Information processing device, control content determination method, waste adjustment method, control content prediction model generation method, reinforcement learning model generation method, control content determination program, and waste adjustment program

Abstract

To stably discharge accommodated objects accommodated in a hopper.SOLUTION: An information processing device (1) includes: a state determination section (101) that determines states of accommodated objects in a hopper; and a control content determination section (103) that determines a control content when accommodated objects are input, on the basis of history information (111) indicating, regarding past input of accommodated objects to the hopper, states of the accommodated objects in the hopper immediately before the input, control contents during the input and a result of the input and a determination result obtained by the state determination section (101).SELECTED DRAWING: Figure 1

Description

[Technical field]

[0001] The present invention relates to an information processing device that determines the contents of control when putting contents into a hopper. [Background technology]

[0002] A typical waste incineration facility is equipped with a funnel-shaped device called a hopper. Waste stored in a storage facility called a waste pit is picked up by a crane or other device and dumped into the hopper. A waste discharge device called a dust feeder is installed downstream of the hopper, and the waste in the hopper is sent into the incinerator by the dust feeder in order of the waste that reaches the bottom of the hopper, where it is incinerated.

[0003] When the amount of waste in an incinerator decreases or runs out, the temperature inside the incinerator drops and it becomes necessary to heat it with a burner, etc., or in some cases, it may become necessary to shut down the incinerator. For this reason, it is important to steadily supply waste to the incinerator using a hopper and dust feeder.

[0004] An example of a technique related to waste supply control for realizing stable combustion in an incinerator is the following Patent Document 1. Patent Document 1 discloses a waste supply calorific value measuring device that calculates the weight of waste moved from the specific gravity of the waste put into a hopper and the volume of the waste moved, and calculates the amount of heat supplied to the incinerator by the weight moved and the calorific value of the waste per unit time. [Prior art documents] [Patent documents]

[0005] [Patent Document 1] Patent No. 3926173 Summary of the Invention [Problem to be solved by the invention]

[0006] The waste supply calorific value measuring device in Patent Document 1 calculates the waste supply calorific value on the assumption that the waste supply from the hopper and dust feeder is stable. Therefore, if the waste supply from these devices becomes unstable, the calculation results will be invalid, and the combustion state of the incinerator may become unstable.

[0007] For example, when the inside of the hopper becomes clogged with garbage (also called a bridge), the garbage above the blockage will not move, and after all the garbage below the blockage has been discharged by the dust feeder, no more garbage will be sent to the incinerator. In this case, the combustion state of the incinerator will become unstable.

[0008] As can be seen from the above-mentioned example of an incineration facility, in order to stably discharge the contents stored in the hopper, it is important to prevent problems such as blockages from occurring in the hopper. However, the technology disclosed in Patent Document 1 cannot prevent the occurrence of such problems, and therefore, it may be difficult to stably discharge the contents stored in the hopper.

[0009] An object of one aspect of the present invention is to provide an information processing device and the like that enables stable discharge of items contained in a hopper. [Means for solving the problem]

[0010] In order to solve the above problems, an information processing device according to one embodiment of the present invention includes a state determination unit that determines the state of the contents contained inside a hopper, and historical information indicating, for past loading of the contents into the hopper, the state of the contents inside the hopper immediately before the loading, the control content at the time of loading, and the result of the loading, and a control content determination unit that determines the control content at the time of loading the contents based on the determination result of the state determination unit.

[0011] In order to solve the above problems, an information processing device according to another aspect of the present invention comprises a state determination unit that determines the state of garbage contained inside a hopper, historical information indicating the state of garbage inside the hopper immediately before garbage was dumped into the hopper in the past, the properties of the garbage dumped, and the results of the dumping, and an input garbage determination unit that determines the properties of the garbage to be dumped into the hopper based on the determination results of the state determination unit, and an input garbage adjustment unit that has a crane that transports garbage within a garbage pit that stores the garbage to be dumped into the hopper adjust the garbage to generate garbage with the properties determined by the input garbage determination unit in the garbage pit.

[0012] In order to solve the above problems, a control content determination method according to one embodiment of the present invention is a control content determination method executed by one or more information processing devices, and includes a state determination step of determining a state of contents contained inside a hopper, and a control content determination step of determining the control content at the time of putting the contents into the hopper based on historical information indicating the state of the contents inside the hopper immediately before the contents were put into the hopper in the past, the control content at the time of putting, and the result of the putting, and the determination result of the state determination step.

[0013] In order to solve the above problems, a garbage adjustment method according to one embodiment of the present invention is a garbage adjustment method executed by one or more information processing devices, and includes: a state determination step of determining the state of garbage contained inside a hopper; an input garbage determination step of determining the properties of the garbage to be input into the hopper based on historical information indicating the state of the garbage inside the hopper immediately before input of garbage into the hopper, the properties of the input garbage, and the results of input, and the results of the determination in the state determination step; and an input garbage adjustment step of having a crane that transports garbage within a garbage pit that stores the garbage to be input into the hopper adjust the garbage, thereby generating garbage with the properties determined in the input garbage determination step within the garbage pit.

[0014] In order to solve the above problems, a method for generating a control content prediction model according to one embodiment of the present invention is a method for generating a control content prediction model executed by one or more information processing devices, and includes a teacher data acquisition step of acquiring teacher data that corresponds, for past loading of contents into a hopper in which the state of the contents inside the hopper immediately before the loading was performed and the control content performed at the time of the loading, and a learning step of generating a control content prediction model that predicts, by learning using the teacher data, the control content to be performed when the contents are loaded into the hopper in order to maintain or improve the state of the hopper, from the state of the contents inside the hopper immediately before the loading of the contents.

[0015] In order to solve the above problems, a method for generating a reinforcement learning model according to one embodiment of the present invention is a method for generating a reinforcement learning model executed by one or more information processing devices, and includes a history acquisition step of acquiring history information indicating the state of the contents inside the hopper immediately before the contents were put into the hopper in the past, the control content at the time of the putting, and the result of the putting, and a learning step of generating a reinforcement learning model by reinforcement learning in which the state of the contents inside the hopper immediately before the contents were put in, as indicated in the history information, is the environment, the control at the time of putting the contents in is the action, and the evaluation value of the result of the putting is the reward. Effect of the Invention

[0016] According to one aspect of the present invention, it is possible to stably discharge items contained in a hopper. [Brief description of the drawings]

[0017] [Figure 1] 1 is a block diagram showing an example of a configuration of a main part of an information processing device according to a first embodiment of the present invention. [Diagram 2] FIG. 2 is a diagram showing an overview of a management system including the information processing device. [Diagram 3] FIG. 13 is a diagram illustrating classification information. [Figure 4]4 is a flowchart illustrating an example of a process executed by the information processing device. [Diagram 5] FIG. 11 is a block diagram showing an example of a configuration of a main part of an information processing device according to a second embodiment of the present invention. [Figure 6] 10 is a flowchart illustrating an example of a method for generating a control content prediction model executed by the information processing device. [Figure 7] FIG. 13 is a diagram illustrating an example of updating a predictive distribution. [Figure 8] FIG. 11 is a block diagram showing an example of a configuration of a main part of an information processing device according to a third embodiment of the present invention. [Figure 9] 10 is a flowchart illustrating an example of a method for generating a control content prediction model executed by the information processing device. [Figure 10] 10 is a flowchart illustrating an example of a method for generating a reinforcement learning model executed by the information processing device. [Figure 11] FIG. 11 is a block diagram showing an example of a configuration of a main part of an information processing device according to a fourth embodiment of the present invention. [Figure 12] 4 is a flowchart illustrating an example of a process executed by the information processing device. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0018] [Embodiment 1] (System Overview) An overview of the management system 5 according to this embodiment will be described with reference to Fig. 2. Fig. 2 is a diagram showing an overview of the management system 5. The management system 5 is a system for managing a waste incineration facility a, and includes an information processing device 1 and a crane control device 2. The waste incineration facility a shown in Fig. 2 includes a waste pit a1, a crane a2, a hopper a3, a dust feeder a4, an incinerator a5, and a fire grate a6.

