Information processing apparatus, state information generation method, determination method, and state information generation program

Abstract

To grasp a state of an object facility in consideration of sensor data even when timing when the state of the object facility is reflected is different for each sensor.SOLUTION: An information processor (1A) includes a relation indicator calculation unit (103) that calculates a relation indicator indicating a relation between sensor datasets for each set of sensors in which a plurality of sensors provided in a target facility are paired, and a state information generation unit (104) that generates state information indicating a state of the target facility using the relation indicator, in which the relation indicator calculation unit (103) calculates the relation indicator for a set of sensor datasets in which sensor datasets obtained at different timings are paired.SELECTED DRAWING: Figure 1

Description

[Technical Field]

[0001] The present invention relates to an information processing device that generates information indicating the status of a facility. [Background technology]

[0002] Generally, in various facilities such as plants, operators grasp the state of the facility based on output values ​​of various sensors installed in the facility and perform necessary control in a timely manner to ensure stable operation. For example, Patent Document 1 listed below is an example of a document disclosing a technology for supporting the management of such facilities.

[0003] Patent Document 1 discloses an information processing device that generates information indicating the state of a mechanical equipment using sensor data measured by multiple sensors installed on the mechanical equipment. More specifically, the information processing device uses the sensor data to calculate a relationship index that indicates the relationship between pairs of two sensors from the multiple sensors, and generates information indicating the state of the mechanical equipment using the relationship index calculated for each pair of sensors. [Prior art documents] [Patent documents]

[0004] [Patent Document 1] Patent Publication No. 2021-043706 Summary of the Invention [Problem to be solved by the invention]

[0005] There is room for improvement in terms of accuracy in the information processing device described in Patent Document 1. More specifically, the information processing device described in Patent Document 1 calculates a relationship index that indicates the relationship between a pair of sensors based on the distribution state of sensor data measured by the pair of sensors during the same period.

[0006] However, the timing at which the facility status is reflected in the sensor data may differ for each sensor. Therefore, when the timing difference is taken into account, the relationship index for a pair of sensors whose sensor data are correlated may show a value that indicates no correlation between the sensor data.

[0007] One aspect of the present invention aims to realize an information processing device, etc. that makes it possible to understand the status of a target facility by taking into account sensor data, even if the timing at which the status of the target facility is reflected in the sensor data differs for each sensor. [Means for solving the problem]

[0008] In order to solve the above problem, an information processing device according to one aspect of the present invention includes a relationship index calculation unit that calculates, for each pair of sensors formed by a plurality of sensors installed in a target facility, a relationship index that indicates the relationship between sensor data obtained by one of the pair of sensors and sensor data obtained by the other of the pair of sensors, and a status information generation unit that generates status information that indicates the status of the target facility using the relationship index calculated for each of the pair of sensors, and the relationship index calculation unit calculates the relationship index for each pair of sensor data formed by a pair of sensor data obtained at different times.

[0009] In order to solve the above problem, an information processing device according to another aspect of the present invention includes a classification unit that performs a process of classifying sensor data obtained by sensors installed in a target facility based on the magnitude of its values, using the sensor data obtained at different times for each of a plurality of sensors installed in the target facility as the target, and a correspondence determination unit that determines the degree of correspondence between the state of the target facility at the time the sensor data classified by the classification unit was obtained and the reference state based on state information that represents the reference state using a relationship index that indicates the relationship, calculated based on the relationship between the magnitude of the values ​​of the sensor data obtained by the plurality of sensors at different times when the target facility is in a predetermined reference state, and the classification result by the classification unit.

[0010] In order to solve the above-mentioned problems, a status information generation method according to one embodiment of the present invention is a status information generation method executed by one or more information processing devices, and includes: a relationship index calculation step of calculating, for each pair of sensors comprising a plurality of sensors installed in a target facility, a relationship index indicating the relationship between sensor data obtained by one of the pair of sensors and sensor data obtained by the other of the pair of sensors; and a status information generation step of generating status information indicating the status of the target facility using the relationship index calculated for each of the pair of sensors, wherein in the relationship index calculation step, the relationship index is calculated for each pair of sensor data comprising sensor data obtained at different times.

[0011] In order to solve the above problem, a determination method according to one embodiment of the present invention is a determination method executed by one or more information processing devices, and includes a classification step in which a process of classifying sensor data obtained by sensors installed in a target facility based on the magnitude of its values ​​is performed on the sensor data obtained at different times for each of multiple sensors installed in the target facility; and a consistency determination step in which, when the target facility is in a predetermined reference state, the state of the target facility at the time the sensor data classified in the classification step was obtained is determined to have a degree of consistency with the reference state based on state information representing the reference state based on the relationship between the magnitude of the values ​​of sensor data obtained by the multiple sensors at different times and the classification result in the classification step. [Effects of the Invention]

[0012] According to one aspect of the present invention, even if the timing at which the state of the target facility is reflected in sensor data differs for each sensor, it is possible to understand the state of the target facility by taking into account the sensor data. [Brief explanation of the drawings]

[0013] [Figure 1]1 is a block diagram showing an example of a configuration of a main part of an information processing device according to a first embodiment of the present invention. [Figure 2] FIG. 2 is a diagram for explaining an overview of a support system including the information processing device. [Figure 3] 10A and 10B are diagrams illustrating examples of classification of sensor data and examples of calculation of relational indices. [Figure 4] FIG. 10 is a diagram showing an example of a classification result of values ​​of sensor data output by two sensors. [Figure 5] FIG. 10 is a diagram showing an example of a correlation map showing the state of a waste incineration plant. [Figure 6] 2 is a flowchart showing an example of processing executed by the information processing device shown in FIG. [Figure 7] FIG. 10 is a block diagram showing an example of a configuration of a main part of an information processing device according to a second embodiment of the present invention. [Figure 8] FIG. 10 is a diagram illustrating an example of calculation of a degree of coincidence. [Figure 9] FIG. 10 is a diagram illustrating an example of calculation of the overall degree of coincidence. [Figure 10] 8 is a flowchart showing an example of processing executed by the information processing device shown in FIG. 7. [Figure 11] FIG. 10 is a block diagram showing an example of a configuration of a main part of an information processing device according to a third embodiment of the present invention. [Figure 12] FIG. 10 is a diagram illustrating the relationship between a weight multiplied by a sensor set unit coincidence degree and a weight multiplied by a status information unit coincidence degree. [Figure 13] 12 is a flowchart showing an example of processing executed by the information processing device shown in FIG. DETAILED DESCRIPTION OF THE INVENTION

[0014] [Embodiment 1] (System Overview) An outline of a support system 100 according to one embodiment of the present invention will be described with reference to Fig. 2. Fig. 2 is a diagram for explaining the outline of the support system 100. The support system 100 is a system for supporting the operation of various facilities, and includes sensors S1 to Sn and an information processing device 1A as shown in the figure. When it is not necessary to distinguish between the sensors S1 to Sn, they will be simply referred to as sensor S.

[0015] In this embodiment, an example will be described in which the support system 100 supports the operation of a waste incineration plant P. The waste incineration plant P is equipped with an incinerator for incinerating waste, as well as power generation equipment that uses the heat generated in the incinerator to generate electricity. Note that the support target of the support system 100 is not limited to mechanical equipment such as the waste incineration plant P, as long as it is a facility whose state can be grasped by sensors or the like and whose operation is controlled (operated) by humans or automatically. For example, the support system 100 can also support the operation of wind power generation equipment, etc. In the following explanations of each embodiment, the term "waste incineration plant P" can be read as any facility that is the target of support (hereinafter also referred to as the target facility).

[0016] The sensor S detects a predetermined physical quantity or a change in the quantity related to the state of the waste incineration plant P, and outputs sensor data indicating the detection results to the information processing device 1A. The sensors S1 to Sn each detect a different object. As described above, in this embodiment, the sensors S are installed at various locations within the waste incineration plant P to support the operation of the waste incineration plant P. For example, the sensor S may include a temperature sensor that indicates the temperature inside the incinerator, a sensor that detects the carbon monoxide concentration in the exhaust gas, etc.

[0017] Furthermore, the sensor S may be provided for a portion of the mechanical equipment of the waste incineration plant P. For example, if multiple sensors S are provided for the incinerator, information indicating the state of the incinerator can be output to the information processing device 1A. More specifically, when outputting information indicating the combustion state, sensors S arranged around the incinerator, such as a thermometer, an air flow meter, or a grate speed measuring instrument, may be used. Furthermore, if multiple sensors S are provided for a portion of the mechanical equipment, information regarding that portion can be output to the information processing device 1A. For example, if multiple sensors S related to exhaust gas from the incinerator (for example, an in-furnace thermometer, a garbage layer thickness measuring instrument, a CO concentration meter, etc.) are provided, information regarding the exhaust gas can be output to the information processing device 1A.

