JP2000029513A - Device and method for monitoring process data - Google Patents

Abstract

PROBLEM TO BE SOLVED: To exactly detect an abnormality of a plant on the stage of symptoms by constructing a configuration capable of comparatively easily setting a monitor reference value based on long-term characteristics in the process amount of the plant without requiring much labor. SOLUTION: This process data monitoring device is provided with sensors 1...1 for detecting the process data of the plant while converting them into electric signals, an A/D converting part 2 for digitizing the electric signals detected by these respective sensors 1...1, a statistic processing part 4 for preparing statistic value data by statistically processing the instantaneous values of process data converted by this converting part 2, a data storage part 3 for storing the statistic values and instantaneous values, a reference value setting part 8 for setting a reference value for process amount monitor, a reference value storage part 6 for storing that reference value, a monitoring part 5 for mutually comparing a monitoring object amount containing at least one of the instantaneous value and statistic value in the data storage part 3 and detecting the change state of the process amount, and a display part 7 for displaying monitor information containing data detected by this monitoring part 5 on a screen.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a process data monitoring apparatus and method for monitoring a plant process using process data such as temperature, pressure, and flow rate of a large-scale plant such as a nuclear power plant. The present invention relates to a device for data processing and analysis capable of detecting a plant abnormality at a symptom stage.

[0002]

2. Description of the Related Art In general, in a nuclear power plant, a part of a detection signal corresponding to process data such as temperature, flow rate, pressure, etc. from various sensors installed at each location is sent to a central operation room and recorded. By outputting to a meter or an indicator, the status of the plant process, such as normal / abnormal, is grasped and monitored. When monitoring such process data, in addition to outputting data to the recorder and indicator, an alarm value is set in advance for important process data in the indicator, and an alert is generated when the data changes to the alarm set value. It is possible.

[0003] In recent nuclear power plants, there is also a monitoring device that takes in important process data into a computer and displays it on a screen of a display device, or executes a comparison process between the data and a pre-input reference value. Has been adopted.

[0004]

However, the above-mentioned conventional process data monitoring apparatus is generally configured to input a design value as a reference value used in the comparison processing.
Since the reference value is always constant, it does not always accurately reflect characteristics such as changes in process data to be monitored.In particular, abnormal situations of the plant are detected at the symptom stage, and It is not always appropriate to take proactive measures.

For example, even if the process data of a plant is in a normal state, it often changes due to its operating conditions, external factors such as temperature, switching of a plurality of systems, and the like.
If a reference value that includes these changes is set, the accuracy of the abnormality detection at the symptom stage is deteriorated, and the detection is delayed.

In order to prevent this, it is necessary to set a reference value in consideration of the change history of the process data to be monitored and its past performance. However, the current monitoring device is configured to monitor only instantaneous value data of process data, and is not conscious of accumulating past process data for a long time. Setting reference values in consideration of past results requires a huge amount of work, such as reading a recorder, and it is virtually impossible to determine the necessary reference values on a human-by-human basis. Was.

The present invention has been made in consideration of such a conventional problem, and has a configuration in which a monitoring reference value based on a long-term characteristic of a process amount of a plant can be set relatively easily without trouble. The purpose of the present invention is to accurately detect an abnormality of a plant at its symptom stage.

[0008]

To achieve the above object, a process data monitoring apparatus according to the first aspect of the present invention converts data relating to a process amount (for example, temperature, pressure, flow rate, etc.) of a plant into an electric signal. A sensor for converting and detecting, an instantaneous value capturing means for digitizing an electric signal detected by the sensor and capturing the same as the instantaneous value data of the process amount, and a statistical value for the instantaneous value data captured by the instantaneous value capturing means. Statistical value creating means for creating statistical value data by processing; data accumulating means for accumulating the statistical value data created by the statistical value creating means and the instantaneous value data acquired by the instantaneous value acquiring means; A reference value setting means for setting reference value data for monitoring the amount;
Reference value storage means for storing reference value data set by the reference value setting means; monitoring target amount data including at least one of instantaneous value data and statistical value data accumulated by the data accumulation means; Monitoring means for detecting the change state of the process amount by comparing the reference value data stored in the value storage means with each other; and monitoring information (for example, process information) including the change state of the process amount detected by the monitoring means. Display means for displaying data, reference values, monitoring results, etc.) on a screen.

Preferably, the instantaneous value acquisition means is an A / D conversion means for converting an output from the sensor into a digital signal.

The data storage means preferably has means for storing the A / D-converted instantaneous value data for several days, while storing the statistical value data calculated from the instantaneous value data by the statistical processing means for a long period of time. I do.

The monitoring means is preferably a means for monitoring the instantaneous value data or the statistic data by comparison with a reference value at a preset time interval. Particularly preferably, the reference value data is always kept constant. A trend monitoring means comprising a level monitoring means for monitoring the data value itself and a gradient monitoring means for monitoring a change gradient of a monitoring target by a reference value which slowly follows a change in process data; And a correlation monitoring means for monitoring the correlation.

The monitoring means can be provided with a monitoring bypass function which detects switching of devices and artificial disturbance and excludes a change caused by those factors from data fluctuation of a monitoring target.

Preferably, the reference value setting means extracts instantaneous value data or statistical data of the past monitoring target from the data storage means, excludes artificial disturbance data from the extracted data, and removes the disturbance data from the extracted data. It is assumed that the monitoring means and the setting method of the reference value are selected on the basis of the distribution shape, and the reference value is automatically set.

According to a second aspect of the present invention, in the first aspect of the invention, the data storage means stores the instantaneous value data so as to be deleted after a predetermined data storage period has elapsed, A statistic storage unit for storing the statistic data so as to be stored for a period longer than the data accumulation period.

According to a third aspect of the present invention, in the first aspect of the present invention, the statistical value creating means extracts the instantaneous value data stored in the data storing means at regular time intervals. And statistical value calculating means for calculating the statistical value data using the instantaneous value data extracted by the instantaneous value extracting means.

According to a fourth aspect of the present invention, in the first aspect of the invention, the monitoring means compares the instantaneous value data with the reference value data at every time interval defined by the sampling. Instantaneous value monitoring means for detecting a change state of the amount; and statistical value monitoring means for detecting the change state of the process amount by comparing the statistical value data with the reference value data at predetermined time intervals. It is characterized by the following.

According to a fifth aspect of the present invention, in the first aspect of the present invention, the monitoring means includes a level monitoring means for detecting a change in the magnitude of the process quantity and a gradient monitoring means for detecting a change gradient in the process quantity. A trend monitoring means provided with at least one of the means, and a correlation monitoring means for detecting a correlation between two different process quantities among the process quantities, and at least one of the correlation monitoring means and the trend monitoring means. It is a means for detecting a change state of the process amount using one of them.

