JP2019121065A - Abnormality sign diagnostic apparatus, abnormal sign diagnostic method and abnormal sign diagnostic program - Google Patents

Abstract

To efficiently diagnose multiple parallel processes.SOLUTION: An abnormality sign diagnostic apparatus according to the present invention comprises: a pre-processing unit that acquires, in time series, measurement values of a plurality of parallel processes in which the same type of apparatus performs the same type of processing, and regards the measured values of multiple parallel processes acquired as measured values in one process; and a classification unit that classifies measurement values regarded as measurement values in one process by the pre-processing unit. More specifically, the pre-processing unit associates a plurality of different virtual times included in a predetermined time width with one another with respect to measured values for which the actual times actually obtained are the same among the measurement values acquired from each of the parallel processes, and the classification unit regards that the measurement value is acquired at virtual time.SELECTED DRAWING: Figure 4

Description

  The present invention relates to an abnormality prognostic diagnosis device, an abnormality prognostic diagnosis method, and an abnormality prognostic diagnosis program.

  If the device is degraded and can not operate normally, it will be repaired / updated. There is no problem if repairs and updates can be performed systematically. However, in the case where equipment incorporated in a large-scale production line suddenly shuts down, its repair and renewal require unexpected labor and cost, which has a great impact on business management. Therefore, it is important to detect the sign before the device falls into a full-fledged abnormality and to perform repair and update preventively.

  Prior to diagnosis of a plant, the diagnostic device of Patent Literature 1 constructs a normal model using state quantities acquired from the plant at a time when it is known that the plant is normal. The diagnostic device determines whether or not the state quantities acquired from the plant at the time of diagnosis are classified into the normal model. The said diagnostic device displays that a plant is normal, when the state quantity acquired from the plant at the diagnosis target time is classified into a normal model. The diagnostic device displays that the plant is in an unknown state which has not been experienced in the past, when the state quantity acquired from the plant at the time of diagnosis is not classified into the normal model.

International Publication No. 2012/073289

  In an actual plant, a plurality of identical steps are often arranged in parallel. For example, a plurality of (several tens to several hundreds in many cases) steps of mixing the raw material A and the raw material B in the reactor to produce the product C are performed in parallel, and some or all of them are operated according to the production plan Do. In such a case, when an abnormality occurs in one of the parallel steps, the same type of abnormality often occurs in the other steps. As a plant maintenance person, I would like to know at a time whether or not an abnormality has occurred in parallel processes, and if an abnormality has occurred, specifically what kind of abnormality has occurred.

However, even if there are a plurality of processes in the plant, the diagnostic device of Patent Document 1 determines which process and which process are in parallel among them, and which process and which process are in series. I do not particularly recognize the relationship. As a result, the diagnostic device individually outputs diagnostic results of a plurality of parallel steps. After all, when a maintenance person tries to diagnose each of a plurality of processes in parallel, the number of diagnoses is multiple, and it is not possible to list each diagnostic result, and it is also possible to know the commonality among those diagnostic results. Can not.
Therefore, an object of the present invention is to efficiently diagnose a plurality of parallel steps.

The abnormality symptom diagnostic device of the present invention acquires measurement values of a plurality of parallel processes in which the same type of apparatus performs the same type of processing in time series for each parallel process, and obtains the acquired measurement values of the plurality of parallel processes in one process And a classification unit that classifies the measurement value regarded as the measurement value in one process by the pre-processing unit.
Other means will be described in the form for carrying out the invention.

  According to the present invention, diagnosis of a plurality of parallel steps can be performed efficiently.

It is a figure explaining the composition of an abnormality sign diagnostic device. It is a figure explaining the composition of a plant. It is a figure explaining classification of measurement value. It is a figure explaining real time and virtual time. It is a figure explaining measurement value information. It is a figure explaining the measurement value information after rearrangement. It is a figure explaining the effect of virtual time and virtual process introduction. It is a figure explaining the effect of virtual time and virtual process introduction. It is a figure explaining clustering using an adaptive resonance theory. It is a figure explaining abnormality and contribution. It is a flowchart of a processing procedure. It is a figure explaining abnormality degree and contribution degree information. It is a figure explaining diagnostic result information.

  Hereinafter, modes for carrying out the present invention (referred to as “the present embodiment”) will be described in detail with reference to the drawings and the like. The present embodiment is an example of diagnosing a plant that produces a chemical product. However, the present invention is also applicable to other objects.

(Configuration of abnormal sign diagnostic equipment)
The configuration of the abnormality symptom diagnosis device 1 will be described with reference to FIG. The abnormality symptom diagnosis device 1 is a general computer, and includes a central control device 11, an input device 12, an output device 13, a main storage device 14, an auxiliary storage device 15, and a communication device 16. These are mutually connected by a bus. The auxiliary storage device 15 stores measurement value information 31, abnormality degree / contribution degree information 32, and diagnosis result information 33.

  The preprocessing unit 21, the classification unit 22, and the post-processing unit 23 in the main storage device 14 are programs. In the following, when an operation subject is described as "○○ part is", it reads information used by the central control unit 11 from the auxiliary storage device 15 for processing in the "○○ part" and stores it in the main storage device 14 In addition, it means performing the processing to be described later. The abnormality symptom diagnosis device 1 is usually disposed in a plant monitoring center (not shown) or the like.

  The plant 2 of the present embodiment is a set of equipment for producing chemical products, and has a plurality of parallel steps (details will be described later). The measurement / control apparatus 3 acquires measurement values such as flow rate, temperature, pressure, etc. from sensors etc. arranged in various places of the plant 2, and directly outputs measurement values through the network 4 or without using the network 4. It transmits to the diagnostic device 1. The measurement and control device 3 controls the opening degree and the like of the valve disposed in the pipe of the plant 2 so that the measured values of the flow rate, the temperature, the pressure and the like become their target values.