[0018] As will be explained in detail below, the information processing device 1A acquires sensor data, which are output values ​​of multiple sensors S over a predetermined period of time. Next, the information processing device 1A calculates a relationship index indicating the relationship between a pair of sensors S consisting of two of the multiple sensors S, based on the distribution status of each sensor data acquired for the sensors S associated with the pair (how each sensor data is distributed). The information processing device 1A then uses the relationship index calculated for each pair of the multiple sensors S to generate and output information indicating the state of the target facility over the predetermined period, i.e., status information. This makes it possible to generate information indicating the status of the target facility, which is useful for supporting the operation of the target facility. Furthermore, the information processing device 1A can generate status information indicating the state of another facility by acquiring sensor data from that facility, making it easy to generalize.

[0019] Furthermore, the above-mentioned relationship index is calculated for pairs of sensor data obtained at different times. This makes it possible to understand the state of the target facility by taking into account the sensor data, even if the timing at which the state of the target facility is reflected in the sensor data differs for each sensor.

[0020] It has been confirmed through experiments by the inventors that the relationship between the distribution patterns of the sensor data in the set of sensors S reflects the state of the facility. Furthermore, it has been confirmed that the above configuration makes it possible to capture nonlinear relationships that are difficult to handle with general analysis methods.

[0021] 2, the status information is correlation maps M111 to M113 that show the status of the waste incineration plant P. The correlation map M111 was generated using sensor data from a stable period when the waste incineration plant P was operating stably, and the correlation map M113 was generated using sensor data from an unstable period when the operating status of the waste incineration plant P became unstable. The correlation map M112 was generated using sensor data from a pre-unstable period, which is the period immediately before the unstable period. The criteria for stability and instability may be set arbitrarily; for example, a period when the amount of steam used for power generation is within a normal range may be defined as a stable period, and a period when the amount of steam exceeds the normal range may be defined as an unstable period.

[0022] As shown in the figure, correlation maps M111-M113 reflect the state of waste incineration plant P from a stable period to an unstable period. More specifically, correlation maps M111-M112 show that the color of the entire correlation map becomes lighter from the stable period to the period just before instability. Furthermore, correlation maps M112-M113 show that the color of the entire correlation map becomes even lighter from the period just before instability to the unstable period.

[0023] Therefore, an operator or the like of the waste incineration plant P can determine whether the state of the waste incineration plant P is closer to the stable period or the unstable period from the overall color intensity and color distribution of the correlation map. In other words, the operator or the like can determine that it is a stable period if the overall color intensity and color distribution of the correlation map generated based on the latest sensor data are closer to the correlation map M111, and can determine that it is an unstable period if they are closer to the correlation map M113.

[0024] Furthermore, the correlation map M112 for the period immediately prior to instability and the correlation map M111 for the stable period can be distinguished by appearance. In other words, the correlation map generated by the information processing device 1A shows signs of instability. Therefore, by referring to the correlation map generated by the information processing device 1A, an operator or the like can take measures to stabilize the waste incineration plant P before the unstable period begins. This makes it possible to prevent the waste incineration plant P from becoming unstable.

[0025] (Device configuration) A more detailed configuration of the information processing device 1A will be described with reference to Fig. 1. Fig. 1 is a block diagram showing an example of the configuration of the main parts of the information processing device 1A. As shown in the figure, the information processing device 1A includes a control unit 10A that controls the various parts of the information processing device 1A, and a storage unit 11 that stores various data used by the information processing device 1A. The control unit 10A also includes a data acquisition unit 101, a classification unit 102, a relationship index calculation unit 103, and a state information generation unit 104.

[0026] Furthermore, the information processing device 1A includes an input unit 12 that accepts input of information to the information processing device 1A, and an output unit 13 that the information processing device 1A outputs information. The input unit 12 and the output unit 13 may be external devices attached to the information processing device 1A. In this embodiment, an example will be described in which the input unit 12 is an input interface unit that accepts input of sensor data output by a sensor S, and the output unit 13 is a display device that displays and outputs an image. Note that the input unit 12 and the output unit 13 are not limited to these examples as long as they have an information input / output function.

[0027] The data acquisition unit 101 acquires sensor data, which is the output value of the sensor S described above. The data acquisition unit 101 may acquire, as sensor data, a numerical value calculated using the output value of the sensor S (for example, a value obtained by normalizing the output value or a value obtained by removing noise components from the output value). Sensor data can also be called process data. In addition to the sensor data, the data acquisition unit 101 may acquire operation setting values ​​(for example, a setting value for the operation speed) of various mechanical equipment included in the waste incineration plant P. Since such operation setting values ​​can be treated in the same way as sensor data, the "sensor data" in this specification is also intended to include the above-mentioned operation setting values.

[0028] The dividing unit 102 divides the sensor data acquired by the data acquiring unit 101 into a plurality of sets according to the magnitude of the values. Details of the method for setting the sets and the method for dividing will be described later with reference to FIG.

[0029] The relationship index calculation unit 103 calculates a relationship index indicating the relationship between sensor data obtained by one of a pair of sensors installed in the target facility and sensor data obtained by the other of the pair of sensors, for each pair of sensors. When the target facility is a waste incineration plant P, the relationship index calculation unit 103 calculates a relationship index indicating the relationship between sensor data obtained by one of a pair of sensors S and sensor data obtained by the other of the pair of sensors S, for each pair of sensors S consisting of two of the multiple sensors S installed in the waste incineration plant P. The relationship index is calculated based on the classification result of the classification unit 102. For example, the relationship index calculation unit 103 may calculate the relationship index from the frequency of sensor data classified into each of a plurality of sets. This allows the relationship between each sensor to be quantified by simple processing. Details of the method for calculating the relationship index will be described later with reference to FIG. 3 etc.

[0030] The status information generating unit 104 generates status information indicating the status of the target facility using the relationship index calculated for each set of sensors. If the target facility is a waste incineration plant P, the status information generating unit 104 generates status information indicating the status of the waste incineration plant P for a predetermined period using the relationship index calculated by the relationship index calculating unit 103 for each set of sensors S.

[0031] In this embodiment, an example will be described in which the status information generating unit 104 generates a correlation map as shown in Fig. 2 as status information. As will be described in detail later, the correlation map generated by the status information generating unit 104 is an image in which a pattern according to the value of the relation index calculated for each pair of sensors S is drawn in each section corresponding to each pair defined on an image plane. The status information generating unit 104 generates such a correlation map and displays it on the output unit 13 or the like, thereby allowing an operator of the waste incineration plant P or the like to visually recognize the status of the waste incineration plant P.

[0032] As described above, the information processing device 1A includes the relationship index calculation unit 103 that calculates a relationship index indicating the relationship between sensor data acquired by one of a plurality of sensors installed in a target facility and sensor data acquired by the other of the paired sensors for each pair of sensors, and the status information generation unit 104 that generates status information indicating the status of the target facility using the relationship index calculated for each pair of sensors. The relationship index calculation unit 103 then calculates the relationship index for each pair of sensor data, each pair being made up of sensor data acquired at different times. Note that the "sensor data" may include at least one data element. For example, the relationship index for a pair of sensor A and sensor B may be calculated using sensor data acquired by sensor A (e.g., including multiple data elements constituting a time series) and sensor data acquired by sensor B (e.g., including multiple data elements constituting a time series).

[0033] The timing at which the state of a facility is reflected in the sensor data from multiple sensors installed in the facility may differ for each sensor. In such cases, if the state of the facility is analyzed using sensor data obtained at the same time, only a portion of the sensor data will be reflected in the analysis results, limiting the improvement of analytical accuracy.

[0034] For example, when a specific abnormality occurs in a certain facility, the effects of the abnormality may appear in the sensor data measured by sensor A installed in the facility at time t immediately after the occurrence of the abnormality, but the effects of the abnormality may appear at a time later than time t (t + Δt) in the sensor data measured by sensor B. In this case, if the sensor data at time t is the subject of analysis, the effects of the abnormality will not appear in either the sensor data measured by sensor A or sensor B at time t, and therefore the combination of sensor A and sensor B will not be taken into consideration in detecting the above abnormality.

[0035] Therefore, the above-described configuration employs a configuration in which a relationship index is calculated for a set of sensor data obtained at different times, and the calculated relationship indexes are used to generate status information indicating the status of the target facility.

[0036] This makes it possible to understand the state of the target facility by taking into account the sensor data, even if the timing at which the state of the target facility is reflected in the sensor data differs for each sensor.

