refinement and conditioning of the wheat, wherein an acquisition unit of the moisture content prediction device first acquires a pre-conditioning wheat moisture content value, which is the moisture content contained in the wheat before refinement and conditioning, a hydration wheat flow rate value, which is the flow rate of the wheat when hydration is added before conditioning, a pre-conditioning hydration flow rate value, which is the flow rate of hydration to be added to the wheat before conditioning, and a pre-conditioning air flow rate value, which is the air volume when air is blown onto the wheat before the end of conditioning, all of which are used when predicting the pre-grinding wheat moisture content; the moisture content prediction device includes a first reading step of reading from a memory unit of the moisture content prediction device a first machine learning model that has been trained by machine learning using training data that uses past actual values ​​including the wheat flow rate when hydrating, the hydration flow rate before tempering, and the air volume before the end of tempering as input values ​​and the past actual value of the wheat moisture content before grinding as output values; and a first prediction step of inputting at least the value of the pre-tempering wheat moisture content, the value of the wheat flow rate when hydrating, the value of the pre-tempering hydration flow rate, and the value of the air volume before the end of tempering acquired in the acquisition step as input values ​​into the first machine learning model read in the first reading step, and acquiring the predicted value of the pre-grinding wheat moisture content as an output value, thereby predicting the pre-grinding wheat moisture content.

[0023] The moisture content prediction method of the present invention is a moisture content prediction method for predicting the moisture content of wheat flour using a moisture content prediction device, and includes a second acquisition step in which an acquisition unit of the moisture content prediction device acquires at least a pre-milling wheat moisture value, which is the moisture content contained in wheat before milling after the completion of refinement of wheat, which is a raw material for the wheat flour, and a milling temperature value, which is the temperature of the ground wheat obtained by grinding the wheat when producing the wheat flour, and a milling relative humidity value, which is the relative humidity, used in predicting the post-milling wheat moisture content, which is the moisture content contained in the wheat flour after milling; The method includes a second reading step of reading from the memory unit of the moisture content prediction device a second machine learning model that has been trained by machine learning using training data that uses past actual values ​​including the wheat moisture content, the milling temperature, and the milling relative humidity as input values ​​and the past actual value of the after-milling flour moisture content as output values; and a second prediction step of predicting the after-milling flour moisture content in the prediction unit of the moisture content prediction device by inputting at least the pre-milling wheat moisture content value, the milling temperature value, and the milling relative humidity value acquired in the second acquisition step as input values ​​to the second machine learning model read in the second reading step, and acquiring the predicted value of the after-milling flour moisture content as an output value.

[0024] Furthermore, the moisture content prediction method of the present invention is a moisture content prediction method that predicts the moisture content contained in wheat flour using a moisture content prediction device, and includes the moisture content prediction method of the present invention described above and an output process that outputs the result predicted in the second prediction process, and is characterized in that the second acquisition process acquires the value of the moisture content of the wheat before grinding predicted in the first prediction process.

[0025] The moisture content prediction method of the present invention is a moisture content prediction method that predicts the moisture content of wheat flour using a moisture content prediction device, and includes an acquisition step in which an acquisition unit of the moisture content prediction device acquires at least a pre-conditioning wheat moisture content value, which is the moisture content contained in the wheat that is the raw material for the wheat flour before refinement and refined, a hydration wheat flow rate value, which is the flow rate of the wheat when hydration is added before tempering, a pre-conditioning hydration flow rate value, which is the flow rate of water added to the wheat before tempering, a pre-conditioning air flow rate value, which is the air volume when air is blown onto the wheat before tempering is completed, and a milling temperature value, which is the temperature of the air that comes into contact with the ground wheat obtained by grinding the wheat and the flour when producing wheat flour from the wheat, and a milling relative humidity value, which is the relative humidity, used in predicting the post-milling wheat moisture content, which is the moisture content contained in the wheat after milling; The method comprises a reading step of reading from a storage unit of the moisture content prediction device a machine learning model trained by machine learning using training data in which past actual values ​​including the wheat moisture content during watering, the hydration flow rate before tempering, the air volume before tempering end, the temperature during milling, and the relative humidity during milling are used as input values, and the past actual values ​​of the after-milling flour moisture content are used as output values; a prediction step in a prediction unit of the moisture content prediction device, in which at least the before-tempering wheat moisture content value, the wheat flow rate during hydration, the hydration flow rate before tempering, the air volume before tempering end, the temperature during milling, and the relative humidity during milling, acquired in the acquisition step, are input as input values ​​to the machine learning model read in the reading step, and the after-milling flour moisture content is predicted by acquiring a predicted value of the after-milling flour moisture content; and an output step of outputting the result predicted in the prediction step from an output unit of the moisture content prediction device.

[0026] Furthermore, the moisture content prediction method of the present invention is characterized in that, when water is added to the wheat multiple times during conditioning, the value of the wheat flow rate during hydration and the value of the hydration flow rate before conditioning are values ​​for each time water is added to the wheat.

[0027] In the moisture content prediction method of the present invention, the pre-conditioning air volume is determined by whether or not air blows onto the wheat before the end of conditioning.

[0028] The moisture content prediction method of the present invention is a moisture content prediction method using a moisture content prediction device to predict the pre-milling wheat moisture content, which is the moisture content contained in wheat, which is a raw material for wheat flour, before milling after the completion of screening and conditioning of the wheat, and includes a first acquisition step in the moisture content prediction device to acquire at least a milling temperature value, which is the temperature of the ground wheat obtained by milling the wheat, and a milling relative humidity value, which is the relative humidity, when producing wheat flour from the wheat, and a post-milling flour moisture content, which is the moisture content contained in the flour after milling, as desired; and a first reading step in the moisture content prediction device to acquire at least the milling temperature value, which is the temperature of the ground wheat obtained by milling the wheat, and a milling relative humidity value, which is the relative humidity, when producing wheat flour from the wheat. The moisture content prediction device is characterized by including a first reading step of reading from a first memory unit of the moisture content prediction device a first machine learning model that has been machine-trained using training data that uses past actual values ​​including the relative humidity at the time of milling and the post-milling wheat moisture content as input values ​​and the past actual value of the pre-milling wheat moisture content as output values, and a first prediction step in a first prediction unit of the moisture content prediction device of inputting at least the value of the milling temperature, the value of the relative humidity at the time of milling, and the desired value of the post-milling wheat moisture content acquired in the first acquisition step as input values ​​to the first machine learning model read in the first reading step, and acquiring the predicted value of the pre-milling wheat moisture content as an output value, thereby predicting the pre-milling wheat moisture content.

[0029] Furthermore, the water addition flow rate prediction method of the present invention is a method for predicting an optimum water addition flow rate to be added to wheat, which is a raw material for wheat flour, during conditioning using a water addition flow rate prediction device, wherein a second acquisition unit of the water addition flow rate prediction device acquires at least a pre-conditioning wheat moisture content value, which is the moisture content contained in the wheat, which is a raw material for wheat flour, before refinement, a water addition wheat flow rate value, which is the flow rate of the wheat when water is added before refinement, a pre-conditioning air volume value, which is the air volume when air is blown onto the wheat before the end of refinement, and a pre-grinding wheat moisture content value, which is the moisture content contained in the wheat, which is a raw material for wheat flour, before grinding after the end of refinement of the wheat, which is the raw material for wheat flour, and which is the desired pre-grinding wheat moisture content value, which are used when predicting the optimum water addition flow rate to make the post-milling wheat moisture content, which is the moisture content contained in the wheat after milling, a second acquisition step and a second reading step in a second reading section of the hydration flow rate prediction device for reading from a second memory section of the hydration flow rate prediction device a second machine learning model trained by machine learning using training data that uses past actual values ​​including at least the pre-conditioning wheat moisture content, the wheat flow rate during hydration, the air volume before the end of tempering, and the pre-grinding wheat moisture content as input values ​​and the past actual value of the pre-conditioning hydration flow rate as an output value; and a second prediction step in a second prediction section of the hydration flow rate prediction device for inputting as input values ​​to the second machine learning model read in the second reading step at least the pre-conditioning wheat moisture content value, the wheat flow rate during hydration, the air volume before the end of tempering, and the desired pre-grinding wheat moisture content value obtained in the second obtaining step, and thereby predicting the optimal hydration flow rate.

[0030] Furthermore, the water addition flow rate prediction method of the present invention is a method for predicting the optimum water addition flow rate to be added to wheat, which is a raw material for wheat flour, when conditioning the wheat using a water addition flow rate prediction device, and includes the moisture content prediction method of the present invention described above and the water addition flow rate prediction method of the present invention described above, and an output step for outputting the result predicted in the second prediction step, and is characterized in that the second acquisition step acquires the value of the pre-grinding wheat moisture content predicted in the first prediction step as the desired value of the pre-grinding wheat moisture content.

