Information processing device, information processing method, and information processing program

Abstract

To provide a technique for classifying, predicting, and estimating a combustion state in an incinerator with higher accuracy.SOLUTION: An information processing apparatus includes: a first image analyzing unit which uses a first learned model having machine-learned first teacher data generated by assigning a classification label in at least one first evaluation axis, which becomes an element for determining a combustion state to evaluate the data in the first evaluation axis using new infrared image data in an incinerator as input, for infrared image data in the incinerator captured by an infrared imaging apparatus; and a combustion state determining unit for determining the current combustion state based on results of the evaluation in the first evaluation axis.SELECTED DRAWING: Figure 2

Description

Technical Field

[0001] The present invention relates to an information processing device, an information processing method, and an information processing program for determining a combustion state in an incinerator.

Background Art

[0002] In waste incineration facilities, in order to not only efficiently recover heat from combustion exhaust gas but also suppress the emission of harmful substances (such as CO, dioxins, NOx, etc.), automatic combustion control (ACC) has been introduced for the purpose of stabilizing the combustion of waste.

[0003] However, when the maintenance of the combustion state in the incinerator cannot be handled by ACC due to significant fluctuations in the quality of the waste, etc., a skilled operator may manually intervene. At this time, the skilled operator makes a judgment of manual intervention based on process values obtained from various sensors and combustion images that capture the combustion state.

[0004] The information of the above combustion images is a very important indicator, and patents for technologies that classify the combustion state and predict process values from combustion images mainly using image recognition technology based on deep learning have been filed by various companies. For example, in Patent Document 1, it is proposed to use a classification model that classifies the combustion state into, for example, eight categories with the combustion image as an input. Also, in Patent Document 2, it is proposed to create an estimation model that predicts estimated values (such as CO concentration, NOx concentration, unburned components in ash, etc.) corresponding to the combustion state with the combustion image as an input.

Prior Art Documents

Patent Documents

[0005]

Patent Document 1

Patent Document 2

Summary of the Invention

Problems to be Solved by the Invention

[0006] The combustion footage primarily shows the flames in the main combustion area of ​​the incinerator (flames generated by the combustion of combustible gases produced by the thermal decomposition and partial oxidation of waste), and the combustion footage provides information such as the intensity of the flames.

[0007] However, the state of combustion is influenced not only by the state of the flame but also by information about the material being burned (for example, the quantity, quality, and type of the material being burned). For example, even if the flame intensity is the same, if the amount of material being burned at that time is different, the actual exhaust gas concentration (CO concentration, etc.) is likely to be different. Furthermore, regarding technologies that classify the state of combustion and predict process values ​​from combustion video, if combustion video only captures the state of the flame, it is highly likely that even if the combustion state is actually different due to differences in the amount of material being burned, it may be difficult to classify them as the same state. Therefore, predicting and estimating process values ​​and classifying, predicting, and estimating the state of combustion may not be successful.

[0008] This invention has been made in consideration of the above points. The object of this invention is to provide a technology for more accurately classifying, predicting, and estimating the combustion state inside an incinerator. [Means for solving the problem]

[0009] An information processing device according to a first aspect of the present invention is: The combustion state is determined from infrared imaging data of the inside of the incinerator captured by an infrared imaging device. A first image analysis unit performs evaluation on the first evaluation axis using new infrared imaging data from inside the incinerator as input, using a first trained model that has been machine-trained on first training data generated by assigning classification labels on at least one first evaluation axis which is an element, The system includes a combustion state determination unit that determines the current combustion state based on the evaluation results on the first evaluation axis.

[0010] In this configuration, by appropriately selecting the imaging wavelength, the infrared imaging device can acquire infrared imaging data from inside the incinerator, including images from the upstream side of the incinerator, where the effects of the flames have been removed, to images of the materials actually being burned. Therefore, by using the infrared imaging data from inside the incinerator to perform an evaluation on at least one first evaluation axis that is a factor in determining the combustion state, and determining the combustion state based on the evaluation results, the combustion state inside the incinerator can be classified, predicted, and estimated with greater accuracy compared to classifying, predicting, and estimating the combustion state and process values ​​from combustion images that only show the flame situation.

[0011] An information processing device according to a second aspect of the present invention is an information processing device according to a first aspect, The first evaluation axis includes at least one of the following: the amount of material to be burned, the quality of the material to be burned, the type of material to be burned, the temperature of the material to be burned, the amount of combustion gas generated (CO, etc.), and the temperature of the incinerator wall.

[0012] An information processing device according to a third aspect of the present invention is an information processing device according to a first or second aspect, The aforementioned infrared imaging data consists of video data lasting 60 seconds or less.

[0013] According to the inventors' findings, the combustion state inside an incinerator changes every 5 to 10 seconds. Therefore, in this configuration, by using motion image data of 60 seconds or less as one unit for infrared imaging data inside the incinerator, the combustion state inside the incinerator can be classified, predicted, and estimated with greater accuracy.

[0014] An information processing device according to a fourth aspect of the present invention is an information processing device according to any one of the first to third aspects, The aforementioned infrared imaging data consists of video data lasting 5 seconds or longer.

[0015] In this configuration, while the combustion state inside the incinerator changes every 5 to 10 seconds, by using motion image data of 5 seconds or more as one unit for infrared imaging data inside the incinerator, the combustion state inside the incinerator can be classified, predicted, and estimated with greater accuracy.

[0016] An information processing device according to a fifth aspect of the present invention is an information processing device according to any of the first to fourth aspects, The classification label is at least one of the following: a label indicating which of the predetermined classification items the data belongs to, and the relative order among multiple infrared imaging data.

[0017] The inventors of this case have conducted actual verifications and confirmed that by using a first trained model, which is generated by assigning a relative order on a first evaluation axis among multiple infrared imaging data points to infrared imaging data taken inside an incinerator using an infrared imaging device, and then performing evaluation on the first evaluation axis using the first trained model (ranking learning) based on the evaluation results, the combustion state inside the incinerator can be classified, predicted, and estimated with greater accuracy. Therefore, according to this embodiment, the combustion state inside the incinerator can be classified, predicted, and estimated with greater accuracy.

[0018] The information processing device according to the sixth aspect of the present invention is an information processing device according to any one of the first to fifth aspects, The first image analysis unit uses a first trained model, which is generated by machine learning from first training data, to which classification labels on the first evaluation axis are assigned to a combination of infrared imaging data from inside the incinerator and process data obtained from sensors installed in the facility and / or computational amounts obtained from said process data. The first image analysis unit then uses this first trained model to perform an evaluation on the first evaluation axis, taking a combination of new infrared imaging data from inside the incinerator and new process data and / or computational amounts obtained from said process data as input.

[0019] The information processing apparatus according to the seventh aspect of the present invention is an information processing apparatus according to the sixth aspect, The process data is data that takes into account the time lag between the imaging time of the infrared imaging data and the response time of the sensor. The time lag may be appropriately determined based on the response speed of the sensor, the installation position of the sensor in the facility, experiments, simulations, and the rules of thumb of the operator.

[0020] The information processing apparatus according to the eighth aspect of the present invention is the information processing apparatus according to any one of the first to seventh aspects, and the first image analysis unit uses a first learned model obtained by machine learning first teacher data generated by assigning classification labels for each of two or more of the first evaluation axes that are elements for determining the combustion state to the infrared imaging data in the incinerator, and performs evaluations for each of the two or more first evaluation axes using new infrared imaging data in the incinerator as an input, the combustion state determination unit determines the current combustion state by mapping the evaluation results for each of the two or more first evaluation axes onto a predetermined combustion state determination map having the two or more first evaluation axes as coordinate axes.

[0021] According to such an aspect, evaluations are performed for each of two or more first evaluation axes that are elements for determining the combustion state using the infrared imaging data in the incinerator, and the combustion state is determined based on the evaluation results for each of the two or more first evaluation axes. Therefore, compared with the case where the combustion state is determined based on the evaluation result of one evaluation axis, the combustion state can be classified, predicted, and estimated with higher accuracy. In addition, since the combustion state is determined by mapping the evaluation results for each of the two or more first evaluation axes onto the combustion state determination map, it is possible to intuitively grasp the relationship between the evaluation results for each of the two or more first evaluation axes and the combustion state, which is the determination result, and it is also possible to quickly determine the combustion state.

[0022] The information processing apparatus according to the ninth aspect of the present invention is the information processing apparatus according to any one of the first to seventh aspects, and By using a second learned model obtained by machine learning on second teacher data generated by assigning classification labels on at least one second evaluation axis that is an element for determining a combustion state to visible light imaging data inside an incinerator imaged by a visible light imaging device, a second image analysis unit that performs evaluation on the second evaluation axis using new visible light imaging data inside the incinerator as an input is further provided. The combustion state determination unit determines the current combustion state by mapping the evaluation results on each of the first evaluation axis and the second evaluation axis onto a predetermined combustion state determination map having the first evaluation axis and the second evaluation axis as coordinate axes.