According to a sixth aspect of the present invention, in the first aspect of the present invention, data relating to a relationship between a change in a contact signal of equipment in the plant and an artificial disturbance such as test operation or switching of the equipment, and An artificial disturbance propagation database that holds data relating to the relationship between the disturbance and the process amount that it propagates in advance, a contact signal capturing unit that captures a contact signal of the device, and a contact signal that is captured by the contact signal capturing unit. Judgment of the presence or absence of an artificial disturbance, and, by referring to the data in the artificial disturbance spread database, determine the artificial disturbance spread signal caused by the spread of the artificial disturbance, and monitoring the artificial disturbance spread signal is not performed. Disturbance judging means for giving an instruction to the monitoring means so as to execute the switching of the reference value data Further comprising the.

According to a seventh aspect of the present invention, in the first aspect of the present invention, there is provided an artificial disturbance propagation database which holds, in a referable manner, data relating to a relationship between an artificial disturbance of equipment in the plant and a process amount to which the disturbance is transmitted. An operation schedule database that holds data relating to the operation schedule of the device in a referable manner, and determines whether or not the artificial disturbance exists by referring to data in the operation schedule database when monitoring the statistical value data, and By referring to the data in the artificial disturbance propagation database to determine the artificial disturbance propagation signal caused by the propagation of the artificial disturbance, to bypass the monitoring of the artificial disturbance propagation signal or to switch the reference value data. And a disturbance determining means for giving an instruction to the monitoring means.

According to an eighth aspect of the present invention, in the first aspect, the reference value setting means includes a trend monitoring reference value setting means for setting reference value data for trend monitoring, and a reference value for correlation monitoring. A correlation monitoring reference value setting means for setting data.

According to a ninth aspect of the present invention, in the invention of the eighth aspect, the trend monitoring reference value setting means extracts the past monitoring target amount data for setting the reference value from the data storage means. A disturbance data excluding unit for excluding disturbance data from the monitoring target amount data extracted by the data extracting unit; and a distribution shape for determining a distribution shape of the monitoring target amount data from which the disturbance data is excluded by the disturbance data excluding unit. Determining means; monitoring method selecting means for selecting a monitoring method of the monitored amount data and a setting method of the reference value data based on the distribution shape of the monitored amount data determined by the distribution shape determining means; Reference value calculation for calculating reference value data from monitoring target amount data excluding the disturbance data based on the selection result by the method selection means Characterized by comprising a stage.

According to a tenth aspect of the present invention, in the ninth aspect of the present invention, the disturbance data excluding means includes a moving average value calculating means for calculating a moving average value of the monitoring target amount data for setting the reference value. A residual calculating means for calculating a residual between the moving average value calculated by the moving average value calculating means and the monitoring target amount data; and disturbance data from the distribution of the residual data calculated by the residual calculating means. Disturbance data determination means for determining, the disturbance data determination means,
When a value obtained by dividing a difference between two adjacent residual data among the residual data by a maximum residual exceeds a preset value, the residual value of the two residual data is calculated. It is a means for excluding large residual data and residual data equal to or larger than the residual value from the monitoring target amount data for setting the reference value as the disturbance data.

According to an eleventh aspect of the present invention, in the invention of the ninth aspect, the distribution shape determining means includes an arithmetic average value calculating means for calculating an arithmetic average value of the monitoring target amount data, and the monitoring target amount data. Moving average value calculating means for calculating a moving average value of the moving average value, and a residual calculating means for calculating a residual from each of the arithmetic average value and the moving average value of the monitoring target amount data calculated by the two calculating means, The distribution of the residual is calculated by calculating the sharpness and distortion of the residual calculated by the residual calculation means, and determining whether the sharpness and distortion are within a preset range. And a residual distribution determining means for determining a shape.

According to a twelfth aspect of the present invention, in accordance with the eleventh aspect of the present invention, the monitoring method selecting means determines a residual value from the arithmetic mean based on a distribution shape result determined by the distribution shape determining means. When the distribution is a normal distribution, a method using level monitoring is selected as a method for monitoring the monitoring target amount data, and a constant of a standard deviation of a residual is added to the arithmetic average value as a method for calculating the reference value data. A method of calculating the reference value data for level monitoring by adding and subtracting a double is selected, and the distribution of the residual from the arithmetic average becomes a non-normal distribution and the distribution of the residual from the moving average value Is a normal distribution, a method that uses both level monitoring and gradient monitoring is selected as a method for monitoring the monitoring target amount data, and the transfer method is used as a method for calculating the reference value data. The reference width of the reference value data for gradient monitoring is calculated as a constant multiple of the standard deviation of the residual from the average value, and the arithmetic average is a constant multiple of the maximum residual from the arithmetic average. A method of calculating the reference value data for level monitoring by addition and subtraction is selected, and the distribution of the residual from the arithmetic average value and the distribution of the residual from the moving average value are non-normal distributions, respectively. In this case, the level monitor and the gradient monitor are selected to be used together, and the maximum residual from the moving average is set to a constant multiple as the method of calculating the reference value data, so that the reference value data for the gradient monitor is calculated. Means for calculating the reference value for level monitoring by adding and subtracting a constant multiple of the maximum residual from the arithmetic mean to the arithmetic mean. It is characterized by .

According to a thirteenth aspect of the present invention, in the eighth aspect of the invention, the correlation monitoring reference value setting means extracts the past monitoring target amount data for setting the reference value from the data storage means. A disturbance data excluding unit for excluding the disturbance data from the monitoring target amount data extracted by the data extracting unit; and determining a distribution shape of the extracted monitoring target amount from which the disturbance data is excluded by the disturbance data excluding unit. Distribution type determining means, a selecting means for selecting a setting method of reference value data based on a distribution shape of the monitoring target amount data determined by the distribution shape determining means, and a setting method selected by the selecting means. And a reference value calculating means for calculating the reference value data.

According to a fourteenth aspect of the present invention, in the thirteenth aspect, the disturbance data elimination means includes a minimum of two.
Fitting formula calculating means for creating fitting formulas of two different process amounts using a multiplication method, and residual calculating means for calculating residual data from the fitting formula created by the fitting formula calculating means And disturbance data determination means for determining disturbance data from the distribution of the residual data calculated by the residual calculation means, wherein the disturbance data determination means comprises two adjacent ones of the residual data. When the value obtained by dividing the difference between the residual data by the maximum residual exceeds a preset value, the residual data having a large residual value and the residual having a residual value greater than the residual value of the two residual data. It is a means for excluding difference data from the monitoring target amount data for setting the reference value as the disturbance data.

According to a fifteenth aspect of the present invention, in accordance with the thirteenth aspect, the distribution shape determining means calculates a fitting equation of two different process quantities using a least squares method. Means, residual calculating means for calculating residual data from the fitting equation calculated by the calculating means, and calculating the degree of sharpness and distortion of the residual data calculated by the residual calculating means. And a residual distribution determining means for determining whether or not the sharpness and the distortion are within predetermined ranges to determine the distribution shape of the residual data.

