Apparatus, method and program

Abstract

To provide a system that supports condition monitoring of a target.SOLUTION: A device is provided which comprises: a measurement data acquisition unit that acquires plural types of measurement data indicating a condition of a target; a group information acquisition unit that acquires group information indicating a group of measurement data having a higher correlation than a reference among the plural types of measurement data; a specifying unit that specifies abnormal measurement data among the measurement data acquired by the measurement data acquisition unit for each group indicated by the group information; a supply unit that supplies abnormal measurement data specified for each group by the specifying unit to a model that outputs a condition index value indicating an acceptability of the condition of the target in response to any one type of the measurement data being input for each group indicated by the group information; and an output unit that outputs a signal according to the condition index value output from the model.SELECTED DRAWING: Figure 1

Description

[Technical Field]

[0001] The present invention relates to an apparatus, a method, and a program. [Background technology]

[0002] Patent Document 1 states that "the output unit 12c receives the sensor data collected by the collection unit 12a as input and outputs the degree of abnormality of each detection target device 20 using each model." [Prior art document] [Patent documents] [Patent Document 1] Patent No. 6453504 Summary of the Invention

[0003] In a first aspect of the present invention, there is provided an apparatus including: a measurement data acquiring unit that acquires multiple types of measurement data indicating a state of an object; a group information acquiring unit that acquires group information indicating groups of measurement data that have a higher correlation than a reference value among the multiple types of measurement data; an identifying unit that identifies, for each group indicated by the group information, abnormal measurement data among the measurement data acquired by the measurement data acquiring unit; a supply unit that supplies the abnormal measurement data identified for each group by the identifying unit to a model that outputs a state index value indicating the quality of the state of the object in response to input of any one type of measurement data for each group indicated by the group information; and an output unit that outputs a signal corresponding to the state index value output from the model.

[0004] The apparatus may further include a designation unit that designates a target time point for the condition index value. The identification unit may identify anomalous measurement data at the target time point. The supply unit may supply the anomalous measurement data at the target time point to the model.

[0005] The identification unit may determine that, among the multiple measurement data in the first group, the measurement data having the largest difference between the average rate of change in a reference period including the target time point and the rate of change at the target time point is abnormal measurement data in the first group.

[0006] The identification unit may determine, as abnormal measurement data in the second group, the measurement data that causes the model to output the worst condition index value when the measurement data in the second group at the target time point are separately supplied to the model by the supply unit.

[0007] The identification unit may determine that one of the multiple measurement data in the third group is abnormal measurement data in the third group when the correlation coefficient between the one measurement data and each of the other measurement data is less than a threshold value during a reference period including the target time point.

[0008] The identification unit may define a group including three or more types of measurement data as the third group, and when a correlation coefficient between one of the plurality of measurement data in the third group and any other measurement data is equal to or greater than a threshold, may define the measurement data having the lowest correlation coefficient with the one measurement data among the plurality of measurement data in the third group as abnormal measurement data in the third group.

[0009] The identification unit may determine, among the plurality of measurement data in the fourth group, the measurement data having the largest autocovariance or the lowest autocorrelation at the target time point as the abnormal measurement data in the fourth group.

[0010] The identification unit may determine, among the measurement data in a fifth group, the measurement data that has the largest difference between a reference time point and the target time point as abnormal measurement data in the fifth group, and may determine, among the acquisition times of data sets of two or more types of measurement data not included in the fifth group, the acquisition time point of the data set that is most similar to the data set at the target time point as the reference time point.

[0011] In any of the above devices, the supply unit may supply the abnormal measurement data identified for each group by the identification unit to the model that outputs the condition index value in response to input of each measurement data of an arbitrary combination pattern in which any one type of measurement data is selected for each group.

[0012] The apparatus may further include a normalization unit that normalizes each of the measurement data included in at least one group, and the supply unit may supply the abnormal measurement data normalized by the normalization unit and identified for each group by the identification unit to the model.

[0013] The above-described device may further include a normalization unit that normalizes each piece of measurement data included in at least one group, and a learning processing unit that performs a learning process for the model using learning data including the measurement data of each group normalized by the normalization unit and the state index value.

[0014] In any of the above devices, the supply unit may supply the abnormal measurement data to the model corresponding to the abnormal measurement data combination pattern identified for each group by the identification unit, among the multiple models provided for each combination pattern of measurement data selected one type from each group, and which output the condition index value in response to input of each measurement data of the corresponding combination pattern.

[0015] The above device may further include a learning processing unit that performs a learning process of the model for each combination pattern of measurement data selected one type from each group, using learning data including each measurement data included in the combination pattern and the state index value, and outputs the state index value in response to input of each measurement data of the combination pattern.

[0016] In a second aspect of the present invention, there is provided an apparatus which may include a measurement data acquiring unit which acquires multiple types of measurement data indicating a state of an object, a group information acquiring unit which acquires group information indicating groups of measurement data among the multiple types of measurement data which have a higher correlation than a reference value, and a learning processing unit which, for each combination pattern of measurement data selected one type from each group, performs a model learning process which uses learning data including each measurement data included in the combination pattern and a state index value indicating whether the state of the object is good or bad in response to input of each measurement data of the combination pattern.

[0017] In any of the above devices, the group information acquisition unit may include a calculation unit that calculates a correlation coefficient for each combination of measurement data in the multiple types of measurement data. The group information acquisition unit may include a grouping unit that groups measurement data combinations having a correlation coefficient higher than a reference correlation coefficient. The group information acquisition unit may include a generation unit that generates the group information indicating the groups of grouped measurement data.

[0018] In any of the above devices, the group information may indicate, as separate groups, each of the plurality of types of measurement data whose correlation with each of the other measurement data is lower than a reference value.

[0019] In a third aspect of the present invention, there is provided a method, which may include: a measurement data acquiring step of acquiring multiple types of measurement data indicating a state of an object; a group information acquiring step of acquiring group information indicating groups of measurement data among the multiple types of measurement data that have a higher correlation than a reference value; an identifying step of identifying, for each group indicated by the group information, abnormal measurement data among the measurement data acquired in the measurement data acquiring step; a supply step of supplying the abnormal measurement data identified for each group in the identifying step to a model that outputs a state index value indicating the quality of the state of the object in response to input of any one type of measurement data for each group indicated by the group information; and an output step of outputting a signal corresponding to the state index value output from the model.

[0020] In a fourth aspect of the present invention, there is provided a method, which may include: a measurement data acquisition step of acquiring multiple types of measurement data indicating a state of an object; a group information acquisition step of acquiring group information indicating groups of measurement data among the multiple types of measurement data that have a higher correlation than a reference value; and a learning processing step of performing, for each combination pattern of measurement data selected one type from each group, a model learning process that uses learning data including each measurement data included in the combination pattern and a state index value indicating whether the state of the object is good or bad, in response to input of each measurement data of the combination pattern.

[0021] In a fifth aspect of the present invention, there is provided a program that causes a computer to function as a measurement data acquiring unit that acquires multiple types of measurement data indicating a state of an object, a group information acquiring unit that acquires group information indicating groups of measurement data that have a higher correlation than a reference value among the multiple types of measurement data, an identifying unit that identifies, for each group indicated by the group information, abnormal measurement data among the measurement data acquired by the measurement data acquiring unit, a supply unit that supplies the abnormal measurement data identified for each group by the identifying unit to a model that outputs a state index value indicating the quality of the state of the object in response to input of any one type of measurement data for each group indicated by the group information, and an output unit that outputs a signal corresponding to the state index value output from the model.

[0022] In a sixth aspect of the present invention, there is provided a program that causes a computer to function as a measurement data acquiring unit that acquires multiple types of measurement data indicating a state of an object, a group information acquiring unit that acquires group information indicating groups of measurement data that have a higher correlation than a reference value among the multiple types of measurement data, and a learning processing unit that performs, for each combination pattern of measurement data selected one type from each group, a model learning process that uses learning data including each measurement data included in the combination pattern and a state index value indicating whether the state of the object is good or bad, in response to input of each measurement data of the combination pattern.

[0023] The above summary of the invention does not list all of the necessary features of the present invention, and subcombinations of these features may also constitute inventions. [Brief explanation of the drawings]

[0024] [Figure 1] 1 shows a system 1 according to an embodiment. [Figure 2] 4 shows the data structure of a group information file 431. [Figure 3]The operation of the device 4 when performing the learning process of the model 435 is shown. [Figure 4] The operation of the device 4 when using the model 435 is shown. [Figure 5] 1 shows a system 1A according to a modified example. [Figure 6] 22 illustrates an example computer 2200 in which aspects of the present invention may be embodied, in whole or in part. DETAILED DESCRIPTION OF THE INVENTION

[0025] The present invention will be described below through embodiments of the invention, but the following embodiments do not limit the scope of the invention according to the claims. Furthermore, not all of the combinations of features described in the embodiments are necessarily essential to the solution of the invention.