[0019] The garbage pit a1 is a storage facility for storing garbage brought into the garbage incineration facility a. The crane a2 is a device for transporting garbage in the garbage pit a1, and includes a bucket a21 for grabbing garbage, a wire a22 for raising and lowering the bucket a21, a girder a23 stretched across the opposing walls of the building of the garbage incineration facility a, and a traverse cart a24 that moves on the girder a23.

[0020] The wire a22 is attached to a traverse carriage a24. The girder a23 moves in the depth direction of FIG. 2 on rails (not shown) installed along the opposing walls of the building of the waste incineration facility a, and the traverse carriage a24 moves in the left-right direction of the figure. Thus, the bucket a21 can be moved to any position in the waste pit a1 by the girder a23 and the traverse carriage a24. The crane control device 2 is a device that controls the operation of the crane a2, and controls the movement, lifting, opening and closing, etc. of the bucket a21 as described above.

[0021] The area in the garbage pit a1 can be divided into a number of sections for management, for example as shown in Fig. 2b. Fig. 2b shows a schematic diagram of the garbage pit a1 as viewed from above, with 30 sections in 5 rows and 6 columns indicated by dashed lines. In this example, each section is identified by a combination of row numbers 1 to 5 and column numbers A to F. The garbage may also be divided in a similar manner in the depth direction, and the state of the garbage in the depth direction may also be managed.

[0022] In the management system 5, the height of the garbage in each section set in this way and the characteristics (nature or state) of the garbage may be recorded as garbage management information, and may be updated as the characteristics of the garbage change. For example, a garbage quality index may be recorded as information indicating the characteristics of the garbage. The garbage quality index is a numerical value indicating the characteristics of the garbage, and can be calculated by evaluating the quality of the garbage with a predetermined evaluation standard. For example, the moisture content of the garbage or the number of times it is stirred may be used as the garbage quality index. In addition, for example, the fractal dimension calculated by a fractal analysis means from an image of the garbage surface, or a numerical value indicating the color characteristics of the garbage obtained by a hue feature analysis of the image, may be used as the garbage quality index. The fractal dimension is an index indicating the size of the garbage. The size of the garbage is correlated with the degree of stirring and homogenization of the garbage, and is also correlated with the occurrence of bridges in the hopper. The garbage management information may be recorded and updated manually or automatically by a device such as the information processing device 1.

[0023] The hopper a3 is a device for temporarily storing the waste to be sent to the incinerator a5. The upper part of the hopper a3 has a funnel-shaped structure with an open top, and the waste is fed from this opening by the crane a2. A dust feeder a4 is provided at the bottom of the hopper a3, and the dust feeder a4 pushes the waste in the hopper a3 into the incinerator a5 and discharges it. The incinerator a5 is a furnace for incinerating the waste, and the fire grate a6 is a device for moving the waste in the incinerator a5. In other words, the waste discharged into the incinerator a5 by the dust feeder a4 is incinerated while being moved by the fire grate a6 in the incinerator a5. The waste incineration facility a may be equipped with a boiler and a generator, and may generate electricity using the heat generated in the incinerator a5.

[0024] The management system 5 is not limited to the waste incineration facility a. The management system 5 can manage any facility including any hopper that stores any contents. For example, even if the management system 5 is used at the waste incineration facility a, the management system can also manage an incineration ash hopper instead of or in addition to the waste hopper a3. Of course, any facility including any hopper other than the waste incineration facility a can be managed. Therefore, in the following description, "hopper" can be read as any hopper, "waste" can be read as any contents, and "dust feeding device" can be read as any discharge device.

[0025] The information processing device 1 judges the state of the waste contained inside the hopper a3. The information processing device 1 then determines the control content when the waste is thrown in based on history information about past throwing of waste into the hopper a3 and the above-mentioned judgment result about the state of the waste. The history information is information that indicates the state of the waste inside the hopper a3 immediately before the waste is thrown in, the control content when the waste is thrown in, and the result of the throwing.

[0026] According to the above configuration, appropriate control content can be determined according to the state of the waste inside the hopper immediately before it is added, thereby making it possible to stably discharge the waste contained in the hopper into the incinerator a5.

[0027] (Configuration of information processing device) The configuration of the information processing device 1 will be described with reference to Fig. 1. Fig. 1 is a block diagram showing an example of a main configuration of the information processing device 1. As shown in the figure, the information processing device 1 includes a control unit 10 that controls each unit of the information processing device 1, and a storage unit 11 that stores various data used by the information processing device 1. The information processing device 1 also includes a communication unit 12 that allows the information processing device 1 to communicate with other devices, an input unit 13 that accepts input of various data to the information processing device 1, and an output unit 14 that allows the information processing device 1 to output various data.

[0028] The control unit 10 includes a state determination unit 101, an input information acquisition unit 102, a control content determination unit 103, a device control unit 104, and a classification unit 105. The storage unit 11 stores history information 111 and classification information 112.

[0029] The history information 111 is information indicating the state of the garbage inside the hopper immediately before the garbage was thrown into the hopper, the control content at the time of the throwing, and the result of the throwing, with respect to the throwing of garbage into the hopper that was performed in the past. For example, the history information 111 may be information in which the height of the garbage inside the hopper immediately before the garbage was thrown in (the hopper level described later), the weight and apparent specific gravity of the thrown in garbage, and information indicating whether a bridge occurred within a predetermined period after the throwing are associated with each other. In this case, the height of the garbage indicates the state of the garbage, the weight and apparent specific gravity of the thrown in garbage indicate the control content of throwing in garbage of such weight and apparent specific gravity, and information indicating whether a bridge occurred indicates the result of the throwing. The apparent specific gravity of the garbage is a value obtained by dividing the weight of the garbage by its volume. A bridge is a trouble in which garbage clogs in the hopper in a bridge-like manner, and is also called a blockage.

[0030] The classification information 112 is information used to determine the control content to be performed on the equipment when the waste in the hopper is in the state indicated by the determination result of the state determination unit 101. The classification information 112 is generated by the classification unit 105, which will be described in detail later.

[0031] The state determination unit 101 determines the state of the garbage stored inside the hopper. The state of the garbage determined by the state determination unit 101 may be related to the discharge state of the garbage. For example, the state determination unit 101 may determine the hopper level as the state of the garbage. The hopper level is an index indicating the height of the garbage stored in the hopper. The hopper level can be determined using a sensor or the like, a laser scanner such as LiDIA, or by analyzing an image of the hopper taken from above. In addition to this, the hopper level can also be determined using a garbage flow model that models the flow of the garbage from the input of the garbage to the discharge of the garbage.

[0032] Furthermore, the height of the garbage in the hopper may vary depending on the position in the hopper. For this reason, the state determination unit 101 may divide the opening of the hopper as viewed from above into a plurality of areas (e.g., 9 areas of 3×3), obtain the height of each area, and calculate the hopper level based on these heights. In this case, for example, the state determination unit 101 may determine the hopper level as the arithmetic average or weighted average of the garbage heights in each area.