According to a sixteenth aspect of the present invention, in the thirteenth aspect of the present invention, when the distribution shape of the residual data determined by the distribution shape determining unit is a normal distribution, the selection unit determines the size of the residual data. While selecting a setting method in which the value obtained by adding and subtracting a constant multiple of the standard deviation of the residual data to the fitting equation is used as the reference value data, when the distribution of the residual data from the fitting equation becomes a non-normal distribution, Is a means for setting a method of setting a value obtained by adding and subtracting a constant multiple of a maximum residual among the residuals of the fitting equation and each data to the fitting equation as reference value data. .

According to a seventeenth aspect of the present invention, in the first aspect of the invention, the reference value setting means changes a process amount which changes due to switching of equipment in the plant among the plant amounts, etc. In such a case, the monitoring target amount data before that time is replaced with the latest monitoring target amount data to obtain the moving average value, and the monitoring reference value data is calculated based on the moving average value. Features.

In the process data monitoring method according to the eighteenth aspect, an electric signal detected by a sensor which converts data concerning a process amount of a plant into an electric signal and detects the electric signal is digitized and taken in as instantaneous value data of the process amount. An instantaneous value capturing step, a statistical value creating step of creating statistical value data by statistically processing the instantaneous value data captured by the instantaneous value capturing means, and data for storing the statistical value data and the instantaneous value data An accumulating step, a reference value setting step of setting reference value data relating to the monitoring of the process amount, a reference value storing step of storing the set reference value data, and a step of storing the accumulated instantaneous value data and statistical value data. By comparing the monitored amount data including at least one of the above and the stored reference value data with each other, A monitoring step of detecting the amount of change state, characterized in that the monitoring information including the changing state of the process variable sensed in the monitoring step and a display step of displaying on the screen.

[0031]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS An embodiment of a process data monitoring device according to the present invention will be described below with reference to the drawings.

The process data monitoring apparatus shown in FIG. 1 is applied to, for example, a nuclear power plant. As shown in FIG. 1, various types of process data such as temperature, pressure, and flow rate of a plant are converted into electric signals and detected. 1, an A / D conversion unit (instantaneous value acquisition means) 2 for digitizing an electric signal from each of the sensors 1 and 1 and acquiring it as instantaneous value data of a process amount, and an A / D conversion unit 2. A data storage unit (data storage means) 3 for storing the monitored amount data including the captured instantaneous value data, statistical processing of the instantaneous value data of the data storage unit 3, and monitoring the statistic (statistical) data Statistical processing unit (statistical value creation means) 4 to be stored in the data storage unit 3 as one of the amount data
And a monitoring unit (monitoring means) 5 that compares the monitoring target amount data with predetermined reference value data to detect a change state thereof, and stores reference value data for comparison processing in the monitoring unit 5. A reference value storage section (reference value storage means) 6 and a reference value setting section (reference value setting means) 8 for calculating and setting reference value data stored in the reference value storage section 6 are provided.

Each of these units is functional only, and if the function can be realized, for example, a CPU
In this case, an existing computer hardware configuration, digital circuit, or the like can be applied.

The process data monitoring apparatus includes, as other components, a display unit (display means) 7 such as a CRT or a liquid crystal for displaying monitoring information such as process data, reference value data, and monitoring result data on a screen, and a user's. An input unit 9 such as a keyboard and a pointing device for receiving a request is provided.

In the above configuration, the electric signals reflecting the respective process amounts from the respective sensors 1...
, And digitized, and the digitized electric signal is stored in the data storage unit 3 as instantaneous value data. The instantaneous value data accumulated in the data accumulating unit 3 is statistically processed at predetermined time intervals by the statistical processing unit 4, and the statistical value data is combined with the instantaneous value data as monitoring target amount data. Is accumulated in

The past data stored in the data storage unit 3 is extracted by the reference value setting unit 8 at the time of a user request from the input unit 9, and the reference value data is calculated from the extracted data by a predetermined calculation method. The reference value data is stored in the reference value storage unit 6.

As a result, the monitoring unit 5 compares the monitoring target amount data stored in the data storage unit 3 with the reference value data stored in the reference value storage unit 6, whereby the change state of the process data is changed. Is detected. Such a detection result is displayed on the screen of the display unit 7 as necessary monitoring information such as a graph of the monitoring target amount, reference value data, and a monitoring result.

Next, an example of the configuration of each section of the above process data monitoring apparatus will be described in detail.

First, a configuration example of the "data storage section 3" will be described with reference to FIG.

The "data storage unit 3" is a storage medium such as a magnetic disk, a magneto-optical disk, and an optical disk.
Functionally, as shown in FIG. 2, the “instantaneous value storage unit 3a”, the “statistical value storage unit 3b”, and the “start / stop data storage unit 3”
c ". Hereinafter, the reason for setting each part in this function will be described.

First, consider the case where a long-term change is detected using this process data monitoring device.
It is necessary to perform monitoring based on the change history of each process amount detected by. However, if all the sampled process amounts are accumulated for a long period of time, the data storage unit 3 requires a huge storage capacity, which is not practical. Therefore, in order to reduce the storage capacity of the data storage unit 3, according to the present invention, statistical processing is performed on sampling data at predetermined time intervals, and all data at the time intervals are aggregated into one point, and the data is collected for several years. A "statistical value storage unit 3b" for storing is adopted.

Considering the case of detecting a sign of an abnormality, all the sampled data may be required to investigate the cause. Therefore, the present invention employs an "instantaneous value storage unit 3a" for accumulating all instantaneous value data before statistical calculation only for a short period of about one week, and automatically deleting the data after a certain period of time.

Further, at the time of starting or stopping the plant, the state of the plant changes every moment. Therefore, according to the present invention, the instantaneous value data for several days is stored as the starting and stopping data for a long period of time. The "start and stop data storage unit 3c" is employed.

Next, an example of the configuration of the statistical processing section 4 will be described with reference to FIGS. The statistical processing unit 4 is, for example, a computer that executes a predetermined algorithm for statistical processing, and functionally includes an “instantaneous value extracting unit 41” and a “statistical value calculating unit 42” as shown in FIG. ing.

The "instantaneous value extraction unit 41" receives the instantaneous value data of the process data from the instantaneous value storage unit 3a of the data storage unit 3 as shown in FIG. The stored instantaneous value data up to the latest value is extracted from the instantaneous value storage unit 41, and the extracted data is supplied to the statistic calculation unit 42.

As shown in FIG. 4, the "statistical value calculation unit 42" converts the instantaneous value data from the instantaneous value extraction unit 41 into statistical value data including an arithmetic mean value, a standard deviation, a maximum value, a minimum value, and the like. The calculated statistical value data is supplied to the statistical value storage unit 3b of the data storage unit 3.

Next, an example of the configuration of the monitoring unit 5 will be described with reference to FIG.