[0026] [1. System 1 Configuration] 1 shows a system 1 according to this embodiment. The system 1 supports monitoring of the state of an object, and includes a facility 2 as an example of the object, and a device 4.

[0027] [1.1. Equipment 2] The facility 2 is provided with a plurality of sensors 20. For example, the facility 2 may be a plant provided with a plurality of devices 21, or may be a composite device combining a plurality of devices 21. Examples of the plant include industrial plants such as chemical and bio plants, plants that manage and control wellheads and surrounding areas of gas fields and oil fields, plants that manage and control power generation such as hydroelectric, thermal and nuclear power plants, plants that manage and control environmental power generation such as solar and wind power plants, and plants that manage and control water supply and sewage systems, dams, etc.

[0028] [1.1-1.Equipment 21] Each device 21 is an instrument, machine, or device, and may be, for example, an actuator such as a valve, pump, heater, fan, motor, or switch that controls at least one physical quantity such as pressure, temperature, pH, speed, or flow rate in the process of the facility 2. The devices 21 may be of different types from each other, or at least some of two or more devices 21 may be of the same type.

[0029] [1.1-2.Sensor 20] Each sensor 20 measures the state of the equipment 2. The sensor 20 may measure at least one physical quantity such as pressure, temperature, pH, speed, or flow rate. The sensor 20 may also measure the yield of the equipment 2, the proportion of impurities mixed in, the operating status of each device 21, or the like. The sensors 20 may be of different types from each other, or at least some of the two or more sensors 20 may be of the same type. As an example, the two or more sensors 20 may be temperature sensors installed at different positions in a furnace in the equipment 2. Of the multiple measurement data measured by the multiple sensors 20, some of the measurement data may show a high correlation with the other measurement data. Each sensor 20 may supply the measurement data to the device 4.

[0030] Communication between the sensor 20 and the device 4 may be performed using, for example, an ISA (International Society of Automation) communication protocol, such as at least one of wireless communication protocols ISA100 or WirelessHART (Highway Addressable Remote Transducer) (registered trademark), or at least one of wired communication protocols BRAIN (registered trademark), FOUNDATION Fieldbus or PROFIBUS.

[0031] [1.2.Device 4] The device 4 may use the measurement data to perform a learning process for the model 435. Additionally or alternatively, the device 4 may support monitoring of the facility 2 using the learned model 435. The device 4 includes a measurement data acquisition unit 401, a preprocessing unit 402, a storage unit 403, an input unit 404, a label addition unit 405, a group information acquisition unit 406, a learning processing unit 407, a designation unit 408, an identification unit 409, a supply unit 410, an output unit 411, and a display unit 412.

[0032] [1.2.1. Measurement data acquisition unit 401] The measurement data acquisition unit 401 acquires multiple types of measurement data that indicate the state of the equipment 2. The measurement data acquisition unit 401 may sequentially acquire each type of measurement data from each sensor 20. The measurement data acquisition unit 401 may also acquire measurement data that has been measured in advance by each sensor 20 collectively from an external storage device (not shown). If a sensor 20 is not performing measurement due to replacement, failure, or the like, the measurement data acquisition unit 401 does not need to acquire measurement data from that sensor 20. In this embodiment, the type of measurement data may differ for each sensor 20, as an example, but may also differ depending on the physical quantity of interest.

[0033] The measurement data acquiring unit 401 may add the measurement time and identification information of the measurement data (measurement data ID, also referred to as a tag) to the acquired measurement data. The measurement data ID may be identification information of the sensor 20 that performed the measurement. The measurement time of the measurement data may be the time when the measurement data was measured, or may be the time when the measurement data was acquired by the measurement data acquiring unit 401. The measurement time and measurement data ID may be added in advance to the measurement data supplied from the sensor 20. The measurement data acquiring unit 401 may store the acquired measurement data in a measurement data file 430 via the preprocessing unit 402.

[0034] [1.2.2. Preprocessing unit 402] The preprocessing unit 402 performs so-called preprocessing on the measurement data. The preprocessing unit 402 may store the preprocessed measurement data in a measurement data file 430 in the storage unit 403. Additionally or alternatively, the preprocessing unit 402 may read out the measurement data stored in the measurement data file 430, perform preprocessing on it, and store it in the measurement data file 430 as preprocessed measurement data. In this embodiment, as an example, the preprocessed measurement data may be read out by a group information acquisition unit 406, a learning processing unit 407, an identification unit 409, and a supply unit 410, which will be described later.

[0035] The preprocessing performed by the preprocessing unit 402 may be at least one of conventionally known outlier processing, normalization processing, and missing value processing, and other preprocessing may be performed in addition to or instead of these. Outlier processing may be processing to remove outliers from the measurement data, or processing to correct outliers to values ​​within a preset range for each type of measurement data. Normalization processing may be processing to correct each type of measurement data to a value within a common range among multiple types of measurement data (for example, scaling to a range of 0 to 1). Normalization processing may include processing to scale the mean of the measurement data to 0 and the variance to 1, i.e., standardization processing. Missing value processing may be processing to delete other measurement data that have the same measurement time as the missing measurement data, or processing to complement the missing measurement data, when the measurement data is missing.

[0036] [1.2.3. Storage section 403] The storage unit 403 stores various information. For example, the storage unit 403 may store a measurement data file 430, a group information file 431, and a plurality of models 435.

[0037] [1.2.3-1. Measurement data file 430] The measurement data file 430 stores the measurement data supplied from the measurement data acquisition unit 401. The measurement data may have been preprocessed by the preprocessing unit 402.

[0038] [1.2.3-2. Group Information File 431] The group information file 431 stores group information indicating groups (also referred to as measurement data groups) of measurement data that have a higher correlation than a standard among the multiple types of measurement data acquired by the measurement data acquisition unit 401. Note that even among measurement data included in the same measurement data group, one of the measurement data may become relatively abnormal, temporarily lowering the correlation.

[0039] The group information may indicate, among the multiple types of measurement data, each measurement data whose correlation with each other measurement data is lower than a reference value as a separate measurement data group, i.e., a single-element measurement data group. The group information may be supplied from a group information acquisition unit 406 described below.

[0040] [1.2.3-3.Model 435] Each model 435 outputs a status index value indicating the quality of the equipment 2 in response to input of any one type of measurement data for each measurement data group indicated by the group information. In this embodiment, as an example, each model 435 may be provided for each combination pattern of measurement data selected one type from each measurement data group, and may output a status index value in response to input of each measurement data of the corresponding combination pattern. Identification information (also referred to as pattern ID) of the corresponding combination pattern may be added to each model 435.

[0041] Here, each model 435 may output a condition index value (also referred to as a health index) that is not binarized into a value indicating that the condition of the equipment 2 is good and a value indicating that the condition of the equipment 2 is bad. For example, the model 435 may be trained using two values, a value indicating that the condition of the equipment 2 is good and a value indicating that the condition of the equipment 2 is bad, and may output a condition index value before binarization based on comparison with a threshold. In the present embodiment, as an example, the condition index value may be a positive value when the condition of the equipment 2 is good (or close to good), and may be a negative value when the condition of the equipment 2 is bad (or close to bad). The model 435 may be, for example, a support vector machine, or may be a trained model using other algorithms such as logistic regression, decision tree, or neural network. The model 435 may be a linear model.

[0042] [1.2.4. Input section 404] The input unit 404 receives operational input from an operator.

[0043] When the measurement data is displayed, the input unit 404 may receive an operation to add to the measurement data a label indicating whether the state of the equipment 2 is good or bad. The input unit 404 may supply the label adding unit 405 with a signal indicating that the operation has been performed.

[0044] The input unit 404 may receive an operation to specify a target time point for the state index value to be output by the model 435. The target time point may be the current time point or a time point in the past. The input unit 404 may supply a signal indicating that the operation has been performed to the specification unit 408.

[0045] [1.2.5. Label Addition Unit 405] In response to an operation by an operator, the label adding unit 405 adds a label to each piece of measurement data indicating whether the condition of the equipment 2 is good or bad. The label adding unit 405 may add a label indicating that the condition was good or bad to the corresponding measurement data in the measurement data file 430. The label added by the label adding unit 405 may be used in the learning process by the learning processing unit 407.

[0046] [1.2.6. Group information acquisition unit 406] The group information acquisition unit 406 acquires group information. The group information acquisition unit 406 may store the acquired group information in the group information file 431 in the storage unit 403.

[0047] The group information acquisition unit 406 may generate group information. In the present embodiment, as an example, the group information acquisition unit 406 may include a calculation unit 461, a grouping unit 462, and a generation unit 463.

