JP2011060168A - Monitoring device - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To monitor a state of a process for use in operation of a plant. <P>SOLUTION: A monitoring device includes: a data acquisition means 101 for acquiring a process value measured as a value that represents the state of the process used to operate the plant, associating the process value with a time when the process value is measured and storing them as measurement data in a storage device; a waveform specifying means 102 for reading the measurement data from the storage device, and specifying a waveform of the process value; and a determining means 103 for determining whether a problem repeatedly occurs in the process from the waveform of the process value specified by the waveform specifying means by using sine theory or cosine theory. <P>COPYRIGHT: (C)2011,JPO&INPIT

Description

  The present invention relates to a monitoring device that monitors the state of each process in plant operation.

  Plants such as water and sewage plants are operated by a plurality of processes. Since problems that occur in any of the processes may affect the entire plant, it is important to monitor the occurrence of problems in each process in the plant. For this reason, various techniques have been developed for plant state diagnosis, process prediction, and the like (see, for example, Patent Documents 1 and 2).

  For example, an instantaneous value (current value) such as the amount of water at a plant location is measured as a process value, and the future behavior of the plant is estimated from a method of diagnosing the current behavior of the plant and the trend of the process value such as the amount of water. In addition, there is a method of diagnosing the current operation of the plant and trying to solve it as soon as possible if a problem occurs.

(Current behavior diagnosis)
For diagnosis of the current behavior in the plant, there is a method of using a process value such as a current water amount obtained from the process and a trend value of the process obtained from a mathematical model or operation data.

  For example, the simplest method is known as shown in FIG. 4, in which an upper threshold, a lower threshold and an allowable time for determining a problem are determined in advance, and the measured process value is the upper limit. This is a method for determining that a problem has occurred in the process when the threshold value or the lower threshold value is deviated from the allowable time or more.

  In addition, when ideal behavior can be simulated by a mathematical model, upper limit threshold and lower limit threshold models are determined in advance from simulation results, as shown in FIG. There is also a method of determining that a problem has occurred in the process when the threshold value is greater than or equal to the threshold value or less than the lower threshold value.

  As the mathematical model used in the simulation, a model that simulates the behavior of the process to be diagnosed or a model that is statistically processed based on the history of past process values is used. For example, the pressure loss is calculated using the head loss model (pipe model) due to the pipe resistance as shown in the formula (A), or the pond water level is calculated using the distribution reservoir model, and the calculated value is Comparison with the measured process value allows diagnosis and monitoring of the current process.

h1−h2 = γQ ^μ (A)
h1: Inlet side pressure h2: Outlet side pressure Q: Flow rate C: Flow velocity coefficient
D: Pipe diameter L: Pipe length λ = 10.666 x C ^-1.85 x D ^-4.87 x L
μ = 1.87
(Estimation and diagnosis of future behavior)
For future behavior and current operation diagnosis at the plant, process values such as the current water volume obtained from the process and trend values obtained based on mathematical models and operational data are used to estimate the future process behavior. In addition, there is a method of diagnosing the quality of the current plant operation using the future process behavior obtained by the estimated behavior.

  Specifically, the current process value is set as the initial value, and the future behavior is simulated using a mathematical model that predicts the behavior of the process, such as a head loss model and a reservoir model. Thereafter, the quality of the current operation method is diagnosed from the future process behavior estimated from the trend value of the process obtained by the simulation. At this time, by using the trend value measured in the past for the construction of the mathematical model, it is possible to estimate and diagnose the future behavior such as a statistical model using a neural network or time series data.

  If the history of past process values is sufficiently accumulated, a past process value similar to the newly obtained process value is searched from the history without creating a new mathematical model by simulation. Trends that include process values may be estimated as future behavior.

  Thus, when diagnosing current behavior and predicting and diagnosing future behavior, there is a feature that process values and trends vary greatly depending on conditions. FIG. 3 (a) shows an example of the trend of the sewerage plant with the amount of wastewater as a process value and the change in the amount of wastewater over time. Normally, as shown in FIG. The trend is fixed. Therefore, such a trend is determined in advance as a mathematical model for a normal day and used for process diagnosis.