The monitoring unit 5 is composed of, for example, a computer that executes a preset monitoring algorithm. Functionally, the monitoring unit 5 monitors the instantaneous value data as shown in FIG. A "statistical value monitoring unit 5b" for monitoring the value data;
Each of the units for executing the three monitoring methods commonly used in 5b includes a “level monitoring unit 51”, a “gradient monitoring unit 52”, and a “correlation monitoring unit 53”.

As shown in FIG. 6, the "instantaneous value monitoring unit 5a" monitors the instantaneous value data by comparing the instantaneous value data with predetermined reference value data every time data is sampled, as shown in FIG. And continue this until a command to end monitoring is input. If the monitoring by the instantaneous value monitoring unit 5a is continued, the load on the monitoring device becomes very high. For example, a signal in which an abnormal sign is detected by the monitoring by the statistical value monitoring unit 5b is used to enhance the subsequent monitoring. Or
It is effective to use this when the process value is changing every moment, such as when the plant is started.

As shown in FIG. 6, when a monitoring start command is input, the "statistical value monitoring unit 5b" sets a reference value preset for the statistic data every time the statistic data is calculated. And then continue monitoring at the above time intervals until a monitoring stop command is input. Even if the monitoring by the statistical value monitoring unit 5b is continued, the load on the monitoring device is not so high as compared with the case of the instantaneous value monitoring unit 5a.

The "level monitoring unit 51" monitors the change of the monitoring target amount data based on certain reference values (upper limit value and lower limit value) as shown in FIG. 7, and is functionally shown in FIG. "Data extraction unit (data input unit) 511",
It comprises a “reference value extraction unit (reference value input) 512” and a “data comparison unit 513”. In the level monitoring section 51, the reference value data in the reference value storage section 6 is extracted by the reference value extraction section 512 at the time of the monitoring, and the data input section 5 is used.
At 11, monitoring target amount data in the data storage unit 3 is input, and monitoring is performed by comparing the magnitude relationship between the monitoring target amount data and the reference value data with each other by the data comparing unit 513.

As shown in FIG. 9, the "gradient monitoring unit 52" monitors a moving average value that slowly follows the temporal change of the monitoring target amount data by adding and subtracting a predetermined margin to and from a reference value obtained by subtraction. In terms of function, as shown in FIG. 10, “data extracting unit 521”, “moving average calculating unit 522”,
“Moving average reference value calculation unit 523”, “data comparison unit 52”
4 ".

The gradient monitor 52 includes a data extractor 52
In step 1, the previous moving average value and the latest data are extracted from the data storage unit 3, and the moving average calculation unit 522 calculates a moving average value up to the latest data using the extracted data.
The moving average value is passed to a moving average reference value calculation unit 522 and a comparison unit 523, and the moving average reference value calculation unit 522 adds and subtracts predetermined margin data extracted from the reference value storage unit 6 to obtain a reference value. Data is calculated, and monitoring is performed by comparing the reference value data and the monitoring target amount data with each other in the data comparing unit 523.

The "correlation monitoring unit 53" includes two preset monitoring target amount data (process X,
The reference value data is obtained by adding and subtracting predetermined margin data to and from the correlation equation determined in the process Y), and monitoring is performed using the reference value data. Functionally, as shown in FIG. It comprises a “correlation value calculation unit 532”, a “correlation reference value calculation unit 533”, and a “data comparison unit 534”.

The correlation monitoring unit 53 includes a data extraction unit 53
In step 1, the latest data of the two monitoring target quantities X and Y for which correlation monitoring is to be performed is extracted from the data storage unit 3, and the extracted data is passed to the correlation reference value calculation unit 532 and the comparison unit 534, and the correlation value calculation is performed. The unit 532 calls the correlation equation of the two monitoring target amount data from the reference value storage unit 6 and sets the value of the correlation value Ys corresponding to the monitoring target amount data X to “Ys = f
(X) ", and using the calculated value, the correlation reference value calculation unit 533 adds and subtracts the predetermined margin data extracted from the reference value storage unit 6 to the correlation value Ys to obtain the reference value. The data is calculated, and the reference value data and the value of the monitoring target amount Y are compared by the data comparison unit 534 to perform monitoring.

In the above process data monitoring apparatus, FIG.
3 or "monitoring bypass units (monitoring bypass means) 50a, 50b" shown in FIG. 14 can be added.

The monitoring bypass unit 50a shown in FIG. 13 includes a "contact signal capturing unit 10" for capturing a contact signal indicating operation or stop of the device, and a relationship between a change in the contact signal and an artificial disturbance such as surveillance of the device. And an "artificial disturbance propagation database 11" which holds data on the relationship between the artificial disturbance and the amount of the process that it propagates, and an artificial disturbance propagation database 11 when there is a change in the contact signal.
, A "man-made disturbance determination unit 12" which extracts a signal indicating the effect of the performed surveillance and its influence and instructs the monitoring unit 5 not to monitor the signal corresponding to the detected surveillance. .

The monitoring bypass unit 50b shown in FIG. 14 includes a “man-made disturbance propagation database 13” that holds data on the relationship between man-made disturbances such as the surveillance of the machine and the amount of the process that the man-made disturbance affects, and the operation of the machine. The "operation schedule database 14" which holds the data relating to the schedule in a referable manner, and a signal which has an influence by referring to the artificial disturbance propagation database 13 when it is determined that the operation is performed from the operation schedule database 14 is monitored. And an “artificial disturbance determination unit 15” that gives an instruction to the monitoring unit 5 so as not to perform the operation. In this case, it is also possible to respond by switching the reference value data only when an artificial disturbance occurs.

Next, an example of the configuration of the "reference value setting unit 8" is shown in FIG.
This will be described with reference to FIGS.

The function of the reference value setting unit 8 is a monitoring method for trend monitoring and correlation monitoring of current data based on the characteristics of the past monitoring target amount data stored in the data storage unit 3 and its reference value. A “trend monitoring reference value setting unit 8a” and a “correlation monitoring reference value setting unit 8b” for automatically setting data are provided.

As shown in FIG. 15, the "trend monitoring reference value setting unit 8a" has a "data extracting unit 81", a "disturbance data excluding unit 82", a "distribution shape determining unit 83", and a "monitoring method selection". And a "reference value calculation unit 85".

The “data extraction unit 81” is used when the user inputs and sets the process value for setting the reference value and the monitoring target amount data, the range of the past data, the data condition, and the like by the input unit 9. , And retrieves the past data from the data storage unit 3 and supplies this data to the disturbance data elimination unit 82 as reference value setting data.

The "disturbance data elimination unit 82" sets unrealistic reference value data due to the influence of the disturbance data when there is disturbance data such as an artificial disturbance in the extracted data from the data extraction unit 81. Therefore, in order to avoid such a noise factor, before calculating the reference, the disturbance data is excluded from the monitoring target amount data and supplied to the distribution shape determination unit 83 and the reference value calculation unit 85 in advance.