[0023] According to such an aspect, in addition to performing evaluation on the first evaluation axis that is an element for determining a combustion state by using infrared imaging data inside the incinerator, evaluation on the second evaluation axis that is an element for determining a combustion state is performed by using visible light imaging data inside the incinerator, and the combustion state is determined based on the evaluation results on each of the first evaluation axis and the second evaluation axis. Therefore, compared with the case of determining the combustion state based on the evaluation result on one evaluation axis, the combustion state can be classified, predicted, and estimated with higher accuracy. Further, by mapping the evaluation results on each of the first evaluation axis and the second evaluation axis onto the combustion state determination map to determine the combustion state, it is possible to intuitively grasp the relationship between the evaluation results on each of the first evaluation axis and the second evaluation axis and the combustion state that is the determination result, and it is also possible to quickly determine the combustion state.

[0024] The information processing apparatus according to the tenth aspect of the present invention is the information processing apparatus according to the ninth aspect, The second evaluation axis includes at least one of a flame state, a position of a combustion completion point, a shape of a combustion completion point, an amount of a combustible material, a quality of a combustible material, a type of a combustible material, an amount of unburned matter, and an amount of incineration ash.

[0025] The information processing apparatus according to the eleventh aspect of the present invention is the information processing apparatus according to any one of the first to tenth aspects, The combustion state determination unit corrects the combustion state determination result according to process data obtained from sensors installed in the facility and / or calculation amounts obtained from said process data.

[0026] An information processing device according to the 12th aspect of the present invention is an information processing device according to any of the 1st to 11th aspects, The combustion state determination unit displays or issues an alert according to the result of the combustion state determination.

[0027] An information processing device according to the 13th aspect of the present invention is an information processing device according to any of the 1st to 12th aspects, The system further includes an instruction unit that transmits operation instructions to the crane control device and / or combustion control device based on the determination result of the combustion state determination unit.

[0028] An information processing device according to the 14th aspect of the present invention is an information processing device according to any of the 1st to 13th aspects, The algorithms used in the aforementioned machine learning include at least one of the following: maximum likelihood classification, Boltzmann machines, neural networks (NN), support vector machines (SVM), Bayesian networks, sparse regression, decision trees, statistical estimation using random forests, boosting, reinforcement learning, and deep learning.

[0029] An information processing device according to the 15th aspect of the present invention is an information processing device according to any of the 1st to 14th aspects, The system further includes a model building unit that generates a first trained model by machine learning first training data, which is generated by assigning classification labels on a first evaluation axis that serves as an element for determining the combustion state to infrared imaging data inside an incinerator captured by an infrared imaging device.

[0030] A system according to a sixteenth aspect of the present invention is: An information processing device relating to the 13th aspect, A crane control device that controls a crane for stirring or transporting waste based on operation instructions transmitted from the information processing device, and / or a combustion control device that controls the combustion of waste in an incinerator, It is equipped with.

[0031] An information processing method according to the 17th aspect of the present invention is an information processing method performed by a computer, The combustion state is determined from infrared imaging data of the inside of the incinerator captured by an infrared imaging device. The steps include: using a first trained model, which is machine-trained on first training data generated by assigning classification labels to at least one first evaluation axis that constitutes an element, to perform evaluation on the first evaluation axis using new infrared imaging data from inside the incinerator as input; A step of determining the current combustion state based on the evaluation results on the first evaluation axis, Includes.

[0032] A program according to the 18th aspect of the present invention is provided for a computer, The process involves using a first trained model, which is created by machine learning from first training data generated by assigning classification labels on at least one first evaluation axis that serves as an element for determining the combustion state to infrared imaging data of the incinerator captured by an infrared imaging device, to perform an evaluation on the first evaluation axis using new infrared imaging data of the incinerator as input. A step of determining the current combustion state based on the evaluation results on the first evaluation axis, Make it run. [Effects of the Invention]

[0033] According to the present invention, the combustion state inside an incinerator can be classified, predicted, and estimated with greater accuracy. [Brief explanation of the drawing]

[0034] [Figure 1] Figure 1 is a schematic diagram showing the configuration of an incineration facility according to the first embodiment. [Figure 2]Figure 2 is a block diagram showing the configuration of an information processing device according to the first embodiment. [Figure 3] Figure 3 is a flowchart showing an example of an information processing method using the information processing device according to the first embodiment. [Figure 4] Figure 4 shows an example of a combustion state determination map. [Figure 5] Figure 5 is a schematic diagram showing the configuration of an incineration facility according to the second embodiment. [Figure 6] Figure 6 is a block diagram showing the configuration of an information processing device according to the second embodiment. [Figure 7] Figure 7 is a flowchart showing an example of an information processing method using the information processing device according to the second embodiment. [Figure 8] Figure 8 is a schematic diagram showing the configuration of an incineration facility according to the third embodiment. [Figure 9] Figure 9 is a block diagram showing the configuration of an information processing device according to the third embodiment. [Figure 10] Figure 10 is a flowchart showing an example of an information processing method using the information processing device according to the third embodiment. [Modes for carrying out the invention]

[0035] Embodiments of the present invention will be described in detail below with reference to the attached drawings. In the following description and the drawings used therein, the same reference numerals will be used for parts that can be identically configured, and redundant explanations will be omitted.

[0036] In this specification, "A and / or B" means either A or B, or both. In this specification, "visible light" means electromagnetic waves with wavelengths from 360 nm to 830 nm, and "infrared radiation" means electromagnetic waves with wavelengths from 830 nm to 14000 nm.

[0037] (First embodiment) Figure 1 is a schematic diagram showing the configuration of the incineration facility 1 according to the first embodiment.

[0038] As shown in Figure 1, the incineration facility 1 includes a platform 21 where transport vehicles (garbage trucks) 22 loaded with waste are parked, a waste pit 3 where waste is stored when it is fed in from the platform 21, and a crane 5 for agitating and transporting the waste stored in the waste pit 3. The facility includes a hopper 4 into which waste transported by crane 5 is fed, an incinerator 6 for burning the waste fed from hopper 4, and a waste heat boiler 2 for recovering waste heat from the exhaust gas generated in the incinerator 6. The type of incinerator 6 is not limited to a stoker furnace as shown in Figure 1, but also includes a fluidized bed furnace. Furthermore, the structure of the waste pit 3 is not limited to a single-stage pit as shown in Figure 1, but also includes a two-stage pit. The incineration facility 1 is also equipped with a crane control device 50 for controlling the operation of crane 5 and a combustion control device 20 for controlling the combustion of waste (material to be burned) in the incinerator 6.

[0039] The waste, loaded onto the transport vehicle 22, is fed from the platform 21 into the waste pit 3 and stored in the waste pit 3. The waste stored in the waste pit 3 is agitated by the crane 5 and then transported by the crane 5 to the hopper 4, which is then fed into the incinerator 6, where it is incinerated and processed.

[0040] As shown in Figure 1, the incineration facility 1 according to this embodiment is equipped with an infrared imaging device 71 and a visible light imaging device 72 for imaging the inside of the incinerator 6, and an information processing device 10 for determining the combustion state inside the incinerator 6.

[0041] In the illustrated example, the infrared imaging device 71 is installed above the downstream side of the incinerator 6. Multiple infrared imaging devices 71 may be installed above the downstream side of the incinerator 6. The installation location of the infrared imaging device 71 is not limited to above the downstream side of the incinerator 6; it may also be installed horizontally downstream or above or horizontally upstream of the incinerator 6. The infrared imaging device 71 has its imaging wavelength (e.g., 3.7 μm to 4.0 μm) appropriately selected in advance to cut out the flame wavelength (e.g., 4.1 μm to 4.5 μm), and is capable of acquiring infrared imaging data (moving image data) that removes the effects of the flame and includes images from the upstream side of the incinerator 6 to the image of the burning material actually being burned. In addition to the flame wavelength, the infrared imaging device 71 also captures the wavelengths of CO2 and water vapor. The imaging wavelength (e.g., 3.7 μm to 4.0 μm) is appropriately selected in advance to further cut out (e.g., 4.0 μm to 4.2 μm), and the infrared imaging data (moving image data) is obtained from the upstream side of the incinerator 6, with the effects of flame, CO2, and water vapor removed. The system may also be capable of acquiring images that include images of the burning material in action. The frame rate of the infrared imaging device 71 does not need to be particularly high; it may be a general frame rate (around 30fps) or a low frame rate (around 5-10fps).

[0042] In the illustrated example, the visible light imaging device 72 is installed above the downstream side of the incinerator 6, and is capable of acquiring visible light imaging data (video data) of the flame state (combustion state) of the materials being burned on the grate. Multiple visible light imaging devices 72 may be installed above the downstream side of the incinerator 6. The installation position of the visible light imaging device 72 is not limited to above the downstream side of the incinerator 6; the visible light imaging device 72 may be installed horizontally downstream of the incinerator 6, or above or horizontally upstream of the incinerator 6. The visible light imaging device 72 may be an RGB camera, a 3D camera, or an RGB-D camera, or a combination of two or more of these. The frame rate of the visible light imaging device 72 does not need to be particularly high; a general frame rate (around 30fps) or a low frame rate (around 5-10fps) may be used.

[0043] Next, the configuration of the information processing device 10 according to this embodiment will be described. Figure 2 is a block diagram showing the configuration of the information processing device 10. The information processing device 10 is composed of one or more computers.

[0044] As shown in Figure 2, the information processing device 10 includes a communication unit 11, a control unit 12, and a storage unit 13. Each unit is connected to the others via a bus or network so that they can communicate with each other.

[0045] Of these, the communication unit 11 is a communication interface between the infrared imaging device 71, the visible light imaging device 72, the crane control device 50, and the combustion control device 20, and the information processing device 10. The communication unit 11 transmits and receives information between the infrared imaging device 71, the visible light imaging device 72, the crane control device 50, and the combustion control device 20 and the information processing device 10.

