JPH073642B2 - Process abnormality diagnosis method - Google Patents

Description

TECHNICAL FIELD The present invention relates to an abnormality diagnosis method for automatically finding the cause of an abnormality in a chemical plant, various production processing processes, etc., that is, the point of failure occurrence.

[Conventional technology]

As a method for diagnosing abnormalities in chemical treatment processes, etc., methods have been developed that use a directed graph with signs that show the correlation between the physical quantity changes of the object to be treated between the points of the process.
Chemical Engineering Papers, Vol. 10, No. 2, pp. 233-239 (198
Details are given in 4).

Further, in the above-mentioned method, a method has been developed which is capable of diagnosing only a single failure point, that is, an abnormality caused by a single cause, and is also capable of diagnosing multiple failure points, that is, an abnormality caused by multiple causes. Volume 3, Issue 343-346 (1985)
It is described in.

It should be noted that a method for making the diagnosis status more accurate by using the change over time of the abnormal pattern by using the data before the diagnosis time and using the multi-layer graph has been put to practical use. No. 5, pp. 609-615 (1984).

However, in the method that enables the diagnosis of anomalies due to a plurality of failure points (hereinafter, a set of these failure points is referred to as a candidate), a large number of candidates may be obtained as a diagnosis result in some cases, and the failure point cannot be identified. However, the method that utilizes the change over time of the abnormal pattern requires several tens of times the memory capacity and the calculation time as compared with the above method when applied to a large-scale process. , Has the drawback that practical use is almost impossible.

As a countermeasure against this, according to Japanese Patent Application No. 61-195540 by the present applicant, the candidate of the failure point obtained by the primary abnormality diagnosis is stored, and the physical quantity different from that at the time of the primary diagnosis is stored. A second abnormality diagnosis is performed, and at this time, further candidates are to be obtained from the candidates obtained by the first diagnosis,
A method for reducing the memory capacity and the calculation time is proposed.

[Problems to be solved by the invention]

However, even if the above method is applied to a large-scale process, the number of candidate points of failure does not decrease sufficiently, and
There may be a problem that the reduction of the calculation time is insufficient, and in particular, the problem that it becomes impossible to identify the failure point occurs in the cyclic process.

[Means for solving problems]

In order to solve the above problems, the present invention is configured by the following means.

That is, a plurality of measurement points for measuring the physical quantity of the object to be processed are provided on the process of processing the object to be processed, and the process in which the physical quantity change propagates from these individual measurement points to other measurement points is stored. , Detecting that the physical quantity has changed in a positive or negative direction from a predetermined reference value at each measurement point, the change in the physical quantity detected at arbitrary multiple measurement points and the ripple process of the physical quantity change between these measurement points. In the abnormality diagnosis method that determines the candidate of the fault point of the process based on and, the time when the physical quantity changed from the reference value to the positive or negative direction at each measurement point is stored multiple times, and the time at each arbitrary measurement point is stored. Considers only those points that do not contradict the spread of changes in physical quantities between these measurement points as candidate points for failure.

[Work]

Therefore, based on the principle that the cause occurs earlier than the result, and the result occurs later than the cause, the abnormality is diagnosed by using only candidates that match the progress of the cause of the abnormality to the result. Even in the above process, the failure point can be determined reliably and promptly.

〔Example〕

Hereinafter, the details of the present invention will be described with reference to the drawings illustrating examples.

FIG. 2 is a schematic diagram of the process. From the tank 2 to which the liquid W as the object to be treated is supplied via the pipeline 1, the liquid W is sequentially transferred to the tanks 4 and 6 via the pipelines 3 and 5. In addition to being supplied, it is fed from the pipe line 7, while it is supplied from the pipe line 8 and 10 by the pump 9 to the tank 6
From the tank to the tank 2 and circulates between the tanks 2, 4, and 6. In the tanks 2, 4, and 6, the heaters 11 to 13 heat the respective tanks, and a predetermined process is performed on each of them.

Therefore, the liquid W flows from the tank 2 to the tank 4 and from the tank 4 to the tank 6.
Flow in a fixed direction, and are sequentially and repeatedly processed in each of the tanks 2, 4, and 6 and then fed to a predetermined site not shown in the drawing by a pipe line 7.

In addition, at each point of each tank 2, 4, and 6, thermometers 14 to 16 are provided as specific multiple points, and the liquid temperature
T ₁ , T ₂ and T ₃ are measured as physical quantities, and an electronic computer (hereinafter CP
The measured value is given to T) 21.

The CPT21 is equipped with a cathode ray tube display device (hereinafter, CRT) 22, keyboard (hereinafter, KB) 23, printer (hereinafter, PRT) 24, etc., and can be operated by operating the KB23 and inputting with a light pen of the CRT22. , Memory 2 in CPT21 is tank 2
The structure of the process by tank 6 is stored and CP
The processor (hereinafter CPU) in T21 executes the program in memory, performs abnormality diagnosis based on the measured values from thermometers 14 to 16 and the process configuration, and outputs this result to CRT2.
2 and PRT24 for display and printout.