[0037] (Classification of sensor data and calculation of related indicators) The classification of sensor data by the classification unit 102 and the calculation of the relationship index by the relationship index calculation unit 103 will be described with reference to Fig. 3. Fig. 3 is a diagram showing an example of classification of sensor data and an example of calculation of the relationship index.

[0038] 121 in FIG. 3 shows the relationship between the sensor data value output by sensor SA, which is one of sensors S1 to Sn, and the sensor data value output by sensor SB, which is another of sensors S1 to Sn. More specifically, 121 in FIG. 3 shows coordinates, which are pairs of the sensor data value output by sensor SA and the sensor data value output by sensor SB, plotted on a coordinate plane. On this coordinate plane, the horizontal direction is the axis indicating the magnitude of the output value of sensor SA, and the vertical direction is the axis indicating the magnitude of the output value of sensor SB. For example, if the sensor data output by sensor SA is a1 to a20 and the sensor data output by sensor SB is b1 to b20, the plotted points will be (a1, b1) to (a20, b20).

[0039] Here, the sensor data paired in the example of Fig. 3 are obtained at different times. Specifically, the sensor data obtained by sensor SA at time t1 (more precisely, obtained during a predetermined period starting from time t1) and the sensor data obtained by sensor SB at time t2 (more precisely, obtained during a predetermined period starting from time t2) are paired. Note that t1 ≠ t2. In this way, in the example of Fig. 3, sensor data obtained at different times are paired.

[0040] In 121 of Fig. 3, the coordinate plane is divided into nine sections. These sections are set based on the classification of the magnitude of the values ​​of the sensor data output by sensors SA and SB (three levels: small, middle, and large). Specifically, nine sections are set, ranging from a section (small-small) where the values ​​of the sensor data output by sensors SA and SB are both "small" to a section (large-large) where the values ​​of the sensor data output by sensors SA and SB are both "large."

[0041] The sorting unit 102 may sort the sensor data based on such divisions. In this case, a threshold for sorting the size of the sensor data is set in advance for each sensor S. For example, when sorting into three sets as in the example of FIG. 3, a threshold for sorting between "Small" and "Middle" and a threshold for sorting between "Middle" and "Large" are set in advance. This allows the sorting unit 102 to sort the sensor data into "Small," "Middle," or "Large" depending on the size of its value. Note that the method for setting the threshold is not particularly limited, and may be set based on information indicating the distribution of the sensor data (e.g., average value, maximum value, minimum value, etc.).

[0042] The dividing unit 102 performs such a division for each combination of sensors S1 to Sn. Note that the sets into which the sensor data is divided may differ for each of sensors S1 to Sn. For example, the sensor data output by sensor S1 may be divided into two sets, and the sensor data output by sensor S2 may be divided into four or more sets.

[0043] The classification unit 102 may also classify the sensor data by dividing the sensor data into multiple fuzzy sets. This allows appropriate classification of sensor data near the boundary between one set and another. For example, the classification unit 102 may use a membership function such as that shown in 123 of FIG. 3 to classify the sensor data into three fuzzy sets, "Small," "Middle," and "Large." The membership function may be created manually or automatically. When automatically created, the classification unit 102 may, for example, calculate the average value and standard deviation σ of the sensor data, and create a membership function based on the maximum value, minimum value, and average value of the sensor data that fall within a range of ±3σ from the average value.

[0044] The relationship index calculation unit 103 calculates a relationship index from the frequency of sensor data classified into each set for each combination of sensors S1 to Sn. In the example of 121 in Fig. 3, of the 20 plotted points, 9 points are included in the "Middle"-"Middle" category, 6 points are included in the "Middle"-"Small" category, and 2 points are included in the "Small"-"Large" category. In addition, one point is included in each of the "Small"-"Small", "Large"-"Small", and "Large"-"Middle" categories, and no points are included in the other categories.

[0045] In this case, the relationship index calculation unit 103 may set the relationship index for each segment to a value such as that shown in 122 of FIG. 3 . In the example of 122 of FIG. 3 , the relationship index for the segment "Middle"-"Middle," which included the most plotted points, is 0.405, and the relationship index for the segment "Middle"-"Small," which included the second most plotted points, is 0.225. Furthermore, the relationship index for the segment "Small"-"Large," which included the third most plotted points, is 0. Since the relationship index for a segment containing only one point is lower than a predetermined threshold, the relationship index is 0, just like a segment containing no points. Thus, in the example of FIG. 3 , the greater the frequency of the sensor data in the segment, the greater the relationship index. A method for calculating such a relationship index will be described below with reference to FIG. 4 .

[0046] (Example of how to calculate related indicators) Fig. 4 is a diagram showing an example of the classification results of the values ​​of the sensor data output by two sensors S. Specifically, in the example of 124 in Fig. 4, the sensor data output by sensor S2 is 120 in total, of which 30 are classified as "Small," 40 as "Middle," and 50 as "Large." In addition, 30 of the sensor data output by sensor S1 are classified as "Small," and of these, 10 of the paired sensor data (output by sensor S2) are classified as "Middle," and 20 of the paired sensor data (output by sensor S2) are classified as "Large."

[0047] On the other hand, in the example 125 of Fig. 4, the classification of the sensor data output by sensor S1 is the same as in the example 124, but it differs from the example 124 in that the number of sensor data output by sensor S2 is 450 in total. Of the sensor data output by sensor S2, 150 are classified as "Small," 250 as "Middle," and 50 as "Large."

[0048] The relationship index calculation unit 103 may calculate the relationship index for each segment using the following formula (1).

[0049] (Relationship index) = (occupancy rate P) × (coverage rate C) × (ratio R) ... Formula (1) The relationship index is calculated for each combination of categories of a set of sensor data. For example, if each sensor data is classified into three types, "Small," "Middle," and "Large," as described above, the relationship index is calculated for each of the nine combinations, from "Small"-"Small" to "Large"-"Large."

[0050] The occupancy rate P is the number of sensor data belonging to one of the sensors in the combination of categories for which the relation index is calculated relative to the total number of sensor data in the category. The coverage rate is the number of sensor data belonging to the combination relative to the total number of sensor data in the category for the other sensor. The ratio R is the number of sensor data belonging to the combination relative to the total number of sensor data in one of the sensors.

[0051] For example, suppose a relationship index is calculated for a combination of sensor data from sensor S1 being classified as "Small" and sensor data from sensor S2 being classified as "Large." In this case, the occupancy rate P is the ratio of sensor data from sensor S2 that belongs to "Large" to sensor data that belongs to "Small" for sensor S1. Therefore, in the example of 124 in FIG. 4, the occupancy rate P=20 / 30.

[0052] Furthermore, the coverage rate C in the above combination is the ratio of the sensor data of sensor S2 that belongs to "Large" among the sensor data of sensor S1 that belongs to "Small" to the number of sensor data of sensor S2 that belongs to "Large". Therefore, in the example of 124 in Figure 4, the coverage rate C = 20 / 50.

[0053] The ratio R in the above combination is the ratio of sensor data belonging to "Large" to the total number of sensor data of the sensor S2. Therefore, in the example of 124 in FIG. 4, the ratio R=50 / 120.

[0054] From the above, in the example of 124 in Fig. 4, the occupancy rate for the combination of sensor S1: "Small" and sensor S2: "Large" is P = 20 / 30, the coverage rate C = 20 / 50, and the ratio R = 50 / 120, so the relationship index for this combination is 1 / 9. On the other hand, in the example of 125 in Fig. 4, the occupancy rate P = 20 / 30, the coverage rate C = 20 / 50, and the ratio R = 50 / 450, so the relationship index is 4 / 135.

[0055] Although it is not essential to multiply by the ratio R in the formula for calculating the relationship index, multiplication by the ratio R is preferable because it makes it possible to make the value of the relationship index valid even when there is a large difference between the number of data in the second category and the number of data in other categories, as in the example of 125. Also, in the above formula (1), at least one of the occupancy rate P, the coverage rate C, and the ratio R may be multiplied by a weight. For example, by setting the weight of the occupancy rate P to a value greater than the weights of the coverage rate C and the ratio R, or by setting the weights of the coverage rate C and the ratio R to a value smaller than the weight of the occupancy rate P, it is possible to calculate a relationship index that emphasizes the occupancy rate P.

[0056] Furthermore, for example, in formula (1), the weight of the coverage rate C may be set to zero to calculate the relationship index. In this case, for the opposite combination of viewpoints (for example, the combination of sensors S1 and S2 versus the combination of sensors S2 and S1), the weight of the occupancy rate P may be set to zero to calculate the relationship index. Furthermore, in these cases, the weights of the other terms in formula (1) may be set to 1. In this way, for two combinations of viewpoints for the same sensor S (for example, for sensors S1 and S2, the combinations S1-S2 and S2-S1), the relationship index may be calculated using different weights (or different calculation formulas). Even with this configuration, a correlation map showing the state of the waste incineration plant P can be generated.