[0048] [1.2.6-1. Calculation unit 461] The calculation unit 461 calculates a correlation coefficient for each combination of measurement data in the multiple types of measurement data acquired by the measurement data acquisition unit 401. The calculation unit 461 may calculate the correlation coefficient using measurement data for at least a portion of a period (also referred to as a period for calculating the correlation coefficient) of the entire period of measurement data stored in the measurement data file 430. The correlation coefficient is an index indicating the correlation between two pieces of data, and may be a value within a range of -1 to +1. The period for calculating the correlation coefficient may be arbitrarily specified by an operator, and may be any period obtained by equally dividing the entire measurement period of the measurement data in the measurement data file 430, for example. The calculation unit 461 may supply the correlation coefficient for each combination to the grouping unit 462.

[0049] [1.2.6-2. Grouping section 462] The grouping unit 462 groups together combinations of measurement data having a higher correlation coefficient than the reference correlation coefficient. The grouping unit 462 may group one piece of measurement data and each piece of measurement data whose correlation coefficient with the one piece of measurement data is higher than the reference correlation coefficient into the same measurement data group. The grouping unit 462 may group each piece of measurement data having no other measurement data with a higher correlation coefficient than the reference correlation coefficient, that is, each piece of measurement data having a correlation coefficient with each other piece of measurement data lower than the reference correlation coefficient, into separate measurement data groups. The reference correlation coefficient may be a fixed value set in advance (for example, a value of 0.9 or more, or a value of 0.95 or more), or may be a value that can be arbitrarily changed by the operator.

[0050] The grouping section 462 may supply the grouping results to the generating section 463. As an example, the grouping section 462 may supply the measurement data IDs of the grouped measurement data to the generating section 463 for each measurement data group.

[0051] [1.2.6-3.Generation unit 463] The generating unit 463 generates group information indicating the measurement data groups grouped by the grouping unit 462. The generating unit 463 may generate the group information by associating the measurement data IDs of the measurement data included in the measurement data group with each other. The generating unit 463 may add identification information (also referred to as group ID) of the measurement data group to the group information. The generating unit 463 may set the group ID of a group of measurement data whose correlation with other measurement data is lower than a standard as the measurement data ID of the measurement data. The generating unit 463 may store the generated group information in the group information file 431.

[0052] [1.2.7. Learning Processing Unit 407] The learning processing unit 407 performs learning processing on the model 435 for each combination pattern of measurement data selected one type from each measurement data group. The learning processing unit 407 may use learning data including each measurement data included in the combination pattern and a state index value to perform learning processing on the model 435 (in the present embodiment, as an example, the model 435 corresponding to the pattern ID of the combination pattern) that outputs a state index value in response to input of each measurement data of the combination pattern. The learning processing unit 407 may perform learning processing on each model 435 so that the difference in state index value due to differences in the combination patterns of measurement data falls within an arbitrary range (as an example, the range of standard error).

[0053] The value of the condition index value included in the training data may be a binary value corresponding to the label added to the measurement data by the label adding unit 405. In this embodiment, as an example, the value of the condition index value included in the training data may be "1" when a label indicating a good condition is added to the measurement data, and may be "-1" when a label indicating a bad condition is added to the measurement data.

[0054] [1.2.8.Specification part 408] The designation unit 408 designates a target time point of the state index value to be output by the model 435. The designation unit 408 may designate the target time point in response to an operation by an operator. The designation unit 408 may supply information indicating the designated target time point to the identification unit 409.

[0055] [1.2.9. Specification part 409] The identifying unit 409 identifies, for each measurement data group indicated by the group information, abnormal measurement data (also referred to as abnormal measurement data) from the measurement data acquired by the measurement data acquiring unit 401. The abnormal measurement data may be measurement data that is relatively abnormal compared to other measurement data in the measurement data group. The identifying unit 409 may identify the abnormal measurement data at the target time point.

[0056] As an example in this embodiment, the identifying unit 409 may identify, among the plurality of measurement data in the first measurement data group, the measurement data having the largest difference between the average rate of change in a reference period including the target time point and the rate of change at the target time point as the abnormal measurement data in the first measurement data group. In this way, the measurement data that exhibits the most abnormal behavior at the target time point is identified as the abnormal measurement data.

[0057] The change rate of the measurement data at a target time point may be the change rate between the measurement data at the target time point and the measurement data at the measurement time point immediately before the target time point. Furthermore, the average change rate of the measurement data for a reference period including the target time point may be the average value of the change rates between the measurement data measured before and after the reference period. The identifying unit 409 may calculate the change rate after smoothing the various measurement data in the time axis direction using a moving average or the like.

[0058] The reference period may be a period of any length that includes multiple measurement timings of the measurement data. The reference period may be a period shorter than the target period for calculating the correlation coefficient. The reference period may include the target time point as its start point, its end point, or any period between the start point and the end point.

[0059] The identifying unit 409 may determine a measurement data group including two or more types of measurement data as a first measurement data group. For a measurement data group including only one type of measurement data, the identifying unit 409 may determine that the measurement data is abnormal measurement data.

[0060] The identifying section 409 may supply the supplying section 410 with the measurement data ID of the abnormal measurement data for each measurement data group and information indicating the target time point.

[0061] [1.2.10. Supply section 410] The supplying section 410 supplies the model 435 with each piece of anomalous measurement data identified for each measurement data group by the identifying section 409. The supplying section 410 may supply the model 435 with the anomalous measurement data at the target time point.

[0062] Here, as an example in this embodiment, the device 4 is provided with a plurality of models 435. Therefore, the supply unit 410 may supply the abnormal measurement data to the model 435 among these plurality of models 435 that corresponds to the combination pattern of the abnormal measurement data identified for each measurement data group by the identification unit 409. As an example in this embodiment, the supply unit 410 may supply the abnormal measurement data to the model 435 that corresponds to the pattern ID of the combination pattern of the abnormal measurement data.

[0063] The supplying unit 410 may further supply the measurement data to the outputting unit 411. The supplying unit 410 may supply the measurement data for a period including the target time point (for example, the above-mentioned reference period) to the outputting unit 411. The measurement data supplied to the outputting unit 411 may be anomalous measurement data or may be measurement data different from the anomalous measurement data.

[0064] [1.2.11. Output Unit 411] The output unit 411 outputs a signal according to the condition index value output from the model 435. The output unit 411 may be a display control unit, and may output a display control signal to the display unit 412. As an example, the output unit 411 may cause the display unit 412 to display the condition index value as it is. Additionally or alternatively, in response to the condition index value indicating that the condition of the equipment 2 is good being output from the model 435, the output unit 411 may cause the display unit 412 to display that the condition of the equipment 2 is good. In response to the condition index value indicating that the condition of the equipment 2 is bad being output from the model 435, the output unit 411 may cause the display unit 412 to display that the condition of the equipment 2 is bad.

[0065] When measurement data is supplied from the supply unit 410, the output unit 411 may display the measurement data together with the state index value on the display unit 412. As an example, the output unit 411 may display the state index value and the measurement data on the same screen of the display unit 412, or may display them on separate screens.

[0066] [1.2.12.Display section 412] The display unit 412 performs display under the control of the output unit 411. In this embodiment, as an example, the display unit 412 is provided in the device 4, but it may be externally connected to the device 4.

[0067] According to the device 4 in the above-described system 1, any abnormal measurement data for each highly correlated measurement data group is input to the model 435, and a signal corresponding to the state index value output from the model 435 is output in response to the abnormal measurement data. Therefore, compared to when all measurement data is input to the model 435 without selecting measurement data for each measurement data group and a state index value is output, the calculation load on the model 435 and the learning process load on the model 435 can be reduced. Furthermore, unlike when measurement data other than the abnormal measurement data from within a measurement data group is input to the model 435, an abnormality in the equipment 2 can be reliably detected. Furthermore, because the abnormal measurement data is identified from the acquired measurement data, even if some measurement data cannot be acquired due to a failure of the sensor 20 or the like, it is possible to avoid a situation where measurement data that should be input to the model 435 cannot be acquired, and the state of the equipment 2 can be reliably detected.

[0068] Furthermore, abnormal measurement data at a specified target time point is identified and supplied to the model 435, so that the state of the equipment 2 at the target time point can be detected.

[0069] Furthermore, among the multiple pieces of measurement data in the first measurement data group, the measurement data with the largest difference between the average rate of change in the reference period including the target time point and the rate of change at the target time point is determined to be abnormal measurement data in the first measurement data group. Therefore, abnormal measurement data that may put the equipment 2 in a defective state can be accurately identified within the first measurement data group.

[0070] Furthermore, since the abnormal measurement data is supplied to the model 435 corresponding to the combination pattern of the abnormal measurement data, the state of the equipment 2 can be detected with high accuracy, unlike when the abnormal measurement data is supplied to an incompatible model 435.

[0071] Furthermore, the model 435 is subjected to a learning process for each combination pattern of measurement data selected one type at a time from each measurement data group. Therefore, the state of the equipment 2 can be detected regardless of the combination pattern of abnormal measurement data.

