WO2016116961A1 - 情報処理装置および情報処理方法 - Google Patents
情報処理装置および情報処理方法 Download PDFInfo
- Publication number
- WO2016116961A1 WO2016116961A1 PCT/JP2015/000243 JP2015000243W WO2016116961A1 WO 2016116961 A1 WO2016116961 A1 WO 2016116961A1 JP 2015000243 W JP2015000243 W JP 2015000243W WO 2016116961 A1 WO2016116961 A1 WO 2016116961A1
- Authority
- WO
- WIPO (PCT)
- Prior art keywords
- monitoring target
- information processing
- data
- time
- target data
- Prior art date
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N5/00—Computing arrangements using knowledge-based models
- G06N5/04—Inference or reasoning models
- G06N5/048—Fuzzy inferencing
-
- G—PHYSICS
- G05—CONTROLLING; REGULATING
- G05B—CONTROL OR REGULATING SYSTEMS IN GENERAL; FUNCTIONAL ELEMENTS OF SUCH SYSTEMS; MONITORING OR TESTING ARRANGEMENTS FOR SUCH SYSTEMS OR ELEMENTS
- G05B23/00—Testing or monitoring of control systems or parts thereof
- G05B23/02—Electric testing or monitoring
- G05B23/0205—Electric testing or monitoring by means of a monitoring system capable of detecting and responding to faults
- G05B23/0218—Electric testing or monitoring by means of a monitoring system capable of detecting and responding to faults characterised by the fault detection method dealing with either existing or incipient faults
- G05B23/0224—Process history based detection method, e.g. whereby history implies the availability of large amounts of data
- G05B23/0227—Qualitative history assessment, whereby the type of data acted upon, e.g. waveforms, images or patterns, is not relevant, e.g. rule based assessment; if-then decisions
- G05B23/0235—Qualitative history assessment, whereby the type of data acted upon, e.g. waveforms, images or patterns, is not relevant, e.g. rule based assessment; if-then decisions based on a comparison with predetermined threshold or range, e.g. "classical methods", carried out during normal operation; threshold adaptation or choice; when or how to compare with the threshold
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N20/00—Machine learning
-
- G—PHYSICS
- G05—CONTROLLING; REGULATING
- G05B—CONTROL OR REGULATING SYSTEMS IN GENERAL; FUNCTIONAL ELEMENTS OF SUCH SYSTEMS; MONITORING OR TESTING ARRANGEMENTS FOR SUCH SYSTEMS OR ELEMENTS
- G05B23/00—Testing or monitoring of control systems or parts thereof
- G05B23/02—Electric testing or monitoring
- G05B23/0205—Electric testing or monitoring by means of a monitoring system capable of detecting and responding to faults
- G05B23/0218—Electric testing or monitoring by means of a monitoring system capable of detecting and responding to faults characterised by the fault detection method dealing with either existing or incipient faults
- G05B23/0224—Process history based detection method, e.g. whereby history implies the availability of large amounts of data
- G05B23/024—Quantitative history assessment, e.g. mathematical relationships between available data; Functions therefor; Principal component analysis [PCA]; Partial least square [PLS]; Statistical classifiers, e.g. Bayesian networks, linear regression or correlation analysis; Neural networks
-
- G—PHYSICS
- G05—CONTROLLING; REGULATING
- G05B—CONTROL OR REGULATING SYSTEMS IN GENERAL; FUNCTIONAL ELEMENTS OF SUCH SYSTEMS; MONITORING OR TESTING ARRANGEMENTS FOR SUCH SYSTEMS OR ELEMENTS
- G05B23/00—Testing or monitoring of control systems or parts thereof
- G05B23/02—Electric testing or monitoring
- G05B23/0205—Electric testing or monitoring by means of a monitoring system capable of detecting and responding to faults
- G05B23/0259—Electric testing or monitoring by means of a monitoring system capable of detecting and responding to faults characterized by the response to fault detection
- G05B23/0297—Reconfiguration of monitoring system, e.g. use of virtual sensors; change monitoring method as a response to monitoring results
Definitions
- the present invention relates to a technique for estimating a time when an abnormality has occurred from a value acquired by a sensor installed in a control system such as an elevator, a plant apparatus, or a machine tool.
- a control system such as an elevator or a machine tool
- a sensor is installed and an abnormality is detected based on a signal acquired by the sensor.
- an abnormality may occur in other signals due to the influence of the signal in which the abnormality has occurred
- a technique for estimating the abnormality cause signal that is the cause of the abnormality based on the signal in which the abnormality has been detected is disclosed.
- a propagation path that is the order of signals through which an abnormality propagates when an abnormality occurs is extracted from a physical causal relationship between signals and listed in advance.
- a propagation path is extracted from the list based on the signal detected as abnormal, and the signal at the beginning of the extracted propagation path is estimated as an abnormality cause signal.
- Prioritization ranks a plurality of signals detected as abnormal in order of preset importance, detection time, or occurrence frequency.
- Patent Documents 1 and 2 have a problem that the signal causing the abnormality may be erroneously determined.
- the present invention has been made to solve the above-described problems, and an object thereof is to obtain an information processing apparatus that more accurately obtains a time when a signal starts to show an abnormality.
- a setting unit that sets the normal range indicating the normal value range of the monitoring target data consisting of time-series signals from the upper limit value and the lower limit value, determines whether the monitoring target data is out of the normal range, and A determination unit that outputs a determination time that is a time when the monitoring target data is out of the normal range when it is determined, and an average value and monitoring target data of a plurality of learning data composed of signals of normal values among known monitoring target data And a detection unit that detects a start time before the determination time input from the determination unit and the monitoring target data starts to show an abnormality based on a degree of deviation that is a difference from.
- the time when the signal starts to show an abnormality can be obtained more accurately.
- FIG. 2 is a block diagram illustrating a configuration of an information processing device according to Embodiment 1.
- FIG. 3 is a graph in which learning data according to Embodiment 1 is drawn.
- FIG. 6 is a diagram illustrating an example of defining a normal range from learning data according to the first embodiment. 6 is a flowchart showing a flow of processing in which the setting unit according to the first embodiment creates a band model.
- the graph which shows the time when the monitoring object data which concern on Embodiment 1 began to deviate from average behavior 6 is a graph showing an example of a deviation degree D (t) according to the first embodiment.
- FIG. 6 is a graph showing an example of a normal range of monitoring target data according to the first embodiment.
- 3 is a graph showing an example of a band model with a constant width according to the first embodiment.
- FIG. 3 is a block diagram illustrating a configuration of an information processing device according to a second embodiment.
- FIG. 4 is a diagram illustrating a configuration example of an information processing system according to a second embodiment.
- FIG. 4 is a block diagram illustrating a configuration of an information processing device according to a third embodiment.
- FIG. 6 shows an example of a screen displayed by a display unit according to Embodiment 3.
- FIG. 2 is a block diagram illustrating a hardware configuration of the information processing apparatus according to the first embodiment.
- FIG. 1 is a block diagram showing the configuration of the information processing apparatus 101 according to the first embodiment.
- a configuration example of the data collection / management apparatus 102 and the monitored object 103 are also described.
- the data collection / management apparatus 102 manages data collected from the monitoring target 103 via the sensor network 111.