[0046] The memory unit 13 is a non-volatile data storage device such as a hard disk or flash memory. The memory unit 13 stores various data handled by the control unit 12. The memory unit 12 also stores the following: the first algorithm 13a1 used for machine learning by the first model construction unit 12c1 (described later), the second algorithm 13a2 used for machine learning by the first model construction unit 12c1, the first imaging data 13b1 acquired by the first imaging data acquisition unit 12a1, the second imaging data 13b2 acquired by the second imaging data acquisition unit 12a2, the first training data 13c1 generated by the first training data generation unit 12b, the second training data 13c2 generated by the second training data generation unit 12b, and the combustion state determination map 13d used by the combustion state determination unit 12e. Details of each piece of information 13a1 to 13d will be described later.

[0047] The control unit 12 is a control means that performs various processing of the information processing device 10. As shown in Figure 2, the control unit 11 includes a first image data acquisition unit 12a1, a second image data acquisition unit 12b1, a first training data generation unit 12b1, a second training data generation unit 12b2, a first model construction unit 12c1, a second model construction unit 12c2, a first image analysis unit 12d1, a second image analysis unit 12d2, a combustion state determination unit 12e, an instruction unit 12f, and a process data acquisition unit 12g. Each of these units may be realized by a processor in the information processing device 10 executing a predetermined program, or it may be implemented in hardware.

[0048] The first imaging data acquisition unit 12a1 acquires the infrared imaging data of the inside of the incinerator 6 captured by the infrared imaging device 71 as the first imaging data 13b1. The first imaging data 13b1 is stored in the storage unit 13.

[0049] The second imaging data acquisition unit 12a2 acquires the visible light imaging data of the incinerator 6 captured by the visible light imaging device 72 as the second imaging data 13b2. The second imaging data 13b2 is stored in the storage unit 13.

[0050] The process data acquisition unit 12g acquires process data (for example, furnace outlet temperature and evaporation rate) measured by various sensors (not shown) installed in the incineration facility 1, and / or calculation data obtained from said process data. The process data and / or calculation data may be stored in the storage unit 13. The calculation data may be the difference between the measured value PV and the set value SV (the absolute value of the difference may also be used) or a predetermined period (for example, 1 minute, 10 minutes, 20 minutes). Moving average, maximum, minimum, median, integral, derivative, and standard deviation over a period of 1 hour. It may include one or more of the difference values ​​(which may also be the absolute values ​​of the difference values) from the value at a predetermined time prior.

[0051] The first training data generation unit 12b1 generates first training data 13c1 by associating the first imaging data 13b1 with classification labels on at least one first evaluation axis that serve as elements for determining the combustion state, which have been artificially assigned by the operator based on empirical rules. The first training data generation unit 12b1 divides the first imaging data 13b1 into multiple blocks (for example, a dry region, a gasification combustion region, and a burning region in the combustion process), and associates the data for each block with classification labels on at least one first evaluation axis that serve as elements for determining the combustion state, which have been artificially assigned by the operator based on empirical rules. The first training data 13c1 may be generated. The first training data 13c1 is stored in the storage unit 13. Here, the classification label may be absolute classification information on the first evaluation axis (a classification determined in relation to a predetermined absolute standard (threshold)) or a relative order on the first evaluation axis among multiple infrared imaging data (a relative order relationship within a group consisting of multiple infrared imaging data). The first evaluation axis may include at least one of the following: the amount of material to be burned, the quality of material to be burned, the type of material to be burned, the temperature of material to be burned, the amount of combustion gas generated (CO, etc.), and the temperature of the incinerator wall. The amount of material to be burned is a value corresponding to, for example, the volume, weight, density, cross-sectional area, etc. of the material to be burned. The quality of material to be burned is a value corresponding to, for example, the flammability, calorific value, moisture content, density, etc. of the material to be burned. The types of materials to be burned include, for example, unbroken garbage bags, paper waste, pruned branches, futons, sludge, bulky crushed waste, cardboard, burlap sacks, paper bags, bottom waste (waste located near the bottom of garbage pit 3 that is compressed into information waste and has a high moisture content), wood chips, textile waste, clothing waste, plastic waste, animal residue, animal carcasses, kitchen waste, vegetation, soil, medical waste, incinerator ash, bicycles, chests of drawers, beds, shelves, desks, chairs, agricultural vinyl, PET bottles, styrofoam, meat and bone meal, agricultural products, ceramics, glass scraps, metal scraps, rubble, concrete scraps, tatami mats, bamboo, straw, and activated carbon. The temperature of the materials to be burned is, for example, a value corresponding to the temperature of the materials to be burned. The amount of combustion gas generated is, for example, a value corresponding to the amount of gas generated during the combustion process, such as CO, hydrogen, hydrocarbons, NOx, SOx, HCl, and dioxins. The temperature of the incinerator walls is, for example, a value corresponding to the temperature of the walls and ceiling of the incinerator.

[0052] For example, if the amount of combustible material is used as the first evaluation axis, the classification labels may be large, normal, or small (absolute information) (the number of label types may be increased), or they may be a ranking of the magnitude of the combustible material amount for multiple first image data (relative information). As another example, if the quality of the combustible material is used as the evaluation axis, the classification labels may be ordinary waste, high-quality waste (i.e., high-calorie waste), or low-quality waste (i.e., low-calorie waste) (absolute information) (the number of label types may be increased), or they may be a ranking of the quality of the combustible material for multiple first image data (relative information). As yet another example, if the type of combustible material is used as the first evaluation axis, the classification labels may be unbroken garbage bags, pruned branches, or futons (absolute information) (the number of label types may be increased). As an alternative example, if the amount of combustion gas generated (CO, etc.) is used as the evaluation axis, the classification labels may be high, normal, or low (absolute information) (the number of label types may be increased), or they may be a ranking of the magnitude of combustion gas generated (CO, etc.) for multiple first imaging data (relative information). As an alternative example, if the temperature of the incinerator wall is used as the first evaluation axis, the classification labels may be high, normal, or low (the number of label types may be increased), or they may be a ranking of the high or low temperatures of the incinerator wall for multiple first imaging data (relative information).

[0053] The first training data generation unit 12b1 may generate first training data 13c1 by associating the first imaging data 13b1 with a combination of process data obtained from various sensors (not shown) installed in the incineration facility 1 and / or computational amounts obtained by calculation from said process data, and assigning classification labels on the first evaluation axis, which are artificially assigned by the operator based on empirical rules. Here, the process data is data that takes into account the time lag between the imaging time of the first imaging data and the response time of the sensor. The time lag may be appropriately determined based on the response speed of the sensor, the installation location of the sensor in the facility, experiments, simulations, and the operator's empirical rules.

[0054] The second training data generation unit 12b2 generates second training data 13c2 by associating the second imaging data 13b1 with classification labels on at least one second evaluation axis that serve as elements for determining the combustion state, which have been artificially assigned by the operator based on empirical rules. The second training data generation unit 12b2 divides the second imaging data 13b1 into multiple blocks (for example, a dry region, a gasification combustion region, and a burning region in the combustion process), and blocks Second training data 13c2 may be generated by associating the unit data with at least one classification label on a first evaluation axis, which is an element for determining the combustion state and is artificially assigned by the operator based on empirical rules. The second training data 13c2 is stored in the storage unit 13. Here, the classification label may be absolute classification information on the second evaluation axis (a classification determined in relation to a predetermined absolute standard (threshold)) or a relative order on the second evaluation axis among multiple visible light imaging data (a relative order relationship within a group consisting of multiple visible light imaging data). The second evaluation axis may include at least one of the following: flame state, location of the combustion completion point (burnout point), shape of the combustion completion point, amount of material to be burned, quality of material to be burned, type of material to be burned, amount of unburned material, and amount of incinerated ash. Flame state is a value corresponding to, for example, the strength of the flame or the brightness of the flame. Location of the combustion completion point is a value corresponding to, for example, the position on the grate that constitutes the incinerator. The shape of the combustion completion point is a value that corresponds to, for example, the curvature or radius of curvature of the combustion completion point. The quantity of the material to be burned is a value that corresponds to, for example, the volume, weight, density, or cross-sectional area of ​​the material to be burned. The quality of the material to be burned is a value that corresponds to, for example, the flammability, calorific value, moisture content, or density of the material to be burned. The types of material to be burned include, for example, unbroken garbage bags, paper waste, pruned branches, futons, sludge, bulky crushed waste, cardboard, burlap sacks, paper bags, bottom waste (waste located near the bottom of garbage pit 3 that is compressed into information waste and has a high moisture content), wood chips, textile waste, clothing waste, plastic waste, animal residue, animal carcasses, kitchen waste, vegetation, soil, medical waste, incinerator ash, bicycles, chests of drawers, beds, shelves, desks, chairs, agricultural vinyl, PET bottles, styrofoam, meat and bone meal, agricultural products, ceramics, glass scraps, metal scraps, rubble, concrete scraps, tatami mats, bamboo, straw, and activated carbon. The amount of unburned material refers to values ​​such as volume, weight, density, cross-sectional area, and quantity of the unburned material. The amount of incinerated ash refers to values ​​such as volume, weight, density, and cross-sectional area of ​​the incinerated ash.

