JP2000259220A - Monitoring device - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a monitoring device which can be applied even to an object with which a deciding standard can be automatically set and an abnormal state can be decided only in a combination condition of plural parameters. SOLUTION: This monitoring device collects in time series N types of primary parameters showing the states of a monitoring object at a specific time and decides whether or not the object is abnormal at the specific time from the difference confirmed between the normal and abnormal states in regard to a combination of the parameter groups at the specific time via the sensors 21-23. A data processing part 40 defines a space near the coordinates of a position shown by the combination of those parameters in a normal state as a normal value range space in an N-dimensional space formed by the parameters and in a fixed time when a system is kept normal. Then it's decided under the definition whether the system is normal or abnormal against a combination of all parameter groups.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a monitoring apparatus for monitoring the status of various plants, devices, machines, plants and animals as monitoring targets, and automatically determining whether the monitoring target is normal or abnormal.

[0002]

2. Description of the Related Art Conventionally, this type of monitoring apparatus measures parameters representing the state of a monitoring target, such as temperature, vibration, component concentration, etc., of each unit, and a normal value range or an upper limit of alarm generation for each parameter. And the lower limit were manually set, and it was determined whether all parameters were within the range of each normal value.

[0003]

In the above-described conventional apparatus, the normal value range has to be manually set for each parameter based on the measurement results in the normal state and the abnormal state, which is very complicated. . In addition, it is not practical to detect an abnormal state that cannot be detected only by a value independent of each parameter and can be detected only by a combination of a plurality of parameters. For example, there is a case where the relationship between the valve opening degree and the flow rate attached to the same pipe is an example.

[0004] It is an object of the present invention to provide a monitoring apparatus which can automatically set a criterion and can be applied to an object whose abnormal state can be determined only by a combination of a plurality of parameters.

[0005]

[MEANS FOR SOLVING THE PROBLEMS] N obtained from the monitored object
In the N-dimensional state space including various kinds of parameters, the vicinity of the position coordinates occupied by the values of the parameters in the normal state is defined as the normal time range space. Then, when a parameter deviating from this space is obtained, an abnormal alarm is generated.

In the present invention, a normal value range space definition mode and a monitoring mode exist as operation modes of the apparatus. When the monitoring target is in a normal state, it operates in a normal value range space definition mode to define a normal value range space. In the monitoring mode, parameters are sampled in real time, it is determined whether or not the coordinates fall within a defined normal value range space, and the result is output as an alarm.

[0007]

DESCRIPTION OF THE PREFERRED EMBODIMENTS One embodiment of the present invention will be described.

FIG. 1 is a configuration diagram of an embodiment of a monitoring apparatus according to the present invention.

The apparatus according to the present embodiment is a system to be monitored 1
A temperature sensor 21, a flow sensor 22, and a pressure sensor 23 are provided as three sensors for measuring 0 to 3 parameters (temperature T, pressure P, flow rate F). In addition, this 3
Interfaces 31, 32, and 33 for reading output signals of the three sensors;
Three parameters (temperature T, pressure P,
It has a data processing unit 40 for obtaining the flow rate F), an operation unit 41 for switching operation modes and setting conditions, and a display unit 42 for displaying an alarm generation state.

Next, the operation of FIG. 1 will be described. The monitoring device of the present embodiment has two operation modes, a normal value range definition mode and a monitoring mode. First, the operation in the normal value range definition mode will be described. The data processing unit 40 controls the sensors 21 to 23 and reads the three parameters T, F, and P from the sensors 21 to 23 once every processing cycle = 1 second.

The data processing unit 40 grasps the range of the normal value by the combination (T, F, P) of these parameters. Specifically, the following processing is performed.

<Process Routine of Fixed Period (at Time t) in Normal Value Range Definition Mode> A temperature measurement signal T (t) is obtained from the sensor 21.

The flow rate measurement signal F (t) is obtained from the sensor 22.

A pressure measurement signal P (t) is obtained from the sensor 23.

In a three-dimensional space (T, F, P) consisting of three parameters T, F, and P, points (T (t), F
The space near (t), P (t)) is added to the domain of the normal value range space.

Specifically, as shown in FIG. 2, the measurement range of T is set to Tmin to Tmax, and the interval is equally divided into _{NT and} divided into _NT sections (for example, 100 equal divisions). Similarly, the measurement range of _F is divided into NF and the measurement range of _P is divided into NP.

As a result, the points (T (t), F (t), P
The three-dimensional space (T, F, P) to which (t)) belongs is _NT ×
This is a group of N _F × N _P partitions. This set of sections is represented by a matrix A (x, y, z).

Here, x, y, and z are integers and satisfy the following.

1 ≦ x ≦ N _T , 1 ≦ y ≦ N _F , 1 ≦ z ≦ N _P One point (T
(T), F (t), P (t)) are a set of partitions A (x,
a (xt, yt, z) which is one element of
t).

Then, the points (T (t), F (t), P
The normal value range in the vicinity of (t)) is defined as a group of points belonging to one of the sections a (x, y, z) satisfying all the following three conditions.