- the information processing apparatus 101 as an abnormal start time estimation apparatus includes a first input unit 104, a setting unit 105, a second input unit 106, a determination unit 107, and a detection unit 108. Note that the first input unit 104 and the second input unit 106 may be realized by one input unit.
- the data collection / management apparatus 102 includes, for example, a normal value learning database (hereinafter referred to as normal value learning DB) 109 and a monitoring target database (hereinafter referred to as monitoring target DB) 110.
- a normal value learning database hereinafter referred to as normal value learning DB
- monitoring target DB monitoring target database
- the normal value learning DB and the monitoring target DB are integrated and managed in one, and the management is distributed in three or more DBs (Data Base), and is managed in the file format instead of the DB format. Good.
- the normal value learning DB 109 stores normal known data among the monitoring target data as learning data.
- the normal value learning DB 109 adds monitoring target data determined to be normal by the information processing apparatus of the present invention as learning data.
- data determined to be normal by an existing method may be stored as learning data.
- the normal value learning DB 109 learns data determined to be normal by the existing method. Save as data.
- an existing method for example, a range below the upper limit value and above the lower limit value of the control system to be monitored may be determined as normal, or a person may determine.
- the normal value learning DB 109 stores data determined to be normal by the existing method as learning data for the first specified period, and then stores the data determined to be normal by the information processing apparatus of the present invention as learning data. You may make it add as.
- the normal value learning DB 109 may delete old data. For example, when the device of the system to be monitored is updated and old data becomes unnecessary, the normal value learning DB 109 deletes old data. Moreover, you may delete, when the capacity
- a data server that holds data determined to be normal is provided in a place other than the data collection / management apparatus 102 such as a monitoring target system, and the normal value learning DB 109 does not store the data itself but holds the data server. You may make it preserve
- the monitored object 103 is a control system such as an elevator, a plant apparatus, or a machine tool, and includes a sensor.
- the monitoring target 103 may be configured by connecting or distributing one or more control systems.
- the monitoring target 103 may not be connected to the sensor network 111 but may be directly connected to the data collection / management apparatus 102.
- a set of signals acquired from the sensors of the monitoring target 103 is continuously or intermittently input to the data collection / management apparatus 101 through the sensor network 111.
- the signal data is a set of signals acquired from the sensors of the monitoring target 103 and is time-series data.
- the data collection / management apparatus 102 may input the signal data input to the monitoring target DB 110 to the normal value learning DB 109. Further, the data collection / management apparatus 102 may input the signal data input to the normal value learning DB 109 to the monitoring target DB 110.
- the data collection / management apparatus 102 outputs, from the normal value learning DB 109, signal data used as a normal value reference when detecting an abnormality, and inputs the signal data to the input unit 104 of the information processing apparatus 101.
- the data collection / management apparatus 102 outputs signal data for estimating the presence / absence of an abnormality and the start time of the abnormality from the monitoring target DB 110 and inputs the signal data to the input unit 106 of the information processing apparatus 101.
- the input unit 104 converts and shapes the signal data input from the normal value learning DB 109 of the data collection / management apparatus 102 and outputs the signal data to the setting unit 105.
- the setting unit 105 sets a normal range that is a range of normal values of signal data when the determination unit 107 determines abnormality.
- the input unit 106 converts and shapes the signal data input from the monitoring target DB 110 and outputs the signal data to the determination unit 107.
- the determination unit 107 determines whether the signal data input from the input unit 106 is out of the normal range input from the setting unit 105.
- the detection unit 108 detects, in the signal data determined to be out of the normal range by the determination unit 107, a start time that started before the normal range and started to show a behavior different from the normal range.
- the input unit 106 converts and shapes the signal data input from the monitoring target DB 110 and outputs the signal data to the determination unit 107.
- the conversion of signal data is a process of converting the format of signal data, for example.
- signal data format conversion there is a process of converting the format of signal data into a predetermined data format.
- the format conversion of the signal data is performed in order to operate each function of the information processing apparatus 101 normally.
- signal data format conversion there is sampling processing of signal data and deletion of signals in unnecessary periods in the signal data for the purpose of speeding up the processing.
- Signal signal shaping includes, for example, processing to classify and aggregate signal data under certain conditions when there is one or more input signal data.
- the input signal data is classified into signal data of similar conditions in the control system settings, external environment such as outside temperature and humidity, and aggregated for each condition. There is processing.
- the abnormality detection accuracy is improved by comparing the signal data to be monitored and the signal data in the normal range with similar conditions. Can be improved.
- FIG. 14 is a hardware configuration example of the information processing apparatus 101 according to the first embodiment.
- An information processing apparatus 101 as an information processing apparatus includes a reception device 1401, a processor 1402, a memory 1403, and a display 1404.
- the first input unit 104 and the second input unit 106 are receiving devices.
- the setting unit 105, the determination unit 107, and the detection unit 108 are realized by a processing circuit such as a CPU that executes a program stored in a memory or a system LSI (Large Scale Integration).
- a plurality of processing circuits may cooperate to execute the above function.
- the detection unit 108 may output the calculated start time to the display 1404.
- the setting unit 105 sets a normal range of signal data when the determination unit 107 determines abnormality.
- FIG. 2 is a graph 201 on which the learning data 202 according to the first embodiment is drawn.
- the vertical axis of the graph 201 indicates the signal value, and the horizontal axis indicates the time.
- the learning data 202 is a plurality of signal data with the same conditions classified and aggregated by the input unit 104.
- the learning data 202 is a set of signals each having a normal value.
- the learning data 202 shows a plurality of signal data, but each signal data may be called learning data.
- Reference numerals 203 and 204 denote arrows indicating the widths of variations in signal data at the respective times.
- the learning data 202 is displayed by superimposing a plurality of signal data.
- the setting unit 105 defines a normal range based on the learning data 202. Since the learning data 202 is a collection of signal data under the same conditions, the learning data 202 roughly shows the same behavior, but there is a difference in variation depending on the time as indicated by the arrows 203 and 204. The variation of 203 is larger than that of 204. This difference in variation for each time is a phenomenon that can occur even in an actual control system.
- FIG. 3 is a diagram illustrating an example of defining a normal range from the learning data 303 according to the first embodiment.
- 301 is a graph in which the learning data 303 is drawn.
- the vertical axis of the graph 301 indicates the signal value, and the horizontal axis indicates the time.
- Reference numeral 303 denotes learning data including a plurality of signal data.
- 304 indicates the time t1.
- 302 is a graph showing a normal region including the learning data 303 in a band model.
- the band model is a model that can define regions having different widths for each time.
- Reference numeral 305 denotes an average value of learning data for each time. In the present embodiment, 305 is referred to as a band model average.
- Reference numeral 306 denotes an upper limit value of the band model.
- Reference numeral 307 denotes a lower limit value of the band model.
- 308 is a difference between the band model average 305 and the upper limit value 306 or the lower limit value 307 of the band model at time t1. In the present embodiment, 308 is referred to as a band model width.
- the difference between the band model average value at each time and the upper limit value of the band model is described as being the same as the difference from the lower limit value, but may be different values.
- FIG. 4 is a flowchart showing a flow of processing in which the setting unit 105 according to the first embodiment creates a band model.