[0033] The input item information acquisition unit 102 acquires input item information, which is information about garbage that has been input or will be input into the hopper. The input item information may be information about garbage that has been input or will be input into the hopper. For example, the input item information may be the apparent specific gravity of the garbage. Since the volume of garbage that can be grasped by one handful of the crane bucket is roughly constant, and the weight of the garbage grasped by the crane is measurable, the input item information acquisition unit 102 can calculate the apparent specific gravity by dividing the weight by the volume.

[0034] The control content determination unit 103 determines the control content when garbage is inserted based on the classification information 112 and the determination result of the status determination unit 101. As will be described later, the classification information 112 is generated from the history information 111, so it can be said that the control content determination unit 103 determines the control content based on the history information 111 and the determination result of the status determination unit 101. The method of determining the control content will be described later.

[0035] The equipment control unit 104 controls the equipment according to the control contents determined by the control content determination unit 103. In this embodiment, an example will be described in which the equipment control unit 104 controls a crane (e.g., crane a2 in FIG. 2) via a crane control device 2 (see FIG. 2). Of course, the equipment to be controlled is not limited to a crane, and may be any equipment that can be expected to maintain or improve the state of waste in a hopper, the state of waste discharged from the hopper, and the state of waste transported in the incinerator by controlling the equipment. For example, the equipment control unit 104 may control a dust feeder (e.g., dust feeder a4 in FIG. 2) or a fire grate (e.g., fire grate a6 in FIG. 2).

[0036] The classification unit 105 generates the classification information 112 from the history information 111. The method of generating the classification information 112 will be described later.

[0037] As described above, the information processing device 1 includes a state determination unit 101 that determines the state of the waste contained in the hopper, and a control content determination unit 103 that determines the control content when the waste is thrown in based on the history information 111 and the determination result of the state determination unit 101. The history information 111 is information that indicates the state of the waste inside the hopper immediately before the waste was thrown in, the control content at the time of throwing in, and the result of the throwing in, for a past throwing of waste into the hopper. With this configuration, it is possible to determine appropriate control content according to the state of the waste inside the hopper immediately before the throwing, and this makes it possible to stably discharge the waste contained in the hopper into the incinerator.

[0038] (Classification information) The classification information 112 and the method for generating it will be described with reference to Fig. 3. Fig. 3 is a diagram for explaining the classification information 112, and shows the classification results of the combination of the state of the waste inside the hopper immediately before the waste was poured in and the control details at the time of the waste being poured in, depending on whether the result of the pouring in shown in the history information 111 was good or bad.

[0039] Specifically, Fig. 3 shows the hopper level, input amount, and apparent specific gravity in a coordinate system consisting of three mutually perpendicular axes in past examples of waste input into the hopper shown in the history information 111. The hopper level indicates the state of the waste immediately before input, and the input amount and apparent specific gravity indicate the control details at the time of input. The input amount is the weight of the waste input into the hopper.

[0040] Figure 3C plots the hopper level and apparent specific gravity of the waste dumped in when a bridge formed in the hopper due to the dumping of 5, 6, and 7 tons. Figure 3c1 to c3 are plots extracted from Figure 3C for the dumping amounts of 5 tons to 7 tons.

[0041] As shown in c1, on plane c11, which corresponds to a loading amount of 7 tons, the plots of cases where bridging occurred appear in a concentrated area. From this, it can be said that the combination of hopper level and apparent specific gravity is correlated with the occurrence of bridging. In other words, there are combinations of hopper level and apparent specific gravity that have a high probability of causing bridging, and combinations of hopper level and apparent specific gravity that have a low probability of causing bridging.

[0042] Therefore, if combinations of hopper level and apparent specific gravity can be classified according to whether or not the probability of bridging occurring is low, it will be possible to perform control to reduce the probability of bridging occurring.

[0043] For example, in plane c11 of c1 in Fig. 3, area c12 is located away from the plots of cases where bridging has occurred. For this reason, the combination of hopper level and apparent specific gravity in area c12 is considered to be a combination with a low probability of causing bridging. Similarly, in plane c21 of c2, area c22 is located away from the plots of cases where bridging has occurred, and in plane c31 of c3, area c32 is located away from the plots of cases where bridging has occurred.

[0044] Therefore, the classification unit 105 may generate classification information 112 indicating each area as described above. For example, the classification unit 105 may estimate a probability distribution of cases in which bridging occurs by various statistical methods from the distribution range of combinations of hopper levels, apparent specific gravity, and input amounts of cases in which bridging occurs among cases included in the history information 111. The classification unit 105 can use this probability distribution to identify areas in which the probability of bridging not occurring is high (for example, equal to or higher than a threshold set by the user), and generate classification information 112 indicating the areas. Note that, if there is insufficient data on cases to be used for estimating the probability distribution, data on cases in other waste incineration facilities equipped with hoppers and cranes with similar specifications, or other waste incineration facilities with similar types and properties of waste to be treated, may be used.

[0045] Furthermore, the classification information 112 may be a classification model that classifies case data (hopper level, apparent specific gravity, and input amount) into cases with a high probability of bridge occurrence and cases with a low probability of bridge occurrence. In this case, the classification unit 105 can generate the above classification model by performing machine learning using training data in which a label indicating a high probability of bridge occurrence is associated with data of cases in which a bridge has occurred. Note that this machine learning can also use training data in which a label indicating a low probability of bridge occurrence is associated with data of cases in which a bridge has not occurred. For example, a neural network model or the like can be applied as the classification model. Furthermore, the classification model can be generated by unsupervised learning, or by a statistical method.

[0046] Furthermore, the classification unit 105 may generate the classification information 112 according to a user's input via the input unit 13. In this case, for example, a veteran crane operator may specify an area that he or she has determined from experience to have a low probability of bridging, and the classification unit 105 may generate the classification information 112 indicating the specified area.

[0047] Furthermore, the classification unit 105 may generate classification information 112 indicating areas with a high probability of bridge occurrence. In this case, the control content determination unit 103 may determine control content that falls outside the areas indicated in the classification information 112.

[0048] Note that the control content determination unit 103 can also determine the control content without using the classification information 112. For example, the control content determination unit 103 may determine the control content so as to satisfy a condition that a plot appears at a position that is a predetermined distance or more away from any of the plots of cases in which a bridge has occurred, which are included in the history information 111.

[0049] (How to determine the control content) When the control content is determined using the hopper level, apparent specific gravity, and input amount, the input information acquisition unit 102 calculates the apparent specific gravity, and the state determination unit 101 determines the hopper level. In this way, the control content determination unit 103 can use the calculated apparent specific gravity, the determined hopper level, and the classification information 112 to determine the control content such that the combination of the hopper level and apparent specific gravity has a low probability of causing a bridge.

[0050] For example, when 7 tons of garbage needs to be put into the hopper, the control content determination unit 103 controls the operation of the crane so that the combination of the hopper level and the apparent specific gravity is within the area c12 indicated in the classification information 112. Specifically, when the crane grabs the garbage to be put into the hopper, the input information acquisition unit 102 calculates the apparent specific gravity of the grabbed garbage. Then, the amount of garbage grabbed by the crane is adjusted to 7 tons. Next, the control content determination unit 103 specifies the range of the hopper level from the calculated apparent specific gravity, in which the combination with the apparent specific gravity falls within the area c12. In other words, the control content determination unit 103 determines the control content to throw garbage when the hopper level falls within the range. Then, the equipment control unit 104 instructs the crane control device 2 to throw garbage into the hopper at the timing when the hopper level determined by the state determination unit 101 from time to time falls within the above range.