The “distribution shape determination unit 83” determines the distribution shape of the monitoring target data excluding the disturbance data, as described later, and sends the determination result to the monitoring method selection unit 84.

The "monitoring method selecting unit 84" selects a monitoring method and a reference value calculating method of the monitoring target amount data based on the judgment result from the distribution shape judging unit 83, and instructs the selection result to the reference value calculating unit. Give to 85.

The “reference value calculation unit 85” calculates reference value data from the monitoring target amount data from the disturbance data exclusion unit 82 by using the monitoring method and the reference value calculation method selected by the monitoring method selection unit 84. This is supplied to the reference value storage 6.

The "disturbance data elimination unit 82" in the trend monitoring described above has a "moving average value calculation unit 821", "residual difference calculation unit 822", and "disturbance determination unit 823" functionally, as shown in FIG. Consists of

The “moving average value calculating section 821” calculates the moving average value of the data from the data extracting section 81 as shown in FIG. 17A and sends it to the residual calculating section 822.

As shown in FIG. 17A, the “residual calculation unit 822” calculates the moving average and the residual εi of the actual data.
Is calculated, and the calculated residual εi is arranged on the “+ side” and the “one side” in ascending order of absolute value as shown in FIG. 17B, and the result is instructed to the disturbance determination unit 823.

The “disturbance determination unit 823” calculates a difference δi between adjacent residual data as shown in FIG.
When the value obtained by dividing the difference δi by the maximum residual εmax exceeds a preset value, the larger of the two adjacent residual data when calculating the difference between the residual data exceeding the preset value. 17C and the residual data equal to or larger than the residual data are excluded from the moving average value data (process values) as excluded data, as shown in FIG. .

As shown in FIG. 18, the “distribution shape judging unit 83” in the above trend monitoring has “arithmetic average calculating unit 831”, “moving average calculating unit 832”, “residual calculating unit 833”, It comprises a “residual distribution determination unit 834”.

The “arithmetic average calculating section 831” is one of FIG.
As shown in FIG. 9, the arithmetic average value is calculated from the monitoring target amount data for setting the reference value excluding the disturbance data, and the result is sent to the residual calculating unit 833.

The “residual calculation unit 833” calculates the residual between the arithmetic mean value and each data, and sends the calculation result of each residual to the residual distribution determination unit 834.

The "residual distribution determination unit 834" determines whether the residual distribution from the arithmetic mean value is a normal distribution, and if it is determined to be a normal distribution, determines "case 1". Then, the determination result is passed to the monitoring method selection unit 833.

Here, when determining whether or not the distribution is a normal distribution, when the skewness of the residual distribution from the arithmetic mean is S and the sharpness is K, the absolute values of the skewness S and the sharpness K are A method is used in which the residual distribution is determined to be a normal distribution when the conditional expression is equal to or smaller than the predetermined values a and b, ie, “lSl <a, 1Kl <b”. The sharpness S and the sharpness K used in this determination method are represented by σ, the standard deviation of the residual around the arithmetic mean,
The residual from the arithmetic average value is (Xi-m), the number of data is N, and the number of data whose residual is (xim) is f
i and μ _k = {(Xi−m) ^k · fi} / N, “S = μ ₃ / σ ³ ” and “K = μ ₄ /
σ ⁴ -3 ”.

When the distribution of the residual from the arithmetic average is not regarded as a normal distribution, the moving average calculating section 832 obtains a moving average of the monitored object amount when the residual distribution is not regarded as a normal distribution. The residual from the average value is calculated, and the result is instructed to the residual distribution determination unit 834. Here, the moving average calculating section 832 and the residual calculating section 833 are provided with the moving average calculating section 821 and the residual calculating section 82 in the disturbance data excluding section 82.
2 may be used.

The “residual distribution determining section 834” determines the residual distribution of the arithmetic average value, and calculates the moving average calculating section 83.
It is determined whether or not the residual around the moving average from No. 2 has a normal distribution. If it is determined that the residual is a normal distribution, “Case 2”; if it is determined that the residual is not a normal distribution, “Case 3” The determination result is instructed to the monitoring method selection unit 833. This determination method is the same as the above-described method of determining the distribution of the residual from the arithmetic average value.

The processing of each unit of the distribution shape determination unit 83 described above can be summarized by a series of processing steps S1 to S9 shown in FIG.

The distribution shape determination unit 83 determines that the distribution shape of the monitoring target amount data is “case 1: distribution of residual from arithmetic mean is normal distribution”, and “case 2: residual from moving average” Difference distribution is normal distribution "and" Case 3:
It determines whether the distribution of the residual from the arithmetic average value and the moving average value corresponds to any of the non-normal distributions, and instructs the monitoring method selection unit 84 of the determination result.

The “monitoring method selecting unit 84” monitors the trend monitoring or correlation monitoring and calculates the monitoring reference value based on the determination result from the distribution shape determining unit 83, that is, “case 1” to “case 3”. Is selected, and the result of the selection is sent to the reference value calculation unit 85. Hereinafter, this monitoring method selection unit 8
The specific processing contents of No. 4 will be described separately for the cases of “Case 1” to “Case 3”.

First, in the case of “case 1”, a method using level monitoring is selected as a monitoring method. As a method of calculating the reference value in this level monitoring, as shown in FIGS. 20A and 20B, the arithmetic average value m1 of the monitoring target amount data is used.
Is multiplied by a specific constant N from the standard deviation σ1 of the reference value, and an arithmetic mean value m1 is added and subtracted to the multiplied value Nσ1 to obtain an upper reference value and a lower reference value as reference value data (reference value = ml ± Nσ1). N in this method
Can be appropriately set by the user. For example, when performing normal monitoring, the value is set in a range of 3 to 5, and when performing stricter monitoring, a value smaller than the range is set. Is used.

In the case of “Case 2”, a method using both the level monitoring and the gradient monitoring is selected as the monitoring method. Among them, the reference value calculation method for gradient monitoring is shown in FIG.
As shown in (a) and (b), the standard deviation σ2 of the residual from the moving average value m2 of the monitoring target amount data is multiplied by a specific constant, and the multiplied value Nσ2 is an upper reference value from the moving average value and A method of obtaining reference value data as the allowable range of the lower limit reference value (= Nσ2) is adopted. In addition, as a reference value calculation method in level monitoring, as shown in FIGS. 22A and 22B, since the residual distribution from the arithmetic mean value m1 is a non-normal distribution, the reference value is calculated on the side larger than the arithmetic mean value m1. The maximum residual ε of
+ Is multiplied by a specific value A, and a value obtained by adding the multiplied value Aε + to the arithmetic average value m1 is defined as an upper limit value and a lower limit value.
Multiply the maximum residual ε− on the smaller side by a specific value A,
A method of obtaining reference value data using a value obtained by subtracting the arithmetic mean value m1 from the multiplication value Aε− as a lower limit reference value is employed.
The value of A in this method can be appropriately set by the user. For example, when performing normal monitoring, 1.
If the monitoring is to be carried out more strictly, a value smaller than the range is used.