- Creation of a band model consists of three stages as shown below.
- the average and standard deviation of the learning data are calculated (step S401)
- the width of the band model is calculated (step S402)
- the upper limit value and the lower limit value of the band model are calculated (step S403).
- step S401 the setting unit 105 calculates the average and standard deviation of learning data as elements for calculating a band model.
- the setting unit 105 calculates the average of the learning data 202 for each time using Formula 1.
- the setting unit 105 calculates the standard deviation for each time using Formula 2.
- step S402 the setting unit 105 calculates the width of the band model.
- the setting unit 105 calculates W (t), which is the vertical width 306 of the band model of the graph 302, using Expression 3.
- n is a value for adjusting the upper and lower widths n ⁇ (t) of the band model.
- W the width 306 of the band model at time t1 of 304 is expressed as W (t1).
- step S403 the setting unit 105 calculates an upper limit value and a lower limit value of the band model, and defines a normal range.
- the setting unit 105 calculates MU (t) indicating the upper limit value 306 for each time of the band model using Equation 4.
- “U” of MU (t) is displayed as a subscript.
- the setting unit 105 calculates ML (t) indicating the lower limit value 307 for each time of the band model using Equation 5.
- ML (t) indicating the lower limit value 307 for each time of the band model using Equation 5.
- “L” in ML (t) is displayed as a subscript.
- the normal range at time t1 of 304 is a range of MU (t1) or more and ML (t1) or less.
- the setting unit 105 determines that the signal is abnormal when the signal is out of the normal range.
- the input unit 106 converts and shapes the signal data input from the monitoring target DB 110 for processing by the determination unit 107.
- the signal data conversion is a process of converting the format of the signal data, for example, as in the input unit 104.
- As an example of signal data format conversion there is processing for converting the format of signal data into a predetermined data format.
- the format conversion of the signal data is performed in order to operate each function of the information processing apparatus 101 normally.
- Examples of signal data format conversion include signal data sampling processing and signal period unnecessary signal deletion in signal data for the purpose of speeding up the processing.
- the input unit 106 may apply the same policy as the input unit 104 regarding sampling of signal data, deletion of signals in unnecessary periods in the signal data, and the like.
- the length of the monitoring target data input to the determination unit 107 may be adjusted by dividing the length of the monitoring target data at a fixed period or sequentially inputting in real time.
- the shaping of the signal data is, for example, classified by the input unit 104 similar to the conditions of the input signal data in order to compare the input signal data under the same conditions as the signal data classified and aggregated by the input unit 104, There is a process for extracting the aggregated signal data. Further, as an example of classification and aggregation of signal data, there is a process of dividing and aggregating signal data for each period of the same type of operation, such as immediately after starting a control system or during steady operation, for a plurality of input signal data .
- the input unit 106 may apply the same policy as the input unit 104 for the period of driving to be divided.
- the determination unit 107 determines whether the signal data input from the input unit 106 is out of the normal range input from the setting unit 105.
- FIG. 5 is a graph showing an example in which the monitoring target data according to the first embodiment is out of the normal range.
- the vertical axis of the graph indicates the signal value, and the horizontal axis indicates the time.
- Reference numeral 501 denotes monitoring target data output from the input unit 106.
- Reference numeral 502 denotes a determination time t2 when the determination unit 107 determines that the monitoring target data 501 is out of the normal range.
- the determination unit 107 determines that the normal range has been exceeded.
- FIG. 5 shows that the monitoring target data 501 exceeds the upper limit value 306 of the band model at time t ⁇ b> 2 of 502.
- a band model is used as an example of the normal range, but a normal range defined by another method may be used as long as it is a method capable of determining the time when the normal range is exceeded.
- the detection unit 108 detects, in the signal data determined to be out of the normal range by the determination unit 107, a start time that starts before deviating from the average behavior and starts before deviating from the normal range, Output.
- FIG. 6 is a graph showing the time when the monitoring target data according to the first embodiment starts to deviate from the average behavior.
- the vertical axis represents the signal value, and the horizontal axis represents the time.
- Reference numeral 601 represents a start time t3 that is a time when the average behavior starts to deviate.
- the start time t3 is a time before the determination time 502.
- the detection unit 108 expresses deviation from the average behavior by the degree of deviation.
- the deviation degree D (t) is calculated by Equation 6.
- FIG. 7 is a graph showing an example of the detachment degree D (t) according to the first embodiment.
- the vertical axis indicates the degree of deviation and the horizontal axis indicates time.
- Reference numeral 701 denotes a degree of detachment D (t).
- Reference numeral 702 denotes an area corresponding to a normal area of the band model.
- Reference numeral 703 denotes a constant n set when calculating the width W (t) of the band model.
- Reference numeral 704 denotes a constant n1 for determining whether or not there is a change.
- the constant n1 is a value different from the constant n.
- the length of time between the start time t3 and the determination time t2 is used only for the determination time t2 without considering that there is a difference for each signal data.
- the start time t3 is calculated. The start time t3 can be used for estimating a signal causing the abnormality.
- the detection unit 108 calculates the start time t3 using the inclination of the detachment degree as a change index.
- the detection unit 108 calculates the change index C (t) using Equation 7. Let t ⁇ 2.
- the detection unit 108 goes back from the determination time t2 and calculates the time when the change index C (t) first falls below the first threshold as the start time t3.
- a setting unit that sets the normal range indicating the normal value range of the monitoring target data composed of time-series signals from the upper limit value and the lower limit value, and whether the monitoring target data is out of the normal range.
- a determination unit that outputs a determination time that is a time when it is determined that the monitoring target data has deviated from the normal range when it is determined that the monitoring target data has deviated from the normal range, and a plurality of signals having normal values among the known monitoring target data Based on the degree of deviation indicating the difference between the average value of the learning data and the monitoring target data, the detection unit detects the start time before the determination time input from the determination unit and when the monitoring target data starts to show an abnormality Thus, the time when the signal starts to show abnormality can be obtained more accurately.
- the setting unit sets the maximum value of the plurality of learning data to the upper limit value at each of the plurality of times, and sets the minimum value of the plurality of learning data to the lower limit value. It is possible to define a normal region having a large spread. Therefore, it is possible to suppress false detection by relaxing the threshold value at a time with large variation, and to suppress omission of detection by tightening the threshold value at a time with small variation.
- the detection unit detects the time before the determination time and the inclination of the deviation degree is equal to or greater than the first threshold as the start time, the abnormality is detected after the occurrence of the abnormality for each signal. It is possible to reduce the influence of time differences.
- FIG. 8 is a graph showing an example of the normal range of the monitoring target data according to the first embodiment.
- Reference numeral 202 denotes the learning data shown in FIG.
- Reference numeral 801 denotes an upper limit value of the normal range.
- Reference numeral 802 denotes a lower limit value of the normal range.
- the monitored control system may have an alarm system that notifies an alarm when the monitored data exceeds the upper limit value or falls below the lower limit value.
- the setting unit 105 may set a range not less than the lower limit value and not more than the upper limit value of the alarm system as a normal range. It is assumed that the setting unit 105 has previously input or set the upper limit value and the lower limit value of the alarm system and holds them. In this case, the information processing apparatus 101 does not hold the input unit 104.