  On the other hand, for example, when a typhoon occurs, the trend of the amount of sewage changes compared to a normal day. In this way, the day when the amount of drainage changes depending on the weather (rainfall, air volume, atmospheric pressure, etc.) and traffic problems is defined as an abnormal day with respect to the normal day, and as shown in FIG. It is necessary to convert and use the trend. In addition, for example, since the amount of sewage changes during periods when water is saved due to water shortages or when the pool is frequently used in the summer, as shown in FIG. It is necessary to convert and use the trend of sewage volume on a normal day.

  In other words, since the trend of the process value varies depending on the situation, it is necessary to obtain the trend of the process value according to the situation and make a diagnosis using the obtained trend.

JP 2008-59270 A JP 2004-62440 A

  As described above, in the conventional method, it is necessary to use the trend obtained according to the situation when diagnosing process values and the like. In this case, since there are many patterns of abnormal days and special days, it is necessary to prepare a lot of data in advance, and there is a problem that the process from acquisition of process values to diagnosis becomes complicated.

  An object of this invention is to provide the monitoring apparatus which monitors the process of a plant easily in view of the said subject.

  In order to solve the above-described problems, the present invention is a monitoring apparatus that is connected to a plant operated by a plurality of processes and monitors the process state of the plant, and is measured as a value representing the process state. Data acquisition means for acquiring the process value, associating the process value and the time at which the process value was measured, and storing it in the storage device as measurement data, reading the measurement data from the storage device, and obtaining the waveform of the process value Waveform specifying means for specifying, and determination means for determining whether or not a problem has repeatedly occurred in the process using the sine theory or cosine theory from the waveform of the process value specified by the waveform specifying means. .

  The present invention can easily monitor plant processes.

It is a block diagram explaining the structure of the monitoring apparatus which concerns on embodiment of this invention. It is a figure explaining the waveform specified with the monitoring apparatus of FIG. It is a figure explaining the data utilized by the diagnosis of a general plant. It is a figure explaining an example of the general diagnostic method of a plant. It is a figure explaining the other example of the general diagnostic method of a plant.

  Below, the monitoring apparatus which concerns on embodiment of this invention is demonstrated using drawing. The monitoring device according to the present invention is connected to a plant such as a water and sewage system, acquires various data (measurement values) measured in the plant, and uses the acquired data to solve problems such as a water hammer in the plant. It is a device that monitors the occurrence.

  As shown in FIG. 1, the monitoring apparatus 1 according to the present invention is an information processing apparatus including a central processing unit (CPU) 10, a storage device 11, an output device 12, an input device 13, and the like. In the monitoring device 1, by executing a monitoring program (not shown) stored in the storage device 11, a data acquisition unit 101, a waveform specifying unit 102, and a determination unit 103 are mounted on the CPU 10 as shown in FIG. Is done.

  The data acquisition unit 101 acquires a measurement value (process value) measured by a sensor of a plant to be monitored and stores it as measurement data D in the measurement value storage device 11. At this time, the data acquisition unit 101 samples the data to be acquired in time series, and stores the data up to the maximum sampling number K as measurement data D in the storage device. That is, the measurement data D is data including the measurement value data di for each time ti.

  The waveform specifying unit 102 reads the measurement data D stored in the storage device 11 and specifies the waveform of the measurement data D. Specifically, the waveform specifying means 102 first reads the measurement data D stored in the storage device 11 and obtains the reference data Z for generating the waveform (sine wave) using the equations (1) and (2). .

Y = (360 × (di / K)) (1)
Z = sin (radian (Y)) (2)
Since an error is likely to occur in the conversion from the frequency distribution to the radians distribution, the waveform specifying unit 102 absorbs the error by using the equations (3) and (4) for the reference data Z obtained by the equation (2). The correction value bias to be obtained is obtained. Here, the reference correction value B used in Expression (3) is a value expressed in% when the maximum value of the measurement value data di is 100 and the minimum value is 0, and is a “dead zone” of a sensor device or the like. . In general, this dead zone is uniquely determined by the device characteristics and the device adaptation system.

b = B / 10 (3)
bias = (b / 10) + (pow (b, 2) / 100/2) + (pow (b, 3) / 100/2) (4)
Further, the waveform specifying means 102 obtains the final data Value from the reference data Z, the correction value bias, the upper limit value dmax, and the reference minimum value dmin using the formula (5). here,
Value = ((dmax−dmin) × ((1.0 + Z) / (2.0 + bias))) + (dmin × (1.0+ (b / 10)) (5)
As described above, the waveform specifying unit 102 can express the measurement data D that is time-series data as a waveform that is the final data value as illustrated in FIG. 2 by using the equations (1) to (5). it can. The fact that it can be expressed as a waveform means that it can be captured as a waveform, that is, the measurement data D can be expressed as an Nth order equation. Therefore, simple time-series measurement value data di can be evaluated as a logical expression.