[0051] Similarly, the control content determination unit 103 may determine the apparent specific gravity of the waste to be dumped. In this case, the apparent specific gravity of the waste at each location in the waste pit (for example, each section shown in FIG. 2b) is measured or predicted in advance and recorded in the waste management information. Then, the equipment control unit 104 specifies the range of apparent specific gravity that falls within the area c12 when combined with the required dump amount (for example, 7 tons) and the latest hopper level determined by the state determination unit 101. In other words, the control content determination unit 103 determines the control content to dump the waste with an apparent specific gravity in that range. Then, the equipment control unit 104 instructs the crane control device 2 to dump the waste into the hopper from the section in which the waste with an apparent specific gravity in that range exists.

[0052] The control content determination unit 103 may also determine the amount of garbage to be put in. In this case, the equipment control unit 104 determines whether the combination of the latest hopper level determined by the state determination unit 101 and the apparent specific gravity calculated by the put-in information acquisition unit 102 is included in area c12, c22, or c32. Then, the equipment control unit 104 specifies the amount of garbage to be put in that corresponds to the area determined to be included. In other words, the control content determination unit 103 determines the control content to put in that amount of garbage. Then, the equipment control unit 104 instructs the crane control device 2 to put the determined amount of garbage into the hopper.

[0053] Of course, the control content determination unit 103 only needs to determine the control content such that the combination of the hopper level immediately before the input, the apparent specific gravity of the waste to be input, and the input amount is included in any of the areas c12, c22, and c32, and the control content is not limited to the above examples. For example, the control content determination unit 103 may determine a plurality of control contents, such as determining both the input timing and the input amount.

[0054] The control content determination unit 103 can also determine the control content of the dust feeder and the fire grate in a similar manner. By appropriately controlling the dust feeder and the fire grate, a stable amount of waste can be sent to the incinerator and the waste can be moved at an appropriate speed to maintain a stable combustion state even when the state of the waste in the hopper changes.

[0055] As described above, the control content determination unit 103 may determine the control content at the time of input based on the classification result of the combination of the state of the garbage inside the hopper immediately before the input and the control content at the time of the input, depending on whether the input result is good or bad as shown in the history information 111.

[0056] Various information on past loading of contents into the hopper, shown in the history information 111, can be classified according to the results of loading, and based on this classification, it is possible to estimate control content that is likely to produce good results or control content that is unlikely to produce bad results. Therefore, with the above configuration, appropriate control content can be determined.

[0057] (Processing flow during operation of waste incineration facilities) The flow of the process (control content determination method) executed by the information processing device 1 during the operation period of the waste incineration facility will be described with reference to Fig. 4. Fig. 4 is a flow chart showing an example of the process executed by the information processing device 1.

[0058] In S11, the state determination unit 101 determines the state of the waste contained inside the hopper. For example, the state determination unit 101 may determine the hopper level described above.

[0059] In S12, the input item information acquisition unit 102 acquires input item information. For example, the input item information acquisition unit 102 may calculate the apparent specific gravity of the garbage input into the hopper as the input item information.

[0060] In S13, the control content determination unit 103 determines the control content at the time of throwing in the waste based on the classification information 112, the judgment result of S11, and the thrown-in item information acquired in S12. Note that since the classification information 112 is generated from the history information 111 as described above, it can be said that the process of S13 is a process performed based on the history information 111.

[0061] For example, it is assumed that the classification information 112 is information indicating areas c12, c22, and c23 where the probability of bridges occurring is low, as shown in Fig. 3. In this case, the control content determination unit 103 determines the control content such that the combination of the hopper level immediately before dumping, the apparent specific gravity of the garbage to be dumped, and the dumping amount is included in any of the areas c12, c22, and c23.

[0062] In S14, the equipment control unit 104 controls the equipment in accordance with the control content determined in S13. For example, the equipment control unit 104 may control a crane via the crane control device 2.

[0063] As described above, the control content determination method according to this embodiment includes a state determination step (S11) for determining the state of the waste contained in the hopper, and a control content determination step (S13) for determining the control content when the waste is thrown into the hopper based on history information 111 indicating the state of the waste inside the hopper immediately before the throwing of waste into the hopper, the control content at the time of throwing, and the result of the throwing, and the determination result of S11. This makes it possible to determine appropriate control content according to the state of the waste inside the hopper and stably discharge the waste contained in the hopper into the incinerator.

[0064] In the first embodiment, the description has been centered on an example in which the state determination unit 101 determines the hopper level, but the state determined by the state determination unit 101 may be anything related to the waste discharge state, and is not limited to this example. For example, the state determination unit 101 may determine the moisture content, degree of mixing, apparent specific gravity, type, appearance, etc. of the waste in the hopper, or may determine the occurrence of a trouble, such as the occurrence of a bridge, a slight blockage, or a normal state. The occurrence of a trouble can be determined from various information related to the trouble (for example, the type of waste input, the speed at which the waste is sucked into the hopper, the speed at which the waste is discharged from the hopper, etc.).

[0065] [Embodiment 2] Other embodiments of the present invention will be described below. For the sake of convenience, the same reference numerals are used to designate components having the same functions as those described in the above embodiments, and the description thereof will not be repeated. This also applies to the third and subsequent embodiments.

[0066] (Configuration of information processing device) The configuration of the information processing device 1A according to this embodiment will be described with reference to FIG. 5. FIG. 5 is a block diagram showing an example of a main configuration of the information processing device 1A according to the second embodiment of the present invention. The control unit 10A of the information processing device 1A is different from the control unit 10 of the information processing device 1 shown in FIG. 1 in that it includes a control content determination unit 103A instead of the control content determination unit 103 and includes a model generation unit 105A instead of the classification unit 105. In addition, the storage unit 11A of the information processing device 1A is different from the storage unit 11 of the information processing device 1 shown in FIG. 1 in that it stores history information 111A instead of history information 111 and stores a control content prediction model 112A instead of classification information 112.

[0067] The control content determination unit 103A determines the control content when garbage is thrown in based on the control content prediction model 112A and the determination result of the state determination unit 101. Although details will be described later, the control content prediction model 112A is generated from the history information 111A, so that the control content determination unit 103A, like the control content determination unit 103, determines the control content based on the history information 111A and the determination result of the state determination unit 101.

[0068] The control content prediction model 112A is a prediction distribution calculated by a stochastic process using the history information 111A, and is generated based on the prediction distribution of a function that indicates the relationship between the control content when garbage is thrown into the hopper and the evaluation value of the result of the throwing. Therefore, it can be said that the control content determination unit 103A determines the control content when garbage is thrown based on the above-mentioned prediction distribution.

[0069] According to the above configuration, the control content is determined based on the prediction distribution calculated by the stochastic process using the history information 111A, so that it is possible to determine appropriate control content. In addition, since the amount of data required to converge the above prediction is relatively small, it is possible to determine appropriate control content even if a sufficient amount of history information 111A is not available.

[0070] The model generating unit 105A generates a control content prediction model 112A by using the history information 111A. A method for generating the control content prediction model 112A will be described later.

[0071] Like the history information 111, the history information 111A is information that indicates the state of the waste inside the hopper immediately before the waste was dumped into the hopper in the past, the control details at the time of dumping, and the results of the dumping. The history information 111A differs from the history information 111 in that an evaluation value indicating the quality of the control result is associated with each case. For example, a low evaluation value is associated with a case in which a problem such as bridging occurred after control, and a high evaluation value is associated with a case in which stable discharge continued after control.

[0072] (How to generate a control content prediction model) As described above, the model generating unit 105A generates the control content prediction model 112A by using the history information 111A. The control content prediction model 112A can be expressed as a function including at least an explanatory variable indicating the determination result of the state determining unit 101 and a target variable indicating the control content, for example. By using this function, the control content can be expressed by the explanatory variable.