- the setting unit sets the upper limit value to the same value at a plurality of times and sets the lower limit value to the same value at a plurality of times. Therefore, the information processing apparatus 101 is an alarm system included in the monitored system. By diverting the upper and lower limit values, the process of setting the normal range from the signal data becomes unnecessary, and labor saving in development becomes possible.
- the setting unit 105 may use a constant-width band model that defines a normal region of the same width centered on the average of the band model at all times.
- a band model with a constant width is effective when the variation of learning data is small and the band model width cannot be set appropriately.
- the band model and the constant width band model may be used separately or in combination.
- FIG. 9 is a graph illustrating an example of a band model having a constant width according to the first embodiment.
- a graph 901 is obtained by superimposing learning data and a band model having a certain width.
- Reference numeral 902 is a graph showing the structure of a band model having a constant width.
- 903 is an upper limit value
- 904 is a lower limit value.
- Reference numeral 905 denotes a width of a band model having a constant width.
- the width 905 of the band model having a constant width may be, for example, a constant multiple of the standard deviation of the band model average 305, or may be set to a constant multiple of the average value of the band model average 305. It is assumed that the setting unit 105 holds a width value or a calculation method in advance. There may be a plurality of width values depending on the monitoring target system and the acquisition conditions of the monitoring target data.
- the setting unit sets the upper limit value and the lower limit value such that the difference between the upper limit value and the average value of the plurality of learning data for the lower limit value becomes the same value at a plurality of times. It is possible to cope with a case where the variation of the learning data is small, the learning data is small, the learning data is a constant value, and the like. In these cases, when a normal area is set from learning data, the upper and lower widths of the normal area are small, and even if it is normal monitoring target data, it is often erroneously determined to be abnormal.
- the setting unit 105 may set a normal region in a space whose dimensions are reduced by a technique such as principal component analysis or independent component analysis. Further, the setting unit 105 may set a normal region in the feature amount using the correlation coefficient, the Mahalanobis distance, and the like.
- a feature amount based on a correlation coefficient or Mahalanobis distance is calculated from a plurality of learning data, and an upper limit value and a lower limit value are set from the range of the feature amount.
- the processing time can be shortened by reducing the dimensions.
- the detection unit 108 calculates the start time at which it starts to deviate from the average behavior, in addition to the case where the change index C (t) exceeds the first threshold value, the deviation degree D ( If t) exceeds the second threshold, it may be considered that there is a change.
- the second threshold is a constant n1.
- the detection unit 108 may use a change index calculated by a known change point detection method from the deviation degree D (t).
- a change point detection method is Bayesian change point detection.
- the start time is detected based on Bayesian change point detection from the degree of deviation
- the detection unit 108 may calculate a start time when the deviation degree D (t) or the change index C (t) is smoothed and then starts to deviate from the average behavior.
- the start time is detected after smoothing the detachment degree or the inclination of the detachment degree, if the value of the detachment degree D (t) often vibrates up and down, the average There is a possibility that miscalculation of the start time that starts to deviate from the behavior can be suppressed.
- Embodiment 2 FIG. In the first embodiment described above, the time when an abnormality has occurred in a signal is estimated. However, in the present embodiment, an embodiment in which a signal causing an abnormality is estimated will be described.
- FIG. 10 is a block diagram illustrating a configuration of the information processing apparatus 1001 according to the second embodiment.
- the information processing apparatus 1001 is an example of an apparatus that utilizes the information processing apparatus 101 as an abnormality start time estimation apparatus, and is an abnormality cause signal estimation apparatus that estimates an abnormality cause signal.
- FIG. 10 as an example of means for collecting data from the control system to be monitored, a configuration example of the data collection / management apparatus 102 and the monitoring target 103 are also described, as in FIG.
- the difference between the information processing apparatus 1001 and the information processing apparatus 10 is that an estimation unit 1002 is added.
- the estimation unit 1002 outputs a signal that is estimated to be the cause of the abnormality based on the start time that has started to deviate from the average behavior from the plurality of signals that are determined to have deviated from the normal range input from the detection unit 108. To do.
- the estimation unit 1002 outputs the signal that first starts to change among the signals input from the detection unit 108 as a signal causing the abnormality. In order to identify the signal that has begun to change first, the estimation unit 1002 rearranges the plurality of input signals that deviate from the normal range in ascending order of the start time, and outputs the first changed signal. You may output the table rearranged in order from the earliest start time.
- the hardware configuration example of the information processing apparatus 1001 is the same as the hardware configuration of the first embodiment shown in FIG.
- the estimation unit 1002 is realized by a processing circuit such as a CPU or a system LSI. A plurality of processing circuits may be executed in cooperation.
- the estimation unit 1002 may output the estimated start time to the display 1404.
- FIG. 11 is a diagram illustrating a configuration example of the information processing system 1100 according to the second embodiment.
- the information processing system 1100 is an abnormality cause signal estimation system that estimates a signal causing an abnormality by using an information processing apparatus 1001 as an abnormality cause signal estimation apparatus.
- the data collection / management apparatus 102 manages data collected from the monitoring target 103.
- the monitoring target 103 may be a control system equipped with sensors, and can be applied to systems such as air conditioners, elevators, plant equipment, automobiles, and railway vehicles.
- the data collection / management apparatus 102 may not be provided, and the information processing apparatus 1001 may hold a functional unit corresponding to the data collection / management apparatus 102.
- the information processing apparatus 1001 is realized using a computer, the data collection / management unit is mounted on the same computer.
- an estimation unit that estimates that the monitoring target data at the start time of the earliest time among the start times of the plurality of monitoring target data input from the detection unit is the signal data causing the abnormality. Since it is provided, it is possible to estimate the signal that caused the abnormality with respect to the abnormality whose causal relationship is unknown.
- the estimation unit 1002 may hold a list indicating a physical causal relationship between signals and use a method for estimating a signal causing an abnormality from this list.
- the estimation unit 1002 holds a list in advance and determines whether or not the signal input from the detection unit 108 is in the list. If it is in the list, the estimation unit 1002 estimates the signal that causes the abnormality from the physical causal relationship. If it is not in the list, the estimation unit 1002 is estimated to be the cause of the abnormality based on the start time at which the average behavior starts to deviate. Signal. With respect to a signal for which a physical cause-and-effect relationship is grasped, a signal that causes an abnormality can be efficiently estimated.
- the estimation unit holds a list of physical causal relationships of a plurality of monitoring target data, and when the monitoring target data input from the detection unit is in the list, the cause of the abnormality based on the list Since the monitoring target data is estimated, it is possible to efficiently estimate the signal causing the abnormality for the signal for which the physical causal relationship is grasped.
- Embodiment 3 In the second embodiment described above, the signal that caused the abnormality is estimated. However, in the present embodiment, the determination time at which the abnormality is determined and the start time that starts to deviate from the average behavior. The embodiment which displays is shown.
- FIG. 12 is a block diagram showing a configuration of the information processing apparatus 1201 according to the third embodiment.
- the difference between the information processing apparatus 1201 and the information processing apparatus 101 is that a display unit 1202 is added.
- the display unit 1202 displays the determination time output from the determination unit 107 and the start time output from the detection unit 108 on the screen.