  The determining means 103 uses the waveform specified by the waveform specifying means 102 to determine whether or not a problem has repeatedly occurred in the process, and monitors the plant process.

  Data required in engineering control (engineering unit data) is often close to the phase of a theoretical sine wave (or cosine wave). Alternatively, such engineering unit data is often a waveform with little variation from a predetermined value, such as a linear waveform balanced on the time axis. The sine theoretical formula and the cosine theoretical formula hold such a theory in terms of the linearity of the waveform. Therefore, by applying sine theory or cosine theory to measurement data D, an estimated value can be obtained relatively easily and correctly.

  Normal distributions (including non-normal distributions that can be dealt with by normalization processing) placed on a plane waveform can be expressed by n-th order expressions. It can be said that correction processing can be performed. That is, sine curve calculated from theoretical value, sine theory formula, sine nth order equation, cosine curve, cosine theory formula or cosine nth order equation, approximate phase analysis by approximate distribution analysis, that is, calculation of unmeasured time axis from measured value And pseudo values on the past time axis can be calculated.

The sine theory (the sine theorem) establishes equation (6),
a / sin A = b / sin B = c / sin C = 2R (6)
Since A = 2R holds when A = π / 2, applying this makes it possible to easily vectorize the measurement data D and evaluate the logical expression. Further, it is possible to avoid the measurement data D from diffusing into higher order expressions. Furthermore, since the sine wave theory is a dimensionless expression (m · m ^-1 ), which is obtained by adding a time t and a unit quantity λ per time to the sine theory, it is sampling data at the time of expansion to the sine wave theory expression. The time formula of measurement data D can be applied.

  Also, when referring to the measurement data D, which is past data, the cosine theory is synonymous with the backward movement of the sine theory, so the measurement data can be referenced by replacing the sine theory with the cosine theory. It becomes. Specifically, the sine theoretical formula may be shifted by π / 2 from the sine theoretical formula to the cosine theoretical formula.

  Here, considering the wave theory, which is a paradox solution (Schrodinger's paradox), the normality of the values that make up the measured or pseudo curve is not essential. Can be generated.

  Therefore, the value on the plane (value on the elevation) itself has little meaning, and attention is paid only to the tendency of the curve. This leads to the generation of an ideal distribution map in anti-phase transformation analysis from a pseudo curve in a theoretical space (regardless of plane or elevation) that the display of the curve leads to an n-order function. become.

  In this way, a solution is obtained for the normal distribution of time distribution data (sampling data in units of time) to which normal theory is applied. Here, the problem is how to obtain the increase amount per time unit of the sine theoretical formula. When this increase amount is a value that deviates significantly from the normal distribution, there is a high possibility that the unit amount λ per time is not normally (accurately) obtained, and it is useless to obtain a solution by the nth order equation. Therefore, when the increase amount is significantly different from the normal distribution, it is necessary to process it as “data that cannot be normalized”. However, since “data that cannot be normalized” always exists in the normalized distribution, it is necessary to eliminate this “data that cannot be normalized” and isolate it from other data.

  As a method of specifying “data that cannot be normalized”, the radians value is calculated from the theoretical three-dimensional sine value obtained from the irreversible normalization formula, and the data out of the angular distribution range is set as “data that cannot be normalized”. It is common. However, in the control theory (control engineering theory), the irreversible normalization formula is applied because the disturbance θ in the control is eliminated reversibly, that is, the control becomes impossible when the disturbance occurs. Not realistic.

  In the control, even when the disturbance θ occurs, it is necessary to maintain continuity, and therefore the disturbance θ needs to be handled as one of the data. For this reason, the correlation phase in the normally scattered distribution arrangement is obtained. The correlation phase to be obtained here is calculated by using recursive analysis based on the sine theory, since it is necessary to extract data significantly deviating from the radians angular distribution conforming to the normal distribution. However, it can be estimated that the reliability of the sinusoidal distribution obtained from only the generation ratio and data value of the disturbance θ within a certain time at this time is quite low. That is, it is because it is estimated that the amount of data samples within a certain time is extremely small or there is no data. Therefore, for a very small amount of “data that cannot be normalized”, it is difficult to perform accurate data analysis only with a normal distribution (sine theorem, cosine theorem, and vector theory).