[0057] (Correlation map generation) The state information generation unit 104 uses the above-mentioned relationship index to generate a correlation map, which is information showing the state of the waste incineration plant P. The generation of the correlation map will be explained based on Fig. 5. Fig. 5 is a diagram showing an example of a correlation map showing the state of the waste incineration plant P. Fig. 5 also shows a normal correlation map M114 and a simplified correlation map M115.

[0058] The correlation map M114 represents the relationship index values ​​for each combination of categories using colors for all combinations of sensors S1 to Sn. Note that Fig. 5 shows the portion corresponding to the combinations of sensors S14 to S19 and sensors S14 to S19 (excluding combinations of the same sensor S) out of all combinations of sensors S1 to Sn. Also, Fig. 5 includes some portions where Small, Middle, and Large are abbreviated as S, M, and L, respectively.

[0059] In the correlation map M114, one section is set for each combination of sensors S, and each section is further divided into nine small sections. The nine small sections correspond to the Small, Medium, and Large categories of sensor data values ​​in a combination of sensors S. For example, the position where the column of sensor S14 intersects with the row of sensor S15 is the section corresponding to the combination of sensors S14-S15. Of the nine small sections included in this section, the upper left small section corresponds to the combination in which the sensor data of sensor S14 and the sensor data of sensor S15 are both "Small." The other small sections also correspond to combinations of sensor data categories in a similar manner.

[0060] The correlation map M114 in FIG. 5 is generated by not coloring small sections whose relationship index is less than a preset threshold, but by coloring the small sections whose relationship index is closer to 1 as a color closer to black, and the small sections whose relationship index is closer to 0 as a color closer to white. In other words, the correlation map M114 represents the relationship index as pixel values ​​in grayscale. Note that the pattern in which the relationship index is represented in the correlation map M114 can be any pattern and is not limited to the example of FIG. 5. For example, the relationship index may be represented by saturation, hue, brightness, or a combination thereof. If a rule for converting the relationship index into a color is determined in advance, the state information generation unit 104 can determine the display color of each section of the correlation map M114 according to the rule.

[0061] On the other hand, the correlation map M115 is an image in which one pattern image is drawn in one section. The pattern for each section is determined using up to nine relationship indices corresponding to that section. Specifically, the state information generation unit 104 calculates a weighted sum of the relationship indices included in one section and sets the average value as the relationship index value for that section. Then, the state information generation unit 104 determines a pattern according to the value of the relationship index for that section (for example, an image in which the entire section is uniformly filled with pixel values ​​according to the value of the relationship index) as the pattern to be drawn in that section. The weight may be set to a value such that the distribution status of the sensor data for each combination of sensors S is reflected in the weighted sum.

[0062] (About the reference map) The status information generating unit 104 may generate a reference map that serves as a standard for an operator or the like using the support system 100 to judge the status of the waste incineration plant P, based on sensor data acquired during a stable period in which the waste incineration plant P is operating stably. The status information generating unit 104 may also generate a correlation map as needed based on sensor data acquired in real time while the waste incineration plant P is operating, and display the generated correlation map together with the reference map on the output unit 13. This allows an operator or the like to compare the reference map with the current correlation map and judge whether the waste incineration plant P is in a stable state or whether there are any signs of instability.

[0063] The state information generating unit 104 may generate the reference map based on sensor data acquired during the unstable period. In this case, an operator or the like compares the reference map with the current correlation map, and if the current correlation map is approaching the reference map, determines that there is a sign of instability.

[0064] (Processing flow) The flow of processing (information processing method) executed by the information processing device 1A will be described with reference to FIG. 6. FIG. 6 is a flowchart showing an example of processing executed by the information processing device 1A. Note that an example of generating a correlation map will be described below, but if sensor data measured at a normally operating waste incineration plant P is acquired in ST11, FIG. 6 becomes a flowchart of processing for creating a reference map. Furthermore, "normal operation" refers to a state in which no abnormalities occur. For example, a state in which the operation of the waste incineration plant P can be continued by automatic control without manual intervention, a state in which the operation of the waste incineration plant P can be continued according to a predetermined operation plan, a state in which the amount of power generation is stable within a predetermined range, etc. are "normal operation" states. Such states can also be called stable states.

[0065] In ST11, the data acquiring unit 101 acquires sensor data from all sensors S. For example, if the time difference between the sensor data to be combined is one minute, the data acquiring unit 101 may acquire sensor data for a time period obtained by adding one minute to a predetermined time period. This provides a set of sensor data for a predetermined time period, each of which is acquired at a time difference of one minute.

[0066] In ST12, the classification unit 102 classifies the sensor data acquired in ST11 according to the magnitude of its value. For example, the classification unit 102 may classify the sensor data into three levels, namely, small, middle, and large, or into two levels or four or more levels. Furthermore, the classification unit 102 may classify the sensor data based on, for example, a threshold value, or may classify the sensor data using a fuzzy set.

[0067] In ST13 (relationship index calculation step), the relationship index calculation unit 103 calculates a relationship index for each category based on the distribution of the sensor data acquired in ST11. More specifically, for each pair of sensors S, which are pairs of sensors S provided in the waste incineration plant P, the relationship index calculation unit 103 calculates a relationship index indicating the relationship between sensor data acquired by one of the pair of sensors S and sensor data acquired by the other of the pair of sensors S. At this time, the relationship index calculation unit 103 calculates a relationship index for a pair of sensor data, which are paired sensor data acquired at different times (for example, sensor data acquired at times with a one-minute difference as described above). At this time, the relationship index calculation unit 103 also calculates a relationship index for each combination of categories and categories according to the number of data included in the category. For example, the relationship index calculation unit 103 may calculate the relationship index using the above-mentioned mathematical formula (1).

[0068] In ST14 (state information generation step), the state information generation unit 104 generates a correlation map, which is state information showing the state of the waste incineration plant P, using the relationship index calculated in ST13 for each set of sensors S. Specifically, the state information generation unit 104 generates the correlation map by determining a drawing pattern according to the relationship index of each combination of Small, Middle, and Large categories in the section for each combination of sensors S.

[0069] In ST15, the status information generating unit 104 displays the correlation map generated in ST14 on the output unit 13. If a correlation map has already been displayed on the output unit 13, the status information generating unit 104 may update the displayed correlation map with the newly generated correlation map. This allows the content of the displayed correlation map to always indicate the latest status of the waste incineration plant P.

[0070] In ST16, the data acquisition unit 101 determines whether or not to end the processing. If the data acquisition unit 101 determines to end the processing (YES in ST16), the processing of FIG. 6 ends. On the other hand, if the data acquisition unit 101 determines to continue the processing (NO in ST16), the processing returns to ST11. The determination condition in ST16 may be set appropriately. For example, the data acquisition unit 101 may determine to end the processing when the waste incineration plant P has stopped operating. This makes it possible to provide operational assistance by displaying the correlation map during the period until the waste incineration plant P stops operating.

[0071] The relationship index calculation unit 103 may calculate a relationship index for each set of sensor data obtained by a different combination of timings.The state information generation unit 104 may then generate a correlation map for each set of sensor data obtained by a different combination of timings.In this way, a plurality of correlation maps with different time shift patterns of the sensor data are generated and output.

[0072] As described above, the method for generating status information according to this embodiment includes a relationship index calculation step (ST13) for calculating, for each pair of sensors S comprising a plurality of sensors S provided in a waste incineration plant P, a relationship index indicating the relationship between sensor data obtained by one of the pair of sensors S and sensor data obtained by the other of the pair of sensors S, and a status information generation step (ST14) for generating status information indicating the status of the waste incineration plant P using the relationship index calculated for each pair of sensors S, in which the relationship index calculation step calculates a relationship index for each pair of sensor data obtained at different times. This makes it possible to grasp the status of the waste incineration plant P by taking into account the sensor data, even if the timing at which the status of the waste incineration plant P is reflected in the sensor data differs for each sensor.

[0073] [Embodiment 2] Other embodiments of the present invention will be described below. For ease of explanation, the same reference numerals will be used to designate components having the same functions as those described in the above embodiments, and their descriptions will not be repeated. This also applies to the third and subsequent embodiments.

[0074] (Device configuration) The information processing device 1B of this embodiment differs from the information processing device 1A of the above embodiment in that it automatically detects whether an abnormality has occurred or if there is a sign of an abnormality in the waste incineration plant P. The information processing device 1B will be described below with reference to Fig. 7. Fig. 7 is a diagram showing the configuration of the information processing device 1B.