[0031] The water addition flow rate prediction method of the present invention is a method for predicting an optimum water addition flow rate to be added to wheat, which is a raw material for wheat flour, during conditioning using a water addition flow rate prediction device, and includes an acquisition step in an acquisition unit of the water addition flow rate prediction device to acquire at least a pre-conditioning wheat moisture content value, which is the moisture content contained in the wheat, which is a raw material for the wheat flour, before refinement, a water addition wheat flow rate value, which is the flow rate of the wheat when water is added before conditioning, a pre-conditioning air flow rate value, which is the air flow rate when air is blown onto the wheat before conditioning is completed, a milling temperature value, which is the temperature of the air that comes into contact with the ground wheat obtained by grinding the wheat and the flour when producing wheat flour from the wheat, and a milling relative humidity value, which is the relative humidity, and a desired post-milling flour moisture content value, all of which are used when predicting the optimum water addition flow rate to make the post-milling flour moisture content, which is the moisture content contained in the wheat after milling, a desired moisture content; The method includes a reading step in which a reading unit reads from a storage unit of the hydration flow rate prediction device a machine learning model trained by machine learning using training data that uses past actual values ​​including at least the pre-tempering wheat moisture content, the wheat flow rate when hydration is added, the air volume before the end of tempering, the milling temperature, the milling relative humidity, and the post-milling flour moisture content as input values ​​and the past actual value of the pre-tempering hydration flow rate as an output value; a prediction step in which a prediction unit of the hydration flow rate prediction device inputs as input values ​​to the machine learning model read in the reading step at least the pre-tempering wheat moisture content value, the wheat flow rate when hydration is added, the air volume before the end of tempering, the milling temperature, the milling relative humidity, and the desired post-milling flour moisture content value acquired in the acquisition step, and predicts the optimal hydration flow rate by obtaining a predicted value of the optimal hydration flow rate as an output value; and an output step in which the result predicted in the prediction step is output.

[0032] Furthermore, in the method for predicting the water addition flow rate of the present invention, when water is added to the wheat multiple times during conditioning, the value of the wheat flow rate during hydration is the value for each time water is added to the wheat, and the value of the water addition flow rate before conditioning, which is the flow rate of water added to the wheat for each time water is added to the wheat other than the optimal water addition flow rate, is included in the input value.

[0033] The method for predicting the amount of water to be added according to the present invention is characterized in that the pre-conditioning air volume is determined by whether or not wind blows on the wheat before the end of conditioning. [Effects of the Invention]

[0034] According to the present invention, there are provided a moisture content prediction device capable of predicting the moisture content contained in wheat or wheat flour, a moisture content prediction method using said moisture content prediction device, a hydration flow rate prediction device capable of predicting the optimum hydration flow rate to be added to wheat during conditioning, and a hydration flow rate prediction method using said hydration flow rate prediction device. [Brief explanation of the drawings]

[0035] [Figure 1] FIG. 1 is a diagram illustrating a process for producing wheat flour from wheat. [Figure 2] 1 is a block diagram showing a system configuration of a moisture content prediction device according to a first embodiment. [Figure 3] 3 is a flowchart illustrating a moisture content prediction method according to the first embodiment. [Figure 4] 10 is a flowchart illustrating a moisture content prediction method according to a second embodiment. DETAILED DESCRIPTION OF THE INVENTION

[0036] A moisture content prediction device and moisture content prediction method according to a first embodiment of the present invention will be described below with reference to the drawings. The moisture content prediction device according to the first embodiment is a device for predicting the moisture content contained in wheat after producing flour from wheat (after milling) at any point in time (hereinafter referred to as the "post-milling flour moisture content"). The post-milling flour moisture content is the value obtained by dividing the weight of water contained in the flour by the weight of the flour. Before describing the configuration of the moisture content prediction device according to the first embodiment, the process of producing flour from wheat will be described with reference to FIG. 1. The wheat milling process to obtain flour from wheat is divided into a pre-process, a refining and conditioning process, and a post-process, a production process. The refining and conditioning section 2, where the refining and conditioning process is performed, includes a coarse wheat tank 4, a first conveying section 6, a refining section 8, a second conveying section 10, and a conditioning section 12, as shown in FIG. 1.

[0037] The coarse barley tank 4 stores wheat, which is a raw material for wheat flour, before being refined, tempered, and blended. The first conveying section 6 blends the wheat in the coarse barley tank 4 and pneumatically transports it to the refining section 8. The refining section 8 refines the wheat that has been blended in the first conveying section 6 and transported by the first conveying section 6. The refining section 8 uses various refining machines to separate the wheat from impurities and seeds of plants other than wheat that have been mixed in the wheat. The second conveying section 10 pneumatically transports the wheat that has finished being refined in the refining section 8 to the refining section 12.

[0038] The conditioning section 12 conditions the wheat transported by the second conveying section 10. As shown in Figure 1, the conditioning section 12 includes a primary hydration section 14, a primary tank 16, a third conveying section 18, a secondary hydration section 20, a secondary tank 22, a fourth conveying section 24, a 1B hydration section 26, and a check bin 28.

[0039] The primary hydration unit 14 pours water supplied at a predetermined flow rate onto the wheat transported by the second transport unit 10 while moving the wheat at a predetermined flow rate. The wheat to which water has been added in the primary hydration unit 14 is stored in the primary tank 16, and the wheat stored in the primary tank 16 is aged for approximately 16 hours. The third transport unit 18 pneumatically transports the wheat after aging in the primary tank 16 to the secondary hydration unit 20.

[0040] Secondary hydration unit 20 moves the wheat transported by third transport unit 18 at a predetermined flow rate while pouring water supplied at a predetermined flow rate onto the wheat. Secondary tank 22 contains the wheat to which water has been added in secondary hydration unit 20, and the wheat contained in secondary tank 22 is left to age for approximately eight hours. Fourth transport unit 24 pneumatically transports the wheat after it has been left to age in secondary tank 22 to 1B hydration unit 26.

[0041] The 1B hydration unit 26 pours water supplied at a predetermined flow rate onto the wheat transported by the fourth transport unit 24 while moving the wheat at a predetermined flow rate. The check bin 28 receives the wheat that has been hydrated by the 1B hydration unit 26 and allows it to rest for approximately 30 minutes. The check bin 28 is equipped with a cooling device (not shown), which blows air onto the wheat moving within the check bin 28 when it is necessary to cool the wheat.

[0042] Next, a description will be given of the manufacturing section 30 where the manufacturing process is carried out. The manufacturing section 30 includes a crushing section 32, a fifth conveying section 34, a grinding section 36, a sixth conveying section 38, and a product tank 40, as shown in FIG.

[0043] The crushing unit 32 crushes wheat that passes through the check bin 38. The fifth conveying unit 34 pneumatically transports the crushed wheat material crushed by the crushing unit 32 to the grinding unit 36. The crushing unit 36 ​​powders the crushed wheat material by repeating the order and number of steps (1) to (4) of (1) grinding by a roller, (2) pneumatic transport, (3) sieving by a sifter, and (4) purification by a purifier, for example, by repeating steps (1), (2), and (3), or by repeating steps (1), (2), (3), and (4), depending on the condition of the crushed wheat material conveyed by the fifth conveying unit 34. The sixth conveying unit 38 pneumatically transports the wheat flour produced by the crushing unit 36 ​​to the product tank 40.

[0044] Next, a moisture content prediction device according to a first embodiment will be described. Fig. 2 is a block diagram showing the system configuration of a moisture content prediction device 50 according to the first embodiment. As shown in Fig. 2, the moisture content prediction device 50 includes a control unit 52 that performs overall control of each unit of the moisture content prediction device 50. A storage unit 54, an input unit 60, and an output unit 64 are connected to the control unit 52. The control unit 52 also includes an acquisition unit 66, a first prediction unit 68, and a second prediction unit 70.

[0045] The memory unit 54 stores a first machine learning model 56 used for prediction by the first prediction unit 68 (described later), and a second machine learning model 58 used for prediction by the second prediction unit 70 (described later). The first machine learning model 56 is a model trained by machine learning in advance to predict the moisture content contained in wheat before grinding after the wheat, which is the raw material for wheat flour, has been refined and tempered, i.e., wheat after passing through the check bin 28 and before being crushed in the crushing unit 32 (hereinafter referred to as the "pre-milling wheat moisture content"). The second machine learning model 58 is a model trained by machine learning in advance to predict the post-milling flour moisture content, which is the moisture content contained in flour after wheat has been produced from wheat (after milling), i.e., flour produced by the crushing unit 36.

[0046] The first machine learning model 56 is a machine learning model trained by regression using a large amount of training data, with various past actual values ​​(measured values ​​measured over a predetermined period of time) necessary to predict the moisture content of wheat before grinding as input values ​​and past actual values ​​(measured values ​​measured over a predetermined period of time) of the pre-grinding wheat moisture content to be predicted as output values. The various past actual values ​​used as input values ​​include at least (1) the moisture content of wheat before screening and tempering, i.e., the wheat in the screening unit 8 (hereinafter referred to as the "pre-conditioning wheat moisture content"), (2-1) the wheat at the time of hydration before the first tempering, i.e., the flow rate per unit time of wheat in the first hydration unit 14 (hereinafter referred to as the "wheat flow rate at first hydration"), (2-2) the wheat at the time of hydration before the second tempering, i.e., the flow rate per unit time of wheat in the second hydration unit 20 (hereinafter referred to as the "wheat flow rate at second hydration"), (3-1 (3-1) The amount of water added per unit time to the wheat before the first conditioning, i.e., the amount of water added in the first hydration section 14 (hereinafter referred to as the "water addition amount before first conditioning"); (3-2) The amount of water added per unit time to the wheat before the second conditioning, i.e., the amount of water added in the second hydration section 20 (hereinafter referred to as the "water addition amount before second conditioning"); and (4) The amount of air blown on the wheat before the end of conditioning to cool it as it moves through the check bin 28 (hereinafter referred to as the "air volume before the end of conditioning").