- FIG. 13 is a diagram illustrating an example of a screen displayed by the display unit 1202 according to the third embodiment.
- a graph 1301 is obtained by superimposing the monitoring target data and the normal range.
- 1302 is a table in which signals determined to be out of the normal range are rearranged in order of start time.
- 1303 is the monitoring target data
- 1304 is the upper limit value of the normal range
- 1305 is the lower limit value of the normal range.
- Reference numeral 1306 denotes a determination time when the determination unit 107 determines that an abnormality has occurred.
- Reference numeral 1307 denotes the start time detected by the detection unit 108.
- Reference numeral 1308 denotes a signal name of a signal determined to be out of the normal range
- 1309 denotes a start time estimated in each monitoring target data
- 1310 denotes a determination time determined to be abnormal in each monitoring target data.
- the hardware configuration example of the information processing apparatus 1201 is the same as the hardware configuration of the first embodiment shown in FIG.
- the display unit 1202 is a display 1404.
- the present embodiment includes a display unit that displays the monitoring target data in a graph and displays the determination time output from the determination unit and the start time detected by the detection unit on the graph. It is possible to visualize the difference between the target data and the difference between the determination time determined to be abnormal and the start time.
- the display unit displays a plurality of pieces of monitoring target data in the order of the start times detected by the detection unit, it is also possible to present the candidates for the monitoring target data causing the abnormality in the order of the high possibility of causing the abnormality. It is.
- Information processing device 101, 1201 Information processing device 102 Data management device 103 Monitored control system 104, 106 Input unit 105 Setting unit 107 Determination unit 108 Detection unit 109 Normal value learning DB 110 Monitoring target DB 111 Sensor network 201, 301, 302, 901, 902, 1301 Graph 202, 303 Signal data 203, 204 Width 304, 502, 601, 1306, 1307 indicating data variation Time 305 Average 306 Upper limit 307 Lower limit 308 Band model Width 501, 1303 Monitoring target data 701 Deviation degree data 702 Band model normal region equivalent area 703, 704 Constant 801, 903, 1304 Upper limit value 802, 904, 1305 Lower limit value 805 Band model width 1001 Information processing apparatus 1002 Estimation unit 1202 Display unit 1302 Table 1308 Signal name 1309 Start time 1310 Determination time
Abstract
Description
例えば、特許文献1では、信号間の物理的な因果関係から、異常が発生した場合に異常が伝搬する信号の順序である伝搬経路を抽出し、予めリスト化している。異常として検出した信号に基づいてリストから伝搬経路を抽出し、抽出された伝搬経路の最初にある信号を異常原因信号として推定している。
図1は、実施の形態1に係る情報処理装置101の構成を示すブロック図である。監視対象のシステムからデータを収集する手段の例として、データ収集・管理装置102の構成例、および監視対象103も併せて記載している。データ収集・管理装置102は、センサネットワーク111を介して監視対象103からデータを収集したデータを管理している。
正常値学習DB109は、古いデータを削除するようにしてもよい。例えば、監視対象のシステムの機器が更新され、古いデータが不要になった場合、正常値学習DB109は古いデータを削除する。また、学習データの容量が必要以上に多くなった場合にも削除してもよい。
監視対象103のセンサから取得した信号の集合が、センサネットワーク111を通して、データ収集・管理装置101に継続的、または断続的に入力される。信号データは、監視対象103のセンサから取得した信号の集合で、時系列データである。ここで、データ収集・管理装置102は、監視対象DB110に入力された信号データを、正常値学習DB109に入力しても構わない。また、データ収集・管理装置102は、正常値学習DB109に入力された信号データを、監視対象DB110に入力しても構わない。
データ収集・管理装置102は、異常を検知するときの正常値の基準とする信号データを、正常値学習DB109から出力し、情報処理装置101の入力部104へ入力する。データ収集・管理装置102は、異常の有無、および異常の開始時刻を推定する信号データを、監視対象DB110から出力し、情報処理装置101の入力部106へ入力する。
入力部104は、データ収集・管理装置102の正常値学習DB109から入力された信号データを変換、整形し、設定部105に出力する。設定部105は、判定部107にて異常を判定するときの、信号データの正常な値の範囲である正常域を設定する。入力部106は、監視対象DB110から入力された信号データを変換、整形し、判定部107に出力する。
図14は、実施の形態1に係る情報処理装置101のハードウェア構成例である。情報処理装置としての情報処理装置101は、受信装置1401、プロセッサ1402、メモリ1403、およびディスプレイ1404により構成される。
第1の入力部104および第2の入力部106は受信装置である。設定部105、判定部107および検出部108は、メモリに記憶されたプログラムを実行するCPU、システムLSI(Large Scale Integration)等の処理回路により、実現される。また、複数の処理回路が連携して上記機能を実行してもよい。検出部108は、算出した開始時刻をディスプレイ1404に出力してもよい。
設定部105は、判定部107において異常を判定するときの、信号データの正常域を設定する。
図4は、実施の形態1に係る設定部105がバンドモデルを作成する処理の流れを示すフローチャートである。バンドモデルの作成は、次に示す通り3段階からなる。学習データの平均および標準偏差を算出し(ステップS401)、バンドモデルの幅を算出し(ステップS402)、バンドモデルの上限値および下限値を算出する(ステップS403)。
ステップS401において、設定部105は、バンドモデルを算出する要素として、学習データの平均および標準偏差を算出する。設定部105は、学習データ202の時刻ごとの平均を数式1で算出する。また、設定部105は、時刻ごとの標準偏差を、数式2で算出する。
設定部105は、バンドモデルの時刻ごとの上限値306を示すMU(t)を数式4で算出する。数式においては、MU(t)の「U」は下付き文字で表示する。
判定部107は、入力部106から入力された信号データが、設定部105から入力された正常域を外れているか判定する。
図8は、実施の形態1に係る監視対象データの正常域の例を示すグラフである。202は、図2で示した学習データである。801は、正常域の上限値である。802は、正常域の下限値である。
一定幅のバンドモデルの幅905は、例えばバンドモデル平均305の標準偏差の定数倍としてもよいし、バンドモデル平均305の平均値の定数倍に設定してもよい。設定部105があらかじめ幅の値または、算出方法を保持しているものとする。幅の値は、監視対象システムや監視対象データの取得条件に応じて、複数あってもよい。