Xt-LT≥x≥xt + LT yt-LF≥y≥yt + LF zt-LP≥z≥zt + LP where LT, LF and LP are small natural number constants for defining the neighborhood. FIG. 3 shows an example in which LT = LF = LP = 2.

By repeating the above processing periodically, a spatial area corresponding to a normal value range is defined in the three-dimensional space as shown in FIG.

Next, the operation in the monitoring mode will be described. The following processing is performed periodically.

<Regular processing routine in monitoring mode> A temperature measurement signal T (t) is obtained from the sensor 21.

The flow rate measurement signal F (t) is obtained from the sensor 22.

The pressure measurement signal P (t) is obtained from the sensor 23.

It is determined whether the section of the three-dimensional space to which the point (T (t), F (t), P (t)) belongs belongs to a space (group of sections) defined as a "normal value range". I do.

If the judgment result is YES (belongs to the normal value space)
In case of ・ No alarm is issued.

<If the result is NO (does not belong to the normal value space)> An alarm is issued.

The system issues an "abnormal alarm" in the sense that "the system has not experienced a normal value definition".

As described above, according to the present embodiment, abnormality diagnosis based on the normal correlation between a plurality of measurement signals, that is, a combination of parameter values never experienced during normal operation has occurred. Can be detected without creating a program or the like based on a condition unique to each system.

In the above-described embodiment, discretization is performed as a "neighbor" when defining a normal value range, and then a certain allowable range is given for each parameter axis. That is, a rectangular parallelepiped “near” exists around a normal measurement value at one point.

However, this is not limited to a rectangular parallelepiped, and may be a sphere or an ellipsoid using a distance in a three-dimensional space.

In the above embodiment, there are only three parameters. However, four, five, or even more parameters may be used. In this case, the normal value range is defined in an N-dimensional space where N is the number of parameters.

Further, as parameters, not only the raw data measured or received, but also their differential values such as first derivative, second derivative,... Nth derivative, primary integration, secondary integration,.
It may be an integral value such as an N-order integral.

Further, the parameter may be not only the current raw data measured or received, but also a value that is temporally earlier by ΔT (the past value that is a certain time earlier) than the time.

In the monitoring mode, not only the presence / absence of an alarm but also the degree of an alarm at the time of the occurrence may be expressed by a distance from a normal value range. FIG. 5 shows an example in a two-dimensional space with two parameters. P1 indicates a combination of parameters measured in the monitoring mode.
And P2. Let L be the distance between these points and the normal value range.
1, L2, if there is a relationship of L1 <L2, P
It is determined that the degree of abnormality is larger in P2 than in P1. As a practical application, the distance L from the normal value range is set to the threshold L
An apparatus is conceivable in which t is determined and an alarm generated depending on whether the distance is longer or smaller is divided into two types.

In the above-described embodiment, the normal value range is automatically defined. However, it is of course possible to manually specify and define the normal value range in an N-dimensional space.

In the case where an abnormal judgment is made but an erroneous alarm is issued, a function of adding the parameter (and its vicinity) at that time in a normal value range space can be added. This is nothing but the addition of a learning function.

It is also possible to add a function for canceling the definition of a normal value domain that has passed a predetermined time (for example, one month) or more from the time of definition. This makes it possible to adapt to a system in which the condition of normal or abnormal changes slowly over time.

As a method of making a decision in the N-dimensional space, a high-speed decision can be made by using the address space of the memory. For example, when using three parameters to divide the measurement range of each parameter into 8 bits and 256 sections, 8 × 3 = 24 bits are allocated to address designation bits of the memory. By doing so,
Each address of the memory corresponds to a section of the three-dimensional space. In the normal value area definition mode, if the section is an abnormal value area, the memory data at the corresponding address = 0. If the section is a normal value area, the memory data at the corresponding address = 1. At times, it is very easy and fast to determine whether the section indicated by the measured value, that is, the memory address is in the abnormal value area or the normal value area.

Furthermore, if 8-bit data is used, the abnormal level can be identified in 255 levels.

In the above example, the measurement signal of the monitoring target system is measured by a sensor built in the monitoring device itself, but a part or all of the measurement signal may be received from the outside by a communication function. In this case, it can be said that a certain sensor comprehensively analyzes not only its own measurement information but also information of other sensors and detects an abnormality of the monitored system. Then, when an abnormality of the monitored system is detected, a measuring unit 5 that measures parameters as shown in FIG.
1. If the system includes the control unit 52 and the interface unit 53, the alarm signal may be transmitted to another device on the network via the interface unit 53.

As a specific monitoring target, there is an abnormality monitoring system for aquatic organisms for monitoring water quality. For example, a system is conceivable in which a plurality of fish are bred in a water tank, and a position in a depth direction and a moving distance per time are used as parameters for capturing the behavior pattern.

[0045] Certain types of fish normally move slowly at relatively deep water depths, but when they suffer from toxic substance inflow, etc., there are differences in behavior patterns such as violent movement near the water surface. According to the basic principle of the present invention, an abnormal behavior pattern can be identified and detected as normal.