[0047] The moisture content of wheat before tempering and the moisture content of wheat before grinding are values ​​obtained by dividing the weight of moisture contained in the wheat by the weight of the wheat. Moisture meters (not shown) are installed in the screening section 8 and between the check bin 28 and the crushing section 32, and the moisture meters measure the moisture content of the wheat. Examples of moisture meters that can be used include an inline infrared moisture meter and an inline microwave moisture meter. Regarding the moisture content of wheat before tempering, instead of measuring it with a moisture meter in the screening section 8, the moisture content of the wheat may be measured before the wheat is stored in the coarse barley tank 4. Instead of installing a moisture meter in the screening section 8 and between the check bin 28 and the crushing section 32, the moisture content of the wheat may be measured using the bone dry method. The wheat flow rate during primary hydration and the wheat flow rate during secondary hydration are the weight of wheat flowing through the wheat flow path of the primary hydration section 14 and the secondary hydration section 20 per unit time, respectively. The water addition flow rate before the first conditioning and the water addition flow rate before the second conditioning are the weight of water (water volume) flowing per unit time through the wheat flow path of the first hydration section 14 and the second hydration section 20. The air volume before the end of conditioning is the volume per unit time of air supplied to and discharged from the wheat flow path in the check bin 28.

[0048] The second machine learning model 58 is a machine learning model trained by regression using a large amount of training data. The input values ​​are various past actual values ​​(measured over a predetermined period) necessary to predict the moisture content of flour after milling, and the output values ​​are past actual values ​​(measured over a predetermined period) of the predicted moisture content of flour after milling. The various past actual values ​​used as input values ​​include at least (1) the moisture content of wheat before grinding, (2) the temperature of the air that comes into contact with the ground wheat produced from wheat and the flour when producing flour from wheat in the production unit 30 (grinding unit 36) (hereinafter referred to as the "milling temperature"), and (3) the relative humidity of the air that comes into contact with the ground wheat produced from wheat and the flour when producing flour from wheat in the production unit 30 (grinding unit 36) (hereinafter referred to as the "milling relative humidity").

[0049] The moisture content of flour after milling is the weight of water contained in the flour divided by the weight of the flour. A moisture meter (not shown) is installed between the grinding unit 36 ​​and the product tank 40, and measures the moisture content of the flour. The moisture meter may be, for example, an in-line infrared moisture meter or an in-line microwave moisture meter. Alternatively, instead of installing a moisture meter between the grinding unit 36 ​​and the product tank 40, the moisture content of the flour may be measured by the bone dry method.

[0050] Furthermore, when creating the first machine learning model 56 and the second machine learning model 58, past actual values ​​from periods with large value fluctuations are excluded from the training data. Furthermore, to obtain stable past actual values, the past actual values ​​may be taken as average values ​​over a certain period, and this average value may be used as the training data. However, if there is large fluctuation in the values ​​of the milling temperature and milling relative humidity, or if there is large fluctuation in the moisture content of the flour after milling, the past actual values ​​are not taken as average values ​​over a certain period, but are taken as time-series values, and this time-series value is used as the training data.

[0051] The input unit 60 inputs the value of the wheat moisture content before tempering, the value of the wheat flow rate at the first hydration addition, the value of the wheat flow rate at the second hydration addition, the value of the hydration flow rate before the first tempering, the value of the hydration flow rate before the second tempering, and the value of the air volume before the end of tempering, which are used when predicting the wheat moisture content before grinding using the first machine learning model 56. The input unit 60 also inputs the value of the temperature during milling and the value of the relative humidity during milling, which are used when predicting the wheat moisture content after milling using the second machine learning model 58.

[0052] The values ​​input to the input unit 60 are actual measurements taken using various measuring instruments, set values ​​set based on past performance, or assumed values ​​such as prior measurements, and are determined depending on the arbitrary point in time at which the moisture content of wheat before grinding is predicted using the first machine learning model 56. That is, during the period from the arbitrary point in time (1) up to the start of transport of the wheat for which moisture is to be predicted to the primary hydration unit 14, an assumed value is input as the value of the moisture content of wheat before tempering, and set values ​​are input as the value of the wheat flow rate at the primary hydration, the value of the wheat flow rate at the secondary hydration, the value of the hydration flow rate before the primary hydration, the value of the hydration flow rate before the secondary hydration, and the value of the air volume before the end of tempering.

[0053] At any given time (2) between the start of transport of the wheat for which moisture is predicted to the primary hydration section 14 and the start of transport to the secondary hydration section 20, actual measured values ​​are input as the value of the wheat moisture content before conditioning, the value of the wheat flow rate during primary hydration, and the value of the hydration flow rate before primary conditioning, and set values ​​are input as the value of the wheat flow rate during secondary hydration, the value of the hydration flow rate before secondary conditioning, and the air volume before the end of conditioning.

[0054] At any point in time (3) between the start of transport of the wheat for which moisture is predicted to the secondary hydration section 20 and the start of transport to the 1B hydration section 26, actual measured values ​​are input as the value of the wheat moisture content before conditioning, the value of the wheat flow rate during primary hydration, the value of the wheat flow rate during secondary hydration, the value of the hydration flow rate before primary conditioning, and the value of the hydration flow rate before secondary conditioning, and a set value is input as the value of the air volume before the end of conditioning.

[0055] At any point in time (4), after the wheat for which moisture is predicted has started to be transported to the 1B hydration section 26, actual measured values ​​are input as the value of the moisture content of the wheat before conditioning, the value of the wheat flow rate during the first hydration, the value of the wheat flow rate during the second hydration, the value of the hydration flow rate before the first conditioning, the value of the hydration flow rate before the second conditioning, and the value of the air volume before the end of conditioning.

[0056] In addition, the values ​​of the milling temperature and milling relative humidity used when predicting the moisture content of flour after milling using the second machine learning model 58 are predicted using the forecast outside temperature and forecast relative humidity based on weather forecasts for a specified period (at least one hour) at the location where the milling unit 30 is installed, obtained at a frequency of at least one hour, the rate of outside air intake into the location where the milling unit 30 is installed, etc., and the set values ​​are input.

[0057] The output unit 64 outputs the prediction results from the first prediction unit 68 and the second prediction unit 70. The acquisition unit 66 of the control unit 52 acquires the values ​​of the moisture content of wheat before tempering, the wheat flow rate at the first hydration addition, the wheat flow rate at the second hydration addition, the hydration flow rate before the first tempering, the hydration flow rate before the second tempering, the air volume before the end of tempering, the temperature during milling, and the relative humidity during milling input to the input unit 60.

[0058] The first prediction unit 68 inputs the value of the pre-conditioning wheat moisture content, the value of the wheat flow rate at the first hydration addition, the value of the wheat flow rate at the second hydration addition, the value of the hydration flow rate before the first hydration addition, the value of the hydration flow rate before the second hydration addition, and the value of the air volume before the end of hydration acquired by the acquisition unit 66 as input values ​​into the first machine learning model 56 stored in the memory unit 54, and predicts the pre-grinding wheat moisture content by acquiring the predicted value of the pre-grinding wheat moisture content as an output value.

[0059] The second prediction unit 70 inputs the predicted value of the pre-milling wheat moisture content predicted by the first prediction unit 68, as well as the milling temperature value and milling relative humidity value acquired by the acquisition unit 66, into the second machine learning model 58 stored in the memory unit 54 as input values, and acquires the predicted value of the post-milling flour moisture content as an output value, thereby predicting the post-milling flour moisture content.

[0060] Next, a moisture content prediction method for predicting the moisture content of flour after milling using the moisture content prediction device 50 according to the first embodiment will be described. Fig. 3 is a flowchart for explaining the process executed by the control unit 52 to predict the moisture content of flour after milling.

[0061] First, the acquisition unit 66 of the control unit 52 acquires the values ​​of the moisture content of wheat before tempering, the wheat flow rate at the first hydration addition, the wheat flow rate at the second hydration addition, the water flow rate before the first tempering, the water flow rate before the second tempering, the air volume before the end of tempering, the temperature during milling, and the relative humidity during milling, all of which have been input to the input unit 60 (step S2).

[0062] Next, the control unit 52 loads the first machine learning model 56 stored in the memory unit 54 (step S3), and the first prediction unit 68 of the control unit 52 predicts the pre-grinding wheat moisture content using the first machine learning model 56 loaded in step S3 (step S4). Specifically, the first prediction unit 68 inputs the pre-conditioning wheat moisture content value, the wheat flow rate during first hydration, the wheat flow rate during second hydration, the hydration flow rate before first hydration, the hydration flow rate before second hydration, and the air volume before the end of hydration obtained in step S2 as input values ​​to the first machine learning model 56, and obtains a predicted pre-grinding wheat moisture content value as an output value, thereby predicting the pre-grinding wheat moisture content of wheat at the above-mentioned arbitrary time point (step S4).

[0063] Next, control unit 52 loads second machine learning model 58 stored in memory unit 54 (step S5), and second prediction unit 70 of control unit 52 predicts the post-milling flour moisture content using second machine learning model 58 loaded in step S5 (step S6). Specifically, second prediction unit 70 inputs the predicted value of the pre-milling wheat moisture content predicted in step S4, the milling temperature value and the milling relative temperature value obtained in step S2 into second machine learning model 58 as input values, and obtains the predicted value of the post-milling flour moisture content as an output value, thereby predicting the post-milling flour moisture content of wheat at the above-mentioned arbitrary time point (step S6).

[0064] After completing the process of step S6, the control unit 52 outputs the results of the predictions made in steps S4 and S6 to the output unit 64 (step S7).

