JPH0572004A - Abnormality-diagnosis device of plant - Google Patents

Abstract

PURPOSE:To enable a cause of abnormality to be estimated quickly and accurately by performing a rough diagnosis based on a matrix regarding an automatic detection event and then performing a detailed diagnosis based on a matrix regarding an operator input event. CONSTITUTION:Knowledge data 6 and 9 are data regarding an automatic detection event which can be detected automatically by a computer 1 and an event which requires judgment by an operator (operator input event). When a signal regarding the automatic detection event is entered 3 periodically from a plant 2, it is judged 4 whether the event is generated or not. Then, if any event is generated, a rough diagnosis is performed at a cause-estimation part 5 by using the data 6, thus enabling an abnormal equipment to be estimated. Then, a cause-estimation part 8 performs a detailed diagnosis by using the data 9 regarding the estimated abnormal equipment and a value indicating degree of establishment of event which is input by the operator from an input device 10 and then displays a cause of abnormality, namely a part which fails and its countermeasures if possible on a display device 7.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a plant abnormality diagnosis apparatus.

[0002]

2. Description of the Related Art Conventionally, as a method of estimating the cause from an abnormal event that has occurred in a plant, as shown in FIG. 3, n events b _j (j = 1, 2, ..., N) and m events have been performed. Cause a _i
There is a method of using knowledge data in which a causal relationship of (i = 1, 2, ..., M) is expressed by a determinant (matrix). For example, as shown in FIG. 3, the relationship between the cause a ₁ and the event b ₁ is r ₁₁
In the case of, the probability that the event b ₁ occurs due to the cause a ₁ is r ₁₁ . If the m × n matrix having r _ij as an element is R and the vectors having a _i and b _j as elements are A and B, respectively, the above causal relationship is expressed by the following equation. A · R = B (1) Here, the symbol “·” means the following calculation. ∨ _i (a _i ∧ r _ij ) = b _j (2) where ∨: max, ∧: min

Conventionally, the cause of abnormality has been estimated by calculating the cause A from the matrix R and the event B that is currently occurring. An example of the device is shown in FIG. As shown in the figure, the computer 1 that executes diagnosis includes a signal input unit 3, a determination unit 4, a cause estimation unit 5, knowledge data 6 and a display unit 7. A signal relating to an event that occurs in the plant 2 is input to the signal input unit 3, and the determination unit 4 determines the occurrence degree of the event from the input signal. The cause estimation unit 5 estimates the cause using the knowledge data 6, and the estimation result is displayed on the display device 7.

[0004]

The event in the plant includes an event that can be automatically detected using a detector and a computer (hereinafter referred to as an automatic detection event) and an event that requires an operator's artificial judgment (hereinafter, referred to as Operator input event)), but the latter is generally more common. Therefore,
In order to make an accurate judgment about all the events, it is not enough to create the knowledge data by only the automatically detected events, and all the operator input events are manually input by the operator. Need to create. But,
In the conventional anomaly detection apparatus shown in FIG. 3, since the knowledge data is created without inputting operator input events, a number of cause candidates are obtained and the cause cannot be narrowed down.

Further, in order to input an operator input event as knowledge data, it is necessary to investigate each event. For that purpose, the operation of the plant is temporarily stopped and an environment capable of investigation is prepared. For example, it took a long time to conduct the survey because preparations such as pulling out the roll had to be made. Further, if the knowledge data in which all the events are expressed by one determinant is used, even if the detection of the automatically detected events is completed until all the investigations of the operator input events are completed, This causes a problem that the cause cannot be estimated using the data. The present invention has been made in view of the above-mentioned prior art, and uses knowledge data by a determinant that distinguishes between an automatically detected event that can be detected automatically and an operator input event that requires artificial judgment. An object is to provide an abnormality diagnosis device for a plant.

[0006]

The configuration of the present invention for achieving such an object is an abnormality in which an abnormality cause of a plant is estimated online using knowledge data composed of a determinant expressing a causal relationship between an abnormal event and a cause. In the diagnostic device, the determinant is created by distinguishing between events that can be automatically detected and events that require artificial judgment, and after making a rough diagnosis based on the determinant of the former event, the latter event It is characterized by making a detailed diagnosis based on a determinant.

[0007]

In the present invention, first, a rough diagnosis is performed using a determinant consisting only of automatically detected events, for example, a device in which an abnormality has occurred is estimated, and then a determinant based on an operator input event, for example, , Detailed diagnosis is performed using a small determinant consisting only of events related to the above causes. Therefore, by narrowing down the operator input events, it is possible to reduce the number of operator investigation items and to make a quick estimation.

The automatic detection event in the present invention is an event that can be carried out by the detector and an event that can be automatically determined by using the data from the detector. For example, whether the cylinder position data is larger than a certain value or not. That is the judgment. The operator input event is an event that cannot be detected by the detector and needs to be investigated and judged artificially, for example, disconnection.

[0009]

DESCRIPTION OF THE PREFERRED EMBODIMENTS The present invention will be described below in detail with reference to the embodiments shown in the drawings. FIG. 1 shows an embodiment of the present invention. As shown in the figure, the computer 1 for executing the diagnosis comprises a signal input unit 3, a judgment unit 4, cause estimation units 5, 8, knowledge data 6, 9, a display unit 7 and an input device 10.