[0055] For example, if the flame state is used as the second evaluation axis, the classification labels may be good, normal, or bad (absolute information) (the number of label types may be increased), or they may be a ranking of the superiority or inferiority of the flame state for multiple (for example, two) second image data (relative information). As another example, if the position of the combustion completion point (burnout point) is used as the evaluation axis, the classification labels may be near, slightly near, normal, slightly far, or far (the number of label types may be increased), or they may be a ranking of the proximity of the combustion completion point for multiple second image data. As yet another example, if the shape of the combustion completion point is used as the evaluation axis, the classification labels may be good, normal, or bad (the number of label types may be increased), or they may be a ranking of the superiority or inferiority of the shape of the combustion completion point for multiple second image data. As yet another example, if the amount of burned material is used as the second evaluation axis, the classification labels may be a lot, normal, or little (the number of label types may be increased), or they may be a ranking of the magnitude of the burned material for multiple second image data. As an alternative example, if the quality of the material being burned is used as the second evaluation axis, the classification labels may be general waste, high-quality waste (i.e., high-calorie waste), and low-quality waste (i.e., low-calorie waste) (the number of label types may be increased), or the quality of the material being burned may be ranked relative to multiple second imaging data. As an alternative example, if the type of material being burned is used as the second evaluation axis, the classification labels may be unbroken garbage bags, pruned branches, and futons (the number of label types may be increased). As an alternative example, if the amount of unburned material is used as the second evaluation axis, the classification labels may be high, normal, and low (the number of label types may be increased), or the amount of unburned material may be ranked relative to multiple second imaging data. As an alternative example, if the amount of incinerated ash is used as the second evaluation axis, the classification labels may be high, normal, and low (the number of label types may be increased), or the amount of incinerated ash may be ranked relative to multiple second imaging data.

[0056] The second training data generation unit 12b2, based on the combination of the second imaging data 13b2 and process data obtained from various sensors (not shown) installed in the incineration facility 1 and / or computational amounts obtained by calculation from said process data, uses the operator's empirical rules to determine the results. Second training data 13c2 may be generated by linking classification labels on a second evaluation axis that have been artificially assigned. Here, the process data is data that takes into account the time lag between the acquisition time of the second imaging data and the response time of the sensor. The time lag may be appropriately determined based on the response speed of the sensor, the installation location of the sensor in the facility, experiments, simulations, and the operator's empirical rules.

[0057] The first imaging data 13b1 and / or the second imaging data 13b2 may be video data of 60 seconds or less, or video data of 30 seconds or less. According to the inventors' findings, the combustion state inside the incinerator 6 changes every 5 to 10 seconds. Therefore, by using, for example, video data of 60 seconds or less (or 30 seconds or less) as one unit for the first imaging data 13b1 and / or the second imaging data 13b2, it becomes possible to classify, predict, and estimate the combustion state inside the incinerator 6 with greater accuracy.

[0058] Furthermore, the first imaging data 13b1 and / or the second imaging data 13b2 may be video data lasting 5 seconds or longer, or 7 seconds or longer. While the combustion state inside the incinerator 6 changes every 5 to 10 seconds, by using video data lasting, for example, 5 seconds or longer (or 7 seconds or longer) as a single unit for the first imaging data 13b1 and / or the second imaging data 13b2, it becomes possible to classify, predict, and estimate the combustion state inside the incinerator 6 with greater accuracy.

[0059] The first model building unit 13c1 generates a first trained model by performing machine learning on the first training data 13b1 generated by the first training data generation unit 12b1 using the first algorithm 13a1. The first algorithm 13a1 used for machine learning may include at least one of the following: maximum likelihood classification, Boltzmann machine, neural network (NN), support vector machine (SVM), Bayesian network, sparse regression, decision tree, statistical estimation using random forest, boosting, reinforcement learning, and deep learning.

[0060] Similarly, the second model building unit 13c2 generates a second trained model by machine learning the second training data 13b2 generated by the second training data generation unit 12b2 using the second algorithm 13a2. The second algorithm 13a2 used for machine learning may include at least one of the following: maximum likelihood classification, Boltzmann machine, neural network (NN), support vector machine (SVM), Bayesian network, sparse regression, decision tree, statistical estimation using random forest, boosting, reinforcement learning, or deep learning.

[0061] The first image analysis unit 13d1 takes the new first imaging data (infrared imaging data) from inside the incinerator 6 acquired by the first imaging data acquisition unit 13a1 as input and uses the first trained model generated by the first model construction unit 13c1 to acquire the evaluation result on the first evaluation axis (for example, the amount of material to be burned) as output data. The first image analysis unit 13d1 may also normalize (score) the evaluation result on the first evaluation axis to a numerical range of, for example, 0 to 100 and acquire a numerical value (score) within that range as output data.

[0062] The second image analysis unit 13d2 takes the new second imaging data (visible light imaging data) of the incinerator 6 acquired by the second imaging data acquisition unit 13a2 as input and uses the second trained model generated by the second model construction unit 13c2 to acquire the evaluation result on the second evaluation axis (for example, flame state) as output data. The second image analysis unit 13d2 may also normalize (score) the evaluation result on the second evaluation axis to a numerical range of, for example, 0 to 100 and acquire a numerical value (score) within that range as output data.

[0063] The combustion state determination unit 12e determines the current combustion state in the incinerator 6 based on the evaluation results (score) on the first evaluation axis by the first image analysis unit 13d1 and the evaluation results (score) on the second evaluation axis by the second image analysis unit 13d2. The combustion state determination result may be a label indicating the characteristics of the combustion state (e.g., over-combustion, thick waste layer combustion, thin waste layer combustion, depleted waste, low-quality waste combustion), or it may be a numerical change of the said label. Over-combustion refers to a combustion state in which combustion is active, for example, when a large amount of high-quality (easily combustible) waste is present in the material to be burned, resulting in a very high combustion temperature. Thick waste layer combustion refers to a combustion state in which the amount of material to be burned is greater than normal. Thin waste layer combustion refers to a combustion state in which the amount of material to be burned is less than normal. Depleted waste refers to a combustion state in which the amount of material to be burned is extremely small compared to normal, resulting in a low combustion temperature. Low-quality waste combustion refers to a combustion state where, for example, a large amount of low-quality waste (such as low-calorie waste) is present in the material being burned, resulting in a low combustion temperature and other conditions of weak combustion.

[0064] The combustion state determination unit 12e may determine the current combustion state in the incinerator 6 by mapping a pair (X,Y) consisting of an evaluation result (X) on the first evaluation axis and an evaluation result (Y) on the second evaluation axis onto a predetermined combustion state determination map 13d where the first evaluation axis is the X coordinate axis and the second evaluation axis is the Y coordinate axis. The combustion state determination unit 12e may also determine the current combustion state in the incinerator 6 by mapping a pair (X,Y,Z,··) consisting of one or more evaluation results on the first evaluation axis and one or more evaluation results on the second evaluation axis onto a predetermined N-dimensional space (N is an integer of 3 or more) combustion state determination map 13d where, for example, one of the first evaluation axes is the X coordinate axis, another of the first evaluation axes is the Y coordinate axis, and one of the second evaluation axes is the Z coordinate axis.

[0065] Figure 4 shows an example of a combustion state determination map 13d. In the combustion state determination map 13d shown in Figure 4, the XY coordinate plane, with the amount of combustible material on the X axis and the flame state on the Y axis, is divided into multiple regions (in the illustrated example, six regions: a depleted waste zone, an over-combustion zone, a thin waste layer combustion zone, a good (normal) combustion zone, a thick waste layer combustion zone, and a low-quality waste combustion zone). The combustion state determination unit 12e maps a pair (X,Y) consisting of the evaluation result on the first evaluation axis (X) and the evaluation result on the second evaluation axis (Y), where the amount of combustible material is the first evaluation axis and the flame state is the second evaluation axis, onto this combustion state determination map 13d. For example, as shown in Figure 4, if point P, which represents the pair (X,Y) consisting of the evaluation result on the first evaluation axis (X) and the evaluation result on the second evaluation axis (Y), is mapped within the good (normal) combustion zone, the combustion state determination unit 12e determines that the current combustion state is "good (normal) combustion". The range of the above region (threshold (position, length, and shape of the boundary line)) may be determined by the operator's empirical rules, or it may be determined by machine learning using an algorithm that includes at least one of the following: maximum likelihood classification, Boltzmann machine, neural network (NN), support vector machine (SVM), Bayesian network, sparse regression, decision tree, statistical estimation using random forest, Kalman filter, autoregression, boosting, reinforcement learning, or deep learning, to determine the correspondence between the evaluation results on the first evaluation axis (X) and the evaluation results on the second evaluation axis (Y) and the process data.

[0066] The combustion state determination unit 12e may correct the combustion state determination result according to process data obtained from various sensors (not shown) installed in the facility and / or calculation amounts obtained by calculation from said process data. For example, if the combustion state determination result is "low-grade waste combustion" and the difference value PV-SV of the furnace outlet temperature or evaporation amount at that time meets predetermined conditions, the combustion state determination result may be corrected to "good combustion (normal)".

[0067] The combustion state determination unit 12e may display an alert on a display (not shown) or an audio alert, depending on the combustion state determination result, indicating that it is difficult to keep the performance management index value within a predetermined range. The alarm may be triggered by vibration, light, or other means.