[0065] According to the moisture content prediction device and moisture content prediction method of the first embodiment, the pre-milling wheat moisture content can be accurately predicted from the minimum necessary variables in the refining and tempering process using the first machine learning model 56. Furthermore, the post-milling flour moisture content can be accurately predicted from the minimum necessary variables in the production process using the second machine learning model 58. Furthermore, the post-milling flour moisture content can be accurately predicted from the minimum necessary variables from the refining and tempering process to the production process using the first machine learning model 56 and the second machine learning model 58.

[0066] In the first embodiment described above, the first machine learning model 56 and the second machine learning model 58 are used to predict the moisture content of wheat flour after milling, but it is also possible to predict the moisture content of wheat before milling using only the first machine learning model 56 and output only the prediction result to the output unit 64. It is also possible to predict the moisture content of wheat after milling using only the second machine learning model 58. In this case, the values ​​of the moisture content of wheat before milling, the temperature during milling, and the relative humidity during milling input to the input unit 60 are actually measured values.

[0067] Furthermore, in the first embodiment described above, the moisture content of flour after milling is predicted using two machine learning models, the first machine learning model 56 and the second machine learning model 58, but it is also possible to predict the moisture content of flour after milling using a single machine learning model. In this case, the machine learning model includes as input values ​​at least the moisture content of wheat before tempering, the wheat flow rate at the first hydration addition, the wheat flow rate at the second hydration addition, the hydration flow rate before the first tempering, the hydration flow rate before the second tempering, the air volume before the end of tempering, the temperature during milling, and the relative humidity during milling, and is trained by regression using a large amount of training data in which these past actual values ​​are used as input values ​​and past actual values ​​of the moisture content of flour after milling are used as output values.

[0068] Furthermore, when one machine learning model is used, the values ​​input into input unit 60 are determined depending on the arbitrary time point for predicting the post-milling flour moisture content. That is, during the period from the arbitrary time point (1) up to the time before the wheat for which moisture is predicted starts to be transported to primary hydration unit 14, an assumed value is input as the value of the pre-tempering wheat moisture content, set values ​​are input as the value of the wheat flow rate during primary hydration, the value of the wheat flow rate during secondary hydration, the value of the value of the hydration flow rate before primary hydration, the value of the hydration flow rate before secondary hydration, and the value of the air volume before the end of tempering, and predicted values ​​from a weather forecast or the like are input as the value of the temperature during milling and the value of the relative humidity during milling.

[0069] At any given time (2) between the time when the wheat for which moisture is predicted starts being transported to the primary hydration section 14 and the time when it starts being transported to the secondary hydration section 20, an actual measured value is input as the value of the moisture content of the wheat before conditioning, actual measured values ​​are input as the value of the wheat flow rate during primary hydration and the value of the hydration flow rate before primary conditioning, set values ​​are input as the value of the wheat flow rate during secondary hydration, the value of the hydration flow rate before secondary conditioning, and the value of the air volume before the end of conditioning, and predicted values ​​from a weather forecast or the like are input as the value of the temperature during milling and the value of the relative humidity during milling.

[0070] At any given time (3) between the time when the wheat for which moisture is predicted starts being transported to the secondary hydration section 20 and the time when it starts being transported to the 1B hydration section 26, an actual measured value is input as the value of the moisture content of the wheat before conditioning, actual measured values ​​are input as the value of the wheat flow rate at the first hydration, the value of the wheat flow rate at the second hydration, the value of the hydration flow rate before the first hydration, and the value of the hydration flow rate before the second hydration, a set value is input as the value of the air volume before the end of conditioning, and predicted values ​​from a weather forecast or the like are input as the value of the temperature during milling and the value of the relative humidity during milling.

[0071] At any point in time (4), after the wheat for which moisture is predicted has started to be transported to 1B hydration section 26, an actual measured value is input as the value of the moisture content of the wheat before conditioning, and actual measured values ​​are input as the value of the wheat flow rate at the first hydration addition, the value of the wheat flow rate at the second hydration addition, the value of the hydration flow rate before first conditioning, the value of the hydration flow rate before second conditioning, and the value of the air volume before the end of conditioning, and predicted values ​​from a weather forecast or the like are input as the value of the temperature during milling and the value of the relative humidity during milling.

[0072] Furthermore, in the first embodiment described above, the minimum necessary variables, i.e., the pre-tempering wheat moisture content, the wheat flow rate at first hydration, the wheat flow rate at second hydration, the hydration flow rate before first hydration, the hydration flow rate before second hydration, and the air volume before the end of tempering, were used as input values ​​to obtain the output value of the pre-grinding wheat moisture content. However, these input values ​​may also be supplemented with the values ​​of the wheat flow rate at pre-1B hydration, i.e., the wheat flow rate in 1B hydration section 26 (hereinafter referred to as the "wheat flow rate at 1B hydration"), and the hydration flow rate per unit time added to wheat before 1B, i.e., the hydration flow rate in 1B hydration section 26 (hereinafter referred to as the "pre-1B hydration flow rate"). In this case, set values ​​are input as the wheat flow rate at 1B hydration and the pre-1B hydration flow rate during the period from (1) when the pre-grinding wheat moisture content is predicted until the wheat for which moisture is predicted starts being transported to primary hydration section 14. At any point in time (2) from the start of transport of wheat for which moisture content is predicted to primary hydration section 14 until the start of transport to secondary hydration section 20, set values ​​are input as the value of the wheat flow rate at 1B hydration and the value of the hydration flow rate before 1B. At any point in time (3) from the start of transport of wheat for which moisture content is predicted to secondary hydration section 20 until the start of transport to 1B hydration section 26, set values ​​are input as the value of the wheat flow rate at 1B hydration and the value of the hydration flow rate before 1B. At any point in time (4) after the start of transport of wheat for which moisture content is predicted to 1B hydration section 26, actual measured values ​​are input as the value of the wheat flow rate at 1B hydration and the value of the hydration flow rate before 1B.

[0073] Next, a water addition flow rate prediction device and method according to a second embodiment of the present invention will be described. Regarding the water addition flow rate prediction device according to the second embodiment, the same components as those shown in Figures 1 and 2 are designated by the same reference numerals, and illustrations and descriptions of those components will be omitted. The water addition flow rate prediction device according to the second embodiment predicts the optimum value of the primary water addition flow rate added in primary hydration section 14, among the water addition flow rates added to wheat, the raw material for wheat flour, during conditioning, i.e., the optimum value of the primary water addition flow rate for adjusting the moisture content of the flour after milling to a desired moisture content (hereinafter referred to as the "primary optimum water addition flow rate").

[0074] The water addition flow rate prediction device according to the second embodiment includes a control unit 52, a memory unit 54, a first machine learning model 56, a second machine learning model 58, an input unit 60, an output unit 64, an acquisition unit 66, a first prediction unit 68, and a second prediction unit 70, similar to the water content prediction device 50 according to the first embodiment shown in Fig. 2. The control unit 52 according to the second embodiment comprehensively controls each unit of the water addition flow rate prediction device according to the second embodiment.

[0075] The first machine learning model 56 according to the second embodiment is a model for predicting the moisture content of wheat before grinding. The second machine learning model 58 according to the second embodiment is a model for predicting the primary optimal water addition flow rate.

[0076] The first machine learning model 56 according to the second embodiment is a machine learning model trained by machine learning using a large amount of training data, in which various past actual values ​​necessary for predicting the moisture content of wheat before grinding are used as input values ​​and the past actual values ​​of the pre-grinding wheat moisture content to be predicted are used as output values. The various past actual values ​​used as input values ​​include at least the temperature during flour milling, the relative humidity during flour milling, and the moisture content of flour after milling.

[0077] The second machine learning model 58 according to the second embodiment is a machine learning model trained by machine learning using a large amount of training data in which various past performance values ​​necessary to predict the hydration flow rate before primary conditioning (primary optimum hydration flow rate) are used as input values ​​and past performance values ​​of the hydration flow rate before primary conditioning are used as output values. The various past performance values ​​used as input values ​​include at least the moisture content of wheat before conditioning, the wheat flow rate during primary hydration, the wheat flow rate during secondary hydration, the hydration flow rate before secondary conditioning, the air volume before the end of conditioning, and the moisture content of wheat before grinding.

[0078] The input unit 60 according to the second embodiment inputs the value of the wheat moisture content before tempering, the value of the wheat flow rate during first hydration, the value of the wheat flow rate during second hydration, the value of the hydration flow rate before second tempering, and the value of the air volume before the end of tempering, which are used when predicting the primary optimum hydration flow rate using the second machine learning model 58 according to the second embodiment. The input unit 60 according to the second embodiment also inputs the value of the temperature during milling, the value of the relative humidity during milling, and the value of the desired post-milling wheat moisture content, which are used when predicting the wheat moisture content before grinding using the first machine learning model 56 according to the second embodiment.

[0079] The values ​​input to the input unit 60 according to the second embodiment are actual measurements taken using various measuring instruments, set values ​​set based on past performance, assumed values ​​such as prior measurements, or desired values. The milling temperature and milling relative humidity values ​​used when predicting the pre-milling wheat moisture content using the first machine learning model 56 according to the second embodiment are set using the predicted outside temperature and relative humidity based on weather forecasts for the location where the milling unit 30 is installed every predetermined time period (at least one hour) obtained at least every hour, the rate of outside air intake into the location where the milling unit 30 is installed, and other such predicted and set values ​​are input, and the desired value (desired post-milling moisture content value) is input as the value of the post-milling flour moisture content.