In case 3 as well, as the monitoring method, a method using both gradient monitoring and level monitoring is selected. Among them, as a reference value calculation method for gradient monitoring, the maximum residual ε + on the side larger than the moving average value m2 of the monitoring target amount data is multiplied by a specific value A, and the multiplied value Aε + is calculated from the moving average value m2. While the maximum residual ε− on the side smaller than the moving average value m2 is a specific value A.
And a method of obtaining reference value data using the multiplied value Aε− as an allowable width of the lower limit reference value from the moving average value m2 is adopted. The value of A in this method can be appropriately set by the user, for example, in the range of 1.2 to 2,
For those that are strictly monitored, use smaller values than within the range. As a reference value calculation method in the level monitoring, the same method as in the case 2 described above is used.

As a method of calculating the reference value in the gradient monitoring, in order to avoid a situation in which the monitoring target amount data changes due to factors such as device switching and a calculation error increases, the data rapidly changes due to these factors. In such a case, it is possible to use a method of replacing the previous data with the latest data, calculating the moving average value, and calculating the monitoring reference value data based on the calculated value.

The monitoring method selection unit 84 selects a monitoring method and a reference value calculation method based on the above processing, and instructs the reference value calculation unit 85 on the reference value calculation method.

The reference value calculation unit 85 includes a monitoring method selection unit 84
The reference value data is calculated using the reference value calculation method selected by (1).

Next, an example of the configuration of the "correlation monitoring reference value setting section 8b" will be described with reference to FIGS.

The "correlation monitoring reference value setting section 8b" is functionally similar to the above-described trend monitoring reference value setting section 8a in that the "data extraction section 81", "disturbance data exclusion section 82",
"Distribution shape determination unit 83", "monitoring method selection unit 84",
It comprises a “reference value calculation unit 85” (see FIG. 15).

The “data extraction unit 81” extracts two types of monitoring target amount data for performing correlation monitoring, and supplies the extracted data to the disturbance data elimination unit 82.

As shown in FIG. 23, the "disturbance data elimination unit 82"
It comprises a “residual difference calculation unit 822” and a “disturbance determination unit 823”. As shown in FIG. 24, the disturbance data elimination unit 82 uses the least-squares method to calculate the fitting equation of the two types of monitoring target amount data (process X and process Y in the figure) in the fitting equation calculation unit 824 as shown in FIG. Then, the residual between the calculated fitting equation and the data is calculated by the residual calculating section 822, and the residual is used by the disturbance determining section 823 as in the case of the trend monitoring reference value setting section 8a. The processing for excluding the disturbance data is performed according to the procedure described above, and the processed data is supplied to the distribution shape determination unit 83.

The “distribution shape determination unit 83” is functionally
As shown in FIG. 5, the data including the “fitting equation calculation unit 835”, the “residual difference calculation unit 833”, and the “residual distribution determination unit 834”, and the disturbance data exclusion unit 82 excludes the disturbance data is used to perform the filtering again. A residual formula is created, a residual between the fitting formula and the data is calculated by a residual calculation unit 833, and whether or not the distribution of the residual is a normal distribution is determined by a residual distribution determination unit 834. I do. Here, the fitting equation calculation section 835 and the residual calculation section 833 of the distribution shape determination section 83 can use the fitting equation calculation section 824 and the residual calculation section 822 of the disturbance data exclusion section 82 described above.

When the variation in the residual distribution shape is a normal distribution as determined by the distribution shape determining unit 83, the “monitoring method selecting unit 84” multiplies the standard deviation from the fitting equation by a specific value, and The value obtained by adding and subtracting the value to the fitting formula is used as the reference value data for the upper reference value and the lower reference value. On the other hand, if the residual distribution shape is non-normal distribution, the maximum residual from the fitting formula is obtained. Is multiplied by a specific value, and a method of setting a value obtained by adding and subtracting the multiplied value to the fitting equation as reference value data of an upper reference value and a lower reference value is selected. Send to

The “reference value calculation unit 85” calculates the reference value data from the monitoring target amount data excluding the disturbance data based on the selection result of the monitoring method selection unit 85.

Therefore, according to this embodiment, the instantaneous value data of various process amounts are accumulated for a relatively short time, and the statistical processing is performed based on the instantaneous value data to accumulate as statistical value data. Therefore, monitoring can be performed in consideration of the long-term characteristics of the monitoring target. In addition, since the monitoring method and the reference value are automatically created in consideration of the characteristics of each monitoring target from the long-term trends, appropriate monitoring reference values can be set relatively without human intervention. Abnormalities can be accurately detected at the symptom stage.

The reference value setting section can be provided with reference value confirmation means for obtaining the confirmation of the user by outputting the set value and the data actually used for the setting to an output device. Further, a configuration is adopted in which the reference value displayed on the output device is changed by an input from the input device, and data for setting the reference value and exclusion data are stored for each monitoring target. It is also possible.

[0096]

As described above, according to the present invention, instantaneous value data of various process amounts are accumulated for a relatively short time, and statistical processing is performed on the basis of the instantaneous value data to obtain statistical value data. Since the storage is possible, monitoring can be performed in consideration of the long-term characteristics of the monitoring target. In addition, since a monitoring method and a reference value that take into account the characteristics of each monitoring target from the long-term tendency are automatically created, it is possible to set an appropriate monitoring reference value with relatively little human intervention. An abnormality in the plant state can be accurately detected at the symptom stage.

[Brief description of the drawings]

FIG. 1 is a schematic block diagram showing the overall configuration of a process data monitoring device according to a first embodiment of the present invention.

FIG. 2 is a schematic block diagram illustrating a configuration example of a data storage unit.

FIG. 3 is a schematic block diagram illustrating a configuration example of a statistical processing unit.

FIG. 4 is a graph illustrating operation timing of a statistical processing unit.

FIG. 5 is a schematic block diagram illustrating a configuration example of a monitoring unit.

FIG. 6 is a graph illustrating operation timing of a monitoring unit.

FIG. 7 is a schematic diagram illustrating a level monitoring method.

FIG. 8 is a schematic block diagram illustrating a configuration example of a level monitoring unit.

FIG. 9 is a schematic diagram illustrating a gradient monitoring method.

FIG. 10 is a schematic block diagram illustrating a configuration example of a gradient monitoring unit.

FIG. 11 is a schematic diagram illustrating a correlation monitoring method.

FIG. 12 is a schematic diagram illustrating a configuration example of a correlation monitoring unit.

FIG. 13 is a schematic diagram illustrating a configuration example of a monitoring bypass unit.

FIG. 14 is a schematic diagram illustrating another configuration example of the monitoring bypass unit.

FIG. 15 is a schematic block diagram illustrating a configuration example of a reference value setting unit.

FIG. 16 is a schematic block diagram illustrating a configuration example of a disturbance data elimination unit.