[0072] Furthermore, a correlation coefficient is calculated for each combination of measurement data among multiple types of measurement data, and measurement data combinations with a correlation coefficient higher than the reference correlation coefficient are grouped to generate group information. Therefore, group information can be generated by grouping measurement data with high correlations.

[0073] Furthermore, the group information indicates, among multiple types of measurement data, measurement data that have a lower correlation with other measurement data than a standard as separate measurement data groups. Therefore, measurement data that constitutes a single-element measurement data group can be included in the input data of model 435, allowing the state of equipment 2 to be detected with high accuracy.

[0074] [2. Group information file] 2 shows the data structure of the group information file 431. In this figure, as an example, there are 12 types of measurement data.

[0075] The group information file 431 may store the measurement data ID of each measurement data as group information by associating it with the measurement data ID of other measurement data (also referred to as strongly correlated measurement data) that has a higher correlation with the measurement data than a reference value. The group information file 431 may also store the group information of each measurement data group by associating it with a group ID. As an example, FIG. 2 shows that measurement data Xa, Xc, Xd, and Xg (also referred to as measurement data with measurement data IDs "Xa", "Xc", "Xd", and "Xg", respectively) are included in measurement data group G1 (also referred to as the measurement data group with group ID "G1"). It also shows that measurement data Xf is included alone in measurement data group G3.

[0076] 2, the number of measurement data included in each measurement data group, in other words, the number of measurement data patterns that can be selected from the measurement data group, is shown. The value "24" obtained by multiplying these numbers of patterns may indicate the total number of combination patterns when arbitrary measurement data is selected for each measurement data group, and thus the number of models 435.

[0077] [3. Operation] [3.1. Learning Phase Operation] 3 shows the operation of the device 4 when performing the learning process for the model 435. The device 4 performs the learning process for each model 435 by performing the processes of steps S11 to S19.

[0078] In step S11, the measurement data acquiring unit 401 acquires multiple types of measurement data indicating the state of the equipment 2. As an example in the present embodiment, the measurement data acquiring unit 401 may acquire all of the measurement data measured in advance by each sensor 20 from an external storage device (not shown).

[0079] In step S13, the preprocessing unit 402 performs preprocessing on the acquired measurement data. The preprocessing unit 402 may perform at least one of conventionally known outlier processing, normalization processing, and missing value processing on the measurement data, and may also perform other preprocessing in addition to or instead of these. The preprocessing unit 402 may store the preprocessed measurement data in the measurement data file 430.

[0080] In step S15, the label adding unit 405, in response to an operation by the operator, adds a label to each piece of measurement data in the measurement data file 430, indicating whether the state of the equipment 2 was good or bad at the time the measurement data was measured. The label adding unit 405 may collectively add a label indicating that the equipment 2 was in a good state to all measurement data measured during a period when the equipment 2 was in a good state, and may collectively add a label indicating that the equipment 2 was in a bad state to all measurement data measured during a period when the equipment 2 was in a bad state.

[0081] In step S17, the group information acquisition unit 406 acquires group information of the multiple types of measurement data stored in the measurement data file 430. The group information acquisition unit 406 may calculate a correlation coefficient for each combination of measurement data in the multiple types of measurement data, and generate group information by grouping measurement data combinations having a correlation coefficient higher than a reference correlation coefficient.

[0082] Furthermore, the group information acquisition unit 406 may generate pattern data by associating a pattern ID with each combination pattern in which one type of measurement data is selected from each measurement data group indicated by the group information. As an example, when the group information shown in FIG. 2 above is acquired, a pattern ID may be associated with each of the 24 combination patterns in total in which one type of measurement data is selected from each of the measurement data groups G1 to G6. The generated pattern data may be stored in the storage unit 403.

[0083] In step S19, the learning processing unit 407 performs a learning process on a model 435 that outputs a state index value in response to input of each measurement data of the combination pattern, using learning data including each measurement data included in the combination pattern and a state index value, for each combination pattern of measurement data selected one type from each measurement data group. The learning processing unit 407 may perform a learning process on a separate model 435 for each combination pattern indicated by the pattern data generated in step S17, using learning data including each measurement data indicated by the pattern data. As an example, when the group information shown in FIG. 2 above is acquired in step S17, the learning processing unit may perform a learning process on 24 models 435 using measurement data of a total of 24 combination patterns.

[0084] This generates a model 435 corresponding to each combination pattern. The learning processing unit 407 may add, to each model 435, the pattern ID of the combination pattern corresponding to the model 435.

[0085] [3.2. Operational Phase] 4 shows the operation of the device 4 when using the model 435. The device 4 supports the monitoring of the facility 2 by the processes of steps S31 to S43. This operation may be started in response to the activation of the facility 2. At the start of the operation, the group information acquired in the above-mentioned step S17 and each model 435 generated in the above-mentioned step S19 may be stored in advance in the storage unit 403.

[0086] In step S31, the measurement data acquisition unit 401 acquires multiple types of measurement data indicating the state of the equipment 2. In this embodiment, as an example, the measurement data acquisition unit 401 may acquire measurement data from each sensor 20 sequentially.

[0087] In step S33, the preprocessing unit 402 performs preprocessing on the acquired measurement data. The preprocessing unit 402 may perform preprocessing in the same manner as in step S13 described above.

[0088] In step S35, the device 4 determines whether or not a designation operation is performed at the target time point. If it is determined that a designation operation is performed (step S35; Yes), the process may proceed to step S35. If it is determined that a designation operation is not performed (step S35; No), the process may proceed to step S31.

[0089] In step S37, the designation unit 408 designates a target time point. The designation unit 408 may designate a target time point in response to an operation by an operator. The designation unit 408 may designate the current time point or a past time point.

[0090] In step S39, the identifying unit 409 identifies abnormal measurement data within the measurement data acquired in the processing of step S31 for each measurement data group indicated by the group information. For example, when the group information of Fig. 2 described above is acquired by the group information acquiring unit 406, the identifying unit 409 may identify any measurement data from each of the measurement data groups with group IDs "G1" to "G6" as abnormal measurement data. The identifying unit 409 may identify abnormal measurement data at the target time point.

[0091] In step S41, the supply unit 410 supplies the abnormal measurement data identified for each measurement data group to each model 435. The supply unit 410 may detect, from among the multiple models 435, a model 435 to which a pattern ID of the combination pattern of the identified abnormal measurement data is added, and supply each abnormal measurement data at the target time point to the model 435. As an example in this embodiment, the supply unit 410 may identify a pattern ID corresponding to the combination of abnormal measurement data by referring to the pattern information stored in the storage unit 403 in the above-mentioned step S17, and supply each abnormal measurement data to the model 435 corresponding to the pattern ID.

[0092] In addition, the supplying unit 410 may supply the measurement data included in other combination patterns to the model 435 corresponding to the other combination patterns. The other combination patterns may be predetermined default combination patterns. The supplying unit 410 may supply each measurement data at a target time point to the model 435 corresponding to the other combination patterns.

[0093] In step S43, the output unit 411 outputs a signal corresponding to the condition index value output from the model 435. As an example, the output unit 411 may cause the display unit 412 to display the condition index value and also cause the display unit 412 to display whether the condition of the equipment 2 is good or bad.

[0094] Note that, in step S41, when measurement data is also supplied to models 435 corresponding to other combination patterns, the output unit 411 may output a signal corresponding to the condition index value output from the model 435. For example, the output unit 411 may output a signal corresponding to the condition index value indicating a more adverse condition between the condition index value output from the model 435 corresponding to the combination pattern of abnormal measurement data and the condition index value output from the model 435 corresponding to the other combination pattern. The output unit 411 may output signals corresponding to the two condition index values ​​in response to a selection operation by the operator.

[0095] According to the above operation, a signal is output according to the condition index value output from model 435 corresponding to the abnormal measurement data for each measurement data group and the condition index value output from model 435 corresponding to the other combination patterns. Therefore, even if the accuracy of identifying the abnormal measurement data is low, an abnormality in equipment 2 can be detected with high accuracy.

[0096] In the above operation stage, it has been described that the group information is stored in advance in the storage unit 403 before the processing of step S31, but the group information may be acquired by the group information acquisition unit 406 at any time between the processing of step S31 and the processing of step S39. In this case, the group information may be generated by the group information acquisition unit 406 based on the past measurement data acquired by the processing of step S31.

[0097] [4. Variation of System 1] 5 shows a system 1A according to a modified example. The system 1A includes a device 4A, which has a preprocessing unit 402A, a storage unit 403A, a learning processing unit 407A, and a supply unit 410A.

[0098] 4.1. Pre-processing unit 402A The preprocessing unit 402A is an example of a normalization unit, and normalizes each measurement data included in at least one measurement data group. Normalization may include standardization to set the mean of each type of measurement data to 0 and the variance to 1.

[0099] Here, if group information has not been acquired by the group information acquisition unit 406, the preprocessing unit 402A may correct (also referred to as scaling) each piece of measurement data to a value within a range set commonly for all types of measurement data. In other words, the preprocessing unit 402A may normalize each piece of measurement data within a range common to all measurement data groups.