[0068] The instruction unit 12f transmits operation instructions to the crane control device 50 and / or combustion control device 20 based on the determination result of the combustion state determination unit 12e. For example, the instruction unit 12f transmits predefined operation instructions for each zone on the combustion state determination map to the crane control device 50 and / or combustion control device 20 based on the determination result of the combustion state determination unit 12e. Specifically, for example, the instruction unit 12f transmits operation instructions to the combustion control device 20 that are appropriate for the current combustion state (increase or decrease combustion temperature, increase or decrease combustion time, increase or decrease air volume, increase or decrease waste feed volume). As an example, if the current combustion state is "over-combustion", the instruction unit 12f may transmit operation instructions to the combustion control device 20 such as decreasing the air volume or decreasing the waste feed volume. As another example, if the current combustion state is "waste layer thickness combustion", the instruction unit 12f may transmit operation instructions to the combustion control device 20 such as decreasing the waste feed volume in order to thin the waste layer. As another example, if the current combustion state is "thin waste layer combustion," the instruction unit 12f may send operational instructions to the combustion control device 20, such as reducing the amount of air or reducing the amount of waste being fed, in order to thicken the waste layer. As yet another example, if the current combustion state is "waste depletion," the instruction unit 12f may send operational instructions to the combustion control device 20, such as reducing the amount of air or increasing the amount of waste being fed, in order to increase the amount of waste entering the incinerator. As yet another example, if the current combustion state is "low-quality waste combustion," the instruction unit 12f may send operational instructions to the combustion control device 20, such as increasing the amount of air or decreasing the amount of waste being fed, in order to thoroughly burn the existing waste.

[0069] Next, an example of an information processing method using the information processing device 10 with the above configuration will be described. Figure 3 is a flowchart of an example of an information processing method.

[0070] As shown in Figure 3, first, before the incineration facility 1 is put into operation, the first training data generation unit 12b1 generates first training data 13c1 by associating classification labels on at least one first evaluation axis, which are elements for determining the combustion state and have been artificially assigned by the operator based on empirical rules, with past infrared imaging data (first imaging data 13b1) of the incinerator 6 captured by the infrared imaging device 71. Then, the second training data generation unit 12b2 generates second training data 13c2 by associating classification labels on at least one second evaluation axis, which are elements for determining the combustion state and have been artificially assigned by the operator based on empirical rules, with past visible light imaging data (second imaging data 13b2) of the incinerator 6 captured by the visible light imaging device 72 (step S11).

[0071] Next, the first model construction unit 13c1 generates a first trained model by machine learning the first training data 13b1 generated by the first training data generation unit 12b1 using the first algorithm 13a1. Then, the second model construction unit 13c2 generates a second trained model by machine learning the second training data 13b2 generated by the second training data generation unit 12b2 using the second algorithm 13a2 (step S12).

[0072] Next, while the incineration facility 1 is in operation, the first imaging data acquisition unit 12a1 acquires new infrared imaging data of the incinerator 6 captured by the infrared imaging device 71 as the first imaging data 13b1, and the second imaging data acquisition unit 12a2 acquires visible light imaging data of the incinerator 6 captured by the visible light imaging device 72 as the second imaging data 13b2 (step S13).

[0073] Next, the first image analysis unit 13d1 takes the new first imaging data (infrared imaging data) of the incinerator 6 acquired by the first imaging data acquisition unit 13a1 as input and uses the first trained model generated by the first model construction unit 13c1 to acquire the evaluation result on the first evaluation axis (for example, the amount of material to be burned) as output data. In addition, the second image analysis unit 13d2 takes the new second imaging data (infrared imaging data) of the incinerator 6 acquired by the second imaging data acquisition unit 13a2 as input and uses the first trained model generated by the first model construction unit 13c1 to acquire the evaluation result on the first evaluation axis (for example, the amount of material to be burned) as output data. Using the optical imaging data as input, the second trained model generated by the second model construction unit 13c2 is used to obtain the evaluation result on the second evaluation axis (for example, flame state) as output data (step S14).

[0074] Then, the combustion state determination unit 12e determines the current combustion state inside the incinerator 6 based on the evaluation results on the first evaluation axis by the first image analysis unit 13d1 and the evaluation results on the second evaluation axis by the second image analysis unit 13d2 (step 15).

[0075] In step S15, the combustion state determination unit 12e may determine the current combustion state in the incinerator 6 by mapping a pair (X,Y) consisting of the evaluation result (X) on the first evaluation axis and the evaluation result (Y) on the second evaluation axis onto a predetermined combustion state determination map 13d where the first evaluation axis is the X coordinate axis and the second evaluation axis is the Y coordinate axis (see Figure 4).

[0076] Subsequently, the instruction unit 12f transmits an operation instruction to the crane control device 50 and / or the combustion control device 20 based on the determination result of the combustion state determination unit 12e (step S16).

[0077] According to this embodiment, the infrared imaging device 71 can acquire an image as infrared imaging data (first imaging data) of the incinerator 6, including an image of the burning material actually burning, from the upstream side of the incinerator 6 with the effects of the flame removed, by appropriately selecting the imaging wavelength. By using the infrared imaging data of the incinerator 6 to perform an evaluation on at least one first evaluation axis that is an element for determining the combustion state, and determining the combustion state based on the evaluation result, the combustion state inside the incinerator 6 can be classified, predicted, and estimated with greater accuracy compared to the case where the classification, prediction, and estimation of the combustion state and the prediction and estimation of process values ​​are performed only from combustion images where only the situation of the flame can be grasped.

[0078] Furthermore, according to this embodiment, by determining the combustion state based on the evaluation results of the first evaluation axis X and the second evaluation axis Y, the combustion state inside the incinerator 6 can be classified, predicted, and estimated with greater accuracy compared to the case where the combustion state is determined based on the evaluation result of a single evaluation axis. In addition, by determining the combustion state by mapping the evaluation results of the first evaluation axis X and the second evaluation axis Y onto the combustion state determination map 13d, it is possible to intuitively grasp the relationship between the evaluation results of the first evaluation axis X and the second evaluation axis Y and the determined combustion state, and it is also possible to determine the combustion state at high speed.

[0079] (Second embodiment) Next, a second embodiment will be described with reference to Figures 5 to 7. Figure 5 is a schematic diagram showing the configuration of the incineration facility 1 according to the second embodiment.

[0080] As shown in Figure 5, the incineration facility 1 according to the second embodiment differs from the first embodiment only in that the visible light imaging device 72 for imaging the inside of the incinerator 6 is omitted; the other configurations are the same as those of the first embodiment.

[0081] Figure 6 is a block diagram showing the configuration of the information processing device 10 according to the second embodiment. The information processing device 10 is composed of one or more computers.

[0082] As shown in Figure 6, the information processing device 10 includes a communication unit 11, a control unit 12, and a storage unit 13. Each unit is connected to the others via a bus or network so that they can communicate with each other.

[0083] Of these, the communication unit 11 is a communication interface between the infrared imaging device 71, the crane control device 50, and the combustion control device 20 and the information processing device 10. The communication unit 11 communicates information between the infrared imaging device 71, the crane control device 50, and the combustion control device 20. Information is sent and received between the processing unit 10 and the device.

[0084] The memory unit 13 is a non-volatile data storage device such as a hard disk or flash memory. The memory unit 13 stores various types of data handled by the control unit 12. The memory unit 12 also stores the first algorithm 13a1 used for machine learning by the first model construction unit 12c1 (described later), the first imaging data 13b1 acquired by the first imaging data acquisition unit 12a1, the first training data 13c1 generated by the first training data generation unit 12b1, and the combustion state determination map 13d used by the combustion state determination unit 12e. Details of each piece of information 13a1 to 13d will be described later.

[0085] The control unit 12 is a control means that performs various processing of the information processing device 10. As shown in Figure 6, the control unit 11 includes a first imaging data acquisition unit 12a1, a first training data generation unit 12b1, a first model construction unit 12c1, a first image analysis unit 12d1, a combustion state determination unit 12e, an instruction unit 12f, and a process data acquisition unit 12g. Each of these units may be realized by a processor in the information processing device 10 executing a predetermined program, or it may be implemented in hardware.

[0086] The configurations of the first imaging data acquisition unit 12a1, the process data acquisition unit 12g, and the instruction unit 12f are the same as in the first embodiment, and therefore their explanation is omitted.

[0087] The first training data generation unit 12b1 generates first training data 13c1 by associating the first imaging data 13b1 with classification labels for each of two or more first evaluation axes that serve as elements for determining the combustion state, which have been artificially assigned by the operator based on empirical rules. The first training data 13c1 is stored in the storage unit 13. Here, the classification labels may be absolute classification information on the first evaluation axis (classification determined in relation to a predetermined absolute standard (threshold)) or relative order on the first evaluation axis among multiple infrared imaging data (relative order relationship within a group consisting of multiple infrared imaging data). The first evaluation axis may include two or more of the following: flame state, amount of material to be burned, quality of material to be burned, type of material to be burned, temperature of material to be burned, amount of combustion gas generated (CO, etc.), and temperature of the incinerator wall. The flame state can be determined from the amount of gaseous fluctuations captured in the first imaging data 13b1 (infrared imaging data).