[0073] Examples of explanatory variables used include various values ​​included in the history information, a waste quality index, and a hopper level. These values ​​may be time-series values ​​with time stamps. The waste quality index is a numerical value indicating the characteristics of the waste, as described above. The waste quality index may be, for example, an evaluation index of the waste surface (for example, a fractal dimension calculated by fractal analysis of an image of the waste surface, or an evaluation index indicating the color characteristics of the waste calculated by hue feature analysis of an image of the waste surface). The waste quality index may also be a value indicating the combustibility of the waste, evaluated by comprehensively considering, for example, the type of waste, the degree of mixing, the moisture content, etc.

[0074] When using explanatory variables such as those described above, the objective variables can be, for example, the timing of dumping, the dump weight, the type and properties of the garbage dumped, etc. In addition, the objective variables can also include the control details of the device that processes the garbage discharged from the hopper, such as the speed of the grate.

[0075] Here, if the history information 111A includes N cases, these cases are

[0076]

number

[0077] It is expressed as The evaluation value for this is

[0078]

number

[0079] This is expressed as:

[0080] The model generation unit 105A uses the history information 111A to calculate, for example, a predictive distribution of a function indicating the relationship between the control content and its evaluation value. This function will be referred to as an evaluation function f(θ) hereinafter. When new case data is added to the history information 111A, the model generation unit 105A updates the predictive distribution to reflect the data.

[0081] The relationship between the control contents and their evaluation values ​​is expressed as Gaussian noise ε n Using ~Ν(0,β)

[0082]

number

[0083] Assuming this, we obtain the following distribution as the predictive distribution of the evaluation function using the Gaussian process:

[0084]

number

[0085]

number

[0086]

number

[0087] where k * = k(θ,θ), and K Θ is [K Θ ] i,j =k(θ i ,θ j ) is the Gram matrix obtained by

[0088]

number

[0089] k Θ , * is [k Θ , * ] i =k(θ i ,θ) and k(·,·) is the kernel function. Here, the parameter of the kernel function is α k Let us assume that.

[0090] The average function μ(θ) indicates the average value of the evaluation function predicted from the history information 111A. The variance function σ(θ) is the variance of the evaluation function predicted from the history information 111A. σ(θ) indicates the uncertainty of the prediction, and its value tends to be large in areas where the history information 111A is insufficient. Note that d12 in FIG. 7, which will be described later, indicates the variance σ. When σ is large, the width of d12 widens, and it can be understood that the prediction is uncertain. In other words, it can be seen that there is a shortage of history information 111A necessary to increase the certainty of the prediction. As is clear from equation (3), the kernel function included in the variance function σ(θ) and the parameter α of the kernel function k affects the calculation of the predictive distribution. Details will be described later, but when calculating the predictive distribution, the parameter α k Optimization is performed.

[0091] In order to generate a control content prediction model 112A capable of predicting the optimal control content, the model generation unit 105A uses a variable θ * Search for candidates for the variable θ * may be a function including the explanatory variables mentioned above, for example.

[0092] In the above search, the model generation unit 105A uses the mean function μ(θ) and the variance function σ(θ) to find the variable θ that maximizes the following acquisition function a(θ): * The variables θ found in this search are * are candidates for variables that represent the optimal control content. This search is based on the UCB (Upper Confidence Bound) strategy. Note that κ in formula (5) is a parameter for adjusting the search and utilization. Of course, it is also possible to search for new control variables using other methods. For example, optimal control variable candidates may be searched for using the PI (Probability of Improvement) strategy or the EI (Expected Improvement) strategy.

[0093]

number

[0094]

number

[0095] The model generation unit 105A searches for the above-mentioned candidates every time a case is added and the predictive distribution is updated, thereby finding the optimal candidate, i.e., the variable θ * θ * The control content prediction model 112A can be derived from the above. The control content prediction model 112A can also be derived from a stochastic process other than the Gaussian process (for example, a Markov process, etc.).

[0096] (Process flow for generating control content prediction model) The flow of processing performed by the model generating unit 105A to generate the control content prediction model 112A will be described with reference to Fig. 6. Fig. 6 is a flowchart showing an example of a method for generating the control content prediction model 112A.

[0097] In S21, the model generation unit 105A sets the data of the case shown in the history information 111A as initial data. At this stage, it is sufficient that the history information 111A includes a control result based on at least one trial (one variable θ and an evaluation value that evaluates the result of the control using the variable θ).

[0098] In S22, the model generation unit 105A optimizes the parameters of the kernel function. As described above, the parameters of the kernel function are α k The optimization method is not particularly limited, and for example, an optimization method applied in general Bayesian optimization can be applied.

[0099] In S23, the model generation unit 105A calculates the predictive distribution of the evaluation function using the initial data set in S21 and the parameters of the kernel function optimized in S22. As described above, this predictive distribution is expressed by the formulas (1) to (3).

[0100] In S24, the model generation unit 105A selects the variable θ* Search for the variable θ * is a candidate for the optimal value of the variable θ. As mentioned above, this process is expressed by the above equations (4) and (5).

[0101] In S25, the model generation unit 105A calculates the variable θ * The method of determining whether or not the variable θ detected in the most recent S24 is the optimal value is not particularly limited. For example, the model generation unit 105A * If the variable θ matches a previously detected variable, * is the optimal value, and if they do not match, it may be determined that the value is not the optimal value. Note that the previously detected variables refer to variables corresponding to the control contents included in the history information 111A, and the evaluation value for the control result has already been calculated.

[0102] If it is determined in S25 that the value is the optimal value (YES in S25), the process proceeds to S30. In S30, the model generation unit 105A determines the optimal value θ * The control content prediction model 112A defined by the above is stored in the storage unit 11A, and the process of FIG. 6 is then terminated.

[0103] On the other hand, in S25, the variable θ * is not the optimal value (NO in S25), the model generation unit 105A * The user of the information processing device 1A is notified of the variable θ based on this notification. * The control result is input to the information processing device 1A via the input unit 13.

[0104] In S26, the model generation unit 105A acquires the evaluation result input as described above. Then, in S27, the model generation unit 105A compares the evaluation result acquired in S26 with the variable θ determined in the most recent S24. * are associated with each other and added to the history information 111A.

[0105] In S28, the model generation unit 105A calculates the evaluation result in S27 and the variable θ * Then, in S29, the model generation unit 105A optimizes the parameters of the kernel function using the history information 111A to which the evaluation result in S27 and the variable θ * The predictive distribution of the evaluation function is calculated using the history information 111A to which the variables have been added and the parameters of the kernel function optimized in S28. After this, the process returns to S24, and a search for variables is performed. In this way, by repeating the addition of variables and the updating of the predictive distribution, it is possible to determine variables that can be expected to produce the desired control result.

[0106] The above-mentioned process procedure is an example, and is not limited to this example. For example, the process of S25 may be performed after S29. In this case, if the result of S25 is YES, the process proceeds to S30, and if the result of S25 is NO, the process returns to S24.

[0107] Here, the update of the predictive distribution will be described with reference to Fig. 7. Fig. 7 is a diagram showing an example of updating the predictive distribution. d1 to d4 in Fig. 7 are variables θ * 7 is a graph showing a prediction distribution with the horizontal axis being the evaluation value and the vertical axis being the evaluation value. In d1 in Fig. 7, d11 is a plot showing variables indicating the control contents of cases included in the history information 111A and the evaluation value of the control content prediction model 112A for those cases. Also, d12 is a prediction distribution, d13 is a true function (a function that correctly expresses the relationship between the evaluation value and the variable), and d14 shows cases where bridging occurred.