[0080] The values ​​input to the input unit 60 according to the second embodiment are determined depending on the arbitrary time point at which the first optimum hydration flow rate is predicted using the second machine learning model 58 according to the second embodiment. That is, during the period from the arbitrary time point (1) until the wheat for which moisture is predicted starts to be transported to the primary hydration unit 14, an assumed value is input as the value of the moisture content of the wheat before tempering, and set values ​​are input as the value of the wheat flow rate during primary hydration, the value of the wheat flow rate during secondary hydration, the value of the hydration flow rate before secondary tempering, and the air volume before the end of tempering.

[0081] At any given time (2) between the start of transport of the wheat for which moisture is predicted to the primary hydration section 14 and the start of transport to the secondary hydration section 20, actual measured values ​​are input as the value of the wheat moisture content before tempering and the value of the wheat flow rate during primary hydration, and set values ​​are input as the value of the wheat flow rate during secondary hydration, the value of the hydration flow rate before secondary tempering, and the value of the air volume before the end of tempering.

[0082] A first prediction unit 68 according to the second embodiment predicts the pre-grinding wheat moisture content by inputting the milling temperature value, milling relative humidity value, and desired post-milling wheat moisture content value entered by the input unit 60 into the first machine learning model 56 as input values ​​and obtaining a predicted pre-grinding wheat moisture content as an output value. A second prediction unit 70 according to the second embodiment predicts the pre-grinding wheat moisture content by inputting the pre-tempering wheat moisture content, the wheat flow rate during first hydration, the wheat flow rate during second hydration, the hydration flow rate before second hydration, and the air volume before the end of tempering entered by the input unit 60 as input values ​​into the second machine learning model 58 and obtaining a predicted primary optimal hydration flow rate as an output value.

[0083] Next, a method for predicting the primary optimum hydration flow rate using the hydration flow rate prediction device according to the second embodiment will be described. Fig. 4 is a flowchart illustrating the processing executed by the control unit 52 according to the second embodiment to predict the primary optimum hydration flow rate.

[0084] First, the acquisition unit 66 of the control unit 52 acquires the value of the moisture content of wheat before tempering, the value of the wheat flow rate at the first water addition, the value of the wheat flow rate at the second water addition, the value of the water addition flow rate before the second tempering, the value of the air volume before the end of tempering, the value of the temperature during milling, the value of the relative humidity during milling, and the value of the desired moisture content of wheat flour after milling, which have been input to the input unit 60 (step S12).

[0085] Next, the control unit 52 loads the first machine learning model 56 stored in the memory unit 54 (step S13), and the first prediction unit 68 of the control unit 52 predicts the moisture content of the wheat before grinding using the first machine learning model 56 loaded in step S13 (step S14). Specifically, the first prediction unit 68 inputs the milling temperature value, milling relative temperature value, and desired post-milling wheat moisture content value obtained in step S12 into the first machine learning model 56 as input values, and obtains the predicted value of the pre-grinding wheat moisture content as an output value, thereby predicting the moisture content of the wheat before grinding (step S14).

[0086] Next, the control unit 52 loads the second machine learning model 58 stored in the memory unit 54 (step S15), and the second prediction unit 70 of the control unit 52 predicts the primary optimal hydration flow rate using the second machine learning model 58 loaded in step S15 (step S16). Specifically, the second prediction unit 70 inputs the pre-tempering wheat moisture content value obtained in step S12, the wheat flow rate during primary hydration, the wheat flow rate during secondary hydration, the hydration flow rate before secondary tempering, and the air volume before the end of tempering as well as the predicted pre-grinding wheat moisture content predicted in step S14 into the second machine learning model 58, and obtains the predicted primary optimal hydration flow rate as an output value, thereby predicting the optimal value of the hydration flow rate in the first hydration unit 14 (step S16).

[0087] After completing the process of step S16, the control unit 52 outputs the results of the predictions made in steps S14 and S16 to the output unit 64 (step S17).

[0088] According to the second embodiment of the hydration flow rate prediction device and hydration flow rate prediction method, the moisture content of wheat before grinding can be accurately predicted from the minimum necessary variables in the production process using the first machine learning model 56. Furthermore, the first optimum hydration flow rate can be accurately predicted from the minimum necessary variables in the refining and conditioning process using the second machine learning model 58. Furthermore, the first optimum hydration flow rate can be accurately predicted from the minimum necessary variables from the refining and conditioning process to the production process using the first machine learning model 56 and the second machine learning model 58.

[0089] In the second embodiment described above, the first optimal hydration flow rate is predicted using the first machine learning model 56 and the second machine learning model 58, but it is also possible to predict the pre-grinding wheat moisture content using only the first machine learning model 56 and output only the prediction result to the output unit 64. It is also possible to predict the first optimal hydration flow rate using only the second machine learning model 58. In this case, the desired pre-grinding wheat moisture content value is input to the input unit 60 in addition to the above-mentioned pre-conditioning wheat moisture content value, wheat flow rate during first hydration addition value, wheat flow rate during second hydration addition value, hydration flow rate before second hydration addition value, and air volume before the end of tempering.

[0090] Furthermore, in the second embodiment described above, the first optimum hydration flow rate is predicted using two machine learning models, the first machine learning model 56 and the second machine learning model 58, but it is also possible to predict the first optimum hydration flow rate using a single machine learning model. In this case, the machine learning model includes as input values ​​at least the moisture content of wheat before tempering, the wheat flow rate during first hydration addition, the wheat flow rate during second hydration addition, the hydration flow rate before second tempering, the air volume before the end of tempering, the temperature during milling, the relative humidity during milling, and the moisture content of flour after milling, and is trained by regression using a large amount of training data in which past actual values ​​of these are used as input values ​​and past actual values ​​of the hydration flow rate before first tempering are used as output values.

[0091] Furthermore, when one machine learning model is used, the values ​​input into input section 60 are determined depending on the arbitrary time point for predicting the hydration flow rate before primary conditioning. That is, during the period from the arbitrary time point (1) up to the time before the wheat for which moisture is predicted starts to be transported to primary hydration section 14, an assumed value is input as the value of the moisture content of the wheat before conditioning, set values ​​are input as the value of the wheat flow rate during primary hydration, the value of the wheat flow rate during secondary hydration, the value of the hydration flow rate before secondary hydration, and the value of the air volume before the end of conditioning, predicted values ​​from a weather forecast or the like are input as the value of the temperature during flour milling and the value of the relative humidity during flour milling, and a desired value is input as the desired moisture content of the flour after milling.

[0092] At any given time (2) between the start of transport of the wheat for which moisture is predicted to the primary hydration section 14 and the start of transport to the secondary hydration section 20, actual measured values ​​are input as the value of the wheat moisture content before tempering and the value of the wheat flow rate during primary hydration, set values ​​are input as the value of the wheat flow rate during secondary hydration, the value of the hydration flow rate before secondary tempering, and the value of the air volume before the end of tempering, predicted values ​​from a weather forecast or the like are input as the value of the temperature during milling and the value of the relative humidity during milling, and a desired value is input as the desired moisture content of the flour after milling.

[0093] Furthermore, in the above-mentioned second embodiment, an example has been described in which the primary optimum hydration flow rate is predicted, but it is also possible to predict the optimal value of the secondary hydration flow rate to be added in secondary hydration section 20, i.e., the optimal value of the secondary hydration flow rate for adjusting the post-milling flour moisture content to a desired moisture content (hereinafter referred to as the "secondary optimum hydration flow rate"). In this case, the second machine learning model includes as input values ​​at least the moisture content of the wheat before tempering, the wheat flow rate during primary hydration, the wheat flow rate during secondary hydration, the hydration flow rate before primary tempering, and the air volume before the end of tempering, and is machine-learned by regression using a large amount of training data in which these past actual values ​​are used as input values ​​and the past actual value of the hydration flow rate before secondary tempering is output value. If the arbitrary point in time at which the secondary optimal hydration flow rate is predicted using this second machine learning model is between the start of transport of the wheat for which moisture is predicted to the secondary hydration section 20 and the start of transport to the 1B hydration section 26, the values ​​input into the input section 60 are the actual measured values ​​for the moisture content of the wheat before tempering, the wheat flow rate during primary hydration, the wheat flow rate during secondary hydration, and the hydration flow rate before primary tempering, and a set value for the air volume before the end of tempering. The second prediction unit 70 then inputs into the second machine learning model the values ​​input into the input unit 60, specifically the actual measured value of the moisture content of wheat before conditioning, the actual measured value of the wheat flow rate at the first hydration addition, the actual measured value of the wheat flow rate at the second hydration addition, the actual measured value of the hydration flow rate before the first conditioning, and the set value of the air volume before the end of conditioning, as well as the desired value of the moisture content of wheat before grinding or the predicted value of the moisture content of wheat before grinding predicted by the first prediction unit 68, and obtains the predicted value of the second optimal hydration flow rate as an output value, thereby predicting the optimal value of the hydration flow rate in the second hydration unit 20.