FIG. 3 is a diagram showing a signed directed graph used when storing the structure of the process in the memory of the CPT 21, and thereby the liquid temperatures T _{1 to} T ₃ at the respective points of the tanks 2, 4 and 6 are shown. It represents the change correlation.

That is, the liquid temperatures T _{1 to} T ₃ are connected to the pipelines ₃ , 5, 8, and 10 by the corresponding arrows, and when the temperature change on the downstream side in the same direction as the liquid temperature change on the upstream side occurs, When the relations between the two are opposite, the signs of − are attached, and in the example of FIG. 2, since all are in the same direction, all the symbols in FIG. It is said.

Further, the CPT21 judges that the measured values of the liquid thermometers 14 to 16 are compared with the reference value as the presence or absence of a change in the positive or negative direction, and this situation is as shown in FIG. .

That is, the reference value 0 is set for each of the measured values T, and the allowable range is doubled in the positive and negative directions.
The determination levels α ⁺ , α ⁻ , β ⁺ , β ⁻ are determined accordingly, and the positive direction change “+”, the negative direction change “−”, and A positive change "+?" And a similar negative change "-?" Are detected.

α ⁻ ≦ T ≦ α ⁺ …… “0” (no change) T> β ⁺ …… “+” T <β ⁻ …… “-” β ⁺ ≧ T> α ⁺ …… “+?” α ⁻ ＞ ^{T ≧ β - ...... "-?} " FIG. 5 is a diagram showing an example of applying the change detection status described above to the code directed graph of Figure 31, in this case, the vessel
All of the temperatures T _{1 to} T _{3 of} 2, 4, 6 are “+”, and if the conventional abnormality diagnosis is performed based on this, as shown in FIG. Temperature T ₁ indicated by Shigemaru
Then, the relationship between this temperature and the other temperatures T ₂ and T ₃ is determined, and then the same determination is performed with the assumed candidate temperature T ₂ as shown in FIG. ), The candidate is assumed to be a tank in the same manner, and tanks 2, 4, and 6 all become candidates.

On the other hand, the times at which the liquid temperatures T _{1 to} T _{3 of the} tanks 2, 4, 6 respectively change are sequentially obtained over, for example, three times, and these times are stored and stored in a memory. If the diagnosis is to be performed by using a tool whose time does not contradict the course of the abnormality, select only one of the relationships from Figures 6 (A) to (C) and Therefore, the diagnosis should be made.

That is, in FIG. 5, the temperatures T _{1 to} T ₃ are "+?"
The time at which the change occurs is ▲ t ^α ₁ ▼, ▲ t ^α ₂ ▼, ▲ t ^α
₃ ▼, and the time when the change of “+” similarly occurs is ▲ t ^β ₁
▼, ▲ t ^β ₂ ▼, ▲ t ^β ₃ ▼, and if the following relation holds, then ▲ t ^α ₂ ▼ = ▲ t ^β ₂ ▼ <▲ t ^α ₁ ▼ = ▲ t ^β ₁ ▼ < ▲
t ^α ₃ ▼ = ▲ t ^β ₃ ▼ (1) Times ▲ t ^α ₂ ▼ and ▲ t ^β ₂ ▼ are earliest, and time ▲ t ^α ₁
▼, ▲ t ^β ₁ ▼ is the next earliest, and time ▲ t ^α ₃ ▼, ▲ t ^β
₃ ▼ is the most delayed, and times ▲ t ^α ₂ ▼ and ▲ t ^β ₂ ▼.
It is highly probable that the temperature T ₂ corresponding to
It suffices to perform only the diagnosis shown in FIG. 6B, which makes it possible to easily and quickly identify the cause.

Further, the above is the case where the process is constituted by the tanks 2, 4, and 6, which is relatively simple, but if this is generalized, the relation of FIG. 7 is established.

That is, a plurality of abnormal propagation routes l _{1 to} lk are assumed, with the hypothesis candidate 31 as the apex and reaching the change detection points 32 _{1 to} 32k where the final abnormal result appears, and each route l _{1 to} lk is in between.
Intermediate measurement point 33 ₁ ~33 _(P-1) mediated in accordance with the.
(However, if 33p = 32k, 33o is the measurement point, 33 ₁ = 33o
And ) FIG. 8 is a diagram in which any one route li is extracted from each route l _{1 to} lk, and the measurement points 33 _{1 to} 33 _P change “+?” Or “−?”. Set each time to ▲ t ^α ₁ ▼ ~ ▲ t ^α _P ▼
And each time when a change of "+" or "-" occurs by ▲ t
^{Let β} ₁ ▼ to ▲ t ^β _P ▼. (However, if the point is a measurement point of "+?" Or "-?", This time is not defined.) By the way, since the abnormal propagation route is found by computer processing, the search is performed over all routes. Done. Therefore, in a very complicated process, when a diagnosis is made without considering the time when the point becomes abnormal, it is possible that a route as shown in FIG. 9 is searched for. The numbers attached in FIG. 9 indicate the order in which changes occurred at that point. Such a phenomenon that the time is reversed is not possible in reality, but when searching for a route by computer processing, an adjacent phenomenon is manually manipulated and the time relationship is not considered, so that it can occur.