[0100] When group information has been acquired by the group information acquisition unit 406, the preprocessing unit 402A may normalize each piece of measurement data included in each measurement data group containing two or more types of measurement data. As an example, the preprocessing unit 402A may correct each piece of measurement data included in each measurement data group containing two or more types of measurement data to a value within a range set for that measurement data group. The preprocessing unit 402A may or may not normalize the measurement data of a measurement data group containing only one type of measurement data within that measurement data group. Alternatively, even when group information has been acquired by the group information acquisition unit 406, the preprocessing unit 402A may perform normalization in the same manner as when group information has not been acquired by the group information acquisition unit 406.

[0101] The preprocessing unit 402A may read out the measurement data stored in the measurement data file 430 of the storage unit 403, normalize the data, and store the normalized measurement data in the measurement data file 430. In this modification, the normalized measurement data may be read out by at least the learning processing unit 407A and the supplying unit 410A. The preprocessing unit 402A may further perform at least one of outlier processing and missing value processing on the measurement data in the same manner as the preprocessing unit 402 described above, or may perform other preprocessing in addition to or instead of this.

[0102] [4.2. Storage section 403A] The storage unit 403A stores a single model 435A. The model 435A outputs a condition index value in response to input of each measurement data of an arbitrary combination pattern in which any one type of measurement data is selected for each measurement data group. Similar to the above-described model 435, the model 435A may output a condition index value that is not binarized into a value indicating that the condition of the equipment 2 is good and a value indicating that the condition is bad. The model 435A may be, for example, a support vector machine, but may also be a trained model using other algorithms such as logistic regression, a decision tree, or a neural network. The model 435 may be a linear model.

[0103] [4.3. Learning Processing Unit 407A] The learning processing unit 407A performs a learning process for the model 435A. The learning processing unit 407A may perform the learning process for the model 435A using learning data including the measurement data and state index values ​​of each measurement data group normalized by the preprocessing unit 402A. Note that the measurement data of a measurement data group including only one type of measurement data does not need to be normalized. The learning processing unit 407A may perform the learning process for the model 435A so that the difference in the state index value due to differences in the combination patterns of the measurement data falls within an arbitrary range (for example, the range of standard error).

[0104] The learning data may include at least one type of measurement data for each measurement data group. As an example, when the group information shown in FIG. 2 is acquired by the group information acquisition unit 406, the learning data may include only one type of measurement data among the measurement data Xa, Xc, Xd, and Xg for the measurement data group G1, or may include multiple types of measurement data. Note that the model 435A that has been trained by the training processing unit 407A using the measurement data of the equipment 2 may be used to determine the state of another equipment 2.

[0105] [4.4. Supply section 410A] The supply unit 410A supplies the model 435A with each piece of anomalous measurement data identified for each measurement data group by the identification unit 409. The supply unit 410A may supply the model 435A with the anomalous measurement data normalized by the preprocessing unit 402A and identified by the identification unit 409. Note that the measurement data of a measurement data group including only one type of measurement data does not need to be normalized. The supply unit 410A may supply the model 435A with the anomalous measurement data at a target time point.

[0106] In addition, the supply unit 410A may separately supply each measurement data of another combination pattern (for example, a default combination pattern) different from the combination pattern of the abnormal measurement data to the model 435A. The supply unit 410A may supply each measurement data at a target time point to the model 435A.

[0107] In addition, when measurement data of another combination pattern different from the combination pattern of abnormal measurement data is separately supplied to the model 435A, the output unit 411 may also output a signal corresponding to the condition index value output from the model 435A using the measurement data of the other combination pattern. For example, the output unit 411 may output a signal corresponding to the condition index value indicating a more poor condition between the condition index value output from the model 435A using the abnormal measurement data for each measurement data group and the condition index value output from the model 435A using the measurement data of the other combination pattern. The output unit 411 may output a signal corresponding to each condition index value in response to a selection operation by the operator. In these cases, a poor condition of the equipment 2 can be accurately detected regardless of the accuracy of identifying the abnormal measurement data.

[0108] According to the device 4A in the above-described system 1A, as with the above-described device 4, any abnormal measurement data for each highly correlated measurement data group is input to the model 435A, and a signal corresponding to the state index value output from the model 435A is output in response to the abnormal measurement data. Therefore, compared to when all measurement data are input to the model 435A without selecting measurement data for each measurement data group and a state index value is output, the computation load on the model 435A and the learning process load on the model 435A can be reduced. Furthermore, unlike when measurement data other than the abnormal measurement data from within a measurement data group is input to the model 435A, an abnormality in the equipment 2 can be reliably detected. Furthermore, because the abnormal measurement data is identified from the acquired measurement data, even if some measurement data cannot be acquired due to a malfunction of the sensor 20, it is possible to avoid a situation where measurement data that should be input to the model 435A cannot be acquired, and the state of the equipment 2 can be reliably detected.

[0109] Furthermore, the identified abnormal measurement data is supplied to model 435A, which outputs a condition index value in response to input of each measurement data of an arbitrary combination pattern in which one type of measurement data is selected for each measurement data group. Therefore, the condition of equipment 2 can be reliably detected regardless of the type of abnormal measurement data.

[0110] Furthermore, each measurement data included in at least one measurement data group is normalized, and abnormal measurement data identified for each measurement data group is supplied to model 435A, so that the state of equipment 2 can be detected with high accuracy even if any measurement data within the measurement data group is abnormal measurement data.

[0111] Furthermore, since the learning process of model 435A is performed using learning data including each measurement data of at least one normalized measurement data group and a state index value, it is possible to generate model 435A that outputs a highly accurate state index value even if any measurement data within the measurement data group is abnormal measurement data.

[0112] [5. Modification of the method for identifying abnormal measurement data] In the above embodiment and modified examples, the identifying unit 409 has been described as identifying, among the plurality of measurement data in the first measurement data group, the measurement data having the largest difference between the average rate of change in a reference period including the target time point and the rate of change at the target time point, as anomalous measurement data in the first measurement data group. In addition to or instead of this, the identifying unit 409 may identify anomalous measurement data using one or more other methods.

[0113] [5.1. Variation (1)] For example, the identification unit 409 may determine that, among the multiple measurement data in the second measurement data group at the target time, the measurement data that causes the model 435, 435A to output the worst condition index value when supplied separately to the model 435, 435A by the supply unit 410 is abnormal measurement data in the second measurement data group.

[0114] As an example, when the group information shown in FIG. 2 is acquired by the group information acquisition unit 406, the identification unit 409 may compare the condition index value output when each measurement data of the measurement data group G5 is input to the model 435, 435A in one or more combination patterns including the measurement data Xi with the condition index value output when each measurement data is input to the model 435, 435A in one or more combination patterns including the measurement data Xj. Of the measurement data of each measurement data group input to the model 435, 435A, the measurement data other than the measurement data group G5 may be the same in the combination pattern including the measurement data Xi and the combination pattern including the measurement data Xj. The identification unit 409 may smooth various measurement data at a target time point along the time axis using a moving average or the like before inputting them to the model 435, 435A. The identification unit 409 may determine, as abnormal measurement data, the measurement data of the measurement data Xi, Xj for which the model 435, 435A outputs a condition index value indicating a more poor condition.

[0115] The identifying unit 409 may identify a measurement data group including two or more types of measurement data as a second measurement data group. For a measurement data group including only one type of measurement data, the identifying unit 409 may identify the measurement data as abnormal measurement data. According to the technique of modification (1), abnormal measurement data that may put the equipment 2 in a defective state can be identified with high accuracy within the second measurement data group.

[0116] [5.2. Variation (2)] Furthermore, if the correlation coefficients between one of the multiple measurement data in the third measurement data group and each of the other measurement data are each less than a threshold value during a reference period including the target time point, the identifying unit 409 may determine the one measurement data as anomalous measurement data in the third measurement data group. The identifying unit 409 may calculate the correlation coefficient after smoothing the various measurement data in the reference period in the time axis direction using a moving average or the like. The identifying unit 409 may determine a measurement data group including three or more types of measurement data as the third measurement data group. As an example, when the group information shown in FIG. 2 above is acquired by the group information acquiring unit 406, the measurement data Xb may be determined as anomalous measurement data in the measurement data group G2 if the correlation coefficients between the measurement data Xb and the measurement data Xe and the correlation coefficient between the measurement data Xb and the measurement data Xl in the measurement data group G2 are each less than a threshold value.

