JP2014035282A - Abnormality diagnostic device - Google Patents

Abstract

PROBLEM TO BE SOLVED: To diagnose abnormality of a plant with high accuracy.SOLUTION: An abnormality diagnostic device includes: accumulated data storage part 20 for storing accumulated data including values of respective variables having input in the past; a determination part 12 for extracting a newly input value, the maximum value and minimum value in a predetermined period included in the accumulated data for each variable to determine an intermediate value as each medium value; a first calculation part 13 for calculating a differential value between the newly input value and the medium value determined by the determination part 12 for each variable; a second calculation part 14 for calculating a Mahalanobis distance using the obtained differential value of each variable and data of a predetermined unit space; and a judgment part 15 for judging whether the obtained Mahalanobis distance is within the range of a prescribed threshold to diagnose abnormality.

Description

  The present invention relates to an abnormality diagnosis device for diagnosing plant abnormality.

  For diagnosis of abnormality in a plant, a plurality of detection values detected in the plant may be used. However, some of these detected values are subject to periodic fluctuations (for example, seasonal fluctuations). Therefore, it is preferable to remove the periodic fluctuation from the detected value in advance in the abnormality diagnosis.

  For example, an average value of detection values obtained during a predetermined period may be obtained as a fluctuation value representing periodic fluctuation, and abnormality may be diagnosed using a value obtained by removing the fluctuation value from a newly input detection value. . In plants, detection values are obtained for a large number of variables, so the MT method (Mahalanobis Taguchi method) is sometimes used as a method for easily diagnosing abnormalities using a large number of detection values. Also in the diagnosis, a method of removing the average value obtained as the fluctuation value may be used.

  There is also a method of diagnosing an abnormality by standardizing detection values and performing moving average processing based on the standardized data (see, for example, Patent Document 1).

  As described above, the average value of the detection values used as the fluctuation value to be removed from the detection values in the abnormality diagnosis is greatly affected by the sudden abnormality. Therefore, if the average value is affected by an abnormality and the average value is removed from the detection value as a fluctuation value, the abnormality value may be removed from the detection value, and the accuracy of abnormality diagnosis may be reduced.

JP 2010-181188 A

  As described above, when an average value of detected values is used as a fluctuation value, it is difficult to diagnose a plant abnormality with high accuracy.

  In view of the above problems, an object of the present invention is to provide an abnormality diagnosis apparatus that diagnoses an abnormality of a plant with high accuracy.

  In order to achieve the above object, an invention according to claim 1 is an abnormality diagnosis device for diagnosing a plant abnormality by comparing values of a plurality of variables newly input from a plant with a predetermined unit space. An accumulated data storage unit for storing accumulated data including the value of each variable input in the above, and for each variable, a newly input value and a maximum value and a minimum value of a predetermined period value included in the accumulated data are extracted, A determination unit that determines each intermediate value as a median value; a first calculation unit that calculates a difference value between the newly input value and the median value determined by the determination unit for each variable; and A second calculation unit that obtains the Mahalanobis distance using the difference value and data of a predetermined unit space, and a determination unit that determines whether the obtained Mahalanobis distance is within a predetermined threshold range and diagnoses an abnormality With.

  According to the present invention, it is possible to improve the accuracy of plant abnormality diagnosis.

It is a block diagram explaining the abnormality diagnosis apparatus concerning an embodiment. It is a data block diagram explaining an example of the accumulation | storage data utilized with an abnormality diagnosis apparatus. It is a flowchart explaining an example of the process in an abnormality diagnosis apparatus. It is a conceptual diagram explaining the abnormality diagnosis using MT method. It is a graph explaining the detected value and fluctuation value which are used with the abnormality diagnosis apparatus.

  The abnormality diagnosis apparatus according to the embodiment is an abnormality diagnosis apparatus that diagnoses an abnormality in plant operation. For example, the plant diagnosed by the abnormality diagnosis device is a power plant. The power plant includes a plurality of devices (pumps, valves, etc.), and values for controlling these devices are set as target values. This target value is, for example, the pressure of the pump, opening / closing of a valve, or the like. In addition, the power plant includes a plurality of sensors, and each sensor measures the intake air amount, the exhaust air amount, the power generation amount, and the like. In the following description, the value of each variable representing the state of the plant such as a target value or a measured value will be described as detection values, and data including these multiple detection values will be described as detection data.