[0108] 7, as the plots indicated by d11 are added, the predictive distribution d12 converges and approaches the true function d13. Therefore, by repeatedly adding cases and updating the predictive distribution d12, it is possible to accurately estimate optimal variables that define the control content prediction model 112A.

[0109] [Embodiment 3] (Configuration of information processing device) The configuration of the information processing device 1B according to this embodiment will be described with reference to FIG. 8. FIG. 8 is a block diagram showing an example of a configuration of a main part of the information processing device 1B according to the third embodiment of the present invention. The control unit 10B of the information processing device 1B is different from the control unit 10 of the information processing device 1 shown in FIG. 1 in that it includes a control content determination unit 103B instead of the control content determination unit 103, and includes a learning unit 105B instead of the classification unit 105. In addition, the storage unit 11B of the information processing device 1B is different from the storage unit 11 of the information processing device 1 shown in FIG. 1 in that it stores teacher data 111B instead of the history information 111, and stores a control content prediction model 112B instead of the classification information 112.

[0110] The control content determination unit 103B determines the control content when garbage is put in using the control content prediction model 112B based on the determination result of the state determination unit 101. The control content determination unit 103B may also use the put-in item information acquired by the put-in item information acquisition unit 102 to determine the control content. As will be described later, the control content prediction model 112B is generated from history information, so the control content determination unit 103B determines the control content based on the history information and the determination result of the state determination unit 101, similar to the control content determination unit 103.

[0111] The learning unit 105B generates the control content prediction model 112B by performing machine learning using the teacher data 111B. A method for generating the control content prediction model 112B by the learning unit 105B will be described later.

[0112] The teacher data 111B is generated from history information indicating the state of the garbage inside the hopper immediately before the garbage was dumped into the hopper in the past, the control details at the time of dumping, and the results of the dumping. More specifically, the teacher data 111B indicates the relationship between the state of the garbage inside the hopper immediately before the garbage was dumped and the control details at the time of dumping when the state of the garbage inside the hopper was maintained or improved after the garbage was dumped.

[0113] The control content prediction model 112B is generated by learning the relationship between the state of the garbage inside the hopper immediately before the garbage is put in and the control content at the time of putting in when the state of the garbage inside the hopper is maintained or improved after the garbage is put in. The objective variable of the control content prediction model 112B is the above control content, and the explanatory variable is information indicating the state of the garbage. As the state of the garbage, various information such as that described in the first embodiment can be applied. The explanatory variable may also include information on the put-in item. As described above, the control content prediction model 112B is generated by the learning unit 105B.

[0114] As described above, the control content determination unit 103B determines the control content when garbage is thrown in using the control content prediction model 112B generated by learning using the teacher data 111B that indicates the relationship between the state of garbage inside the hopper immediately before garbage is thrown in and the control content when the garbage state inside the hopper is maintained or improved after the garbage is thrown in. This makes it possible to determine the control content that can be expected to maintain or improve the hopper state.

[0115] (Flow of the method for generating a control content prediction model) The flow of a method for generating the control content prediction model 112B by the learning unit 105B will be described with reference to Fig. 9. Fig. 9 is a flowchart showing an example of a method for generating the control content prediction model 112B.

[0116] In S31 (teacher data acquisition step), the learning unit 105B acquires the teacher data 111B from the memory unit 11B. As described above, the teacher data 111B is data that associates the state of the garbage inside the hopper immediately before the garbage was dumped into the hopper in the past, where the state of the hopper was maintained or improved after the garbage was dumped, with the control content performed at the time of the garbage dump.

[0117] In S32 (learning step), the learning unit 105B learns using the teacher data 111B acquired in S31 to generate a control content prediction model 112B that predicts the control content to be performed when garbage is loaded into the hopper in order to maintain or improve the state of the hopper, based on the state of garbage inside the hopper immediately before garbage is loaded.

[0118] In S33, the learning unit 105B judges whether or not to end the learning. If the judgment in S33 is YES, the process proceeds to S34, where the learning unit 105B stores the generated control content prediction model 112B in the storage unit 11 and ends the process. The condition for ending the learning may be determined in advance. For example, the learning unit 105B may end the learning when the generation and update of the control content prediction model 112B using a predetermined number of teacher data 111B is completed (in other words, when the loop of S31 to S33 is repeated a predetermined number of times).

[0119] If the result of S33 is NO, the process returns to S31. When the process returns to S31, the learning unit 105B acquires new teacher data 111B, performs learning again, and updates the control content prediction model 112B.

[0120] As described above, the method of generating the control content prediction model 112B in this embodiment includes a teacher data acquisition step (S31) of acquiring teacher data 111B that associates the state of the garbage inside the hopper immediately before the garbage was deposited into the hopper in the past, for which the hopper state was maintained or improved after the deposit, with the control content performed at the time of the deposit, and a learning step (S32) of generating a control content prediction model 112B that predicts, by learning using the teacher data 111B, the control content to be performed when the garbage is deposited into the hopper in order to maintain or improve the hopper state, from the state of the garbage inside the hopper immediately before the garbage is deposited.

[0121] According to the above configuration, it is possible to generate a control content prediction model 112B that can predict the content of control to be performed when garbage is poured into the hopper in order to maintain or improve the state of the hopper, based on the state of garbage inside the hopper immediately before garbage is poured in. Then, by performing control with the content predicted by this control content prediction model 112B, it becomes possible to stably discharge garbage contained in the hopper.

[0122] (About reinforcement learning) The control content determination unit 103B may determine the control content when garbage is deposited using a reinforcement learning model generated by reinforcement learning in which the state of garbage inside the hopper immediately before garbage is deposited, as shown in the history information, is the environment, the control content when garbage is deposited is the behavior, and the evaluation value of the result of deposit is the reward.

[0123] According to the above configuration, it is possible to determine an action that is expected to produce good results, i.e., the control content when the contents are added, depending on the environment immediately before the contents are added, i.e., the state of the contents inside the hopper.

[0124] The reinforcement learning model may be generated by another device, or the learning unit 105B may generate the reinforcement learning model. In the latter case, the history information 111A of the second embodiment may be stored in the storage unit 11B instead of the teacher data 111B.

[0125] The reinforcement learning model can be generated, for example, by a flow as shown in Fig. 10. Fig. 10 is a flowchart showing an example of a method for generating a reinforcement learning model.

[0126] In S31A (history acquisition step), the learning unit 105B acquires history information. The acquired history information is information indicating the state of the waste inside the hopper immediately before the waste was dumped into the hopper in the past, the control details at the time of dumping, the result of the dumping, and its evaluation value.

[0127] In S32A (learning step), the learning unit 105B generates a reinforcement learning model by reinforcement learning in which the state of the garbage inside the hopper immediately before the garbage is dumped, as shown in the historical information acquired in S31A, is used as the environment, the control at the time of dumping the garbage is used as the action, and the evaluation value of the result of dumping is used as the reward.

[0128] In S33A, the learning unit 105B judges whether or not to end the learning. If the judgment in S33A is YES, the process proceeds to S34A, where the learning unit 105B stores the generated reinforcement learning model in the storage unit 11B and ends the process. The condition for ending the learning may be determined in advance. For example, the learning unit 105B may end the learning when the generation and update of the reinforcement learning model using a predetermined number of teacher data 111B is completed (in other words, when the loop of S31A to S33A is repeated a predetermined number of times).

[0129] On the other hand, if the determination in S33A is NO, the process returns to S31A. When the process returns to S31A, the learning unit 105B performs learning again based on the history information, and updates the reinforcement learning model.