[0094] Instead of using two machine learning models, the first and second machine learning models, to predict the second optimal hydration flow rate, a single machine learning model may be used to predict the second optimal hydration flow rate. In this case, the machine learning model is trained by regression using a large amount of training data, including as input values ​​at least the moisture content of wheat before tempering, the wheat flow rate during first hydration addition, the wheat flow rate during second hydration addition, the hydration flow rate before first tempering, the air volume before the end of tempering, the milling temperature, the milling relative humidity, and the post-milling flour moisture content. These input values ​​are past actual values ​​for these values, and the output value is past actual values ​​for the hydration flow rate before second tempering. The values ​​input to the input unit 60 include actually measured values ​​for the moisture content of wheat before tempering, the wheat flow rate during first hydration addition, the wheat flow rate during second hydration addition, and the hydration flow rate before first tempering, a set value for the air volume before the end of tempering, predicted values ​​from a weather forecast or the like for the milling temperature and the milling relative humidity, and a desired value for the post-milling flour moisture content. The machine learning model receives as input values ​​the actual measured value of the moisture content of wheat before tempering, the actual measured value of the wheat flow rate during the first hydration, the actual measured value of the wheat flow rate during the second hydration, the actual measured value of the hydration flow rate before the first tempering, the set value of the air volume before the end of tempering, the predicted value of the temperature during milling, the predicted value of the relative humidity during milling, and the desired value of the moisture content of the flour after milling, all of which are input to the input unit 60, and outputs as an output value the predicted value of the second optimal hydration flow rate.

[0095] Furthermore, in the second embodiment described above, the minimum necessary variables for inputting the output value of the primary optimal hydration flow rate were used as input values ​​to obtain the output value of the primary optimal hydration flow rate: the moisture content of wheat before tempering, the wheat flow rate during primary hydration, the wheat flow rate during secondary hydration, the hydration flow rate before secondary tempering, the air volume before tempering completion, and the moisture content of wheat before grinding. However, the values ​​of the wheat flow rate during 1B hydration and the pre-1B hydration flow rate may be added to these input values. In this case, set values ​​are input as the values ​​of the wheat flow rate during 1B hydration and the pre-1B hydration flow rate for the arbitrary time point for predicting the primary optimal hydration flow rate (1) before the wheat for which moisture is predicted starts to be transported to primary hydration unit 14. Set values ​​are input as the values ​​of the wheat flow rate during 1B hydration and the pre-1B hydration flow rate for the arbitrary time point (2) after the wheat for which moisture is predicted starts to be transported to primary hydration unit 14 and before the wheat for which moisture is predicted starts to be transported to secondary hydration unit 20. When predicting the secondary optimal hydration flow rate, the values ​​of the wheat flow rate during 1B hydration and the hydration flow rate before 1B may be added to the variables described above as input values ​​to obtain the output value of the secondary optimal hydration flow rate. In this case, set values ​​are input as the values ​​of the wheat flow rate during 1B hydration and the hydration flow rate before 1B for any point in time when the secondary optimal hydration flow rate is predicted, between the start of transport of the wheat for which moisture is predicted to secondary hydration section 20 and the start of transport to 1B hydration section 26.

[0096] Furthermore, in each of the above-described embodiments, a set value or an actual measured value is input as the value of the air volume before the end of conditioning, but the air volume before the end of conditioning may simply be the presence or absence of air volume, that is, whether or not air is blown onto the wheat to cool it as it moves through the check bin 28. Furthermore, in each of the above-described embodiments, by calculating a moving average of the input values, it is possible to reduce the influence on the predicted value caused by the mixing of various types of wheat during transport in the screening and conditioning section 2, and the mixing of various crushed wheat flour products and as-ground powder during branching, transport, and processing in the grinding section 36.

[0097] In the above-described embodiments, the wheat moisture content before conditioning, wheat flow rate during hydration, hydration flow rate before conditioning, and air volume before conditioning are used as input values ​​to obtain the output value of the wheat moisture content before grinding or the primary optimum hydration flow rate. However, these variables may be changed to include, for example, (1) the air volume, air temperature, and air relative humidity in the air transport in each conveying section 6, 10, 18, and 24, (2) the temperature around the primary tank 16 and secondary tank 20 during conditioning, and (3) the temperature around the primary tank 16 and secondary tank 20 during conditioning. At least one of the following may be added as variables: (1) the relative humidity at the top of tank 16 and secondary tank 20; (2) the temperature of water added in the primary water addition section 14 and secondary water addition section 20; (3) the temperature of the wheat before water is added in the primary water addition section 14 and secondary water addition section 20; (4) the temperature of the wheat before water is added in the primary water addition section 14 and secondary water addition section 20; (5) the temperature of the wheat before and after cooling the wheat before the end of screening and conditioning (before and after passing through check bin 28); (6) the temperature and relative humidity before the use of the air used to cool the wheat before the end of screening and conditioning; and (7) the value of the wheat flow rate and temperature after screening and conditioning but before grinding.

[0098] In the above-described embodiments, the values ​​of the temperature after milling and the relative humidity after milling are used as examples of the minimum variables required as input values ​​to obtain the output values ​​of the moisture content of wheat flour after milling and the moisture content of wheat flour before grinding. However, these variables may include (1) the air volume, air temperature, and air relative humidity in the air transport in each conveying section 34, 38, (2) the temperature and relative humidity of each section (each floor) in the manufacturing section 30, (3) the flow rate of crushed wheat flour and the rising powder in the crushed wheat flour and rising powder flow path, (4) the temperature of crushed wheat flour and the rising powder in the crushed wheat flour and rising powder flow path, (5) the temperature in each device in the grinding section 36, (6) the temperature of each device in the grinding section 36, At least one of the following may be added as variables: (7) the destination of the crushed wheat flour and flour after they are discharged from the vessel; (8) the volume of air blown upwards toward each purification device in the crushing section 36; (9) the volume of air blown upwards during air transport and the flow rate of crushed wheat flour at each location in the crushing section 36; (10) the destination of the flour product type, for example, which process it goes through in the crushing section 36 and which product it is sorted into (first grade flour, second grade flour, third grade flour, flour powder, bran); and (11) flour added from outside to the flour, for example, the flow rate and moisture value of the flour when another flour is mixed in the process of transporting first grade flour, second grade flour, etc. [Explanation of symbols]

[0099] 2...refining and conditioning section, 4...coarse barley tank, 6...first conveying section, 8...refining section, 10...second conveying section, 12...conditioning section, 14...first hydration section, 16...first tank, 18...third conveying section, 20...second hydration section, 22...second tank, 24...fourth conveying section, 26...1B hydration section, 28...check bin, 30...production section, 32...crushing section, 34...fifth conveying section, 36...grinding section, 38...sixth conveying section, 40...product tank, 50...moisture content prediction device, 52...control section, 54...memory section, 56...first machine learning model, 58...second machine learning model, 60...input section, 64...output section, 66...acquisition section, 68...first prediction section, 70...second prediction section.

Claims

1. A moisture content prediction device for predicting the moisture content of wheat before grinding, which is the moisture content contained in wheat before grinding after the completion of screening and conditioning of wheat, which is a raw material for wheat flour, a first storage unit for storing a first machine learning model trained by machine learning using training data in which input values ​​are past actual values ​​including at least a pre-conditioning wheat moisture content, which is the moisture content of the wheat before refinement, a hydration wheat flow rate, which is the flow rate of the wheat when hydration is added before conditioning, a pre-conditioning hydration flow rate, which is the flow rate of hydration added to the wheat before conditioning, and a pre-conditioning air volume, which is the air volume when air is blown onto the wheat before conditioning is completed, and an output value is the past actual value of the pre-grinding wheat moisture content; a first acquisition unit that acquires at least the value of the wheat moisture content before conditioning, the value of the wheat flow rate during hydration, the value of the hydration flow rate before conditioning, and the value of the air volume before the end of conditioning, which are used when predicting the wheat moisture content before grinding; a first prediction unit that predicts the moisture content of wheat before grinding by inputting, as input values, at least the moisture content of wheat before tempering, the wheat flow rate during hydration, the moisture flow rate of wheat before tempering, and the air volume before the end of tempering, which are acquired by the first acquisition unit, into the first machine learning model stored in the first storage unit, and acquiring, as an output value, a predicted value of the moisture content of wheat before grinding; A moisture content prediction device comprising:

2. A moisture content prediction device for predicting the moisture content of wheat flour, a second storage unit for storing a second machine learning model trained by machine learning using training data in which input values ​​are past actual values ​​including at least the pre-milling wheat moisture content, which is the moisture content contained in the wheat before milling after the wheat, which is the raw material of the wheat flour, has been refined and refined, and the milling temperature, which is the temperature of the air that comes into contact with the ground wheat obtained by grinding the wheat and the wheat flour when producing the wheat flour from the wheat, and the milling relative humidity, which is the relative humidity, and output values ​​are past actual values ​​of the post-milling flour moisture content, which is the moisture content contained in the wheat flour after milling; A second acquisition unit that acquires at least the value of the wheat moisture content before milling, the value of the temperature during milling, and the value of the relative humidity during milling, which are used when predicting the wheat flour moisture content after milling; a second prediction unit that predicts the flour moisture content after milling by inputting at least the pre-milling wheat moisture content value, the milling temperature value, and the milling relative humidity value acquired by the second acquisition unit as input values ​​into the second machine learning model stored in the second storage unit and acquiring a predicted value of the flour moisture content after milling as an output value; A moisture content prediction device comprising:

3. A moisture content prediction device for predicting the moisture content of wheat flour, The moisture content prediction device according to claim 1 and the moisture content prediction device according to claim 2; an output unit that outputs a prediction result by the second prediction unit, The second acquisition unit is a moisture content prediction device that acquires the predicted value of the pre-grinding wheat moisture content predicted by the first prediction unit.