As the plant 2 to be diagnosed, for example, the hydraulic pressure reducing device of the rolling mill shown in FIG. 5 is used.
As shown in the figure, this hydraulic pressure reduction device sandwiches a rolled material 11 between upper and lower work rolls 12 and 13, and further sandwiches upper and lower work rolls 12 and 13 between upper and lower backup rolls 14 and 15. is there. Work roll 1
The rolls 2, 13 are controlled by roll bending cylinders 16, 17, and the positions of the backup rolls 14, 15 are controlled by hydraulic pressure reduction cylinders 18, 19. The position of the hydraulic pressure reduction cylinders 18 and 19 is the servo valve 2
Controlled by 0, 21 and hydraulic pressure reduction cylinders 18, 19
The position of is detected by the position detectors 22 and 23. The load of the hydraulic pressure reduction cylinders 18 and 19 is the load cell 2
4,25 detected.

A signal relating to an automatic detection event that has occurred in the plant 2 is periodically input to the signal input unit 3 and the determination unit 4
Determines the degree of occurrence of an automatic detection event from the input signal. The knowledge data 6 is knowledge data about an automatically detected event that can be automatically detected by a computer, and the cause estimation unit 5 uses the knowledge data 6 to estimate the cause. Further, the knowledge data 9 is knowledge data about an event that requires an operator's judgment, and the cause estimating unit 8 determines an event input by the operator through the input device 10, that is,
Estimate the cause based on operator input events. The estimation results of the cause estimation units 5 and 8 are displayed on the display device 7.

The relationship between the knowledge data 6 and 9 is shown in FIG. As shown in the figure, the knowledge data 6 stores the determinants of the automatically detected events b _{1 to} b ₄ and the causes a _{1 to} a ₄ , and in the figure, the event b ₂ has two causes a ₁ , a ₃ Is supported. As a specific example, in the above example, the position detectors 22 and 23
Due to the cause a ₁ of failure of the position detector, an event that an abnormality is detected by the self-diagnosis function of the position detector itself and a signal indicating that the position detectors 22 and 23 are abnormal is output to the outside b ₁ probability r ₁₁ 90% of occurrence of the probability r ₁₂ 70% of occurrence of event b ₂ of the cylinder position actual value limit if undetectable cylinder position data, a work side (in FIG. 5 left) or drive side (in FIG. 5 If either one of (right side) fails, the probability b ₁₃ of the event b ₃ of the cylinder position actual value unbalance occurs 70%, and the event b of the actual load value upper limit b because the cylinder position data cannot be detected and the cylinder is pushed too much. The probability r _{14 of 4} is 7
It becomes 0%.

On the other hand, the knowledge data 9 is provided with determinants 9-1 to 9-4 corresponding to the causes a _{1 to} a ₄ of the knowledge data 6, respectively. Indicates the detailed cause and operator input event. For example,
When the cause estimation unit 5 estimates that the cause is a ₁ , the determinant 9-1 of the knowledge data 9 is selected, and the operator selects the determinant 9-1.
After investigating and determining each event of −1 and inputting the event satisfaction degree value, the cause estimating unit 8 uses the determinant 9-1 to perform detailed estimation of the cause. In this way, by making the determinant correspond to each cause of the knowledge data, the number of input items of the operator can be reduced and efficient diagnosis can be performed. With such a device configuration, the cause of abnormality is estimated as follows. First, a signal regarding an automatically detected event is input from the plant 2 to the signal input unit 3 periodically, and the determination unit 4 determines whether or not an event has occurred. If any event occurs, the cause estimation unit 5 makes a rough judgment using the knowledge data 6 about the automatically detected event,
For example, the device in which an abnormality has occurred is estimated.

Next, the cause estimating unit 8 performs a detailed diagnosis by using the estimated knowledge data 9 about the abnormality-causing device and the event success degree value input by the operator from the input device 10 to determine the cause of the abnormality, that is, The location where the abnormality has occurred is displayed on the display device 7. Further, it is desirable to display the countermeasure on the display device 7. In the above embodiment, the determinant of the operator input event corresponding to each cause of the automatically detected event was created, but the present invention is not limited to this, and the determinant of the operator input event for these causes is comprehensive. May be created.

[0015]

As described above in detail with reference to the embodiments, the abnormality diagnosing device of the present invention distinguishes an event into an automatic detection event and an operator input event. Based on the determinant based on the determinant, the determinant based on the operator input event is used to make a detailed determination, so that the operator input event can be diagnosed, and as a result, the cause of the abnormality can be quickly and accurately determined. It is possible to estimate.

[Brief description of drawings]

FIG. 1 is a block diagram of a plant abnormality diagnosis apparatus according to an embodiment of the present invention.

FIG. 2 is an explanatory diagram showing a relationship between automatically detected events and knowledge data regarding operator input events.

FIG. 3 is an explanatory diagram showing knowledge data.

FIG. 4 is a block diagram of a conventional plant abnormality diagnosis device.

FIG. 5 is a front view of a hydraulic pressure reducing device of a rolling mill that is an example of a plant.

[Explanation of symbols]

 1 computer 2 plant 3 signal input unit 4 judgment unit 5,8 cause estimation unit 6,9 knowledge data 7 display device 10 input device 11 rolled material 12,13 work roll 14,15 backup roll 16,17 roll bending cylinder 18,19 Hydraulic pressure reduction cylinder 20,21 Servo valve 22,23 Position detector 24,25 Load cell

Claims (1)

[Claims] 1. An abnormality diagnosing apparatus for estimating a plant abnormality cause online using knowledge data consisting of a determinant expressing a causal relationship between an abnormality event and a cause, and an event which can be automatically detected and an artificial judgment. It is characterized in that the determinant is created by distinguishing it from the event requiring the above, and after performing a rough diagnosis based on the determinant regarding the former event, a detailed diagnosis is performed based on the determinant regarding the latter event. Abnormality diagnostic equipment for plants.