[0130] As described above, the method for generating a reinforcement learning model in this embodiment includes a history acquisition step (S31A) for acquiring historical information indicating the state of the garbage inside the hopper immediately before the garbage was dumped into the hopper in the past, the control details at the time of dumping, and the result of the dumping, and a learning step (S32A) for generating a reinforcement learning model by reinforcement learning in which the state of the garbage inside the hopper immediately before the garbage was dumped, as indicated in the historical information, is the environment, the control at the time of garbage dumping is the action, and the evaluation value of the result of the dumping is the reward.

[0131] According to the above configuration, it is possible to generate a reinforcement learning model that can determine the control content that will optimize the result of waste dumping, depending on the state of the waste inside the hopper just before the waste is dumped. Then, by controlling the content predicted by this reinforcement learning model, it becomes possible to stably discharge the waste contained in the hopper.

[0132] [Embodiment 4] The configuration of an information processing device 1C according to this embodiment will be described with reference to Fig. 11. Fig. 11 is a block diagram showing an example of a configuration of a main part of an information processing device 1C according to a fourth embodiment of the present invention. A control unit 10C of the information processing device 1C differs from the control unit 10 of the information processing device 1 shown in Fig. 1 in that it does not include an input item information acquisition unit 102, that it includes a control content determination unit 103C instead of the control content determination unit 103, and that it is equipped with a dust adjustment unit 106C.

[0133] The control content determination unit 103C determines the properties of the waste to be put into the hopper. Then, the put-in waste adjustment unit 106C has a crane adjust the waste in the waste pit to generate waste with the determined properties in the waste pit. With this configuration, the crane adjusts the waste in the waste pit, automatically resolving a situation in which the waste pit does not contain waste with the properties that should be put in, and making it possible to put in waste with appropriate properties according to the state of the waste inside the hopper.

[0134] The above-mentioned properties may be anything related to the state of waste discharged from the hopper. For example, the control content determining unit 103C may determine at least one of the apparent specific gravity, the degree of agitation, and the type of the waste.

[0135] The characteristics of the waste to be dumped can be determined based on historical information about previous dumping of waste into the hopper, which indicates the state of the waste inside the hopper immediately before the dumping, the characteristics of the waste dumped, and the results of the dumping.

[0136] For example, the control content determination unit 103C may determine the apparent specific gravity of the garbage to be put into the hopper based on the relationship between the hopper level, the amount of garbage put in, and the apparent specific gravity as shown in FIG. 3. In this case, the control content determination unit 103C may determine the apparent specific gravity using the classification information 112. For example, the control content determination unit 103C may determine the garbage properties using the control content prediction models 112A and 112B or the reinforcement learning model as shown in the second and third embodiments. In this case, a model that has been learned as the garbage properties may be used. The classification information 112, the control content prediction models 112A and 112B, and the reinforcement learning model are all based on historical information, so that the garbage properties are determined based on historical information in any of the above methods. Also, determining the properties of the garbage to be put in can be said to determine the control content of putting garbage with the properties into the hopper.

[0137] The control content determination unit 103C can also be described as a garbage input determination unit that determines the characteristics of garbage to be input into the hopper based on historical information indicating the state of the garbage inside the hopper immediately before input of garbage into the hopper in the past, the characteristics of the input garbage, and the results of input, and the judgment result of the status judgment unit 101.

[0138] The method of adjusting the garbage by the garbage adjustment unit 106C is not particularly limited. For example, when the degree of agitation of the garbage to be input is determined, the input garbage adjustment unit 106C may cause a crane to agitate the garbage to generate garbage with the determined degree of agitation. The degree of agitation of the garbage at each location in the garbage pit may be managed as garbage management information. In addition, if the apparent specific gravity and moisture content are recorded in the garbage management information, the input garbage adjustment unit 106C can adjust the garbage with the determined apparent specific gravity and moisture content by referring to the garbage management information. For example, when it is determined to input garbage with a standard moisture content, the input garbage adjustment unit 106C may generate garbage with a standard moisture content by mixing garbage with a moisture content higher than the standard and garbage with a moisture content lower than the standard.

[0139] As described above, the information processing device 1C includes a state determination unit 101 that determines the state of the garbage stored inside the hopper, a control content determination unit 103C (garbage to be put in determination unit) that determines the properties of the garbage to be put into the hopper based on history information indicating the state of the garbage inside the hopper immediately before the past putting of garbage into the hopper, the properties of the put-in garbage, and the results of the putting, and the determination results of the state determination unit 101, and a garbage adjustment unit 106C that causes a crane that transports garbage in a garbage pit that stores the garbage to be put into the hopper to adjust the garbage, thereby generating garbage with the determined properties in the garbage pit. This automatically resolves the state where garbage with the properties to be put in is not in the garbage pit, and makes it possible to put in garbage with appropriate properties according to the state of the garbage inside the hopper.

[0140] (Processing flow during operation of waste incineration facilities) The flow of the processing (control content determination method / waste adjustment method) executed by the information processing device 1C during the operation of the waste incineration facility will be described with reference to Fig. 12. Fig. 12 is a flowchart showing an example of the processing executed by the information processing device 1C. The processing of Fig. 12 may be executed in accordance with the cycle of dumping waste into the hopper, for example. For example, when dumping waste every 20 to 30 minutes, the processing of S41 to S44 described below may be executed every 20 to 30 minutes.

[0141] In S41 (state determination step), the state determination unit 101 determines the state of the waste contained inside the hopper. For example, the state determination unit 101 may determine the hopper level described above.

[0142] In S42 (a step of determining the type of garbage to be thrown in), the control content determination unit 103C determines the properties of the garbage to be thrown in based on the result of the determination in S41. As described above, the properties of the garbage can be determined based on history information. For example, the control content determination unit 103C may use the classification information 112 to determine an apparent specific gravity that has a low probability of bridging and corresponds to the hopper level and the amount of garbage thrown in determined in S41. The control content determination unit 103C may also determine the timing of throwing in the garbage.

[0143] In S43 (input waste adjustment step), the waste adjustment unit 106C causes the crane to adjust the waste via the equipment control unit 104, generating waste with the properties determined in S42 in the waste pit. For example, if the apparent specific gravity is determined in S42, the waste adjustment unit 106C may generate waste with the determined apparent specific gravity in a specified section in the waste pit. Note that, if waste with the properties determined in S42 is already present in the waste pit, the processing of S43 may be omitted.

[0144] In S44, the equipment control unit 104 controls the dumping of the waste with the properties determined in S42 into a hopper. Specifically, the equipment control unit 104 may have a crane grab the waste in the section where the waste with the properties determined in S42 was generated, and dump the waste into the hopper. If the dumping timing was also determined in S42, the equipment control unit 104 will have the crane dump the waste into the hopper at the determined timing. After this, the process returns to S41.

[0145] In S42, the properties of the waste to be thrown in may be determined collectively for each of multiple throwings. For example, when the properties of the waste for the next throwing and the throwing after that are determined in S42, the waste is adjusted until the next throwing (S43), and throwing control is performed when the next throwing timing arrives (S44). Then, the waste is adjusted until the throwing after that (S43), and throwing control is performed when the throwing timing after that arrives (S44). In addition, when the properties of the waste are determined collectively for multiple throwings, the waste may also be adjusted collectively.

[0146] As described above, the waste adjustment method according to this embodiment includes a state determination step (S41) for determining the state of the waste stored inside the hopper, a waste input determination step (S42) for determining the properties of the waste to be input into the hopper based on the history information indicating the state of the waste inside the hopper immediately before the input of waste, the properties of the input waste, and the input result, and the determination result of S41, and a waste input adjustment step (S43) for making a crane that transports waste in a waste pit that stores the waste to be input into the hopper adjust the waste to generate waste with the properties determined in S42 in the waste pit. This automatically resolves the state in which waste with the properties to be input is not present in the waste pit, and makes it possible to input waste with appropriate properties according to the state of the waste inside the hopper.