  In such a plant, each detected value may fluctuate periodically depending on time and season even when it is operating normally. For example, in the case of a power plant, the amount of power generated in the daytime is larger than that in the nighttime, the amount of power generation in summer is larger than that in spring, and the temperature in the surroundings is higher in summer than in winter. This is because the operating conditions and the environment around the plant change according to time and season. In the abnormality diagnosis device according to the embodiment, when there is such a periodic variation, the periodic variation can be adjusted, and the abnormality can be diagnosed without being affected by the variation.

  As illustrated in FIG. 1, the abnormality diagnosis apparatus 1 according to the embodiment includes an addition unit 11 that adds a detection value input from a plant to accumulated data 21, and a maximum value and a minimum value of a predetermined period for each variable. A determination unit 12 that obtains a median value that is an intermediate value, a first calculation unit 13 that obtains a difference between a newly input detection value and a median value for each variable, and a difference between each obtained variable and unit space data 23, a second calculation unit 14 for determining the Mahalanobis distance, and a determination unit 15 for determining whether the calculated Mahalanobis distance is within a predetermined threshold range and diagnosing an abnormality.

  The abnormality diagnosis apparatus 1 is, for example, an information processing apparatus including a central processing unit (CPU) 10 and a storage device 20, and the abnormality diagnosis program P stored in the storage device 20 is read and executed. As illustrated in FIG. 1, an adding unit 11, a determination unit 12, a first calculation unit 13, a second calculation unit 14, and a determination unit 15 are mounted on the CPU 10.

  In the example of the abnormality diagnosis device 1 shown in FIG. 1, the storage device 20 stores accumulated data 21, fluctuation value data 22, and unit space data 23 in addition to the abnormality diagnosis program P. Further, the abnormality diagnosis device 1 is connected to an input device 2 such as a keyboard, a mouse, an operation button, and a touch panel used for inputting operation signals. It is connected to the device 3.

  The accumulated data 21 is data obtained by accumulating past detection data. In the example shown in FIG. 2, the accumulated data 21 is a target value (variables 1 and 2) set in each device of the plant, which is a condition value for plant operation, or a measured value (variable 3) measured in the case of this target value. 4) and the like. In addition, in the accumulated data 21 shown in FIG. 2, a sample number in time order is given to each record.

  When the addition unit 11 newly receives detection data including detection values of a plurality of variables from the plant, the addition unit 11 generates a record including the input detection data and a sample number, and generates the accumulated data 21 stored in the storage device 20. Add a record.

  The determination unit 12 reads the accumulated data 21 from the storage device 20 at a predetermined timing for obtaining the median, and extracts a record that satisfies the extraction condition. Further, the determination unit 12 extracts the maximum value and the minimum value for each variable from the detection values included in the extracted record, and obtains an intermediate value ((maximum value + minimum value) / 2) as a median value. The determination unit 12 uses the obtained median value for each variable as a fluctuation value representing a periodic fluctuation, and causes the storage device 20 to store data including the obtained fluctuation value of each variable as the fluctuation value data 22.

  Here, the predetermined timing at which the determining unit 12 obtains the median value is, for example, the timing at which an operation signal requesting to obtain the median value is input via the input device 2, every few minutes, every several hours, or several It is a predetermined periodic timing such as every day.

  The extraction condition used by the determination unit 12 for extracting records from the accumulated data 21 is a period such as several hours or several days. For example, when “one month” is defined as the extraction condition, the determination unit 12 extracts the latest record for one month from the accumulated data 21, and sets the maximum value and the minimum value in the past month for each variable. The value is obtained, and the median value of each variable is determined as the variation value. This extraction condition is determined in an optimum period according to the characteristics of periodic fluctuations. For example, if the fluctuation is long-term, the period of the extraction condition is long, and if the fluctuation is short-term, the period of the extraction condition is short. In addition, it is not always necessary to use an extraction condition for extracting continuous detection values, and an extraction condition for extracting only a daytime detection value or a nighttime detection value may be used.