  For a small amount of “data that cannot be normalized” that cannot be ignored, the δ value is obtained from the nth-order curve by differential operation over a certain time range, and the relationship with the occurrence ratio of disturbance θ is obtained by a linear expression. In addition, it is necessary to obtain a value for each disturbance θ. By giving the value obtained from the primary expression as α in the inverse operation expression, it becomes possible to obtain the probability of occurrence of disturbance θ per unit time, and the consistency comparison difference value with the distribution map of the obtained data group Theoretically, it is possible to obtain a distribution curve from the following Kn equation. In other words, the non-normality estimated from the result obtained by the differential equation of time unit, the difference area between the area drawn by the normal distribution curve based on the sine wave theory and the area drawn by the curve obtained by the Kn order curve The vector value of the distribution is the probability of occurrence of the disturbance θ per fixed time, and the actual value of the disturbance at that time can be a correction value of the increase amount per time unit in the sine theoretical formula.

  With the above theoretical value (if it is an approximate value of actual sampling data), a correction value per time unit can be obtained.

The final problem is that, as in Fermat's final theorem, the formula for calculating the solution obtained by the ^nth -order formula is Z ⁿ = X ⁿ + Y ⁿ and n is a prime number greater than 3. In other words, the solution cannot be obtained. However, the data distribution chart and curve in the sine theorem and cosine theorem are originally possible only by development from theoretical formulas, and do not actually exist in nature. Therefore, even if correction is performed using an approximate value of the prime number n in order to avoid a state without a solution, it is considered that an error with respect to the obtained solution is very small.

  For example, when the water hammer is estimated and determined, a specified waveform is used as shown in FIG. The solution α obtained here is the vertex of a theoretical sine waveform in a normal distribution, and shows a tendency of increase or decrease depending on time t.

  When the value α obtained as a solution is on the sine waveform, the determination unit 103 determines that there is a possibility that the water hammer is repeatedly generated periodically.

  On the other hand, when the value α is not on the sine wave, the determination unit 103 moves the time axis t using the cosine theorem and determines whether or not it depends on the transition of the time axis t of the sine wave. That is, when the obtained solution α is on the cosine wave, it is determined that there is a possibility that the water hammer is repeatedly generated.

  When there is no value α on either the sine waveform or the cosine waveform, it is considered that the determination means 103 has no water hammer occurrence point, and the water hammer is regarded as a disturbance θ, and the normality is determined from the wave waveform theory. By calculating, the transition of water hammer is predicted.

  If the value α does not appear as a point of the normal distribution area of the wave waveform from the estimated point in the wave waveform, that is, the normalized relative time (t−1, t−1) of the time axis t, the determination unit 103 A water hammer is defined as a temporary occurrence.

  The wave theory is the theory that “the particle is stationary and does not move, but when the particle moves, the particle wave moves”, where “moving particle wave” is set as time t, and the wave theory By obtaining the phase relationship between the point obtained from the equation and the point obtained by the sine or cosine theorem, the water hammer generation point can be developed into a theoretical equation, and its tendency and transition can be grasped as a normal distribution diagram.

  According to the monitoring apparatus according to the embodiment of the present invention described above, since sine theory or cosine theory is applied to the process value to be acquired, it is not necessary to prepare model data or the like for determining the process value in advance, It can be easily determined whether or not the occurrence has occurred repeatedly.

1 ... Monitoring device 10 ... CPU
DESCRIPTION OF SYMBOLS 11 ... Memory | storage device 12 ... Output device 13 ... Input device 101 ... Data acquisition means 102 ... Waveform specification means 103 ... Determination means

Claims (1)

A monitoring device connected to a plant operated by a plurality of processes and monitoring the state of the process of the plant,
Data acquisition means for acquiring a process value measured as a value representing the state of the process, and associating the process value with the time at which the process value was measured and storing it in a storage device as measurement data;
Waveform specifying means for reading measurement data from the storage device and specifying a waveform of a process value;
From the waveform of the process value specified by the waveform specifying means, using the sine theory or cosine theory, a determination means for determining whether or not a problem has repeatedly occurred in the process,
A monitoring device comprising:

Images