以上の実施の形態1では、信号に異常が発生した時刻を推定するようにしたものであるが、本実施の形態においては、異常の原因の信号を推定する実施の形態を示す。
以上の実施の形態2では、異常が発生した原因の信号を推定するようにしたものであるが、本実施の形態においては、異常を判定した判定時刻と平均的な振る舞いから外れ始めた開始時刻を表示する実施の形態を示す。
102 データ管理装置
103 監視対象の制御システム
104、106 入力部
105 設定部
107 判定部
108 検出部
109 正常値学習DB
110 監視対象DB
111 センサネットワーク
201、301、302、901、902、1301 グラフ
202、303 信号データ
203、204 データのばらつきを示す幅
304、502、601、1306、1307 時刻
305 平均
306 上限値
307 下限値
308 バンドモデル幅
501、1303 監視対象データ
701 外れ度合いデータ
702 バンドモデル正常域相当領域
703、704 定数
801、903、1304 上限値
802、904、1305 下限値
805 バンドモデル幅
1001 情報処理装置
1002 推定部
1202 表示部
1302 表
1308 信号名
1309 開始時刻
1310 判定時刻
Claims (16)
- 時系列の信号から成る監視対象データの正常な値の範囲を示す正常域を上限値と下限値とから設定する設定部と、
前記監視対象データが前記正常域を外れたか否かを判定し、外れたと判定した場合に前記監視対象データが前記正常域から外れたと判定した時刻である判定時刻を出力する判定部と、
既知の監視対象データのうち正常な値の信号から成る複数の学習データの平均値と前記監視対象データとの差を示す外れ度合いに基づいて、前記判定部から入力された前記判定時刻より前であって前記監視対象データが異常を示し始めた開始時刻を検出する検出部と、
を備えることを特徴とする情報処理装置。 - 前記設定部は、複数の時刻それぞれで前記複数の学習データの最大値を前記上限値に設定するとともに、前記複数の学習データの最小値を前記下限値に設定することを特徴とする請求項1に記載の情報処理装置。
- 前記設定部は、前記上限値を前記複数の時刻で同じ値に設定するとともに、前記下限値を前記複数の時刻で同じ値に設定することを特徴とする請求項1に記載の情報処理装置。
- 前記設定部は、前記上限値と前記下限値について前記複数の学習データの平均値との差が前記複数の時刻で同じ値になるように前記上限値と前記下限値とを設定することを特徴とする請求項1に記載の情報処理装置。
- 前記設定部は、前記複数の学習データから相関係数に基づく特徴量を算出し、前記特徴量の範囲から前記上限値と前記下限値とを設定することを特徴とする請求項1に記載の情報処理装置。
- 前記設定部は、前記複数の学習データからマハラノビス距離に基づく特徴量を算出し、前記特徴量の範囲から前記上限値と前記下限値とを設定することを特徴とする請求項1に記載の情報処理装置。
- 前記検出部は、前記判定時刻より前であって前記外れ度合いの傾きが第1の閾値以上となる時刻を前記開始時刻として検出することを特徴とする請求項1から6のいずれか一項に記載の情報処理装置。
- 前記検出部は、前記外れ度合いの傾きを平滑化処理した後、前記開始時刻を検出することを特徴とする請求項7に記載の情報処理装置。
- 前記検出部は、前記判定時刻より前であって前記外れ度合いの傾きが第1の閾値以上となるとともに、前記外れ度合いが第2の閾値以上となる時刻を前記開始時刻として検出することを特徴とする請求項1から6のいずれか一項に記載の情報処理装置。
- 前記検出部は、前記外れ度合いまたは前記外れ度合いの傾きを平滑化処理した後、前記開始時刻を検出することを特徴とする請求項9に記載の情報処理装置。
- 前記検出部は、前記外れ度合いからベイズの変化点検出に基づいて前記開始時刻を検出することを特徴とする請求項1から6のいずれか一項に記載の情報処理装置。
- 前記検出部から入力された複数の監視対象データの前記開始時刻のうち、最も早い時刻の前記開始時刻の監視対象データが異常の原因の信号データであると推定する推定部を備えたことを特徴とする請求項1から11のいずれか一項に記載の情報処理装置。
- 前記推定部は、複数の監視対象データの物理的な因果関係のリストを保持し、前記検出部から入力された監視対象データが前記リストにある場合、前記リストに基づいて異常の原因の監視対象データを推定することを特徴とする請求項12に記載の情報処理装置。
- 監視対象データをグラフで表示するとともに前記判定部から出力される前記判定時刻と前記検出部が検出した前記開始時刻とを前記グラフ上に表示する表示部を備えたことを特徴とする請求項1から11のいずれか一項に記載の情報処理装置。
- 複数の監視対象データについて前記検出部が検出した前記開始時刻を表示する表示部を備えたことを特徴とする請求項1から11のいずれか一項に記載の情報処理装置。
- 時系列の信号から成る監視対象データの正常な値の範囲を示す正常域を上限値と下限値とから設定する設定ステップと、
前記監視対象データが前記正常域を外れたか否かを判定し、外れたと判定した場合に前記監視対象データが前記正常域から外れたと判定した時刻である判定時刻を出力する判定ステップと、
既知の監視対象データのうち正常な値の信号から成る複数の学習データの平均値と前記監視対象データとの差を示す外れ度合いに基づいて、前記判定時刻より前であって前記監視対象データが異常を示し始めた開始時刻を検出する検出ステップと、
を有する情報処理方法。
Priority Applications (5)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
US15/525,549 US20170316329A1 (en) | 2015-01-21 | 2015-01-21 | Information processing system and information processing method |
EP15878662.4A EP3249483B1 (en) | 2015-01-21 | 2015-01-21 | Information processing device and information processing method |
JP2016570205A JP6330922B2 (ja) | 2015-01-21 | 2015-01-21 | 情報処理装置および情報処理方法 |
PCT/JP2015/000243 WO2016116961A1 (ja) | 2015-01-21 | 2015-01-21 | 情報処理装置および情報処理方法 |
CN201580073632.3A CN107209508B (zh) | 2015-01-21 | 2015-01-21 | 信息处理装置及信息处理方法 |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
PCT/JP2015/000243 WO2016116961A1 (ja) | 2015-01-21 | 2015-01-21 | 情報処理装置および情報処理方法 |
Publications (1)
Publication Number | Publication Date |
---|---|
WO2016116961A1 true WO2016116961A1 (ja) | 2016-07-28 |
Family
ID=56416534
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
PCT/JP2015/000243 WO2016116961A1 (ja) | 2015-01-21 | 2015-01-21 | 情報処理装置および情報処理方法 |
Country Status (5)
Country | Link |
---|---|
US (1) | US20170316329A1 (ja) |
EP (1) | EP3249483B1 (ja) |
JP (1) | JP6330922B2 (ja) |
CN (1) | CN107209508B (ja) |
WO (1) | WO2016116961A1 (ja) |
Cited By (12)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN106656837A (zh) * | 2016-10-14 | 2017-05-10 | 东软集团股份有限公司 | 网络拥塞问题的定位方法以及装置 |
WO2018061842A1 (ja) * | 2016-09-27 | 2018-04-05 | 東京エレクトロン株式会社 | 異常検知プログラム、異常検知方法および異常検知装置 |
JP2018077764A (ja) * | 2016-11-11 | 2018-05-17 | 東京エレクトロン株式会社 | 異常検知装置 |
JP2018147255A (ja) * | 2017-03-06 | 2018-09-20 | 株式会社大林組 | 異常判定システム |
WO2018207350A1 (ja) | 2017-05-12 | 2018-11-15 | 三菱電機株式会社 | 時系列データ処理装置、時系列データ処理システムおよび時系列データ処理方法 |
WO2019159227A1 (ja) * | 2018-02-13 | 2019-08-22 | 三菱電機株式会社 | 鉄道車両の状態監視システム |
JP2019537123A (ja) * | 2017-08-11 | 2019-12-19 | アイティーエス カンパニー リミテッドIts Co., Ltd. | 駆動部の精密予知保全方法 |
JP2019537122A (ja) * | 2017-08-11 | 2019-12-19 | アイティーエス カンパニー リミテッドIts Co., Ltd. | 駆動部の精密予知保全方法 |
JP2020038590A (ja) * | 2018-09-06 | 2020-03-12 | 日立オートモティブシステムズ株式会社 | データ検証装置、状態監視装置、及びデータ検証方法 |
JP2020038945A (ja) * | 2018-09-06 | 2020-03-12 | パナソニックIpマネジメント株式会社 | 実装システム、異常判断装置及び異常判断方法 |