  The fluctuation value data 22 is data including the fluctuation value of each variable determined by the determination unit 12 in this way. The fluctuation value data 22 only needs to include at least the latest fluctuation value of each variable.

  The first calculator 13 reads the fluctuation value data 22 from the storage device 20 at a predetermined timing for diagnosing abnormality. Moreover, the 1st calculation part 13 calculates | required the difference of the variation value contained in the detected value newly input from the plant and the read variation value data 22 about each variable, and calculated | required the difference value (detection value-variation value). Is used as a diagnostic value obtained by removing the fluctuation value from the detected value, and is output to the second calculator 14. Here, the predetermined timing for diagnosing abnormality is, for example, a periodic timing such as every few minutes.

  The second calculation unit 14 reads the unit space data 23 from the storage device 20 when the diagnosis value of each variable obtained by the first calculation unit 13 is input. In addition, the second calculation unit 14 obtains the Mahalanobis distance for the unit space defined by the unit space data 23 by using the input diagnostic value for each variable and the read unit space data 23. The obtained value is output to the determination unit 15.

  The unit space data 23 is data representing a unit space used for abnormality diagnosis. Specifically, the unit space data 23 includes values of a plurality of variables used to generate the Mahalanobis space, which is a unit space, an average value of each variable, a standard deviation of each variable, and a correlation matrix for each variable. This is data in which an inverse matrix is associated with a threshold value set in the Mahalanobis space. The value of the variable included in the unit space data 23 is a value selected as a unit space generation target from the value of the variable included in the accumulated data 21.

  Specifically, (1) The second calculation unit 14 first obtains an average value and a standard deviation for each variable. (2) After that, the second calculation unit 14 standardizes the data using the value of each variable, the average value and the standard deviation obtained for each variable, and obtains a normalized value for each variable. (3) Subsequently, the second calculation unit 14 obtains a correlation matrix for each variable by using the normalized value obtained for each variable and obtains an inverse matrix of the correlation matrix. (4) Finally, the second calculation unit 14 obtains the Mahalanobis distance using the obtained inverse matrix.

  The determination unit 15 reads the unit space data 23 from the storage device 20 when the Mahalanobis distance is input from the second calculation unit 14. Further, the determination unit 15 determines the possibility of abnormality in the plant by comparing the Mahalanobis distance input from the second calculation unit 14 with a threshold value determined for the unit space defined by the unit space data 23. That is, when the calculated Mahalanobis distance is within the threshold, it is determined that there is no possibility of abnormality, and when it is outside the threshold, it is determined that there is a possibility of abnormality. Thereafter, the determination unit 15 outputs the determination result to the output device 3.

  For example, as shown in FIG. 4, when the abnormality is determined using the MT method, the determination is made when the Mahalanobis distance D1 obtained by the second calculation unit 14 is smaller than a set threshold value, that is, within the Mahalanobis space. The unit 15 determines that the current state of the plant is the same as the normal state, and the plant is operating normally. On the other hand, when the Mahalanobis distance D2 obtained by the second calculating unit 14 is larger than the set threshold value, that is, when the Mahalanobis space is out of the Mahalanobis space, the determination unit 15 is in a state where the current plant state is different from the usual state. It is determined that there is a possibility of abnormality in the plant.

  In addition, the abnormality diagnosis apparatus 1 may be comprised from several information processing apparatus, for example, only the addition part 11 may be contained in the information processing apparatus different from the other process parts 12-15. Further, only a part of the data stored in the storage device 20 may be stored in an external storage device.

  Next, a process for diagnosing an abnormality in the abnormality diagnosis apparatus 1 will be described using the flowchart shown in FIG.

  First, the determination unit 12 reads the accumulated data 21 from the storage device 20 at a predetermined timing to extract the record value satisfying the extraction condition, obtains a median value for each variable, and obtains the obtained median value for each variable. Fluctuation value data 22 is generated as a variation value and stored in the storage device 20 (S1).