[0075] 7, information processing device 1B includes a control unit 10B instead of control unit 10A. Control unit 10B differs from control unit 10A of information processing device 1A in that it includes a coincidence determination unit 201 and an abnormality detection unit 202 instead of state information generation unit 104, and in that it does not include relationship index calculation unit 103. Note that control unit 10B may also include relationship index calculation unit 103 and state information generation unit 104.

[0076] The coincidence determination unit 201 determines the degree to which the state of the waste incineration plant P, which is the target facility, coincides with a predetermined reference state, i.e., the degree of coincidence. The degree of coincidence is an index that indicates the degree to which the distribution of sensor data when the waste incineration plant P is in the reference state coincides with the distribution of sensor data acquired by the data acquisition unit 101. The degree of coincidence can also be referred to as similarity. Note that the degree of coincidence is information that indicates the state of the waste incineration plant P, so the coincidence determination unit 201 can also be called a state information generation unit.

[0077] The reference state is represented by a relationship index that indicates the relationship, calculated based on the relationship between the magnitudes of the values ​​of sensor data obtained by the multiple sensors S at different times when the waste incineration plant P is in the reference state. The degree of coincidence is determined based on the classification result obtained by the classification unit 102 of the sensor data obtained in the state for which the degree of coincidence is to be determined. Note that, as in the first embodiment, the classification unit 102 performs a process of classifying the sensor data obtained by the sensors S provided in the waste incineration plant P by the magnitude of the values, for each of the multiple sensors S provided in the waste incineration plant P, with the sensor data obtained at different times as the target.

[0078] By including the coincidence determination unit 201, the information processing device 1B can numerically indicate whether the operating state of the waste incineration plant P is close to or far from the reference state. For example, if a stable state is taken as the reference state, a high degree of coincidence indicates a high probability that the operating state is stable. On the other hand, if an abnormal state is taken as the reference state, a high degree of coincidence indicates a high probability that an abnormality has occurred. The degree of coincidence can be displayed and output in the same way as a correlation map and presented to an operator, or can be used to detect abnormalities, as will be described below.

[0079] The abnormality detection unit 202 detects abnormalities in the waste incineration plant P based on the degree of coincidence determined by the degree of coincidence determination unit 201. Note that the abnormalities that the abnormality detection unit 202 detects include unstable states of the waste incineration plant P as well as states in which the state of the waste incineration plant P is showing a tendency to become unstable.

[0080] When the stable state is defined as the reference state, the anomaly detection unit 202 may determine that an anomaly has occurred when the degree of match is equal to or less than a predetermined threshold. Furthermore, the anomaly detection unit 202 may determine whether or not an anomaly has occurred based on the rate of change of the degree of match. For example, the anomaly detection unit 202 may determine that an anomaly has occurred when the degree of match drops sharply in a short period of time, or when the degree of match continues to drop for a predetermined period of time. Furthermore, when the abnormal state is defined as the reference state, the anomaly detection unit 202 may determine that an anomaly has occurred when the degree of match is equal to or greater than a predetermined threshold, when the degree of match rises sharply in a short period of time, or when the degree of match continues to rise.

[0081] If the abnormality detection unit 202 determines that there is a possibility that an abnormality has occurred, it notifies an operator or the like. The manner of notification is not particularly limited, and for example, the abnormality detection unit 202 may notify by displaying information indicating that an abnormality has been detected on the output unit 13. Note that the determination of the presence or absence of an abnormality and the notification may be performed in separate processing blocks.

[0082] As described above, the information processing device 1B includes a classification unit 102 that performs a process of classifying sensor data obtained by sensors installed in the target facility based on the magnitude of its values, using sensor data obtained at different times for each of multiple sensors installed in the target facility; and a consistency determination unit 201 that determines the degree of consistency between the state of the target facility at the time the sensor data classified by the classification unit 102 was obtained and the reference state, based on state information that represents the reference state using a relationship index that indicates the relationship, calculated based on the relationship between the magnitude of the values ​​of the sensor data obtained by the multiple sensors at different times when the target facility is in a predetermined reference state, and the classification result by the classification unit 102.

[0083] According to the above configuration, even when the timing at which the state of the target facility is reflected in the sensor data is different for each sensor, it is possible to grasp the state of the target facility in consideration of those sensor data. Note that, as described above, "sensor data obtained at different timings" can also be rephrased as "sensor data with a deviation in the obtained timing". For example, sensor data obtained at times t1 to t p and sensor data obtained at times (t1 + Δt) to (t p + Δt) are "sensor data obtained at different timings", and are also "sensor data with a deviation in the obtained timing".

[0084] (State information used for determination of degree of coincidence) As described above, for determination of the degree of coincidence, state information represented by a relationship index calculated based on the relationship of the magnitudes of sensor data values obtained at different timings by a plurality of sensors S when the waste incineration plant P is in the reference state is used. Note that the method for calculating the relationship index is as described in Embodiment 1. This state information can be generated in the same manner as the reference map of Embodiment 1. The state information when the waste incineration plant P is in the reference state may be generated in advance and stored in the storage unit 11 or the like.

[0085] For determination of the degree of coincidence, it is sufficient to have at least one piece of state information, but in this embodiment, an example of using a plurality of pieces of state information will be described. In this case, the plurality of pieces of state information are assumed to be generated using sets of sensor data with different combinations of the obtained timings (in other words, patterns of time deviations of sensor data).

[0086] For example, state information generated for each combination (k k0 <k k1 ~t kp )(k p <k p-1 <…<k1<k0) of sensor data obtained at time t by sensors S1 to Sn and sensor data obtained at time t k0 and time t k1From the combination of sensor data, time t k0 and time t kp It is only necessary to store p pieces of state information corresponding to the combinations of sensor data up to the combination of sensor data 11 in the storage unit 11 or the like.

[0087] (Method of determining the degree of match) The degree of coincidence may be calculated, for example, as follows. In the following description, the relation index calculated using the sensor data when the waste incineration plant P is in a reference state is referred to as the reference-time relation index. The reference map described above can be created using the reference-time relation index. The state that is the target for determining the degree of coincidence with the reference state is referred to as the target state.

[0088] Prior to determining the degree of match, the classification unit 102 first classifies the sensor data acquired by the data acquisition unit 101 when the waste incineration plant P is in the target state based on the magnitude of its value. Next, the degree of match determination unit 201 calculates the degree of match that indicates the degree to which the sensor data acquired by the data acquisition unit 101 matches each classification. For example, the degree of match determination unit 201 may determine the degree of match using a membership function such as that shown in FIG. 3.

[0089] Next, the match determination unit 201 calculates the degree of match for each segment based on the reference-time relation index for each segment and the degree of fit calculated for that segment.The match determination unit 201 then calculates the degree of match for each combination of sensors S from the degree of match for each segment, and determines the final degree of match as the sum of the degrees of match calculated for each combination.In the following, to distinguish between these three types of match, each degree of match may be referred to as the "segment-group-based match degree," "sensor-group-based match degree," and "status-information-based match degree," respectively.

[0090] An example of calculating the degree of coincidence will be explained below with reference to Fig. 8. Fig. 8 is a diagram showing an example of calculating the degree of coincidence. It is assumed that the reference time-related index generated for sensors S1 and S2 using sensor data with a time difference of Δt is the value shown in 131 in Fig. 8. 131 is a portion of the state information for the reference state (the portion corresponding to the combination of sensors S1 and S2). It is also assumed that the value of the sensor data for sensor S1 at time t0 during a period in which the waste incineration plant P is in the target state is v1, and the value of the sensor data for sensor S2 at time (t0 + Δt) during the same period is v2.

[0091] In this case, as shown in 141 of FIG. 8, the sorting unit 102 calculates the degree of conformance of v1 with respect to each category (Small, Middle, Large) of the magnitude of the value from the sensor S1. To calculate the degree of conformance, a membership function for each category of the magnitude of the value from the sensor S1 may be used. In the example of 141 of FIG. 8, the degrees of conformance of v1 with each category of "Small", "Middle", and "Large" are calculated to be 0.0, 0.7, and 0.3, respectively. These numerical values ​​indicate the classification results of the sensor data v1 by the sorting unit 102.

[0092] Similarly to the above, the sorting unit 102 calculates the degree of conformance of v2, which is the sensor data value of the sensor S2, with each of the value magnitude categories (Small, Middle, Large) of the sensor S2 value. In the example of 141 in Fig. 8, the degrees of conformance of v2 with each of the categories "Small", "Middle", and "Large" are calculated to be 0.4, 0.6, and 0.0, respectively. These numerical values ​​indicate the classification results of the sensor data v2 by the sorting unit 102.