[0088] For example, if the flame state is used as one of the primary evaluation axes, the classification labels may be good, normal, or bad (absolute information) (the number of label types may be increased), or they may be a ranking of the superiority or inferiority of the flame state for multiple first imaging data (relative information). As another example, if the amount of burned material is used as one of the primary evaluation axes, the classification labels may be a lot, normal, or little (absolute information) (the number of label types may be increased), or they may be a ranking of the magnitude of the burned material for multiple first imaging data (relative information). As yet another example, if the quality of the burned material is used as one of the primary evaluation axes, the classification labels may be ordinary waste, high-quality waste (i.e., high-calorie waste), or low-quality waste (i.e., low-calorie waste) (absolute information) (the number of label types may be increased), or they may be a ranking of the superiority or inferiority of the burned material quality for multiple first imaging data (relative information). As an alternative example, if the type of material being burned is used as one of the primary evaluation axes, the classification labels may be garbage bags, unbroken bags, garbage, pruned branches, futons (absolute information) (the number of label types may be increased). As an alternative example, if the amount of combustion gas generated (CO, etc.) is used as one of the primary evaluation axes, the classification labels may be high, normal, low (absolute information) (the number of label types may be increased), or the order of the magnitude of combustion gas generated (CO, etc.) for multiple primary imaging data (relative information). As an alternative example, if the temperature of the incinerator wall is used as one of the primary evaluation axes, the classification labels may be high, normal, low (the number of label types may be increased), or the order of the high and low temperatures of the incinerator wall for multiple primary imaging data may be increased. (Relative information) is also acceptable.

[0089] The first training data generation unit 12b1 may generate first training data 13c1 by associating the first imaging data 13b1 with a combination of process data obtained from various sensors (not shown) installed in the incineration facility 1 and / or computational amounts obtained by calculation from said process data, and with classification labels for each of two or more first evaluation axes that have been artificially assigned by the operator based on empirical rules. Here, the process data is data that takes into account the time lag between the imaging time of the first imaging data and the response time of the sensor. The time lag may be appropriately determined based on the response speed of the sensor, the installation location of the sensor in the facility, experiments, simulations, and the operator's empirical rules.

[0090] The first imaging data 13b1 may be video data of 60 seconds or less, or video data of 30 seconds or less. According to the inventors' findings, the combustion state inside the incinerator 6 changes every 5 to 10 seconds. Therefore, by using, for example, video data of 60 seconds or less (or 30 seconds or less) as one unit for the first imaging data 13b1, it becomes possible to classify, predict, and estimate the combustion state inside the incinerator 6 with greater accuracy.

[0091] Furthermore, the first imaging data 13b1 may be video data lasting 5 seconds or longer, or 7 seconds or longer. While the combustion state inside the incinerator 6 changes every 5 to 10 seconds, by using, for example, video data lasting 5 seconds or longer (or 7 seconds or longer) as one segment for the first imaging data 13b1, it becomes possible to classify, predict, and estimate the combustion state inside the incinerator 6 with greater accuracy.

[0092] The first model building unit 13c1 generates a first trained model by performing machine learning on the first training data 13b1 generated by the first training data generation unit 12b1 using the first algorithm 13a1. The first algorithm 13a1 used for machine learning may include at least one of the following: maximum likelihood classification, Boltzmann machine, neural network (NN), support vector machine (SVM), Bayesian network, sparse regression, decision tree, statistical estimation using random forest, boosting, reinforcement learning, and deep learning.

[0093] The first image analysis unit 13d1 takes the new first imaging data (infrared imaging data) inside the incinerator 6 acquired by the first imaging data acquisition unit 13a1 as input and uses the first trained model generated by the first model construction unit 13c1 to acquire evaluation results for each of two or more first evaluation axes (for example, flame state and amount of burned material) as output data. The first image analysis unit 13d1 may also normalize (score) the evaluation results for each of the two or more first evaluation axes to a numerical range of, for example, 0 to 100 and acquire the numerical value (score) within that range as output data.

[0094] The combustion state determination unit 12e determines the current combustion state in the incinerator 6 based on the evaluation results (scores) from each of the two or more first evaluation axes performed by the first image analysis unit 13d1. The combustion state determination result may be a label indicating the characteristics of the combustion state (e.g., over-combustion, thick waste layer combustion, thin waste layer combustion, waste depletion, low-quality waste combustion), or it may be a numerical version of the said label.

[0095] The combustion state determination unit 12e may determine the current combustion state in the incinerator 6 by mapping a set of evaluation results from each of two or more first evaluation axes onto a predetermined combustion state determination map 13d that uses the two or more first evaluation axes as coordinate axes.

[0096] Figure 4 shows an example of a combustion state determination map 13d. The combustion state determination map shown in Figure 4 In map 13d, an XY coordinate plane, with the amount of combustible material on the X axis and the flame state on the Y axis, is divided into multiple regions (in the illustrated example, six regions: a depleted waste zone, an over-combustion zone, a thin waste layer combustion zone, a good (normal) combustion zone, a thick waste layer combustion zone, and a low-quality waste combustion zone). The combustion state determination unit 12e maps a pair (X,Y) consisting of evaluation results for each of the two different first evaluation axes, where the amount of combustible material and the flame state are used as the first evaluation axes, onto this combustion state determination map 13d. For example, as shown in Figure 4, if point P, which represents a pair (X,Y) consisting of an evaluation result (X) when the amount of combustible material is the first evaluation axis and an evaluation result (Y) when the flame state is the first evaluation axis, is mapped within the good (normal) combustion zone, the combustion state determination unit 12e determines that the current combustion state is "good (normal) combustion".

[0097] The combustion state determination unit 12e may correct the combustion state determination result according to process data obtained from various sensors (not shown) installed in the facility and / or calculation amounts obtained by calculation from said process data. For example, if the combustion state determination result is "low-grade waste combustion" and the difference value PV-SV of the furnace outlet temperature or evaporation amount at that time meets predetermined conditions, the combustion state determination result may be corrected to "good combustion (normal)".

[0098] The combustion state determination unit 12e may, depending on the combustion state determination result, display an alert on a display (not shown) indicating that it is difficult to keep the performance management index value within a predetermined range, or it may emit an alert by sound, vibration, light, etc.

[0099] Next, an example of an information processing method using the information processing device 10 with the above configuration will be described. Figure 7 is a flowchart of an example of an information processing method.

[0100] As shown in Figure 7, first, before the incineration facility 1 is put into operation, the first training data generation unit 12b1 generates the first training data 13c1 by associating classification labels on two or more first evaluation axes, which are elements for determining the combustion state and which have been artificially assigned by the operator based on empirical rules, with past infrared imaging data (first imaging data 13b1) taken inside the incinerator 6 by the infrared imaging device 71 (step S21).

[0101] Next, the first model construction unit 13c1 generates a first trained model by machine learning the first training data 13b1 generated by the first training data generation unit 12b1 using the first algorithm 13a1 (step S22).

[0102] Next, while the incineration facility 1 is in operation, the first imaging data acquisition unit 12a1 acquires new infrared imaging data of the inside of the incinerator 6 captured by the infrared imaging device 71 as the first imaging data 13b1 (step S23).

[0103] Next, the first image analysis unit 13d1 takes the new first imaging data (infrared imaging data) from inside the incinerator 6 acquired by the first imaging data acquisition unit 13a1 as input and uses the first trained model generated by the first model construction unit 13c1 to acquire evaluation results for each of two or more first evaluation axes (for example, amount of material to be burned and flame state) as output data (step S24).

[0104] Then, the combustion state determination unit 12e determines the current combustion state inside the incinerator 6 based on the evaluation result on the first evaluation axis by the first image analysis unit 13d1 (step 25).

[0105] In step S25, the combustion state determination unit 12e determines a set of evaluation results from each of the two or more first evaluation axes, using the two or more first evaluation axes as coordinate axes, based on a predetermined combustion state determination unit. The current combustion state inside the incinerator 6 can be determined by mapping it onto the fixed map 13d (see Figure 4).

[0106] Subsequently, the instruction unit 12f transmits an operation instruction to the crane control device 50 and / or the combustion control device 20 based on the determination result of the combustion state determination unit 12e (step S16).

[0107] According to this embodiment, the infrared imaging device 71 can acquire an image as infrared imaging data (first imaging data) of the incinerator 6, including an image of the material being burned, from the upstream side of the incinerator 6 with the effects of the flame removed, by appropriately selecting the imaging wavelength. By using the infrared imaging data of the incinerator 6 to perform evaluations on two or more first evaluation axes that are elements for determining the combustion state, and determining the combustion state based on the evaluation results, the combustion state inside the incinerator 6 can be classified, predicted, and estimated with greater accuracy compared to classifying, predicting, and estimating the combustion state and process values ​​from combustion images that only show the flame situation.

[0108] Furthermore, according to this embodiment, by determining the combustion state based on the evaluation results of each of the two or more first evaluation axes, the combustion state inside the incinerator 6 can be classified, predicted, and estimated with greater accuracy compared to the case where the combustion state is determined based on the evaluation results of only one evaluation axis. In addition, by determining the combustion state by mapping the evaluation results of each of the two or more first evaluation axes onto the combustion state determination map 13d, it is possible to intuitively grasp the relationship between the evaluation results of each of the two or more first evaluation axes and the determined combustion state, and it is also possible to determine the combustion state at high speed.

[0109] (Third embodiment) Next, a third embodiment will be described with reference to Figures 8-10. Figure 8 is a schematic diagram showing the configuration of the incineration facility 1 according to the third embodiment.