4. A moisture content prediction device for predicting the moisture content of wheat flour, a storage unit for storing a machine learning model trained by machine learning using training data in which input values ​​are past actual values ​​including at least the pre-conditioning wheat moisture content, which is the moisture content contained in the wheat that is the raw material for the wheat flour before refinement, the wheat flow rate during hydration, which is the flow rate of the wheat when hydration is added before hydration, the pre-conditioning hydration flow rate, which is the flow rate of hydration to be added to the wheat before hydration, the pre-conditioning air volume, which is the air volume when air is blown onto the wheat before the end of hydration, and the milling temperature, which is the temperature of the air that comes into contact with the ground wheat obtained by grinding the wheat and the wheat flour when producing the wheat flour from the wheat, and the milling relative humidity, which is the relative humidity, and the output value is the past actual value of the post-milling flour moisture content, which is the moisture content contained in the wheat flour after milling; an acquisition unit that acquires at least the value of the wheat moisture content before tempering, the value of the wheat flow rate during hydration, the value of the hydration flow rate before tempering, the value of the air volume before the end of tempering, the value of the temperature during milling, and the value of the relative humidity during milling, which are used when predicting the wheat flour moisture content after milling; a prediction unit that predicts the moisture content of the flour after milling by inputting, as input values, at least the moisture content of the wheat before tempering, the wheat flow rate during hydration, the moisture flow rate before tempering, the air volume before the end of tempering, the temperature during milling, and the relative humidity during milling, which are acquired by the acquisition unit, into the machine learning model stored in the storage unit, and acquiring, as an output value, a predicted value of the moisture content of the flour after milling; an output unit that outputs a prediction result by the prediction unit; A moisture content prediction device comprising:

5. When water is added to the wheat multiple times during tempering, 5. The moisture content predicting device according to claim 1, wherein the value of the wheat flow rate at the time of hydration and the value of the hydration flow rate before tempering are values ​​each time hydration is added to the wheat.

6. 5. The moisture content predicting device according to claim 1, 3 or 4, wherein the air volume before the end of conditioning is determined by whether or not air is blown onto the wheat before the end of conditioning.

7. A moisture content prediction device for predicting the moisture content of wheat before grinding, which is the moisture content contained in wheat before grinding after the completion of screening and conditioning of wheat, which is a raw material for wheat flour, a first storage unit for storing a first machine learning model that has been machine-learned using training data in which input values ​​are past actual values ​​including, as input values, the milling temperature, which is the temperature of the air that comes into contact with the ground wheat obtained by milling the wheat when producing wheat flour from the wheat, and the milling relative humidity, which is the relative humidity, and the post-milling flour moisture content, which is the amount of moisture contained in the wheat flour after milling, and the pre-milling wheat moisture content, which is the output value; and A first acquisition unit that acquires at least the value of the milling temperature, the value of the milling relative humidity, and the value of the desired post-milling flour moisture content, which are used when predicting the pre-milling wheat moisture content to set the post-milling flour moisture content to the desired moisture content; a first prediction unit that predicts the moisture content of wheat before grinding by inputting, as input values, at least the value of the temperature during flour milling, the value of the relative humidity during flour milling, and the desired value of the moisture content of wheat flour after milling, which are acquired by the first acquisition unit, into the first machine learning model stored in the first storage unit, and acquiring, as an output value, a predicted value of the moisture content of wheat before grinding; A moisture content prediction device comprising:

8. A water addition flow rate prediction device for predicting an optimum water addition flow rate to be added to wheat, which is a raw material for wheat flour, when conditioning the wheat, comprising: a second storage unit for storing a second machine learning model trained by machine learning using training data in which input values ​​are past actual values ​​including at least a pre-conditioning wheat moisture content, which is the moisture content contained in the wheat that is the raw material for the wheat flour before refinement, a hydration wheat flow rate, which is the flow rate of the wheat when hydration is added before refinement, a pre-conditioning air flow rate, which is the air volume when air is blown onto the wheat before the end of refinement, and a pre-grinding wheat moisture content, which is the moisture content contained in the wheat before grinding after the end of refinement of the wheat, and an output value is a past actual value of a pre-conditioning hydration flow rate, which is the hydration flow rate at which water is added to the wheat before refinement; a second acquisition unit that acquires at least the value of the wheat moisture content before tempering, the value of the wheat flow rate during hydration, the value of the air volume before tempering is completed, and the desired value of the wheat moisture content before grinding, which are used when predicting the optimal water addition flow rate for adjusting the wheat moisture content after milling to the desired moisture content; a second prediction unit that predicts the optimal hydration flow rate by inputting, as input values, at least the pre-conditioning wheat moisture content value, the wheat flow rate during hydration, the air volume before the end of conditioning, and the desired pre-grinding wheat moisture content value acquired by the second acquisition unit into the second machine learning model stored in the second storage unit, and acquiring, as an output value, a predicted value of the optimal hydration flow rate; A water addition flow rate prediction device comprising:

9. A water addition flow rate prediction device for predicting an optimum water addition flow rate to be added to wheat, which is a raw material for wheat flour, when conditioning the wheat, comprising: The water content prediction device according to claim 7 and the water addition flow rate prediction device according to claim 8; an output unit that outputs a prediction result by the second prediction unit, The second acquisition unit acquires the predicted value of the pre-grinding wheat moisture content predicted by the first prediction unit as the desired pre-grinding wheat moisture content value.

10. A water addition flow rate prediction device for predicting an optimum water addition flow rate to be added to wheat, which is a raw material for wheat flour, when conditioning the wheat, comprising: a storage unit for storing a machine learning model trained by machine learning using training data in which input values ​​are past actual values ​​including at least the pre-conditioning wheat moisture content, which is the moisture content contained in the wheat that is the raw material for the wheat flour before refinement, the wheat flow rate during hydration, which is the flow rate of the wheat when hydration is added before hydration, the air volume before end of hydration, which is the air volume when air is blown on the wheat before end of hydration, the milling temperature, which is the temperature and milling relative humidity of the air that comes into contact with the ground wheat obtained by grinding the wheat and the wheat when producing wheat flour from the wheat, and the post-milling flour moisture content, which is the moisture content contained in the wheat after milling, and the output value is the past actual value of the pre-conditioning hydration flow rate, which is the flow rate of hydration added to the wheat before hydration; an acquisition unit that acquires at least the value of the wheat moisture content before tempering, the value of the wheat flow rate during hydration, the value of the air volume before the end of tempering, the value of the temperature during milling, the value of the relative humidity during milling, and the value of the desired moisture content of the flour after milling, which are used when predicting the optimal water addition flow rate for adjusting the moisture content of the flour after milling to the desired moisture content; a prediction unit that predicts the optimum water addition flow rate by inputting, as input values, at least the value of the moisture content of the wheat before conditioning, the value of the wheat flow rate during hydration, the value of the air volume before the end of conditioning, the value of the temperature during milling, the value of the relative humidity during milling, and the desired value of the moisture content of the wheat after milling, which are acquired by the acquisition unit, into the machine learning model stored in the storage unit, and acquiring, as an output value, a predicted value of the optimum water addition flow rate; an output unit that outputs a prediction result by the prediction unit; A water addition flow rate prediction device comprising:

11. When water is added to the wheat multiple times during tempering, The value of the wheat flow rate when adding water is a value for each time water is added to the wheat, The water addition flow rate prediction device according to any one of claims 8 to 10, wherein the input values ​​include a value of a pre-conditioning water addition flow rate, which is a flow rate other than the optimum water addition flow rate to be added to the wheat each time water is added to the wheat.

12. The water addition flow rate prediction device according to any one of claims 8 to 10, wherein the air volume before the end of conditioning is determined by whether or not there is air blowing on the wheat before the end of conditioning.

13. A moisture content prediction method for predicting a pre-grinding wheat moisture content, which is the moisture content contained in wheat before grinding after the completion of refinement of wheat, which is a raw material for wheat flour, using a moisture content prediction device, comprising: a first acquisition step in which an acquisition unit of the moisture content prediction device acquires, in order to predict the pre-grinding wheat moisture content, at least a pre-conditioning wheat moisture content value, which is the moisture content contained in the wheat before refinement, a hydration wheat flow rate value, which is the flow rate of the wheat when hydration is added before hydration, a pre-conditioning hydration flow rate value, which is the flow rate of hydration added to the wheat before hydration, and a pre-conditioning air volume value, which is the volume of air blown onto the wheat before hydration is completed; a first reading step in which a reading unit of the moisture content prediction device reads from a storage unit of the moisture content prediction device a first machine learning model that has been machine-learned using training data in which past actual values ​​including at least the pre-conditioning wheat moisture content, the wheat flow rate during hydration, the pre-conditioning hydration flow rate, and the air volume before the end of the conditioning are used as input values, and the past actual value of the pre-grinding wheat moisture content is used as an output value; a first prediction step in a prediction unit of the moisture content prediction device, in which at least the pre-conditioning wheat moisture content value, the value of the wheat flow rate during hydration, the value of the pre-conditioning hydration flow rate, and the value of the air volume before the end of hydration, which are acquired in the acquisition step, are input as input values ​​to the first machine learning model read in the first reading step, and the predicted value of the pre-grinding wheat moisture content is acquired as an output value, thereby predicting the pre-grinding wheat moisture content; A moisture content prediction method including:

14. A moisture content prediction method for predicting the moisture content of wheat flour using a moisture content prediction device, comprising: a second acquisition step in which the acquisition unit of the moisture content prediction device acquires at least a pre-milling wheat moisture value, which is the moisture content contained in wheat before milling after the completion of refinement of wheat, which is the raw material of the wheat flour, and a milling temperature value, which is the temperature of the air that comes into contact with the ground wheat obtained by milling the wheat and the flour when producing the wheat flour, and a milling relative humidity value, which is the relative humidity, used when predicting the post-milling wheat moisture content, which is the moisture content contained in the wheat flour after milling; a second reading step in which a reading unit of the moisture content prediction device reads from a storage unit of the moisture content prediction device a second machine learning model that has been machine-learned using training data in which past actual values ​​including at least the pre-milling wheat moisture content, the milling temperature, and the milling relative humidity are used as input values, and the past actual values ​​of the post-milling wheat flour moisture content are used as output values; a second prediction step in which, in the prediction unit of the moisture content prediction device, at least the pre-milling wheat moisture content value, the milling temperature value, and the milling relative humidity value obtained in the second acquisition step are input as input values ​​to the second machine learning model read in the second reading step, and the predicted value of the post-milling flour moisture content is obtained as an output value, thereby predicting the post-milling flour moisture content; A moisture content prediction method including:

15. A moisture content prediction method for predicting the moisture content contained in wheat flour using a moisture content prediction device, comprising: A moisture content prediction method according to claim 13 and a moisture content prediction method according to claim 14; an output step of outputting the result predicted in the second prediction step, The second obtaining step is a moisture content prediction method for obtaining the value of the moisture content of the wheat before grinding predicted in the first prediction step.