[0093] The matching degree determination unit 201 then multiplies the matching degree of v1 by the matching degree of v2 to calculate the matching degree for each of the categories of the value size combinations. In the example 141 in Fig. 8, the matching degree for the "Middle"-"Small" category is 0.7 x 0.4 = 0.28, and the matching degree for the "Middle"-"Middle" category is 0.7 x 0.6 = 0.42. The matching degree for the "Large"-"Small" category is 0.3 x 0.4 = 0.12, and the matching degree for the "Large"-"Middle" category is 0.3 x 0.6 = 0.18. The matching degrees for the other categories are 0.0.

[0094] Next, the matching degree determination unit 201 calculates the segment-pair unit matching degree for each segment of the combination of magnitudes by multiplying the reference-time relation index by the degree of conformance calculated for that segment. In the example 131 of Fig. 6, the reference-time relation index for the "Middle"-"Middle" segment is 0.35, the reference-time relation index for the "Large"-"Small" segment is 0.8, and the reference-time relation index for the other segments is 0.0. Therefore, as shown in 142 of Fig. 8, the segment-pair unit matching degree for the "Middle"-"Middle" segment is 0.35 x 0.42 = 0.147, the segment-pair unit matching degree for the "Large"-"Small" segment is 0.12 x 0.8 = 0.096, and the segment-pair unit matching degrees for the other segments are 0.0.

[0095] The match determination unit 201 then sums the group-unit match degrees and sets this as the sensor-unit match degree when the sensor data of v1 and v2 were measured. That is, the sensor-unit match degree in the example 142 in Fig. 8 is 0.147 + 0.096 = 0.243. The match determination unit 201 performs the above process for all combinations of sensors S, and sets the sum of the sensor-unit match degrees calculated for each combination as the state information unit match degree between the target state and the reference state.

[0096] The degree of coincidence of status information units calculated in this manner becomes larger the more similar the distribution of the sensor data acquired by the data acquisition unit 101 is to the distribution of the sensor data when the waste incineration plant P is in the reference state. Therefore, the degree of coincidence of status information units can be said to indicate the degree of similarity between the distribution of the sensor data when the waste incineration plant P is in the reference state and the distribution of the sensor data acquired by the data acquisition unit 101.

[0097] (About using multiple state information) In this embodiment, multiple pieces of state information (all of which are state information for the reference state) are used, which are generated using sets of sensor data each having a different combination of acquired timings (in other words, the above-mentioned Δt value indicating the time offset of the sensor data). For this reason, the coincidence determination unit 201 performs a process for determining the degree of coincidence of each piece of state information for each piece of state information based on the state information for the reference state and the classification result by the classification unit 102 of the sensor data acquired in the target state. The coincidence determination unit 201 then calculates an overall degree of coincidence by combining the degrees of coincidence of each piece of state information determined in each process. This configuration makes it possible to calculate an overall degree of coincidence that takes into account multiple combinations of timings at which sensor data were acquired, thereby enabling even more accurate state understanding.

[0098] Alternatively, the coincidence determination unit 201 may calculate the overall coincidence as a weighted sum of the status information unit coincidences determined for each of the plurality of status information. In this case, the weight by which each status information unit coincidence is multiplied is preferably calculated by repeatedly calculating the overall coincidence for a target facility in a predetermined state, updating the weight by which each of the plurality of status information unit coincidences is multiplied so as to reduce the error in the calculated overall coincidence, while changing the sensor data used to calculate the overall coincidence. This makes it possible to calculate a highly accurate overall coincidence that reflects the relative importance of multiple combinations of timings at which sensor data was obtained. The method of calculating the weights will be described in detail later.

[0099] 9 is a diagram showing an example of calculation of the overall degree of agreement. More specifically, in the illustrated example, the overall degree of agreement between the state (target state) of the waste incineration plant P at time t and the reference state is calculated. The overall degree of agreement is calculated using the sensor data v1(t) to v n (t) is used. As shown in the figure, t∈{t k0 ,t k1 ,…,t kp}. Then, from these sensor data, correlation maps M0 to M p The correlation map M0~M p indicates the degree of coincidence between the target state and the reference state for each combination of sensors (sensor group unit coincidence degree). p From the state information unit matching degree e0~e p In this embodiment, it is not essential to present the state information unit coincidence degree to the user, so the correlation maps M0 to M p There is no need to generate an image.

[0100] As mentioned above, the correlation maps M0 to M p indicates the degree of match (sensor group unit match) for each combination of sensors between the target state and the reference state. As described above, the sensor group unit match is the sum of multiple group unit matches calculated by multiplying the reference time relation index of each group for a sensor group by the conformance of that group. Therefore, each correlation map M0 to M p contains the same number of sensor pair unit matches as there are sensor combinations.

[0101] Correlation map M0~M p is generated using sensor data with different timing combinations (in other words, different time lag patterns of sensor data). For example, the correlation map M p is (t kp -t k0 ) (obtained during the target state). Similarly, the correlation map M1 is generated using time-delayed sensor data (obtained during the target state). k1 -t k09, multiple correlation maps M1 to M2 are generated using sensor data sets with different timing combinations (in other words, different time lag patterns of the sensor data). p is generated.

[0102] Here, the correlation map M0 is obtained at timing t k0 The correlation map M0 is generated using the sensor data without any time lag. Therefore, the status information unit coincidence e0 calculated from the correlation map M0 is also calculated using the sensor data without any time lag. In this way, the multiple status information unit coincidences used to calculate the overall coincidence may include status information unit coincidences generated using sensor data without any time lag.

[0103] The coincidence determination unit 201 uses the correlation map M p The sum of the matching scores for each sensor pair shown in is used as the correlation map M p The degree of agreement between the state information units e p The coincidence determination unit 201 calculates the correlation map M p-1 Similarly, for ~M0, the state information unit matching degree e p-1 Then, the coincidence determination unit 201 calculates the calculated state information unit coincidence e p ~e0 each has weight w p The overall match score e is calculated by multiplying the scores by ∼w0 and adding them together. The overall match score e calculated in this way includes (t kp -t k0 )~(t k1 -t k0 ) reflects the correlation between the sensor data with a time lag, and also reflects the correlation between the sensor data with no time lag. Therefore, the overall degree of agreement e makes it possible to accurately grasp the state of the waste incineration plant P.

[0104] (Processing flow) The flow of the process (determination method) executed by information processing device 1B will be described with reference to Fig. 10. Fig. 10 is a flowchart showing an example of the process executed by information processing device 1B. Note that the process in ST27 is similar to ST16 in Fig. 6, and therefore description thereof will not be repeated here.

[0105] In ST21, the data acquisition unit 101 acquires sensor data from all sensors S. For example, if the time lag of the sensor data to be combined is one minute, the data acquisition unit 101 may acquire sensor data for a period obtained by adding one minute to a predetermined time. This results in a set of sensor data for a predetermined period of time, each of which is acquired at a one-minute lag. In this case, the state of the waste incineration plant P during the period of the predetermined time + one minute is the target state.

[0106] In ST22 (classification step), the classification unit 102 classifies the sensor data acquired in ST21 according to the magnitude of its values. For example, the classification unit 102 may classify the sensor data into three levels, namely, small, middle, and large, or into two levels or four or more levels. The classification unit 102 may also classify the sensor data based on, for example, a threshold value, or may classify the sensor data using a fuzzy set. As described above, the classification result of the classification unit 102 is expressed in the form of a degree of conformance of the sensor data to each level.

[0107] In ST23 (matching degree determination step), the matching degree determination unit 201 determines the degree of match between the target state and the reference state, i.e., the degree of match per state information, based on the state information of the reference state and the classification result in ST22. This process is performed for each of the multiple state information. Specifically, the determination of the degree of match per state information is performed by (1) calculating the degree of match for each combination of value magnitude classifications in the sensor data sets, (2) multiplying the calculated degree of match by the reference time relation index included in the state information of the reference state by the calculated degree of match, (3) calculating the sum of the calculated degrees of match per sensor data set as the degree of match per sensor set, and (4) calculating the sum of the calculated degrees of match per sensor set as the degree of match per state information.