[0147] [Modifications] The executing entity of each process described in each of the above-mentioned embodiments is arbitrary and is not limited to the above-mentioned examples. In other words, as long as the executing entity of each process described in each of the above-mentioned embodiments is capable of executing, the devices constituting the management system 5 can be changed as appropriate. For example, among the processes shown in FIG. 4, the processes of S11 to S12 may be executed by one information processing device, and the processes of S13 to S14 may be executed by another information processing device. In this way, the executing entity of each process described in each of the above-mentioned embodiments may be one information processing device or multiple information processing devices.

[0148] [Software implementation example] The functions of the information processing device 1, 1A to 1C (hereinafter referred to as the "device") can be realized by a program for causing a computer to function as the device, and a program (control content determination program / dust adjustment program) for causing a computer to function as each control block of the device (particularly each part included in the control unit 10, 10A to 10C).

[0149] In this case, the device includes a computer having at least one control device (e.g., a processor) and at least one storage device (e.g., a memory) as hardware for executing the program. The control device and storage device execute the program to realize each function described in each of the above embodiments.

[0150] The program may be non-transitory and may be recorded in one or more computer-readable recording media. The recording media may or may not be included in the device. In the latter case, the program may be provided to the device via any wired or wireless transmission medium.

[0151] In addition, some or all of the functions of each of the control blocks can be realized by a logic circuit. For example, an integrated circuit in which a logic circuit that functions as each of the control blocks is formed is also included in the scope of the present invention. In addition, the functions of each of the control blocks can be realized by, for example, a quantum computer.

[0152] The present invention is not limited to the above-described embodiments, and various modifications are possible within the scope of the claims. Embodiments obtained by appropriately combining the technical means disclosed in different embodiments are also included in the technical scope of the present invention. [Explanation of symbols]

[0153] 1. Information processing device 101 state determination unit, 103 control content determination unit, 111 history information 1A Information processing equipment 101 state determination unit, 103A control content determination unit, 111A history information, 112A Control Content Prediction Model 1B Information processing equipment 101 state determination unit, 103B control content determination unit, 111B teacher data, 112B Control Content Prediction Model 1C Information processing equipment 101 state determination unit, 103C control content determination unit (throw-in garbage determination unit), 106C Garbage Control Unit

Claims

1. a state determination unit that determines a state of the contents contained in the hopper; An information processing device comprising: a control content determination unit that determines the control content when the contents are added to the hopper based on historical information indicating the state of the contents inside the hopper immediately before the contents were added to the hopper in the past, the control content at the time of the addition, and whether or not a bridge occurred within a specified period after the addition, and the judgment result of the state judgment unit.

2. 2. The information processing device according to claim 1, wherein the control content determination unit determines the control content at the time of putting the contents into the hopper based on a predictive distribution of a function calculated by a stochastic process using the historical information, the predictive distribution indicating the relationship between the control content at the time of putting the contents into the hopper and an evaluation value that evaluates the control depending on whether or not a bridge has occurred within a specified period after the putting in.

3. The information processing device described in claim 1, wherein the control content determination unit determines the control content when the contents are added using a control content prediction model generated by learning, using teacher data based on the historical information, the relationship between the state of the contents inside the hopper immediately before the contents are added and the control content when the contents are added when the state of the contents inside the hopper is maintained or improved after the contents are added.

4. The information processing device according to claim 1, wherein the control content determination unit determines the control content when the contents are added based on a classification result of a combination of the state of the contents inside the hopper immediately before the addition and the control content at the time of the addition, depending on whether or not a bridge has occurred, as shown in the history information.

5. The information processing device of claim 1, wherein the control content determination unit determines the control content when the contents are added using a reinforcement learning model generated by reinforcement learning in which the state of the contents inside the hopper immediately before the contents are added, as shown in the history information, is the environment, the control content when the contents are added is the action, and the reward is an evaluation value that evaluates the control depending on whether a bridge has occurred within a predetermined period after the addition.

6. The contents are garbage, The control content determination unit determines the properties of the waste to be put into the hopper, An information processing device as described in any one of claims 1 to 5, further comprising an input waste adjustment unit that adjusts the waste by a crane that transports waste within a waste pit that stores the waste to be input into the hopper, thereby generating waste of the determined characteristics within the waste pit.

7. A state determination unit that determines the state of the waste contained inside the hopper; a waste input determination unit that determines the characteristics of the waste to be input into the hopper based on history information indicating the state of the waste inside the hopper immediately before the input of waste into the hopper in the past, the characteristics of the input waste, and whether or not a bridge has occurred within a specified period after the input, and on the determination result of the state determination unit; An information processing device comprising: an input garbage adjustment unit that adjusts the garbage by a crane that transports garbage within a garbage pit that stores the garbage to be input into the hopper, thereby generating garbage with the determined characteristics within the garbage pit.

8. A control content determination method executed by one or more information processing devices, comprising: A state determination step of determining a state of the contents contained in the hopper; A control content determination method including: a control content determination step for determining the control content when the contents are added to the hopper based on historical information indicating the state of the contents inside the hopper immediately before the contents were added to the hopper in the past, the control content at the time of addition, and whether or not a bridge occurred within a specified period after the contents were added, and on the determination result of the state determination step.

9. A dust adjustment method executed by one or more information processing devices, comprising: A state determination step for determining the state of the waste contained inside the hopper; a waste input determination step for determining the characteristics of the waste to be input into the hopper based on history information indicating the state of the waste inside the hopper immediately before the input of waste into the hopper in the past, the characteristics of the input waste, and whether or not a bridge has occurred within a specified period after input, and on the determination result of the state determination step; and an input waste adjustment step of adjusting the waste by using a crane that transports the waste within a waste pit that stores the waste to be input into the hopper, thereby generating waste with the determined characteristics within the waste pit.

10. A method for generating a control content prediction model, executed by one or more information processing devices, comprising: a teacher data acquisition step of acquiring teacher data that associates the state of the contents inside the hopper immediately before a past loading of contents into the hopper in which the state of the hopper was maintained or improved after the loading with the control content performed at the time of the loading; A learning step of generating a control content prediction model that predicts a control content to be performed when the contents are poured into the hopper in order to maintain or improve the state of the hopper, based on the state of the contents inside the hopper immediately before the contents are poured into the hopper, by learning using the teacher data; A method for generating a control content prediction model, wherein the training data is generated from historical information indicating the state of the contents inside the hopper immediately before the contents were placed into the hopper in the past, the control content performed at the time of the placement, and whether or not a bridge occurred within a specified period after the placement.

11. A method for generating a reinforcement learning model, executed by one or more information processing devices, comprising: a history acquisition step of acquiring history information indicating a state of the contents inside the hopper immediately before the contents were put into the hopper in the past, the control content at the time of the putting, and whether or not a bridge occurred within a predetermined period after the putting; a learning step of generating a reinforcement learning model by reinforcement learning in which the state of the contents inside the hopper immediately before the contents are added, as shown in the history information, is the environment, the control at the time of adding the contents is the action, and an evaluation value evaluating the control depending on whether or not a bridge has occurred within a specified period after the addition is used as a reward.

12. 2. A control content determination program for causing a computer to function as the information processing device according to claim 1, the control content determination program causing a computer to function as the state determination section and the control content determination section.

13. A garbage adjustment program for causing a computer to function as the information processing device according to claim 7, the garbage adjustment program causing a computer to function as the state determination unit, the input garbage determination unit, and the input garbage adjustment unit.

Images