[0117] When the correlation coefficient between one piece of measurement data and any other piece of measurement data in the third measurement data group is equal to or greater than a threshold, the identifying unit 409 may identify the measurement data with the lowest correlation coefficient as the anomalous measurement data in the third measurement data group. For example, when the correlation coefficient between measurement data Xb and measurement data Xe in measurement data group G2 is higher than a threshold and the correlation coefficient between measurement data Xb and measurement data Xl is lower than the threshold, the identifying unit 409 may identify measurement data Xl as the anomalous measurement data in measurement data group G2. This eliminates the need to sequentially set each piece of measurement data in the third measurement data group as the anomalous measurement data and calculate the correlation coefficient with other measurement data. Therefore, anomalous measurement data that may cause equipment 2 to be in a defective state can be easily identified in the third measurement data group. According to the method of variation (2), anomalous measurement data that may cause equipment 2 to be in a defective state can be accurately identified in the third measurement data group.

[0118] The threshold value of the correlation coefficient may be set arbitrarily and may be the same as or different from the reference correlation coefficient. The threshold value of the correlation coefficient may be set for each combination of measurement data. For example, when the mean value of the correlation coefficient calculated from measurement data Xa and measurement data Xb stored in the measurement data file 430 is R and the variance is σ, the threshold value of the correlation coefficient between the measurement data Xa and Xb may be R-σ. The period for calculating the mean value R and the variance σ may be the period for which the correlation coefficient is calculated, or a part of that period.

[0119] [5.3. Variation (3)] Furthermore, the identifying unit 409 may determine, among the plurality of measurement data in the fourth measurement data group, the measurement data with the largest autocovariance at the target time point or the measurement data with the lowest autocorrelation as the anomalous measurement data in the fourth measurement data group. The identifying unit 409 may also determine, as the anomalous measurement data, the measurement data with the largest autocovariance and the lowest autocorrelation at the target time point. The identifying unit 409 may determine, as the fourth measurement data group, a measurement data group including one or more types of measurement data. According to the technique of modification (3), it is possible to accurately identify, within the fourth measurement data group, anomalous measurement data that may put the equipment 2 in a defective state.

[0120] [5.4. Variation (4)] Furthermore, the identifying unit 409 may determine, among the measurement data in the fifth measurement data group, the measurement data that has the largest difference between the reference time point and the target time point as the abnormal measurement data in the fifth measurement data group. Here, the identifying unit 409 may determine, among the acquisition times of two or more types of measurement data datasets not included in the fifth measurement data group, the acquisition time point of the dataset that is most similar to the dataset at the target time point as the reference time point. The identifying unit 409 may determine, as the fifth measurement data group, a measurement data group including one or more types of measurement data. According to the method of variation (4), abnormal measurement data that may put the equipment 2 in a defective state can be accurately identified within the fifth measurement data group.

[0121] Note that various types of measurement data included in a data set may be included in the same measurement data group or different measurement data groups. A high degree of similarity between data sets may mean that the distance between the data sets is short.

[0122] [6. Other Modifications] In the above embodiment and modified examples, the identifying unit 409 has been described as identifying one type of abnormal measurement data within a measurement data group, but two or more types of abnormal measurement data may be identified. For example, in the identifying methods described in the embodiment and modified examples, if there are n or more types (n is a natural number greater than or equal to 2) of corresponding measurement data, the identifying unit 409 may treat each of these measurement data as abnormal measurement data of the same measurement data group. As an example, in the identifying method of the embodiment, if there are two types of measurement data among the multiple measurement data in the first measurement data group that have the largest difference between the average rate of change in a reference period including the target time point and the rate of change at the target time point, the identifying unit 409 may treat each of these measurement data as abnormal measurement data in the first measurement data group.

[0123] There are multiple abnormal measurement data in the same measurement data group (abnormal measurement data X1, ... X n If there is abnormal measurement data X1, . . . X2, . . . X3, . . . X4, . . . X5, . . . X6, . . . X7, . . . X8, . . . X9, . . . X10, . . . X11, . . . X12, . . . X13, . . . n Any one of the abnormal measurement data X N (where N is a natural number of 1≦N≦n) and abnormal measurement data of each of the other measurement data groups, the abnormal measurement data of the combination pattern may be supplied to each model 435 corresponding to each combination pattern. Also, the supply unit 410A of the device 4A may supply the abnormal measurement data X1, ...X2 of the same measurement data group to each model 435 corresponding to each combination pattern of the abnormal measurement data X1, ...X2 of the same measurement data group. n Any one of the abnormal measurement data X N and abnormal measurement data of each of the other measurement data groups, the abnormal measurement data of the combination pattern may be supplied to model 435A. In these cases, output unit 411 may output a signal corresponding to the condition index value indicating the worst condition among the condition index values ​​output from models 435 and 435A, or may output a signal corresponding to each condition index value in response to a selection operation by the operator.

[0124] Furthermore, although the group information acquisition unit 406 has been described as generating group information, it may also acquire group information generated by another device.

[0125] In addition, although the group information acquisition unit 406, learning processing unit 407, identification unit 409, and supply units 410, 410A have been described as reading measurement data preprocessed by preprocessing units 402, 402A from storage units 403, 403A, they may also read measurement data that has not been preprocessed. Also, preprocessing units 402, 402A may read measurement data that has not been preprocessed from storage units 403, 403A, preprocess the data, and supply the data to any of group information acquisition unit 406, learning processing unit 407, identification unit 409, or supply units 410, 410A.

[0126] Although the devices 4 and 4A have been described as separate devices, they may be the same device. For example, the device 4 may function as the device 4A when the number of combination patterns of measurement data selected from each measurement data group exceeds a reference number. As an example, the preprocessing unit 402 of the device 4 may function as the preprocessing unit 402A, the learning processing unit 407 of the device 4 may function as the learning processing unit 407A, and the supply unit 410 of the device 4 may function as the supply unit 410A. This allows for training of a separate model 435 for each combination pattern when the number of combination patterns of measurement data is small, and training of a common model 435A when the number of combination patterns of measurement data is large, thereby reducing the training load. The reference number may be determined according to the specifications of the device 4 (for example, parallel processing performance) or may be determined arbitrarily by the operator.

[0127] Furthermore, although the device 4, 4A has been described as including the preprocessing units 402, 402A, the storage units 403, 403A, the input unit 404, the label adding unit 405, the learning processing units 407, 407A, the designation unit 408, the identification unit 409, the supply units 410, 410A, the output unit 411, and the display unit 412, any of these may be omitted. For example, if the device 4, 4A does not include the storage units 403, 403A, it may be connected to an external storage device. If the device 4, 4A does not include the label adding unit 405, it may acquire measurement data to which labels have been added in advance using the measurement data acquisition unit 401. If the device 4, 4A does not include the learning processing units 407, 407A, it may store a model 435 that has been previously subjected to learning processing in the storage units 403, 403A. The device 4, 4A may set a default time point (for example, the current time point) as the target time point if it does not have the designation unit 408. If the device 4, 4A does not have the identification unit 409, the supply units 410, 410A, or the output unit 411, it may perform learning processing of the model 435.

[0128] Furthermore, the output unit 411 has been described as outputting a display control signal according to the state index value, but it may also output other types of signals, such as an audio signal or an alert signal, according to the state index value.

[0129] Various embodiments of the present invention may also be described with reference to flowcharts and block diagrams, where the blocks may represent (1) stages of a process in which operations are performed or (2) sections of an apparatus responsible for performing the operations. Particular stages and sections may be implemented by dedicated circuitry, programmable circuitry provided with computer-readable instructions stored on a computer-readable medium, and / or a processor provided with computer-readable instructions stored on a computer-readable medium. Dedicated circuitry may include digital and / or analog hardware circuitry, and may include integrated circuits (ICs) and / or discrete circuits. Programmable circuitry may include reconfigurable hardware circuitry, including logical AND, OR, XOR, NAND, NOR, and other logic operations, flip-flops, registers, memory elements such as field programmable gate arrays (FPGAs), programmable logic arrays (PLAs), and the like.

[0130] A computer-readable medium may include any tangible device capable of storing instructions that are executed by an appropriate device, such that the computer-readable medium having instructions stored thereon comprises an article of manufacture containing instructions that can be executed to create means for performing the operations specified in the flowcharts or block diagrams. Examples of computer-readable media may include electronic, magnetic, optical, electromagnetic, and semiconductor storage media. More specific examples of computer-readable media may include floppy disks, diskettes, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), electrically erasable programmable read-only memory (EEPROM), static random access memory (SRAM), compact disc read-only memory (CD-ROM), digital versatile disc (DVD), Blu-ray (RTM) disc, memory stick, integrated circuit card, and the like.

[0131] The computer readable instructions may include either assembler instructions, Instruction Set Architecture (ISA) instructions, machine instructions, machine-dependent instructions, microcode, firmware instructions, state-setting data, or source or object code written in any combination of one or more programming languages, including object-oriented programming languages ​​such as Smalltalk®, JAVA®, C++, etc., and conventional procedural programming languages ​​such as the “C” programming language or similar programming languages.