[0110] As shown in Figure 8, the incineration facility 1 according to the third embodiment differs from the first embodiment only in that the infrared imaging device 71 for imaging the inside of the incinerator 6 is omitted; the other configurations are the same as those of the first embodiment.

[0111] Figure 9 is a block diagram showing the configuration of the information processing device 10 according to the third embodiment. The information processing device 10 is composed of one or more computers.

[0112] As shown in Figure 9, the information processing device 10 includes a communication unit 11, a control unit 12, and a storage unit 13. Each unit is connected to the others via a bus or network so that they can communicate with each other.

[0113] Of these, the communication unit 11 is a communication interface between the visible light imaging device 72, the crane control device 50, and the combustion control device 20, and the information processing device 10. The communication unit 11 transmits and receives information between the visible light imaging device 72, the crane control device 50, and the combustion control device 20 and the information processing device 10.

[0114] The memory unit 13 is a non-volatile data storage device such as a hard disk or flash memory. The memory unit 13 stores various data handled by the control unit 12. The memory unit 12 also stores the second algorithm 13a2 used for machine learning by the second model construction unit 12c2 (described later), the second imaging data 13b2 acquired by the second imaging data acquisition unit 12a2, the second training data 13c2 generated by the second training data generation unit 12b2, and the combustion state determination map 13d used by the combustion state determination unit 12e. Details of each piece of information 13a2 to 13d will be described later.

[0115] The control unit 12 is a control means that performs various processing of the information processing device 10. As shown in Figure 9, the control unit 11 includes a second imaging data acquisition unit 12a2, a second training data generation unit 12b2, a second model construction unit 12c2, a second image analysis unit 12d2, a combustion state determination unit 12e, an instruction unit 12f, and a process data acquisition unit 12g. Each of these units may be realized by a processor in the information processing device 10 executing a predetermined program, or it may be implemented in hardware.

[0116] The configurations of the second imaging data acquisition unit 12a2, the process data acquisition unit 12g, and the instruction unit 12f are the same as in the first embodiment, and therefore their explanation will be omitted.

[0117] The second training data generation unit 12b2 generates second training data 13c2 by associating classification labels for each of two or more second evaluation axes, which are elements for determining the combustion state and which are artificially assigned by the operator based on empirical rules, with the second imaging data 13b2. The second training data 13c2 is stored in the storage unit 13. Here, the classification labels may be absolute classification information on the second evaluation axis (classification determined in relation to a predetermined absolute standard (threshold)) or relative order on the second evaluation axis among multiple visible light imaging data (relative order relationship within a group consisting of multiple visible light imaging data). The second evaluation axis may include two or more of the following: flame state, location of the combustion completion point, shape of the combustion completion point, amount of burned material, quality of burned material, type of burned material, amount of unburned material, and amount of incinerated ash.

[0118] For example, if the flame state is used as one of the second evaluation axes, the classification labels may be good, normal, or bad (absolute information) (the number of label types may be increased), or they may be a ranking of the superiority or inferiority of the flame state for multiple second imaging data (relative information). As another example, if the position of the combustion completion point (burnout point) is used as one of the second evaluation axes, the classification labels may be near, slightly near, normal, slightly far, or far (the number of label types may be increased), or they may be a ranking of the proximity of the combustion completion point position for multiple second imaging data. As yet another example, if the shape of the combustion completion point is used as one of the second evaluation axes, the classification labels may be good, normal, or bad (the number of label types may be increased), or they may be a ranking of the superiority or inferiority of the combustion completion point shape for multiple second imaging data. As an alternative example, if the amount of combustible material is used as one of the second evaluation axes, the classification labels may be large, normal, or small (absolute information) (the number of label types may be increased), or they may be a ranking of the magnitude of the combustible material amount for multiple second imaging data (relative information). As an alternative example, if the quality of the combustible material is used as one of the second evaluation axes, the classification labels may be ordinary waste, high-quality waste (i.e., high-calorie waste), or low-quality waste (i.e., low-calorie waste) (the number of label types may be increased), or they may be a ranking of the quality of the combustible material for multiple second imaging data. As an alternative example, if the type of combustible material is used as one of the second evaluation axes, the classification labels may be unbroken garbage bags, pruned branches, or futons (the number of label types may be increased). As an alternative example, if the amount of uncombustible material is used as the second evaluation axis, the classification labels may be large, normal, or small (the number of label types may be increased), or they may be a ranking of the magnitude of the uncombustible material amount for multiple second imaging data. As an alternative example, if the amount of incinerated ash is used as one of the second evaluation axes, the classification labels may be large, normal, or small (the number of label types may be increased), or the amount of incinerated ash may be ordered relative to the amount of multiple second imaging data.

[0119] The second training data generation unit 12b2 may generate second training data 13c2 by associating the second imaging data 13b2 with a combination of process data obtained from various sensors (not shown) installed in the incineration facility 1 and / or computational amounts obtained by calculation from said process data, and for each of two or more second evaluation axes artificially assigned by the operator based on empirical rules. Here, the process data is the first imaging data This data takes into account the time lag between the image acquisition time and the sensor's response time. The time lag may be determined as appropriate based on the sensor's response speed, the sensor's installation location within the facility, experiments, simulations, and the operator's empirical rules.

[0120] The second imaging data 13b2 may be video data of 60 seconds or less, or video data of 30 seconds or less. According to the inventors' findings, the combustion state inside the incinerator 6 changes every 5 to 10 seconds. Therefore, by using, for example, video data of 60 seconds or less (or 30 seconds or less) as one unit for the second imaging data 13b1, it becomes possible to classify, predict, and estimate the combustion state inside the incinerator 6 with greater accuracy.

[0121] Furthermore, the second imaging data 13b2 may be video data lasting 5 seconds or longer, or 7 seconds or longer. While the combustion state inside the incinerator 6 changes every 5 to 10 seconds, by using, for example, video data lasting 5 seconds or longer (or 7 seconds or longer) as one segment for the second imaging data 13b2, it becomes possible to classify, predict, and estimate the combustion state inside the incinerator 6 with greater accuracy.

[0122] The second model building unit 13c2 generates a second trained model by machine learning the second training data 13b2 generated by the second training data generation unit 12b2 using the second algorithm 13a2. The second algorithm 13a2 used for machine learning may include at least one of the following: maximum likelihood classification, Boltzmann machine, neural network (NN), support vector machine (SVM), Bayesian network, sparse regression, decision tree, statistical estimation using random forest, boosting, reinforcement learning, or deep learning.

[0123] The second image analysis unit 13d2 takes the new second imaging data (visible light imaging data) of the incinerator 6 acquired by the second imaging data acquisition unit 13a2 as input and uses the second trained model generated by the second model construction unit 13c2 to acquire evaluation results for each of two or more second evaluation axes (for example, flame state and amount of burned material) as output data. The second image analysis unit 13d2 may also normalize (score) the evaluation results for each of the two or more second evaluation axes to a numerical range of, for example, 0 to 100 and acquire the numerical value (score) within that range as output data.

[0124] The combustion state determination unit 12e determines the current combustion state in the incinerator 6 based on the evaluation results (scores) from each of the two or more second evaluation axes performed by the second image analysis unit 13d2. The combustion state determination result may be a label indicating the characteristics of the combustion state (e.g., over-combustion, thick waste layer combustion, thin waste layer combustion, waste depletion, low-quality waste combustion), or it may be a numerical version of the said label.

[0125] The combustion state determination unit 12e may determine the current combustion state in the incinerator 6 by mapping a set of evaluation results from each of the two or more second evaluation axes onto a predetermined combustion state determination map 13d that uses the two or more second evaluation axes as coordinate axes.

[0126] Figure 4 shows an example of a combustion state determination map 13d. In the combustion state determination map 13d shown in Figure 4, the XY coordinate plane, with the amount of combustible material as the X-axis and the flame state as the Y-axis, is divided into multiple regions (in the illustrated example, six regions: a depleted waste zone, an over-combustion zone, a thin waste layer combustion zone, a good (normal) combustion zone, a thick waste layer combustion zone, and a low-quality waste combustion zone). The combustion state determination unit 12e maps a pair (X,Y) consisting of evaluation results for each of the two second evaluation axes, where the amount of combustible material and the flame state are two different second evaluation axes, onto this combustion state determination map 13d. For example, as shown in Figure 4, the evaluation result (X) when the amount of combustible material is the second evaluation axis and the evaluation result (Y) when the flame state is the second evaluation axis. If point P, which represents the pair (X,Y), is mapped within the good combustion (normal) zone, the combustion state determination unit 12e determines that the current combustion state is "good combustion (normal)".

[0127] The combustion state determination unit 12e may correct the combustion state determination result according to process data obtained from various sensors (not shown) installed in the facility and / or calculation amounts obtained by calculation from said process data. For example, if the combustion state determination result is "low-grade waste combustion" and the difference value PV-SV of the furnace outlet temperature or evaporation amount at that time meets predetermined conditions, the combustion state determination result may be corrected to "good combustion (normal)".

[0128] The combustion state determination unit 12e may, depending on the combustion state determination result, display an alert on a display (not shown) indicating that it is difficult to keep the performance management index value within a predetermined range, or it may emit an alert by sound, vibration, light, etc.

[0129] Next, an example of an information processing method using the information processing device 10 with the above configuration will be described. Figure 10 is a flowchart of an example of an information processing method.