[0108] For example, assume that in ST22, the degree of conformance of each sensor data set for the three categories of “Small,” “Middle,” and “Large” is calculated. In this case, the matching determination unit 201 calculates the degree of conformance for each of the nine category combinations for each sensor data set by multiplying the degree of conformance for each category by the corresponding reference-time relation index. Then, the matching determination unit 201 calculates the degree of conformance for each of the nine category combinations by multiplying the corresponding reference-time relation index. For example, the degree of conformance for the category combination “Small”-“Small” in the combination of sensor data from sensor S1 and sensor data from sensor S2 is multiplied by the reference-time relation index for that category combination included in the state information of the reference state. In this way, the value of “conformance × reference-time relation index” calculated for each category combination is summed to calculate the degree of conformance for each sensor set for the combination of sensor data from sensor S1 and sensor data from sensor S2. In the same manner, the degrees of conformance for each sensor set are calculated for the other sensor data sets, and the sum of these values ​​is the degree of conformance for each state information unit.

[0109] In ST24, the coincidence determination unit 201 calculates an overall coincidence by summing up the status information unit coincidences for each of the plurality of status information calculated in ST23. As described above, the coincidence determination unit 201 may calculate the overall coincidence as a weighted sum of the status information unit coincidences corresponding to each of the plurality of status information.

[0110] In ST25, the anomaly detection unit 202 determines whether or not there is an anomaly based on the overall coincidence generated in ST24. If it is determined that there is an anomaly in ST25 (YES in ST25), the process proceeds to ST26, where the anomaly detection unit 202 notifies an operator or the like that it has detected an anomaly. On the other hand, if it is determined that there is no anomaly in ST25 (NO in ST25), the process proceeds to ST27.

[0111] As described above, the determination method according to this embodiment includes a classification step (ST22) in which a process of classifying sensor data obtained by sensors S installed in the waste incineration plant P according to the magnitude of the values ​​is performed on sensor data obtained at different times for each of multiple sensors S installed in the waste incineration plant P, and a coincidence determination step (ST23) in which a degree of coincidence (degree of coincidence per state information unit) between the state of the waste incineration plant P at the time the sensor data classified in the classification step was acquired and the reference state is determined based on state information representing the reference state based on the relationship between the magnitudes of the values ​​of the sensor data obtained by the multiple sensors S at different times when the waste incineration plant P is in a predetermined reference state and the classification result in the classification step. Thus, even if the timing at which the state of the waste incineration plant P is reflected in the sensor data differs for each sensor S, it is possible to understand the state of the waste incineration plant P by taking the sensor data into consideration.

[0112] [Embodiment 3] (Device configuration) An information processing device 1C of this embodiment will be described with reference to Fig. 11. Fig. 11 is a diagram showing the configuration of the information processing device 1C. The information processing device 1C differs from the information processing device 1A of the first embodiment in that it has a function for optimizing weights when calculating the overall degree of coincidence. As shown in the figure, the information processing device 1C has a control unit 10C, which includes a data acquisition unit 101, a classification unit 102, a degree of coincidence determination unit 201, and a weight update unit 301.

[0113] As explained in the second embodiment, the coincidence determination unit 201 determines the degree of coincidence between a predetermined state and a reference state (the above-mentioned state information unit coincidence) based on a plurality of state information representing the reference state of the waste incineration plant P and a classification result obtained by classifying, by magnitude of values, sensor data obtained by a plurality of sensors S at different times when the waste incineration plant P is in a predetermined state. Then, the coincidence determination unit 201 calculates an overall coincidence by combining the plurality of state information unit coincidences.

[0114] As explained in the second embodiment, the plurality of status information are generated using sets of sensor data with different combinations of acquired timing (in other words, different values ​​of the time difference Δt between when the sensor data was acquired). Also, "sensor data acquired at different timings" can be rephrased as "sensor data with a timing difference (Δt) between when the sensor data was acquired." The degree of agreement is determined using the status information and classification results generated using sets of sensor data with the same time difference Δt. In other words, in a waste incineration plant P in a reference state, p The sensor data obtained at time (t1+Δt)~(t p +Δt) and the state information (information showing the relationship index for each combination of sensors S) generated using the sensor data obtained at time t1' to t p The sensor data obtained at time (t1'+Δt)~(t p The degree of match is determined using the classification results of the sensor data obtained at time t'+Δt.

[0115] The weight update unit 301 updates the weights by which each of the plurality of status information unit coincidences is multiplied so as to reduce the error in the overall coincidence, which is the weighted sum of the plurality of status information unit coincidences determined for the plurality of status information by the coincidence determination unit 201. In the example of FIG. 9, the weights updated by the weight update unit 301 are w p ~w0.

[0116] According to the information processing device 1C having the above configuration, it is possible to automatically optimize weights according to the relative importance of a plurality of combinations of timings at which sensor data is obtained.

[0117] For example, the weight update unit 301 may calculate the updated weight w' using the following formula: where w is the weight before update, L is the error, and η is the learning rate. w'=w-η×(∂L / ∂w) The method for calculating the error is not particularly limited. For example, the weight update unit 301 may calculate the error by: 2Alternatively, the error may be calculated as e / 2. Note that e is the calculated state information unit matching degree, and e' is the correct value of the state information unit matching degree.

[0118] Here, the multiple status information unit coincidences used in calculating the overall coincidence may be the sum of the products of the relationship indices indicating the relationship between each set of multiple sensor data, calculated from the sensor data when the waste incineration plant P is in a reference state, and the conformance of each set of sensor data to each category, i.e., the weighted sum of the sensor group unit coincidences. In this case, when updating the weights by which each of the multiple status information unit coincidences is multiplied, the weight update unit 301 preferably also updates the weights by which each sensor group unit coincidence is multiplied. This makes it possible to automatically optimize the weights by which the sensor group unit coincidences are multiplied.

[0119] Here, the relationship between the weight multiplied by the sensor group unit coincidence degree and the weight multiplied by the status information unit coincidence degree will be explained based on Fig. 12. Fig. 12 is a diagram showing the relationship between the weight multiplied by the sensor group unit coincidence degree and the weight multiplied by the status information unit coincidence degree. Note that in Fig. 12, similarly to Fig. 9, correlation maps M0 to M p The state information unit agreement calculated from e0 to e p The weights to be multiplied by these state information unit matching degrees are w0 to w p It states that:

[0120] Also, on the right side of FIG. 12, the sensor pair unit matching degree e included in the correlation map M0 is 0 11 ~e 0 nn As shown in the figure, the sensor pair unit match is calculated for each combination of sensors S. In FIG. 12, the sensor pair unit match is shown with the number of the corresponding correlation map superscripted and a number indicating the corresponding sensor combination subscripted. For example, the sensor pair unit match for the combination of sensors S1 and S2 in correlation map M0 is e 0 12 This becomes:

[0121] The correlation map M1 in FIG. 12 is k0 and t k1 The classification result of the sensor data by the classification unit 102 is k1 -t k0 The correlation map M1 is calculated using the sensor data (obtained under the reference condition) and the reference time relation index generated using the sensor data (obtained under the reference condition). 1 11 ~e 1 nn These sensor pair unit matching scores e 1 11 ~e 1 nn In Figure 12, the weight to be multiplied is w 1 11 ~w 1 nn is shown as

[0122] In this way, the sensor group unit matching degree e 0 11 ~e p nn is calculated from the classification result of the sensor data and the reference state relation index. Also, as can be seen from FIG. 12, the sensor group unit matching degree e 0 11 ~e p nn Once the state information unit match degree e0~e p is determined, and the state information unit matching degree e0~e p Once the overall agreement score e is determined, the weight w 1 11 ~w p nn is the degree of agreement of each state information unit e0 to e p It affects the value of each state information unit match e0~e p Weights w0~w to be multiplied p affects the overall agreement value e.

[0123] Therefore, for the given sensor data, the weight w is set so that the calculated value of the overall agreement e approaches the correct value, in other words, so that the error of the calculated value of the overall agreement e becomes small. 11 ~w nn and w0~wp By repeating this process of updating the weights while changing the sensor data, it is possible to optimize each of the weights. 11 ~w nn and w0~w p These weights do not necessarily have to be updated simultaneously, and may be updated individually.

[0124] (Processing flow) The flow of the process (weight optimization method) executed by information processing device 1C will be described with reference to Fig. 13. Fig. 13 is a flowchart showing an example of the process executed by information processing device 1C. Note that the processes of ST32 to ST34 are generally similar to the processes of ST22 to ST24 in Fig. 10, and therefore will not be described again here.

[0125] In ST31, the data acquisition unit 101 acquires sensor data to be used for optimizing the weights. The sensor data to be used for optimizing the weights is sensor data obtained when the waste incineration plant P is in a predetermined state. The predetermined state is a state in which the state is known, in other words, a state in which the correct value of the overall degree of agreement in that state is known. For example, since the sensor data used when generating the reference map is sensor data obtained when the waste incineration plant P is in a reference state, in ST31 the data acquisition unit 101 may acquire the sensor data used when generating the reference map. Note that in ST31, as in ST21, the data acquisition unit 101 acquires time-series sensor data so that sets of sensor data with a time lag can be generated.