Further, in the above-described abnormality monitoring system for aquatic organisms for monitoring water quality, it is possible to use, as a parameter, an area occupied by the color of a target fish in a specific area of an image viewed from the side of the aquarium. .

[0047]

As described above, according to the present invention, the N parameters in the normal state in the N-dimensional space formed by the N parameters within a certain period of time during which the monitoring target is in the normal state. A space near the coordinates of the position indicated by the combination is defined as a normal value range space, and the monitoring target is normal or abnormal for all combinations of parameter groups in the N-dimensional space under the definition. Since the determination is made as to whether or not there is a monitoring device, it is possible to provide a monitoring device that can automatically set a determination criterion and can be applied to a target in which an abnormal state can be determined for the first time based on a combination of a plurality of parameters. .

[Brief description of the drawings]

FIG. 1 is a configuration diagram of an embodiment of the present invention.

FIG. 2 is an explanatory diagram of defining a normal value range in the operation of the present invention.

FIG. 3 is an explanatory diagram of defining a normal value range in the operation of the present invention.

FIG. 4 is a diagram showing an example of a defined normal value range.

FIG. 5 is an explanatory diagram regarding a case where the degree of abnormality is identified based on a distance from a normal value range.

FIG. 6 is an explanatory diagram regarding a case where the degree of abnormality is identified based on a distance from a normal value range.

[Explanation of symbols]

 DESCRIPTION OF SYMBOLS 10 ... Monitoring system 21 ... Temperature sensor 22 ... Flow rate sensor 23 ... Pressure sensor 31-33 ... Interface 40 ... Data processing part 41 ... Operation part 42 ... Display part

Claims (13)

[Claims] 1. An N representing a state of a monitoring target at a specific time.
Collects one-dimensional parameters in a time series and determines whether the monitoring target is in an abnormal state at the specific time based on a difference between a normal time and an abnormal time regarding the combination of these parameter groups at the specific time. In a monitoring device that performs a normal operation, within a certain period of time during which the monitoring target is in a normal state, in the N-dimensional space formed by the N kinds of parameters, the vicinity of the coordinates of the position indicated by the combination of the normal N kinds of parameters is Means for defining a space as a normal value range space, and means for determining whether the monitoring target is normal or abnormal for a combination of all parameter groups in the N-dimensional space under the definition. A monitoring device comprising: 2. The correlation between a plurality of types of parameters is calculated not only by an instantaneous value from time to time but also by temporal differentiation, primary differentiation, secondary differentiation,
... The monitoring apparatus according to claim 1, further comprising means for obtaining a correlation based on a correlation including a value up to the N-th derivative. 3. The correlation of a plurality of types of parameters is calculated not only by an instantaneous value and a differential value from time to time but also by temporal integration, primary integration,
3. The monitoring apparatus according to claim 1, further comprising means for obtaining a correlation based on a correlation including values up to the second-order integration,..., The N-th integration. 4. The apparatus according to claim 1, further comprising means for obtaining a correlation between a plurality of types of parameters using not only instantaneous values at times but also values which are temporally ΔT earlier for each signal. 4. The monitoring device according to claim 3. 5. The monitoring apparatus according to claim 1, further comprising means for indicating a degree of abnormality of the monitoring target by a distance from the normal value domain. 6. The monitoring apparatus according to claim 1, further comprising means for manually defining at least one of the abnormal value area and the normal value area instead of automatically. 7. A system according to claim 1, further comprising means for adding a parameter and a nearby value to a domain of said normal value range space when an abnormality is determined but a false alarm is given. The monitoring device according to claim 1. 8. The monitoring device according to claim 1, further comprising means for canceling the normal value domain over a predetermined time from the definition time. 9. The monitoring apparatus according to claim 1, wherein an address space of a memory is used as a method of performing the difference in the determination in the N-dimensional space. 10. A network, a measuring unit for measuring a state of a monitoring target, an interface unit for transmitting a signal on the network, control in the measuring unit and the interface unit,
A control unit that performs data processing, and captures a normal correlation between a measurement signal obtained from the measurement unit and another signal to be monitored obtained from a network via an interface unit, and correlates these signals. And means for transmitting an alarm signal to another device on the network via the interface unit when the value deviates from the normal range. The monitoring device according to claim 1. 11. A method for collecting N kinds of one-dimensional parameters representing a state of an aquatic organism at a specific time in a time-series manner, and obtaining a difference between a normal state and an abnormal state regarding a combination of these parameter groups at a specific time. In a monitoring device for detecting an abnormal state, a position indicated by a combination of N normal parameters in an N-dimensional space formed by N parameters within a certain period of time when the aquatic organism is in a normal state. Means for defining a space in the vicinity of coordinates as a normal value range space, and whether the aquatic organism is normal or abnormal for a combination of all parameter groups in the N-dimensional space under the definition. A monitoring device for aquatic organisms, comprising: means for performing determination. 12. The aquatic organism monitoring device according to claim 11, wherein a position in a depth direction and a moving distance per time are used as the parameters. 13. The aquatic organism state monitoring device according to claim 11, wherein an area occupied by the color of the aquatic organism in a specific area of the image viewed from the side of the aquarium is used as the parameter.