WO2020152741A1 (ja) * | 2019-01-21 | 2020-07-30 | 三菱電機株式会社 | 異常要因推定装置、異常要因推定方法、及びプログラム |
JPWO2022049701A1 (ja) * | 2020-09-03 | 2022-03-10 |
Families Citing this family (18)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JP5845374B1 (ja) * | 2015-08-05 | 2016-01-20 | 株式会社日立パワーソリューションズ | 異常予兆診断システム及び異常予兆診断方法 |
DE102017219549A1 (de) * | 2017-11-03 | 2019-05-09 | Volkswagen Aktiengesellschaft | Verfahren zur Zustandsüberwachung einer Fertigungsanlage |
JP6721563B2 (ja) * | 2017-11-28 | 2020-07-15 | ファナック株式会社 | 数値制御装置 |
WO2019123510A1 (ja) * | 2017-12-18 | 2019-06-27 | 三菱電機株式会社 | 表示制御装置、表示システム、表示装置、表示方法および表示プログラム |
KR102240993B1 (ko) * | 2018-03-20 | 2021-04-15 | 미쓰비시덴키 가부시키가이샤 | 표시 장치, 표시 시스템, 및 표시 화면 생성 방법 |
WO2019239542A1 (ja) * | 2018-06-14 | 2019-12-19 | 三菱電機株式会社 | 異常検知装置、異常検知方法及び異常検知プログラム |
JP7031512B2 (ja) * | 2018-06-25 | 2022-03-08 | 東芝三菱電機産業システム株式会社 | 鉄鋼プラント用監視作業支援システム |
WO2020044533A1 (ja) * | 2018-08-31 | 2020-03-05 | 東芝三菱電機産業システム株式会社 | 製造プロセス監視装置 |
WO2020049615A1 (ja) * | 2018-09-03 | 2020-03-12 | 三菱電機株式会社 | 信号表示制御装置および信号表示制御プログラム |
JP6631746B1 (ja) * | 2018-09-28 | 2020-01-15 | ダイキン工業株式会社 | クラスタ分類装置、環境生成装置及び環境生成システム |
JP6760348B2 (ja) | 2018-10-11 | 2020-09-23 | 株式会社富士通ゼネラル | 空気調和機、データ送信方法及び空気調和システム |
WO2020115827A1 (ja) * | 2018-12-05 | 2020-06-11 | 三菱電機株式会社 | 異常検知装置及び異常検知方法 |
JP7115346B2 (ja) * | 2019-02-07 | 2022-08-09 | 株式会社デンソー | 異常検知装置 |
JP6790154B2 (ja) * | 2019-03-07 | 2020-11-25 | 東芝デジタルソリューションズ株式会社 | 協調型学習システム及び監視システム |
FR3099596B1 (fr) * | 2019-07-30 | 2023-09-01 | Commissariat Energie Atomique | Procédé d’analyse et procédé de détermination et de prédiction du régime de fonctionnement d’un système énergétique |
WO2021065068A1 (ja) * | 2019-09-30 | 2021-04-08 | 株式会社オートネットワーク技術研究所 | 検知装置、車両、検知方法および検知プログラム |
JP7205514B2 (ja) * | 2020-03-31 | 2023-01-17 | 横河電機株式会社 | 学習データ処理装置、学習データ処理方法、学習データ処理プログラム、及び非一時的なコンピュータ読み取り可能な媒体 |
US11620162B2 (en) * | 2021-05-24 | 2023-04-04 | Capital One Services, Llc | Resource allocation optimization for multi-dimensional machine learning environments |
Citations (7)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JPH0916250A (ja) * | 1995-06-26 | 1997-01-17 | Toshiba Eng Co Ltd | 監視制御装置 |
JP2002341909A (ja) * | 2001-05-18 | 2002-11-29 | Sofutorokkusu:Kk | 工作機器の監視方法 |
JP2009075692A (ja) * | 2007-09-19 | 2009-04-09 | Toshiba Corp | プラント警報装置およびプラント警報方法 |
JP2010211440A (ja) * | 2009-03-10 | 2010-09-24 | Railway Technical Res Inst | 異常予測装置、異常予測システム、異常予測方法、及びプログラム |
JP2011170518A (ja) * | 2010-02-17 | 2011-09-01 | Nec Corp | 状態監視装置及び方法 |
WO2012073289A1 (ja) * | 2010-12-02 | 2012-06-07 | 株式会社日立製作所 | プラントの診断装置及びプラントの診断方法 |
JP2013008092A (ja) * | 2011-06-22 | 2013-01-10 | Hitachi Engineering & Services Co Ltd | 異常予兆診断装置及び異常予兆診断方法 |
Family Cites Families (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JP2006252229A (ja) * | 2005-03-11 | 2006-09-21 | Nec Electronics Corp | 異常検出システムおよび異常検出方法 |
US8285513B2 (en) * | 2007-02-27 | 2012-10-09 | Exxonmobil Research And Engineering Company | Method and system of using inferential measurements for abnormal event detection in continuous industrial processes |
US9405291B2 (en) * | 2012-07-31 | 2016-08-02 | Fisher-Rosemount Systems, Inc. | Systems and methods to monitor an asset in an operating process unit |
GB201215649D0 (en) * | 2012-09-03 | 2012-10-17 | Isis Innovation | System monitoring |
-
2015
- 2015-01-21 CN CN201580073632.3A patent/CN107209508B/zh active Active
- 2015-01-21 JP JP2016570205A patent/JP6330922B2/ja active Active
- 2015-01-21 WO PCT/JP2015/000243 patent/WO2016116961A1/ja active Application Filing
- 2015-01-21 US US15/525,549 patent/US20170316329A1/en not_active Abandoned
- 2015-01-21 EP EP15878662.4A patent/EP3249483B1/en active Active
Patent Citations (7)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JPH0916250A (ja) * | 1995-06-26 | 1997-01-17 | Toshiba Eng Co Ltd | 監視制御装置 |
JP2002341909A (ja) * | 2001-05-18 | 2002-11-29 | Sofutorokkusu:Kk | 工作機器の監視方法 |
JP2009075692A (ja) * | 2007-09-19 | 2009-04-09 | Toshiba Corp | プラント警報装置およびプラント警報方法 |
JP2010211440A (ja) * | 2009-03-10 | 2010-09-24 | Railway Technical Res Inst | 異常予測装置、異常予測システム、異常予測方法、及びプログラム |
JP2011170518A (ja) * | 2010-02-17 | 2011-09-01 | Nec Corp | 状態監視装置及び方法 |
WO2012073289A1 (ja) * | 2010-12-02 | 2012-06-07 | 株式会社日立製作所 | プラントの診断装置及びプラントの診断方法 |
JP2013008092A (ja) * | 2011-06-22 | 2013-01-10 | Hitachi Engineering & Services Co Ltd | 異常予兆診断装置及び異常予兆診断方法 |
Cited By (23)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
TWI737816B (zh) * | 2016-09-27 | 2021-09-01 | 日商東京威力科創股份有限公司 | 異常檢測程式、異常檢測方法及異常檢測裝置 |
WO2018061842A1 (ja) * | 2016-09-27 | 2018-04-05 | 東京エレクトロン株式会社 | 異常検知プログラム、異常検知方法および異常検知装置 |