  When new detection data is input from the plant (S2), the first calculation unit 13 reads the fluctuation value data 22 from the storage device 20 at a predetermined timing, and for each variable, detection included in the newly input detection data. The difference between the value and the fluctuation value included in the fluctuation value data 22 is obtained (S3).

  Subsequently, the second calculation unit 14 reads the unit space data 23 from the storage device 20, and obtains the variation value of each variable that is the calculation result of the first calculation unit 13 and the value included in the read unit space data 23. Utilizing this, the Mahalanobis distance is obtained (S4). After that, the determination unit 15 determines the possibility of abnormality by comparing the Mahalanobis distance calculated by the second calculation unit 14 with the threshold value included in the unit space data 23 (S5).

  When there is no possibility of abnormality (NO in S6), the determination unit 15 determines that the plant is operating normally, and returns to the process of step S2. The determination unit 15 may output the determination result to the output device 3.

  On the other hand, when it is determined that there is a possibility of abnormality (YES in S6), the determination unit 15 outputs the determination result to the output device 3 (S7). Thereby, since the possibility of abnormality can be grasped | ascertained, the specific process of an abnormality cause is performed by an operator after that.

  In the flowchart illustrated in FIG. 3, the process of adding the detection value newly input by the adding unit 11 to the accumulated data 21 is omitted.

  For example, it is assumed that a detected value is periodically input from a plant for a variable as shown in FIG. Also, it is assumed that the detected value includes an abnormal value that is partly abnormal. In this case, the average value obtained using the detection value shown in FIG. 5A is as shown in FIG. Here, although the degree of influence varies depending on the extraction condition used to calculate the average value, the average value is greatly affected by the abnormal value as shown in FIG. 5B. In other words, the longer the period used for calculating the average value (the more samples of detection values used for calculating the average value), the longer the period affected by the abnormal value in the average value, and the average value calculation. The shorter the period used for (the smaller the number of detected value samples used for calculating the average value), the greater the influence of each average value on the abnormal value. Therefore, if the average value is used as the fluctuation value, the abnormal value is also removed from the detected value by the fluctuation value.

  On the other hand, the median value obtained using the detected value shown in FIG. 5A is as shown in FIG. That is, the median value is an intermediate value between the maximum value and the minimum value in the specified period, and thus the influence of the abnormal value is less than the average value. Therefore, even if such an intermediate value is excluded from the detected value, the abnormal value remains, so that the abnormality can be detected.

  As described above, the abnormality diagnosis apparatus according to the present invention uses a median value that is less susceptible to anomalies in a fluctuation value that represents a periodic fluctuation such as a seasonal fluctuation. Therefore, in the abnormality diagnosis apparatus according to the present invention, the influence of abnormality is likely to appear in the required Mahalanobis distance, the abnormality can be detected quickly, and the abnormality of the plant can be diagnosed with high accuracy.

  As mentioned above, although this invention was demonstrated in detail using embodiment, this invention is not limited to embodiment described in this specification. The scope of the present invention is determined by the description of the claims and the scope equivalent to the description of the claims.

1 ... Abnormality diagnosis device 10 ... CPU
DESCRIPTION OF SYMBOLS 11 ... Addition part 12 ... Determination part 13 ... 1st calculation part 14 ... 2nd calculation part 15 ... Determination part 20 ... Memory | storage device 21 ... Accumulated data 22 ... Fluctuation value data 23 ... Unit space data P ... Abnormality diagnosis program 2 ... Input Device 3 ... Output device

Claims (1)

An abnormality diagnosis device for diagnosing an abnormality in the plant by comparing values of a plurality of variables newly input from the plant with a predetermined unit space,
An accumulated data storage unit for storing accumulated data including values of variables input in the past;
For each variable, a newly entered value and a maximum value and a minimum value of a predetermined period value included in the accumulated data, and a determination unit that determines an intermediate value as a median value,
For each variable, a first calculation unit for obtaining a difference value between a newly input value and the median value determined by the determination unit;
A second calculation unit for obtaining a Mahalanobis distance using the obtained difference value of each variable and data of a predetermined unit space;
An abnormality diagnosis apparatus comprising: a determination unit that determines whether the calculated Mahalanobis distance is within a predetermined threshold range and diagnoses the abnormality.

Images