[0132] The computer-readable instructions may be provided to a processor or programmable circuitry of a general-purpose computer, special-purpose computer, or other programmable data processing apparatus, either locally or over a wide-area network (WAN) such as a local area network (LAN), the Internet, etc., which executes the computer-readable instructions to create means for performing the operations specified in the flowcharts or block diagrams. Examples of processors include computer processors, processing units, microprocessors, digital signal processors, controllers, microcontrollers, etc.

[0133] 6 illustrates an example of a computer 2200 in which aspects of the present invention may be embodied, in whole or in part. Programs installed on the computer 2200 may cause the computer 2200 to function as or perform operations associated with an apparatus or one or more sections of the apparatus according to embodiments of the present invention, and / or to perform a process or steps of a process according to embodiments of the present invention. Such programs may be executed by the CPU 2212 to cause the computer 2200 to perform specific operations associated with some or all of the blocks in the flowcharts and block diagrams described herein.

[0134] A computer 2200 according to this embodiment includes a CPU 2212, a RAM 2214, a graphics controller 2216, and a display device 2218, which are interconnected by a host controller 2210. The computer 2200 also includes input / output units such as a communication interface 2222, a hard disk drive 2224, a DVD-ROM drive 2226, and an IC card drive, which are connected to the host controller 2210 via an input / output controller 2220. The computer also includes legacy input / output units such as a ROM 2230 and a keyboard 2242, which are connected to the input / output controller 2220 via an input / output chip 2240.

[0135] The CPU 2212 operates according to programs stored in the ROM 2230 and RAM 2214, thereby controlling each unit. The graphics controller 2216 acquires image data generated by the CPU 2212 into a frame buffer or the like provided in the RAM 2214 or into the graphics controller 2216 itself, and causes the image data to be displayed on the display device 2218.

[0136] The communication interface 2222 communicates with other electronic devices via a network. The hard disk drive 2224 stores programs and data used by the CPU 2212 in the computer 2200. The DVD-ROM drive 2226 reads programs or data from the DVD-ROM 2201 and provides the programs or data to the hard disk drive 2224 via the RAM 2214. The IC card drive reads programs and data from an IC card and / or writes programs and data to an IC card.

[0137] The ROM 2230 stores therein a boot program or the like that is executed by the computer 2200 upon activation, and / or programs that depend on the hardware of the computer 2200. The input / output chip 2240 may also connect various input / output units to the input / output controller 2220 via a parallel port, a serial port, a keyboard port, a mouse port, etc.

[0138] The programs are provided by a computer-readable medium such as a DVD-ROM 2201 or an IC card. The programs are read from the computer-readable medium, installed in the hard disk drive 2224, RAM 2214, or ROM 2230, which are also examples of computer-readable media, and executed by the CPU 2212. Information processing described in these programs is read by the computer 2200, and brings about cooperation between the programs and the various types of hardware resources described above. An apparatus or method may be configured by realizing information manipulation or processing in accordance with the use of the computer 2200.

[0139] For example, when communication is performed between the computer 2200 and an external device, the CPU 2212 may execute a communication program loaded into the RAM 2214 and instruct the communication interface 2222 to perform communication processing based on the processing described in the communication program. Under the control of the CPU 2212, the communication interface 2222 reads transmission data stored in a transmission buffer processing area provided in the RAM 2214, the hard disk drive 2224, the DVD-ROM 2201, or a recording medium such as an IC card, and transmits the read transmission data to the network, or writes reception data received from the network to a reception buffer processing area or the like provided on the recording medium.

[0140] The CPU 2212 may also cause all or a necessary portion of a file or database stored on an external recording medium such as the hard disk drive 2224, the DVD-ROM drive 2226 (DVD-ROM 2201), an IC card, etc. to be read into the RAM 2214, and perform various types of processing on the data on the RAM 2214. The CPU 2212 then writes back the processed data to the external recording medium.

[0141] Various types of information, such as various types of programs, data, tables, and databases, may be stored on the recording medium and may undergo information processing. The CPU 2212 may perform various types of processing on data read from the RAM 2214, including various types of operations, information processing, conditional judgment, conditional branching, unconditional branching, information search / replacement, etc., as described throughout this disclosure and specified by the instruction sequences of the programs, and write the results back to the RAM 2214. The CPU 2212 may also search for information in a file, database, etc. on the recording medium. For example, if multiple entries each having an attribute value of a first attribute associated with an attribute value of a second attribute are stored on the recording medium, the CPU 2212 may search for an entry that matches a condition specified by the attribute value of the first attribute from among the multiple entries, read the attribute value of the second attribute stored in the entry, and thereby obtain the attribute value of the second attribute associated with the first attribute that satisfies a predetermined condition.

[0142] The above-described programs or software modules may be stored in a computer-readable medium on or near the computer 2200. A recording medium such as a hard disk or RAM provided in a server system connected to a dedicated communication network or the Internet can also be used as a computer-readable medium, thereby providing the programs to the computer 2200 via the network.

[0143] Although the present invention has been described above using embodiments, the technical scope of the present invention is not limited to the scope described in the above embodiments. It will be apparent to those skilled in the art that various modifications and improvements can be made to the above embodiments. It is clear from the claims that such modifications and improvements can also be included within the technical scope of the present invention.

[0144] It should be noted that the execution order of each process, such as operations, procedures, steps, and stages, in the devices, systems, programs, and methods shown in the claims, specifications, and drawings is not specifically stated as "before," "prior to," etc., and that the processes can be performed in any order unless the output of a previous process is used in a subsequent process. Even if the operational flow in the claims, specifications, and drawings is described using "first," "next," etc. for convenience, this does not mean that the processes must be performed in this order. [Explanation of symbols]

[0145] 1 System 2 Equipment 4 equipment 20 sensors 21 Equipment 401 Measurement data acquisition unit 402 Preprocessing section 403 Storage section 404 Input section 405 Label Addition Section 406 Group Information Acquisition Unit 407 Learning processing unit 408 Specified section 409 Specific section 410 Supply section 411 Output section 412 Display section 430 measurement data files 431 Group Information File 435 model 461 Calculation Unit 462 Grouping Section 463 Generation part 2200 Computer 2201 DVD-ROM 2210 host controller 2212 CPU 2214 RAM 2216 Graphics Controller 2218 Display Device 2220 Input / Output Controller 2222 communication interface 2224 hard disk drive 2226 DVD-ROM drive 2230 ROM 2240 I / O chip 2242 keyboard

Claims

1. a measurement data acquisition unit that acquires multiple types of measurement data indicating the state of the target; a group information acquisition unit that acquires group information indicating a group of measurement data having a higher correlation than a reference value from among the plurality of types of measurement data; an identifying unit that identifies abnormal measurement data from the measurement data included in each group indicated by the group information, the measurement data being acquired by the measurement data acquiring unit; a supply unit that supplies the abnormal measurement data identified for each group by the identification unit to a model that outputs a condition index value indicating whether the condition of the object is good or bad in response to input of any one type of measurement data for each group indicated by the group information; an output unit that outputs a signal corresponding to the state index value output from the model; Equipped with The identification unit identifies, among the multiple measurement data in the first group, the measurement data that has the largest difference between the average rate of change in a reference period including the target time point and the rate of change at the target time point as abnormal measurement data in the first group.

2. a measurement data acquisition unit that acquires multiple types of measurement data indicating the state of the target; a group information acquisition unit that acquires group information indicating a group of measurement data having a higher correlation than a reference value from among the plurality of types of measurement data; an identifying unit that identifies abnormal measurement data from the measurement data included in each group indicated by the group information, the measurement data being acquired by the measurement data acquiring unit; a supply unit that supplies the abnormal measurement data identified for each group by the identification unit to a model that outputs a condition index value indicating whether the condition of the object is good or bad in response to input of any one type of measurement data for each group indicated by the group information; an output unit that outputs a signal corresponding to the state index value output from the model; Equipped with The identification unit identifies, as abnormal measurement data in the second group, measurement data that causes the model to output the worst condition index value when the measurement data in the second group at the target time point are separately supplied to the model by the supply unit.

3. a measurement data acquisition unit that acquires multiple types of measurement data indicating the state of the target; a group information acquisition unit that acquires group information indicating a group of measurement data having a higher correlation than a reference value from among the plurality of types of measurement data; an identifying unit that identifies abnormal measurement data from the measurement data included in each group indicated by the group information, the measurement data being acquired by the measurement data acquiring unit; a supply unit that supplies the abnormal measurement data identified for each group by the identification unit to a model that outputs a condition index value indicating whether the condition of the object is good or bad in response to input of any one type of measurement data for each group indicated by the group information; an output unit that outputs a signal corresponding to the state index value output from the model; Equipped with The identification unit is configured to identify one of the multiple measurement data in a third group as abnormal measurement data in the third group when the correlation coefficients between the one measurement data and each of the other measurement data are each less than a threshold value during a reference period including a target time point.