17 (a) to (c) are graphs for explaining a disturbance data exclusion method.

FIG. 18 is a schematic block diagram illustrating a configuration example of a distribution shape determination unit.

FIG. 19 is a schematic flowchart illustrating processing of a distribution shape determination unit.

FIGS. 20A and 20B are graphs illustrating a reference value setting method when the distribution of the residual from the arithmetic mean is a normal distribution (case 1).

FIGS. 21A and 21B are graphs illustrating a reference value setting method for gradient monitoring when the distribution of the residual from the moving average value is a normal distribution (Case 2).

22A and 22B are graphs for explaining a method of setting a reference value for level monitoring in case 2. FIG.

FIG. 23 is a schematic block diagram illustrating a configuration example of a disturbance data exclusion unit in the correlation monitoring reference value setting unit.

FIG. 24 is a graph illustrating a calculation method of a fitting equation calculation unit in a correlation monitoring reference value setting unit.

FIG. 25 is a schematic block diagram illustrating a configuration example of a distribution shape determination unit in a correlation monitoring reference value setting unit.

[Explanation of symbols]

 DESCRIPTION OF SYMBOLS 1 Sensor 2 A / D conversion part 3 Data accumulation part 3a Instantaneous value storage part 3b Statistical value storage part 3c Start / stop data storage part 4 Statistical processing part 41 Instantaneous value extraction part 42 Statistical amount calculation part 5 Monitoring part 5a Instantaneous value monitoring Unit 5b statistical value monitoring unit 50a, 50b monitoring bypass unit 51 level monitoring unit 511 data extraction unit 512 data comparison unit 513 reference value extraction unit 52 gradient monitoring unit 521 data extraction unit 522 moving average calculation unit 523 moving average reference value calculation unit 524 Data comparison unit 53 Correlation monitoring unit 531 Data extraction unit 532 Correlation average calculation unit 533 Moving average reference value calculation unit 534 Data comparison unit 6 Reference value storage unit 7 Display unit (output unit) 8 Reference value setting unit 81 Data extraction unit 82 Disturbance Data exclusion section 821 Moving average calculation section 822 Residual calculation section 823 Disturbance determination section 824 Fitin Expression calculation unit 83 Distribution shape determination unit 831 Arithmetic average calculation unit 832 Moving average calculation unit 833 Residual calculation unit 834 Distribution determination unit 835 Fitting expression calculation unit 84 Monitoring method selection unit 85 Reference value calculation unit 9 Input unit 10 Contact signal capture Unit 11, 13 Artificial disturbance database 12, 14 Artificial disturbance determination unit 15 Operation schedule database

 ────────────────────────────────────────────────── ─── Continuing on the front page (72) Koji Hima, Inventor F-term (reference) 5H223 AA03 BB02 CC08 DD03 DD09 EE06 FF03 at Toshiba Yokohama Office, 8 Shinsugita-cho, Isogo-ku, Yokohama-shi, Kanagawa

Claims (18)