16. A moisture content prediction method for predicting the moisture content of wheat flour using a moisture content prediction device, comprising: an acquisition step in which the acquisition unit of the moisture content prediction device acquires, in order to predict the post-milling wheat moisture content, which is the amount of moisture contained in the wheat that is the raw material for the wheat flour before refinement and refinement, a wheat flow rate during hydration, which is the flow rate of the wheat when hydration is added before hydration, a hydration flow rate before hydration, which is the flow rate of hydration added to the wheat before hydration, a pre-conditioning air volume, which is the volume of air blown onto the wheat before hydration is completed, and a milling temperature, which is the temperature of the air that comes into contact with the ground wheat and the wheat flour when the wheat is ground, and a milling relative humidity, which is the relative humidity, used in predicting the post-milling wheat moisture content, which is the amount of moisture contained in the wheat flour after hydration and refinement; a reading step in which a reading unit of the moisture content prediction device reads from a storage unit of the moisture content prediction device a machine learning model trained by machine learning using training data in which past actual values ​​including at least the pre-tempering wheat moisture content, the wheat flow rate during hydration, the pre-tempering hydration flow rate, the air volume before the end of tempering, the milling temperature, and the milling relative humidity are used as input values, and the past actual values ​​of the post-milling wheat moisture content are used as output values; a prediction step in which, in a prediction unit of the moisture content prediction device, at least the value of the wheat moisture content before tempering, the value of the wheat flow rate during hydration, the value of the hydration flow rate before tempering, the value of the air volume before the end of tempering, the value of the temperature during milling, and the value of the relative humidity during milling, which are acquired in the acquisition step, are input as input values ​​to the machine learning model read in the reading step, and the predicted value of the flour moisture content after milling is acquired as an output value, thereby predicting the flour moisture content after milling; an output step of outputting the result predicted in the prediction step from an output unit of the moisture content prediction device; A moisture content prediction method including:

17. When water is added to the wheat multiple times during tempering, 17. The moisture content prediction method according to claim 13, 15 or 16, wherein the value of the wheat flow rate during hydration and the value of the hydration flow rate before conditioning are values ​​each time hydration is added to the wheat.

18. 17. The moisture content prediction method according to claim 13, claim 15 or claim 16, wherein the air volume before the end of conditioning is determined by whether or not air is blown onto the wheat before the end of conditioning.

19. A moisture content prediction method for predicting a pre-grinding wheat moisture content, which is the moisture content contained in wheat before grinding after the completion of refinement of wheat, which is a raw material for wheat flour, using a moisture content prediction device, comprising: a first acquisition step in the moisture content prediction device for acquiring, in a first acquisition unit thereof, at least a milling temperature value, which is the temperature of the air contacting the ground wheat obtained by milling the wheat and the wheat flour when producing the wheat flour from the wheat, a milling relative humidity value, which is the relative humidity, and a post-milling flour moisture content value, which is the amount of moisture contained in the wheat flour after milling; a first reading step in which a first reading unit of the moisture content prediction device reads from a first memory unit of the moisture content prediction device a first machine learning model that has been machine-learned using training data in which past actual values ​​including at least the milling temperature, the milling relative humidity, and the post-milling flour moisture content are used as input values, and the past actual value of the pre-grinding wheat moisture content is used as an output value; a first prediction step in which at least the milling temperature value, the milling relative humidity value, and the desired post-milling wheat moisture content value acquired in the first acquisition step are input as input values ​​to the first machine learning model read in the first reading step in a first prediction unit of the moisture content prediction device, and the predicted value of the pre-milling wheat moisture content is acquired as an output value, thereby predicting the pre-milling wheat moisture content; A moisture content prediction method including:

20. A water addition flow rate prediction method for predicting an optimum water addition flow rate to be added to wheat, which is a raw material for wheat flour, during conditioning using a water addition flow rate prediction device, comprising: a second acquisition step in the water addition flow rate prediction device for acquiring, in a second acquisition unit thereof, a value of at least a pre-conditioning wheat moisture content, which is the moisture content contained in the wheat that is the raw material for the wheat flour before refinement, a value of a hydration wheat flow rate, which is the flow rate of the wheat when hydration is added before refinement, a value of a pre-conditioning air flow rate, which is the air volume when air is blown onto the wheat before the end of refinement, and a value of a pre-grinding wheat moisture content, which is the moisture content contained in the wheat that is the raw material for the wheat flour after refinement and before grinding, which is the desired value, to be used when predicting the optimal water addition flow rate for adjusting the post-milling wheat moisture content, which is the moisture content contained in the wheat flour after milling, which is the moisture content contained in the wheat flour after milling, to a desired moisture content; a second reading step in which a second reading unit of the hydration flow rate prediction device reads from a second memory unit of the hydration flow rate prediction device a second machine learning model that has been machine-learned using training data in which past actual values ​​including at least the moisture content of the wheat before tempering, the wheat flow rate during hydration, the air volume before the end of tempering, and the moisture content of the wheat before grinding are used as input values, and the past actual value of the hydration flow rate before tempering is used as an output value; a second prediction step in a second prediction section of the hydration flow rate prediction device, in which at least the value of the pre-conditioning wheat moisture content, the value of the wheat flow rate during hydration, the value of the air volume before the end of hydration, and the desired value of the pre-grinding wheat moisture content obtained in the second acquisition step are input as input values ​​to the second machine learning model read in the second reading step, and a predicted value of the optimal hydration flow rate is obtained as an output value, thereby predicting the optimal hydration flow rate; A method for predicting water addition flow rate, including:

21. A water addition flow rate prediction method for predicting an optimum water addition flow rate to be added to wheat, which is a raw material for wheat flour, during conditioning using a water addition flow rate prediction device, comprising: A water content prediction method according to claim 19 and a water addition flow rate prediction method according to claim 20; an output step of outputting the result predicted in the second prediction step, The moisture content prediction method includes a second obtaining step of obtaining the value of the pre-grinding wheat moisture content predicted in the first predicting step as the desired value of the pre-grinding wheat moisture content.

22. A water addition flow rate prediction method for predicting an optimum water addition flow rate to be added to wheat, which is a raw material for wheat flour, during conditioning using a water addition flow rate prediction device, comprising: an acquisition step in which an acquisition unit of the water addition flow rate prediction device acquires at least a pre-conditioning wheat moisture content value, which is the moisture content contained in wheat, the raw material of the wheat flour before refinement, a hydration wheat flow rate value, which is the flow rate of the wheat when hydration is added before hydration, a pre-conditioning air volume value, which is the air volume when air is blown onto the wheat before hydration is completed, a milling temperature value, which is the temperature of the air that comes into contact with the ground wheat obtained by grinding the wheat when producing wheat flour from the wheat, and a milling relative humidity value, which is the relative humidity, and a desired post-milling flour moisture content value, all of which are used when predicting the optimal water addition flow rate for adjusting the post-milling flour moisture content, which is the moisture content contained in the wheat flour after milling, to a desired moisture content; a reading step in which a reading unit of the hydration flow rate prediction device reads from a storage unit of the hydration flow rate prediction device a machine learning model trained by machine learning using training data in which past actual values ​​including at least the moisture content of the wheat before tempering, the wheat flow rate during hydration, the air volume before the end of tempering, the temperature during milling, the relative humidity during milling, and the moisture content of the flour after milling are used as input values, and the past actual value of the hydration flow rate before tempering is used as output value; a prediction step in which at least the pre-conditioning wheat moisture content value, the hydration wheat flow rate value, the pre-conditioning air volume value, the milling temperature value, the milling relative humidity value, and the desired post-milling wheat moisture content value, which are acquired in the acquisition step, are input to the machine learning model read in the reading step, and the prediction unit of the hydration flow rate prediction device predicts the optimum hydration flow rate by acquiring a predicted optimum hydration flow rate as an output value; an output step of outputting the result predicted in the prediction step; A method for predicting water addition flow rate, including:

23. When water is added to the wheat multiple times during tempering, The value of the wheat flow rate when adding water is a value for each time water is added to the wheat, The method for predicting a water addition flow rate according to any one of claims 20 to 22, wherein a value of a pre-conditioning water addition flow rate, which is a flow rate other than the optimum water addition flow rate to be added to the wheat each time water is added to the wheat, is included in the input values.

24. The method for predicting the amount of water added according to any one of claims 20 to 22, wherein the air volume before the end of conditioning is determined by whether or not there is air blowing on the wheat before the end of conditioning.

Images