4. 4. The device according to claim 3, wherein the identification unit defines a group including three or more types of measurement data as the third group, and when a correlation coefficient between one of the plurality of measurement data in the third group and any other measurement data is equal to or greater than a threshold, the identification unit defines the measurement data in the third group that has the lowest correlation coefficient with the one measurement data as abnormal measurement data in the third group.

5. a measurement data acquisition unit that acquires multiple types of measurement data indicating the state of the target; a group information acquisition unit that acquires group information indicating a group of measurement data having a higher correlation than a reference value from among the plurality of types of measurement data; an identifying unit that identifies abnormal measurement data from the measurement data included in each group indicated by the group information, the measurement data being acquired by the measurement data acquiring unit; a supply unit that supplies the abnormal measurement data identified for each group by the identification unit to a model that outputs a condition index value indicating whether the condition of the object is good or bad in response to input of any one type of measurement data for each group indicated by the group information; an output unit that outputs a signal corresponding to the state index value output from the model; Equipped with The identification unit determines, among the plurality of measurement data in a fourth group, the measurement data having the largest autocovariance at a target time point or the measurement data having the lowest autocorrelation as the abnormal measurement data in the fourth group.

6. a measurement data acquisition unit that acquires multiple types of measurement data indicating the state of the target; a group information acquisition unit that acquires group information indicating a group of measurement data having a higher correlation than a reference value from among the plurality of types of measurement data; an identifying unit that identifies abnormal measurement data from the measurement data included in each group indicated by the group information, the measurement data being acquired by the measurement data acquiring unit; a supply unit that supplies the abnormal measurement data identified for each group by the identification unit to a model that outputs a condition index value indicating whether the condition of the object is good or bad in response to input of any one type of measurement data for each group indicated by the group information; an output unit that outputs a signal corresponding to the state index value output from the model; Equipped with The identification unit determines, among the measurement data in a fifth group, the measurement data that has the largest difference between a reference time point and a target time point as abnormal measurement data in the fifth group, and determines, among the acquisition times of data sets of two or more types of measurement data not included in the fifth group, the acquisition time point of the data set that is most similar to the data set at the target time point as the reference time point.

7. a designation unit for designating a target time point of the state index value, The identification unit identifies abnormal measurement data at the target time point, The apparatus according to claim 1 , wherein the supplying unit supplies the model with anomalous measurement data at the time points of interest.

8. 7. The device according to claim 1, wherein the supplying unit supplies the abnormal measurement data identified for each group by the identifying unit to the model that outputs the state index value in response to input of each measurement data of an arbitrary combination pattern in which any one type of measurement data is selected for each group.

9. a normalization unit that normalizes each measurement data included in at least one group; The apparatus according to claim 8 , wherein the supplying unit supplies the abnormal measurement data normalized by the normalizing unit and identified for each group by the identifying unit to the model.

10. a normalization unit that normalizes each measurement data included in at least one group; a learning processing unit that performs a learning process for the model using learning data including the measurement data of each group normalized by the normalization unit and the state index value; The apparatus of claim 8 further comprising:

11. 7. The device according to claim 1, wherein the supply unit supplies the abnormal measurement data to a model corresponding to the abnormal measurement data combination pattern identified for each group by the identification unit, among a plurality of models provided for each combination pattern of measurement data selected one type from each group and outputting the state index value in response to input of each measurement data of the corresponding combination pattern.

12. 12. The device according to claim 11, further comprising: a learning processing unit that performs, for each combination pattern of measurement data selected one type from each group, a learning process of the model that uses learning data including each measurement data included in the combination pattern and the state index value, and outputs the state index value in response to input of each measurement data of the combination pattern.

13. a measurement data acquisition unit that acquires multiple types of measurement data indicating the state of the target; a group information acquisition unit that acquires group information indicating a group of measurement data having a higher correlation than a reference value from among the plurality of types of measurement data; a learning processing unit that performs a model learning process for each combination pattern, using learning data including each measurement data included in the combination pattern and a state index value indicating whether the state of the object is good or bad, for each combination pattern of measurement data selected one type from each group, and outputs the state index value in response to input of each measurement data of the combination pattern; An apparatus comprising:

14. The group information acquisition unit a calculation unit that calculates a correlation coefficient for each combination of measurement data in the plurality of types of measurement data; a grouping unit that groups together measurement data sets having a correlation coefficient higher than a reference correlation coefficient; a generation unit that generates the group information indicating a group of grouped measurement data; 7. The device according to claim 1, further comprising:

15. The apparatus according to claim 1 , wherein the group information indicates, as separate groups, each of the plurality of types of measurement data whose correlation with each of the other measurement data is lower than a reference value.

16. a measurement data acquisition stage for acquiring multiple types of measurement data indicating the state of the target; a group information acquisition step of acquiring group information indicating a group of measurement data having a higher correlation than a reference value among the plurality of types of measurement data; an identifying step of identifying, for each group indicated by the group information, abnormal measurement data among the measurement data included in the group acquired in the measurement data acquiring step; a supply step of supplying the abnormal measurement data identified for each group in the identification step to a model that outputs a condition index value indicating whether the condition of the object is good or bad in response to input of any one type of measurement data for each group indicated by the group information; an output step of outputting a signal corresponding to the state index value output from the model; Equipped with In the identification step, Among the plurality of measurement data in the first group, the measurement data having the largest difference between the average rate of change in a reference period including the target time point and the rate of change at the target time point is determined as abnormal measurement data in the first group; Among the plurality of measurement data in the second group at the target time point, the measurement data that causes the model to output the worst state index value when separately supplied to the model in the supply step is determined to be the abnormal measurement data in the second group; In a reference period including the target time point, when a correlation coefficient between one of the plurality of measurement data in the third group and each of the other measurement data is less than a threshold, the one measurement data is determined to be abnormal measurement data in the third group; Among the plurality of measurement data in the fourth group, the measurement data having the largest autocovariance at the target time point or the measurement data having the lowest autocorrelation is determined as the abnormal measurement data in the fourth group; A method in which, among the measurement data in a fifth group, the measurement data that has the largest difference between the reference time point and the target time point is defined as the abnormal measurement data in the fifth group, and, among the acquisition times of data sets of two or more types of measurement data not included in the fifth group, the acquisition time point of the data set that is most similar to the data set at the target time point is defined as the reference time point.

17. a measurement data acquisition stage for acquiring multiple types of measurement data indicating the state of the target; a group information acquisition step of acquiring group information indicating a group of measurement data having a higher correlation than a reference value among the plurality of types of measurement data; a learning processing stage for performing a model learning process for each combination pattern, using learning data including each measurement data included in the combination pattern and a condition index value indicating whether the condition of the object is good or bad, for each combination pattern of measurement data selected one type from each group, and outputting the condition index value in response to input of each measurement data of the combination pattern; A method for providing the above.

18. Computer, a measurement data acquisition unit that acquires multiple types of measurement data indicating the state of the target; a group information acquisition unit that acquires group information indicating a group of measurement data having a higher correlation than a reference value from among the plurality of types of measurement data; an identifying unit that identifies abnormal measurement data from the measurement data included in each group indicated by the group information, the measurement data being acquired by the measurement data acquiring unit; a supply unit that supplies the abnormal measurement data identified for each group by the identification unit to a model that outputs a condition index value indicating whether the condition of the object is good or bad in response to input of any one type of measurement data for each group indicated by the group information; an output unit that outputs a signal corresponding to the state index value output from the model; It functions as The identification unit Among the plurality of measurement data in the first group, the measurement data having the largest difference between the average rate of change in a reference period including the target time point and the rate of change at the target time point is determined as abnormal measurement data in the first group; Among the plurality of measurement data in the second group at the target time, the measurement data that causes the model to output the worst state index value when separately supplied to the model by the supply unit is determined to be abnormal measurement data in the second group; In a reference period including the target time point, when a correlation coefficient between one of the plurality of measurement data in the third group and each of the other measurement data is less than a threshold, the one measurement data is determined to be abnormal measurement data in the third group; Among the plurality of measurement data in the fourth group, the measurement data having the largest autocovariance at the target time point or the measurement data having the lowest autocorrelation is determined as the abnormal measurement data in the fourth group; A program that defines, among the measurement data in a fifth group, the measurement data that has the largest difference between a reference time point and a target time point as abnormal measurement data in the fifth group, and defines, as the reference time point, the acquisition time point of the dataset that is most similar to the dataset at the target time point among the acquisition times of two or more types of measurement data datasets that are not included in the fifth group.

19. Computer, a measurement data acquisition unit that acquires multiple types of measurement data indicating the state of the target; a group information acquisition unit that acquires group information indicating a group of measurement data having a higher correlation than a reference value from among the plurality of types of measurement data; a learning processing unit that performs a learning process of a model for each combination pattern, using learning data including each measurement data included in the combination pattern and a state index value indicating whether the state of the object is good or bad, for each combination pattern of measurement data selected one type from each group, and outputs the state index value in response to input of each measurement data of the combination pattern; A program that functions as a

Images