[Claims] 1. A sensor for converting data relating to a process amount of a plant into an electric signal and detecting the electric signal, and an instantaneous value capturing means for digitizing the electric signal detected by the sensor and capturing the digitalized signal as instantaneous value data of the process amount. Statistical value creating means for creating statistical value data by statistically processing the instantaneous value data captured by the instantaneous value capturing means; and statistical value data and the instantaneous value capturing means created by the statistical value creating means. Data storage means for storing the captured instantaneous value data; reference value setting means for setting reference value data for monitoring the process amount; and reference value storage means for storing the reference value data set by the reference value setting means. Monitoring amount data including at least one of instantaneous value data and statistical value data accumulated by the data accumulating means; Monitoring means for detecting a change state of the process amount by comparing the reference value data stored in the value storage means with each other; and monitoring information including the change state of the process amount detected by the monitoring means on a screen. A process data monitoring device comprising: a display unit for displaying. 2. The invention according to claim 1, wherein said data storage means stores said instantaneous value data such that said instantaneous value data is deleted after a predetermined data storage period has elapsed, and said statistical value data is stored in said data storage means. A process data monitoring device comprising: a statistic storage unit that stores data so as to be stored for a period longer than a data accumulation period. 3. The instantaneous value extracting means according to claim 1, wherein said statistical value creating means extracts said instantaneous value data stored in said data storing means at regular time intervals, and said instantaneous value extracting means. A process data monitoring device comprising: a statistic value calculating unit that calculates the statistic value data using the instantaneous value data extracted by the extracting unit. 4. The invention according to claim 1, wherein said monitoring means detects a change state of said process amount by comparing said instantaneous value data with said reference value data at each time interval defined by sampling. An instantaneous value monitoring means for detecting a change in the process amount by comparing the statistical value data with the reference value data at predetermined time intervals. Data monitoring device. 5. The invention according to claim 1, wherein the monitoring means includes at least one of a level monitoring means for detecting a change in the magnitude of the process amount and a gradient monitoring means for detecting a change gradient of the process amount. Trend monitoring means, and correlation monitoring means for detecting a correlation between two different process quantities among the process quantities. The process is performed by using at least one of the correlation monitoring means and the trend monitoring means. A process data monitoring device, which is means for detecting a change state of a quantity. 6. The invention according to claim 1, wherein data relating to a relationship between a change in a contact signal of a device in the plant and an artificial disturbance such as a test operation or switching of the device, and the artificial disturbance and the ripple. An artificial disturbance propagation database that holds data relating to the relationship with the process amount in advance, contact signal capturing means for capturing the contact signal of the device, and the presence or absence of the artificial disturbance based on the contact signal captured by the contact signal capturing means And
By referring to the data in the artificial disturbance propagation database to determine the artificial disturbance propagation signal caused by the propagation of the artificial disturbance, the monitoring of the artificial disturbance propagation signal is not performed, or the reference value data is switched. And an artificial disturbance determination unit for giving an instruction to the monitoring unit to execute the process data. 7. The apparatus according to claim 1, wherein an artificial disturbance propagation database that holds data on a relationship between an artificial disturbance of the equipment in the plant and a process amount to which the equipment disperses, and an operation schedule of the equipment. An operation schedule database that holds data relating to the operation reference database, and when monitoring the statistic data, judges the presence or absence of the artificial disturbance by referring to data in the operation schedule database, and in the artificial disturbance propagation database. An instruction is given to the monitoring means so as to bypass the monitoring of the artificial disturbance propagation signal or to switch the reference value data by judging an artificial disturbance propagation signal caused by the propagation of the artificial disturbance with reference to the data of A process data monitoring device comprising: an artificial disturbance determination unit. 8. The invention according to claim 1, wherein the reference value setting means includes a trend monitoring reference value setting means for setting trend monitoring reference value data and a correlation monitoring setting for setting correlation monitoring reference value data. A process data monitoring device comprising: a reference value setting unit. 9. The data processing device according to claim 8, wherein the trend monitoring reference value setting means extracts past monitoring target amount data for setting a reference value from the data storage means, and the data extraction means. Disturbance data exclusion means for excluding disturbance data from the monitoring target amount data extracted by the method, distribution shape determination means for determining the distribution shape of the monitoring target amount data from which the disturbance data has been excluded by the disturbance data exclusion means, A monitoring method selecting means for selecting a monitoring method of the monitoring object amount data and a setting method of the reference value data based on a distribution shape of the monitoring object amount data determined by the shape determining means, and a selection result by the monitoring method selecting means Reference value calculating means for calculating reference value data from the monitoring target amount data excluding the disturbance data based on the reference value. Process data monitoring device. 10. The moving average value calculating means according to claim 9, wherein said disturbance data excluding means calculates a moving average value of said reference value setting monitoring target amount data, and said moving average value calculating means. Calculation means for calculating the residual between the moving average value calculated by the above and the monitored amount data, and disturbance data determination means for determining the disturbance data from the distribution of the residual data calculated by the residual calculation means The disturbance data determination means is provided when the value obtained by dividing the difference between two adjacent residual data among the residual data by the maximum residual exceeds a preset value,
Means for excluding residual data having a large residual value among the residual data and residual data having a residual value equal to or larger than the residual value from the monitoring target amount data for setting the reference value as the disturbance data. Process data monitoring device. 11. The invention according to claim 9, wherein the distribution shape determining means calculates an arithmetic average value of the monitoring target amount data, and calculates a moving average value of the monitoring target amount data. Moving average value calculating means, a residual calculating means for calculating a residual from each of an arithmetic average value and a moving average value of the monitoring target amount data calculated by the two calculating means, and the residual calculating means. Calculating the sharpness and distortion degree of the calculated residual, and determining the distribution shape of the residual by determining whether the sharpness and distortion degree are within a preset range. A process data monitoring device comprising: a distribution determining unit. 12. The monitoring method selecting means according to claim 11, wherein the monitoring method selecting means makes the residual distribution from the arithmetic mean a normal distribution based on the distribution shape result determined by the distribution shape determining means. In this case, a method using level monitoring is selected as a method of monitoring the monitoring target amount data, and a constant multiple of a standard deviation of a residual is added to and subtracted from the arithmetic average value as a method of calculating the reference value data. By selecting a method of calculating the reference value data for the level monitoring, the distribution of the residual from the arithmetic average becomes a non-normal distribution, and the distribution of the residual from the moving average becomes a normal distribution. In addition, a method of using both level monitoring and gradient monitoring as a method of monitoring the monitoring target amount data is selected, and a method of calculating the reference value data is a method of measuring a residual from the moving average value. The constant width of the deviation is used to calculate the reference width of the reference value data for gradient monitoring, and the arithmetic mean value is added to and subtracted from the arithmetic mean value by a constant multiple of the maximum residual from the arithmetic mean value, thereby obtaining the level. If the method of calculating the reference value data for monitoring is selected, the distribution of the residual from the arithmetic average and the distribution of the residual from the moving average are non-normal distributions, respectively.
By selecting to use both the level monitoring and the gradient monitoring, and by setting the maximum residual from the moving average to be a constant multiple as the method of calculating the reference value data, the reference width of the reference value data for gradient monitoring can be reduced. Means for calculating and calculating a method of calculating the reference value data for level monitoring by adding and subtracting a constant multiple of the maximum residual from the arithmetic average value to the arithmetic average value. Process data monitoring device. 13. The data processing apparatus according to claim 8, wherein the correlation monitoring reference value setting means extracts past monitoring target amount data for setting a reference value from the data storage means, and the data extraction means. Disturbance data exclusion means for eliminating disturbance data from the monitoring target amount data extracted by the distribution data determination means, and a distribution shape determination means for determining a distribution shape of the extracted monitoring target quantity from which the disturbance data has been excluded by the disturbance data exclusion means. Selecting means for selecting a setting method of the reference value data based on the distribution shape of the monitoring target amount data determined by the distribution shape determining means; and selecting the reference value data based on the setting method selected by the selecting means. A process data monitoring device, comprising: a reference value calculating means for calculating. 14. The fitting data calculating means according to claim 13, wherein said disturbance data rejecting means creates a fitting equation of two different process quantities using a least squares method, and said fitting equation calculating means. A residual calculating means for calculating residual data from the fitting equation created by the equation calculating means, and a disturbance data determining means for determining disturbance data from a distribution of the residual data calculated by the residual calculating means. The disturbance data determination means, when a value obtained by dividing a difference between two adjacent residual data among the residual data by a maximum residual exceeds a preset value, the two residual data are determined. A means for excluding, from the difference data, residual data having a large residual value and residual data not less than the residual value from the monitoring target amount data for setting the reference value as the disturbance data. A process data monitoring device, characterized in that: 15. The fitting shape calculating means according to claim 13, wherein said distribution shape judging means creates fitting formulas of two different process amounts using a least squares method, and said calculating means Calculating the residual data from the fitting equation calculated by the following equation; calculating the sharpness and distortion of the residual data calculated by the residual calculation means; A process data monitoring apparatus, comprising: a residual distribution determining unit that determines whether the degree is within a preset range to determine a distribution shape of the residual data. 16. The fitting method according to claim 13, wherein said selecting means determines the fit of the residual data when the distribution shape of said residual data determined by said distribution shape determining means is a normal distribution. While selecting a setting method in which the value obtained by adding and subtracting a constant multiple of the standard deviation of the standard deviation is used as the reference value data, if the distribution of the residual data from the fitting equation is a non-normal distribution, A process data monitoring device, characterized by means for setting a setting method in which a value obtained by adding and subtracting the fitting formula and a constant multiple of a maximum residual among residuals of each data is used as reference value data. 17. The invention as set forth in claim 1, wherein said reference value setting means, when a process amount which changes due to switching of equipment in said plant among said plant amounts suddenly changes, before the time point, Process data monitoring, comprising means for replacing the monitoring target amount data with the latest monitoring target amount data to obtain a moving average value thereof, and calculating the monitoring reference value data based on the moving average value. apparatus. 18. An instantaneous value capturing step of digitizing an electrical signal detected by a sensor for converting data relating to a process amount of a plant into an electrical signal and detecting the electrical signal, and capturing the digitized data as instantaneous value data of the process amount; A statistical value creating step of creating statistical value data by statistically processing the instantaneous value data captured by the capturing means; a data accumulating step of accumulating the statistical value data and the instantaneous value data;
A reference value setting step of setting reference value data relating to the monitoring of the process amount; a reference value storing step of storing the set reference value data; and at least one of the accumulated instantaneous value data and statistical value data A monitoring step of detecting a change state of the process amount by comparing the monitoring target amount data including the reference value data and the stored reference value data with each other; and monitoring including a change state of the process amount detected in the monitoring step. And a display step of displaying information on a screen.