[0126] In ST32 to ST34, the overall degree of coincidence is calculated using the sensor data acquired in ST31. Then, in ST35, the weight update unit 301 updates the weight by which each of the multiple state information unit degrees of coincidence is multiplied so as to reduce the error in the overall degree of coincidence calculated in ST34. At this time, the weight update unit 301 may also update the weight by which each sensor set unit degree of coincidence is multiplied.

[0127] In ST36, the weight update unit 301 determines whether or not to end the weight update. If the determination in ST36 is YES, the illustrated processing ends. On the other hand, if the determination in ST36 is NO, the processing returns to ST31. In ST31 after transition from ST36, new sensor data is acquired, and the processing from ST32 onwards is repeated using the sensor data.

[0128] The condition for terminating the weight update in ST36 (hereinafter referred to as the termination condition) may be determined as appropriate. For example, the termination condition may be that ST31 to ST35 have been repeated a predetermined number of times. Alternatively, the termination condition may be that the error in the overall degree of agreement calculated in ST34 is equal to or less than a predetermined threshold. In this case, the process of ST36 is performed before the process of ST35.

[0129] As described above, the weight optimization method according to this embodiment includes a match determination step (ST33) for determining the degree of match between a predetermined state and a reference state based on a plurality of state information pieces representing the reference state using a relationship index indicating the relationship between the magnitudes of sensor data values ​​obtained by a plurality of sensors S installed in the waste incineration plant P at different times when the waste incineration plant P is in a predetermined reference state, and a classification result obtained by classifying the sensor data pieces obtained by the plurality of sensors S at different times when the waste incineration plant P is in the predetermined state based on the magnitudes of the sensor data values, and a weight update step (ST35) for updating the weight by which each of the plurality of state information unit matches is multiplied so as to minimize the error in the overall match, which is the weighted sum of the plurality of state information unit matches determined for the plurality of state information pieces in the match determination step. Thus, weights can be automatically optimized for a plurality of combinations of timings at which sensor data was obtained, according to their relative importance.

[0130] [Modification] The entity that executes each process described in each of the above embodiments can be changed as appropriate. That is, a plurality of information processing devices (which can also be called processors) that can communicate with each other can realize functions similar to those of information processing device 1A, 1B, or 1C. For example, the processes shown in FIGS. 6, 10, and 13 may be shared and executed by a plurality of information processing devices. That is, the entity that executes each method according to the above embodiments may be one information processing device or multiple information processing devices.

[0131] Furthermore, the configuration of dividing sensor data into sets such as large, medium, and small is one method for expressing the distribution of sensor data, and can be replaced with other methods capable of expressing the distribution of sensor data.

[0132] [Software implementation example] The functions of information processing devices 1A, 1B, and 1C can be realized by programs (status information generation program / judgment program / weight determination program) that cause a computer to function as information processing device 1A, 1B, or 1C, and that cause a computer to function as each control block of information processing device 1A, 1B, or 1C (particularly each part included in control unit 10A, 10B, or 10C).

[0133] In this case, the information processing device 1A, 1B, or 1C includes a computer having at least one control device (e.g., a processor) and at least one storage device (e.g., a memory) as hardware for executing the program. The control device and storage device execute the program, thereby realizing the functions described in each of the above embodiments.

[0134] The program may be non-transitory and may be recorded on one or more computer-readable recording media. The recording media may or may not be included in the information processing device 1A, 1B, or 1C. In the latter case, the program may be supplied to the information processing device 1A, 1B, or 1C via any wired or wireless transmission medium.

[0135] Furthermore, some or all of the functions of the control blocks can be realized by logic circuits. For example, an integrated circuit in which a logic circuit that functions as each of the control blocks is formed is also included in the scope of the present invention. In addition, the functions of the control blocks can also be realized by, for example, a quantum computer.

[0136] The present invention is not limited to the above-described embodiments, and various modifications are possible within the scope of the claims. Embodiments obtained by appropriately combining the technical means disclosed in different embodiments are also included in the technical scope of the present invention. [Explanation of symbols]

[0137] 1A Information processing equipment 102 Division 103 Relationship Index Calculation Unit 104 Status information generation unit 1B Information processing equipment 102 Division 201 Matching degree judgment section 1C Information processing equipment 102 Division 201 Matching degree judgment section 301 Weight Update Unit

Claims

1. a relationship index calculation unit that calculates, for each pair of sensors formed by a plurality of sensors installed in the target facility, a relationship index that indicates a relationship between sensor data obtained by one of the pair of sensors and sensor data obtained by the other of the pair of sensors; a status information generating unit that generates status information indicating a status of the target facility using the relationship index calculated for each of the sets of sensors, The information processing device, wherein the relationship index calculation unit calculates the relationship index for a pair of sensor data obtained at different times.

2. 2. The information processing device according to claim 1, wherein the state information is an image in which a pattern according to the value of the relation index calculated for each set of the plurality of sensors is drawn in each section corresponding to each set defined on an image plane.

3. a classification unit that performs a process of classifying sensor data obtained by sensors installed in a target facility according to the magnitude of the values ​​of the sensor data, the process being performed on the sensor data obtained at different times for each of a plurality of sensors installed in the target facility; an information processing device comprising: state information representing the reference state using a relationship index indicating the relationship calculated based on the relationship between the magnitudes of the sensor data values ​​obtained by the multiple sensors at different times when the target facility is in a predetermined reference state; and a consistency determination unit that determines the consistency between the state of the target facility when the sensor data classified by the classification unit was obtained and the reference state based on the classification result by the classification unit.

4. 4. The information processing device according to claim 3, wherein the matching degree determination unit performs a process of determining a degree of matching based on the status information and the classification results by the classification unit for each of a plurality of pieces of status information generated using sets of sensor data each having a different combination of timing obtained, and calculates an overall degree of matching by combining the degrees of matching determined in each process.

5. the matching degree determination unit calculates a weighted sum of the matching degrees determined for each of the plurality of pieces of status information as the overall matching degree; 5. The information processing device according to claim 4, wherein the weights are calculated by repeating a process of calculating the overall degree of coincidence for the target facility in a predetermined state, updating a weight by which each of the multiple degrees of coincidence is multiplied so as to reduce an error in the calculated overall degree of coincidence, while changing the sensor data used to calculate the overall degree of coincidence.

6. a coincidence determination unit that determines the degree of coincidence between a predetermined state and a reference state based on a plurality of state information pieces that represent the reference state using a relationship index that indicates a relationship calculated based on the relationship between the magnitudes of values ​​of sensor data obtained by a plurality of sensors installed in the facility at different times when the facility is in the predetermined state, and a classification result that classifies the sensor data obtained by the plurality of sensors at different times when the facility is in the predetermined state based on the magnitudes of the values; and a weight update unit that updates the weights by which each of the multiple degrees of agreement is multiplied so as to reduce an error in the overall degree of agreement, which is the weighted sum of the degrees of agreement determined by the degree of agreement determination unit for the multiple pieces of status information.

7. the degree of coincidence is a weighted sum of products of relationship indices indicating a relationship between each set of the plurality of sensor data, calculated from the sensor data when the target facility is in the reference state, and a degree of conformance of each set of sensor data to each category; The information processing apparatus according to claim 6 , wherein the weight update unit also updates a weight by which a product of the relation index and the degree of matching is multiplied when updating the weight by which each of the plurality of degrees of matching is multiplied.

8. A method for generating state information executed by one or more information processing devices, comprising: a relationship index calculation step of calculating, for each pair of sensors comprising a plurality of sensors installed in the target facility, a relationship index indicating a relationship between sensor data obtained by one of the pair of sensors and sensor data obtained by the other of the pair of sensors; a status information generating step of generating status information indicating a status of the target facility using the relationship index calculated for each of the sets of sensors, The method for generating state information, wherein the relationship index calculating step calculates the relationship index for a pair of sensor data obtained at different times.

9. A determination method executed by one or more information processing devices, a classification step of classifying sensor data obtained by sensors installed in the target facility according to the magnitude of the values ​​of the sensor data, the classification step being performed on the sensor data obtained at different times for each of a plurality of sensors installed in the target facility; a consistency determination step of determining the consistency between the state of the target facility when the sensor data classified in the classification step was acquired and the reference state, based on state information representing the reference state based on the relationship between the magnitudes of the sensor data values ​​obtained by the plurality of sensors at different times when the target facility is in a predetermined reference state, and the classification result in the classification step.

10. 2. A state information generating program for causing a computer to function as the information processing device according to claim 1, the state information generating program causing a computer to function as the relationship index calculation unit and the state information generating unit.

Images