Therefore, in the route li, set the measurement point of "+" or "-" to k,
As j, l ..., ▲ t ^α _i ▼ ≦ ▲ t ^β _k ▼ (i = 1,2,3, ... k-1) ...
(2a) ▲ t ^α _i ▼ ≤ ▲ t ^β _j ▼ (i = 1,2,3, ... j-1) ...
(2b) ▲ t ^α _i ▼ ≦ ▲ t ^β _h ▼ (i = 1,2,3, ... h-1) ...
(2c) If such an operation is performed and all of the above are satisfied, it is possible to improve the accuracy of the abnormality diagnosis by determining that the data of the route does not conflict.

Therefore, it is sufficient to check whether or not the relationships of the expressions (2a) to (2c) are established for each of the routes l _{1 to} lk in FIG. 7, and obtain the route li that satisfies this time condition. .

Fig. 1 is a flowchart of the abnormal diagnosis situation by the program in CPT21 based on the above principle.
The preprocessing programs 1 and 102 are constantly executed, and the abnormality diagnosis program after step 111 is to be executed in response to the above-described change detection.

That is, the liquid temperature gauge 14
"0", "+", "-", for the measured values from ~ 16,
"+?" And "-?" Are judged respectively, and if any of them contains "+" or "-", "+" or
Is there a "-" mark? "102 becomes Y (YES), and the process moves to" cause search "111.

In the step 111, the time when "+", "-", "+?", "-?" Occurs is found by the clock function configured by the CPU of the CPT21, and this time is stored in the memory and is also stored in the memory. Using the stored routes shown in FIG. 7, the cause of the abnormality is sequentially searched, and then the time condition check according to the equations (2a) to (2d) 112 is performed, and this result is satisfied.
"?" 113 is judged, and according to Y of this, CRT22 performs "candidate display" 121, and then checks "All candidates / output done?" 122. While this is N (NO), step 111 or later. Is repeated.

If step 122 becomes Y, normally step 101,
Return to iteration 102.

Therefore, even in the process of circulating the object to be processed,
The failure point can be specified, the time required for calculation can be shortened, and the abnormality can be diagnosed reliably and promptly.

However, the structure shown in FIG. 2 is determined according to the state of the process. As the object to be processed, not only liquid but also lumps of ore, coal, concrete block, or the like, or
It may be an object such as a powder or a granular body, and various modifications can be freely made such that the physical quantity to be detected may be selected from temperature, humidity, flow rate, pressure, weight and the like.

〔The invention's effect〕

As is clear from the above description, according to the present invention, the calculation time required for process abnormality diagnosis is shortened, and
Since the failure point can be specified even in the circulating process of the object to be processed, it is advantageous especially when applied to a large-scale process, and a remarkable effect is obtained in automatic abnormality diagnosis of various processes.

[Brief description of drawings]

FIG. 1 shows an embodiment of the present invention, FIG. 1 is a flow chart of abnormality diagnosis processing, FIG. 2 is a schematic view of the process, and FIG. 3 is a directed graph with signs showing the correlation of liquid temperature changes in FIG.
FIG. 4 is a diagram of judgment levels used for detecting changes in liquid temperature, and FIG.
Fig. 6 is a diagram showing an example of actual measurement of change detection conditions, Fig. 6 is a diagram showing conditions of abnormality diagnosis, Fig. 7 is a diagram showing abnormal transmission routes, and Fig. 8 is any one route extracted from Fig. 7. 9 and FIG. 9 are diagrams for explaining the inconsistent route. 1,3,5,7,8,10 …… Pipe, 2,4,6 …… Bath, 14-16 …… Temperature, 21 …… CPT (computer), 22 …… CRT (CRT display) ), 23 …… KB (keyboard), 24 …… PRT (printer), W …… liquid, T _{1 to} T ₃ …… liquid temperature.

Claims (1)

[Claims] 1. A process in which a plurality of measurement points for measuring a physical quantity of the object to be processed are provided on a process of processing the object to be processed, and the change in the physical quantity spreads from these individual measurement points to other measurement points. Is stored, and it is detected that the physical quantity changes in a positive or negative direction from a predetermined reference value at each of the measurement points, and a change in the physical quantity detected at any of a plurality of measurement points and the mutual measurement points are detected. In the abnormality diagnosis method for determining a candidate of a failure point of the process based on the propagation process of the change in the physical quantity between the plurality of times, the physical quantity at each measurement point changes from the reference value in the positive or negative direction a plurality of times. The difference in the process characterized in that only the ones in which the respective times at any of a plurality of measurement points are stored and which are consistent with the spillover process of the change in the physical quantity between the measurement points are set as the candidates for the failure point. Diagnostic methods.