[0130] As shown in Figure 10, first, before the incineration facility 1 is put into operation, the second training data generation unit 12b2 generates second training data 13c2 by associating classification labels on two or more second evaluation axes, which are elements for determining the combustion state and which have been artificially assigned by the operator based on empirical rules, with past visible light imaging data (second imaging data 13b2) inside the incinerator 6 captured by the visible light imaging device 72 (step S31).

[0131] Next, the second model construction unit 13c2 generates a second trained model by machine learning the second training data 13b2 generated by the second training data generation unit 12b2 using the second algorithm 13a2 (step S32).

[0132] Next, while the incineration facility 1 is in operation, the second imaging data acquisition unit 12a2 acquires new visible light imaging data of the incinerator 6 captured by the visible light imaging device 72 as the second imaging data 13b2 (step S33).

[0133] Next, the second image analysis unit 13d2 takes the new second imaging data (visible light imaging data) from inside the incinerator 6 acquired by the second imaging data acquisition unit 13a2 as input and uses the second trained model generated by the second model construction unit 13c2 to acquire evaluation results for each of two or more second evaluation axes (for example, amount of material to be burned and flame state) as output data (step S34).

[0134] Then, the combustion state determination unit 12e determines the current combustion state inside the incinerator 6 based on the evaluation results on the second evaluation axis by the second image analysis unit 13d2 (step 35).

[0135] In step S35, the combustion state determination unit 12e may determine the current combustion state in the incinerator 6 by mapping a set of evaluation results from each of the two or more second evaluation axes onto a predetermined combustion state determination map 13d that uses the two or more second evaluation axes as coordinate axes (see Figure 4).

[0136] Subsequently, the instruction unit 12f transmits an operation instruction to the crane control device 50 and / or the combustion control device 20 based on the determination result of the combustion state determination unit 12e (step S16).

[0137] According to the embodiment described above, based on the evaluation results in each of the two or more second evaluation axes, By determining the combustion state, the combustion state within the incinerator 6 can be classified, predicted, and estimated with greater accuracy compared to determining the combustion state based on the evaluation result of a single evaluation axis. Furthermore, by mapping the evaluation results of each of the two or more second evaluation axes onto the combustion state determination map 13d, the relationship between the evaluation results of each of the two or more second evaluation axes and the determined combustion state can be intuitively grasped, and the combustion state can be determined at high speed.

[0138] In the above-described embodiment, some of the processing of the control unit 12 may be performed on a cloud server separate from the information processing device 10, rather than on the information processing device 10. Part of the storage unit 13 may also be located on a cloud server separate from the information processing device 10, rather than on the information processing device 10.

[0139] For example, the processing of the first training data generation unit 12d1 and / or the second training data generation unit 12d2 may be executed on a cloud server to generate the first training data 13c1 and / or the second training data 13c2, or the processing of the first model construction unit 12c1 and / or the second model construction unit 12c2 may be executed on a cloud server to construct a trained model. Alternatively, the processing of the first image analysis unit 12d1 and / or the second image analysis unit 12d2 may be executed on a cloud server using the trained model constructed on the cloud server, or the information processing device 10 may download the trained model (trained parameters) from the cloud server and use it within the information processing device 10 to execute the processing of the first image analysis unit 12d1 and / or the second image analysis unit 12d2.

[0140] Although embodiments and modifications of the present invention have been described above by example, the scope of the present invention is not limited thereto, and it is possible to modify and transform it according to the purpose within the scope described in the claims. Furthermore, each embodiment and modification can be appropriately combined as long as the processing content is not contradictory.

[0141] Furthermore, although the information processing device 10 according to an embodiment of the present invention may be composed of one or more computers, the program for implementing the information processing device 10 on one or more computers and the computer-readable recording medium on which the program is stored non-temporarily are also subject to protection in this invention. [Explanation of symbols]

[0142] 1 Incineration facility 2 Boilers 3. Garbage pit 4 Hoppers 5 Cranes 6 Incinerator 71 Infrared imaging device 72 Visible light imaging device 10 Information Processing Devices 11 Communications Department 12 Control Unit 12a1 First imaging data acquisition unit 12a2 Second imaging data acquisition unit 12b1 First Training Data Generation Unit 12b2 Second Training Data Generation Unit 12c1 First Model Construction Section 12c2 Second Model Construction Section 12d1 1st Image Analysis Department 12d2 2nd image analysis section 12e Combustion state determination unit 12f Instruction section 12g Process data acquisition unit 13 Storage section 13a1 First Algorithm 13a2 Second Algorithm 13b1 First imaging data 13b2 Second imaging data 13c1 First training data 13c2 Second Training Data 13d Combustion State Determination Map 20 Combustion control device 21 Platforms 22 Transport Vehicles 50 Crane control device

Claims

1. A first image analysis unit performs an evaluation on the first evaluation axis using new image data from inside the incinerator as input, using a first trained model that has been machine-trained on first training data generated by assigning classification labels on at least one first evaluation axis that serves as an element for determining the combustion state to imaging data inside the incinerator, A combustion state determination unit that determines the current combustion state based on the evaluation results on the first evaluation axis, Equipped with, The aforementioned imaging data is characterized by being video data of 60 seconds or less.

2. The aforementioned imaging data is visible light imaging data of the inside of an incinerator captured by a visible light imaging device. The information processing apparatus according to feature 1.

3. The first evaluation axis includes at least one of the following: flame state, location of combustion completion point, shape of combustion completion point, amount of material to be burned, quality of material to be burned, type of material to be burned, amount of unburned material, and amount of incinerated ash. The information processing apparatus according to feature 2.

4. The aforementioned imaging data is video data lasting 5 seconds or longer. The information processing apparatus according to any one of the features 1 to 3.

5. The classification label is at least one of the following: a label indicating which of a predetermined set of classification items the data belongs to, and the relative order among multiple imaging data. An information processing apparatus according to any one of the features 1 to 4.

6. The first image analysis unit uses a first trained model, which has been trained on first training data generated by assigning classification labels on the first evaluation axis to a combination of imaging data from inside the incinerator and process data obtained from sensors installed in the facility and / or computational amounts obtained from said process data, to perform an evaluation on the first evaluation axis using a combination of new imaging data from inside the incinerator and new process data and / or computational amounts obtained from said process data as input. An information processing apparatus according to any one of the features 1 to 5.

7. The process data is data that takes into account the time lag between the imaging time of the imaging data and the response time of the sensor. The aforementioned time lag is determined based on at least one of the following: the response speed of the sensor, the installation location of the sensor within the facility, experiments, simulations, or the operator's empirical rules. The information processing apparatus according to feature 6.

8. The first image analysis unit uses a first trained model, which is generated by machine learning from first training data, that is created by assigning classification labels for each of the two or more first evaluation axes that serve as elements for determining the combustion state to the imaging data inside the incinerator, to perform evaluations on each of the two or more first evaluation axes using new imaging data inside the incinerator as input. The combustion state determination unit determines the current combustion state by mapping the evaluation results from each of the two or more first evaluation axes onto a predetermined combustion state determination map that uses the two or more first evaluation axes as coordinate axes. An information processing apparatus according to any one of the features 1 to 7.

9. The combustion state determination unit corrects the combustion state determination result according to process data obtained from sensors installed in the facility and / or calculation amounts obtained from said process data. The information processing apparatus according to any one of the features 1 to 8.

10. The combustion state determination unit displays or issues an alert according to the combustion state determination result. An information processing apparatus according to any one of the features 1 to 9.

11. The system further includes an instruction unit that transmits operation instructions to the crane control device and / or combustion control device based on the determination result of the combustion state determination unit. An information processing apparatus according to any one of the features 1 to 10.

12. The algorithms used in the aforementioned machine learning include at least one of the following: maximum likelihood classification, Boltzmann machines, neural networks (NN), support vector machines (SVM), Bayesian networks, sparse regression, decision trees, statistical estimation using random forests, boosting, reinforcement learning, and deep learning. The information processing apparatus according to any one of the features 1 to 11.

13. The system further includes a model building unit that generates a first trained model by machine learning first training data, which is generated by assigning classification labels on at least one first evaluation axis that serves as an element for determining the combustion state to imaging data inside the incinerator. The information processing apparatus according to any one of the features 1 to 12.

14. The information processing apparatus according to claim 11, A crane control device that controls a crane for stirring or transporting waste based on operation instructions transmitted from the information processing device, and / or a combustion control device that controls the combustion of waste in an incinerator, A system characterized by having the following features.

15. A method of information processing performed by a computer, The steps include: using a first trained model, which is created by machine learning on first training data generated by assigning classification labels on at least one first evaluation axis that serves as an element for determining the combustion state to imaging data inside the incinerator, to perform an evaluation on the first evaluation axis using new imaging data inside the incinerator as input; The step includes determining the current combustion state based on the evaluation results on the first evaluation axis, The aforementioned imaging data is video data with a duration of 60 seconds or less. An information processing method characterized by the following:

16. On the computer, The steps include: using a first trained model, which is created by machine learning on first training data generated by assigning classification labels on at least one first evaluation axis that serves as an element for determining the combustion state to imaging data inside the incinerator, to perform an evaluation on the first evaluation axis using new imaging data inside the incinerator as input; A step of determining the current combustion state based on the evaluation results on the first evaluation axis, Make it run, The aforementioned imaging data is video data with a duration of 60 seconds or less. An information processing program characterized by the following.

Images