JPWO2018061842A1 (ja) * | 2016-09-27 | 2019-06-27 | 東京エレクトロン株式会社 | 異常検知プログラム、異常検知方法および異常検知装置 |
CN106656837A (zh) * | 2016-10-14 | 2017-05-10 | 东软集团股份有限公司 | 网络拥塞问题的定位方法以及装置 |
JP2018077764A (ja) * | 2016-11-11 | 2018-05-17 | 東京エレクトロン株式会社 | 異常検知装置 |
CN108074832A (zh) * | 2016-11-11 | 2018-05-25 | 东京毅力科创株式会社 | 异常探测装置 |
JP2018147255A (ja) * | 2017-03-06 | 2018-09-20 | 株式会社大林組 | 異常判定システム |
WO2018207350A1 (ja) | 2017-05-12 | 2018-11-15 | 三菱電機株式会社 | 時系列データ処理装置、時系列データ処理システムおよび時系列データ処理方法 |
JPWO2018207350A1 (ja) * | 2017-05-12 | 2019-11-07 | 三菱電機株式会社 | 時系列データ処理装置、時系列データ処理システムおよび時系列データ処理方法 |
US11531688B2 (en) | 2017-05-12 | 2022-12-20 | Mitsubishi Electric Corporation | Time-series data processing device, time-series data processing system, and time-series data processing method |
JP2019537123A (ja) * | 2017-08-11 | 2019-12-19 | アイティーエス カンパニー リミテッドIts Co., Ltd. | 駆動部の精密予知保全方法 |
JP2019537122A (ja) * | 2017-08-11 | 2019-12-19 | アイティーエス カンパニー リミテッドIts Co., Ltd. | 駆動部の精密予知保全方法 |
WO2019159227A1 (ja) * | 2018-02-13 | 2019-08-22 | 三菱電機株式会社 | 鉄道車両の状態監視システム |
JP7090655B2 (ja) | 2018-02-13 | 2022-06-24 | 三菱電機株式会社 | 鉄道車両の状態監視システム |
JPWO2019159227A1 (ja) * | 2018-02-13 | 2021-03-11 | 三菱電機株式会社 | 鉄道車両の状態監視システム |
JP2020038590A (ja) * | 2018-09-06 | 2020-03-12 | 日立オートモティブシステムズ株式会社 | データ検証装置、状態監視装置、及びデータ検証方法 |
JP7034038B2 (ja) | 2018-09-06 | 2022-03-11 | 日立Astemo株式会社 | データ検証装置、状態監視装置、及びデータ検証方法 |
JP7153855B2 (ja) | 2018-09-06 | 2022-10-17 | パナソニックIpマネジメント株式会社 | 実装システム、異常判断装置及び異常判断方法 |
JP2020038945A (ja) * | 2018-09-06 | 2020-03-12 | パナソニックIpマネジメント株式会社 | 実装システム、異常判断装置及び異常判断方法 |
WO2020152741A1 (ja) * | 2019-01-21 | 2020-07-30 | 三菱電機株式会社 | 異常要因推定装置、異常要因推定方法、及びプログラム |
JPWO2022049701A1 (ja) * | 2020-09-03 | 2022-03-10 | ||
WO2022049701A1 (ja) * | 2020-09-03 | 2022-03-10 | 三菱電機株式会社 | 機器分析装置、機器分析方法および機器分析プログラム |
JP7214054B2 (ja) | 2020-09-03 | 2023-01-27 | 三菱電機株式会社 | 機器分析装置、機器分析方法および機器分析プログラム |
Also Published As
Publication number | Publication date |
---|---|
EP3249483A1 (en) | 2017-11-29 |
CN107209508B (zh) | 2018-08-28 |
CN107209508A (zh) | 2017-09-26 |
JPWO2016116961A1 (ja) | 2017-08-10 |
US20170316329A1 (en) | 2017-11-02 |
EP3249483B1 (en) | 2020-05-20 |
JP6330922B2 (ja) | 2018-05-30 |
EP3249483A4 (en) | 2018-09-12 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
JP6330922B2 (ja) | 情報処理装置および情報処理方法 | |
US9658916B2 (en) | System analysis device, system analysis method and system analysis program | |
US9753801B2 (en) | Detection method and information processing device | |
US11067973B2 (en) | Data collection system, abnormality detection method, and gateway device | |
JP6351862B2 (ja) | 運転状態分類装置 | |
US10346758B2 (en) | System analysis device and system analysis method | |
WO2013042789A1 (ja) | 運用管理装置、運用管理方法、及びプログラム | |
US9524223B2 (en) | Performance metrics of a computer system | |
JP7040851B2 (ja) | 異常検知装置、異常検知方法及び異常検知プログラム | |
JPWO2018104985A1 (ja) | 異常分析方法、プログラムおよびシステム | |
US20160004620A1 (en) | Detection apparatus, detection method, and recording medium | |
JPWO2018073960A1 (ja) | 表示方法、表示装置、および、プログラム | |
JP6489235B2 (ja) | システム分析方法、システム分析装置、および、プログラム | |
US20150378806A1 (en) | System analysis device and system analysis method | |
JP5928104B2 (ja) | 性能監視装置、性能監視方法、及びそのプログラム | |
US20190064789A1 (en) | System analyzing device, system analyzing method, and computer-readable recording medium | |
WO2018122890A1 (ja) | ログ分析方法、システムおよびプログラム | |
JPWO2017033443A1 (ja) | 渋滞予測システム、渋滞予測方法、及び、プログラム | |
EP2527981A2 (en) | Process allocation apparatus and process allocation method | |
US10140577B2 (en) | Data processing method and apparatus | |
CN109976986B (zh) | 异常设备的检测方法及装置 | |
US10845794B2 (en) | State identification device, state identification method and mechanical device | |
TWI636419B (zh) | Information processing device, information processing method and information processing program product | |
JP2012037991A (ja) | 予測装置、予測システム及びプログラム | |
US20210183528A1 (en) | Information processing apparatus, information processing method, and non-transitory computer readable medium |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
121 | Ep: the epo has been informed by wipo that ep was designated in this application |
Ref document number: 15878662 Country of ref document: EP Kind code of ref document: A1 |
|
ENP | Entry into the national phase |
Ref document number: 2016570205 Country of ref document: JP Kind code of ref document: A |
|
WWE | Wipo information: entry into national phase |
Ref document number: 15525549 Country of ref document: US |
|
REEP | Request for entry into the european phase |
Ref document number: 2015878662 Country of ref document: EP |
|
NENP | Non-entry into the national phase |
Ref country code: DE |