(plant)
The configuration of the plant 2 will be described with reference to FIG. The plant 2 reacts the raw material A and the raw material B as a whole to produce a product C. In more detail, the process PR1, the process PR2, and the process PR3 share this production in parallel. A tank 41 for storing the raw material A and a tank 42 for storing the raw material B are present upstream of the process PR1, the process PR2 and the process PR3, and a tank 43 for storing the product C is present downstream.

  Focus on the process PR1. The raw material A is sent from the tank 41 to the reactor 46 via the pump 44A and the valve 45A. The raw material B is sent from the tank 42 to the reactor 46 via the pump 44B and the valve 45B. The reactor 46 reacts the raw material A and the raw material B to produce a product C. The product C is sent to the tank 43. The reaction rate of the reactor 46 varies with the temperature of the reactor 46. The reactor 46 has a cooling device (not shown), and the flow rate of the cooling water is controlled by the opening degree of the valve 47.

A flow meter F1A is disposed in a pipe between the valve 45A and the reactor 46. The flow meter F1A measures the flow rate of the raw material A, and based on the measured value, the measurement / control device 3 (FIG. 1) controls the opening degree of the valve 45A. A flow meter F1B is disposed in a pipe between the valve 45B and the reactor 46. The flow meter F1B measures the flow rate of the raw material B, and based on the measured value, the measurement / control device 3 controls the opening degree of the valve 45B. The pressure gauge P1 measures the pressure of the reactor 46. A thermometer T1 is disposed at the outlet of the reactor 46. The thermometer T1 measures the temperature at the outlet of the reaction device 46, and the measurement / control device 3 controls the opening degree of the valve 47 based on the measured value. Then, the flow rate of the cooling water flowing into the reactor 46 is changed.
The same applies to steps PR2 and PR3.

(Identical multiple parallel processes)
As is apparent from the above description, the paralleling process in the present embodiment is a plurality of processes having the following features, and is also referred to as "parallel process" hereinafter.
Each of the steps performs the same type of treatment (for example, treatment of reacting the raw material A and the raw material B to produce the product C).
Each of the processes has the same type of equipment (eg, pumps, valves and reactors).
Each of the steps is arranged in parallel with the other steps.
-In each of the steps, the sensor obtains the same state quantities (e.g. flow rate, pressure and temperature).

(Classification of measurement values)
Classification of measurement values will be described with reference to FIG. The abnormality predictor diagnostic apparatus 1 according to the present embodiment classifies measurement values into a plurality of categories using an adaptive resonance theory which is one of clustering techniques. Classification using adaptive resonance theory is also described in Patent Document 1. The measurement values are generally time series vectors. Each component of the vector is a measured value of flow rate, pressure, temperature, and the like. The abnormality symptom diagnostic device 1 draws points indicating these vectors in a multi-dimensional space, and classifies those points whose positions are close to each other into the same category.

  The horizontal axis in FIG. 3 is time in the upper view 51 and the lower view 52. The vertical axis in FIG. 3 is a measured value in the upper figure 51, and is a category in the lower figure 52. The time on the horizontal axis is divided into a normal period in which the plant state is known to be normal, and a diagnostic period in which it is not known whether the plant state is normal. The normal period is also a period for collecting a sample of measurement values that is a basis for comparing measurement values in a diagnosis period. In that sense, the normal period is also referred to as the "learning" period. In the upper diagram 51, a solid line graph FA showing changes in the flow rate of the raw material A in a certain process and a broken line graph FB showing the change in the flow rate of the raw material B in the relevant process are drawn. Reactor pressure and outlet temperature are also measured simultaneously. However, in order to simplify the explanation, these are omitted here.

The graph FA and the graph FB go up and down between a warning upper limit value and a warning lower limit value common to both. When the range from the warning lower limit value to the warning upper limit value is divided into “large” and “small” with the intermediate value as the boundary line, four combinations of “large” and “small” are assumed as follows: It can be done.
(FA, FB) = (Large, Small), (Large, Small), (Large, Large), (Large, Large)

  The combination was “(FA, FB) = (large, small)” at six points in period 53 of the normal period. Therefore, the abnormality symptom diagnosis device 1 classifies this combination into "category 1", and draws six "◆" which are classification results in the following figure 52. The combination was “(FA, FB) = (small, small)” at four points in period 54 of the normal period. Therefore, the abnormality symptom diagnosis device 1 classifies this combination into "category 2", and draws four classifications "◆" as the classification result in the following diagram 52. The combination was “(FA, FB) = (small, large)” at two time points 55 of the normal period. Therefore, the abnormality symptom diagnosis device 1 classifies this combination into "category 3", and draws two classifications "◆" in the following figure 52.

  In the normal period, the abnormal sign diagnostic device 1 classifies the measurement value into one of category 1, category 2 and category 3. The difference in measured values among these categories is attributable to differences in operating conditions, operating environment, and the like.

  The combination was “(FA, FB) = (small, small)” at three points in period 56 of the diagnosis period. Therefore, the abnormality symptom diagnosis device 1 classifies this combination into the known (learned) "category 2", and draws three classifications "◆" which are classification results in the following diagram 52. The combination was “(FA, FB) = (large, large)” at three points in period 57 of the diagnosis period. This combination has not appeared in the normal period. Therefore, the abnormality symptom diagnostic device 1 classifies this combination into a new "category 4", and draws three "■" as classification results in the following diagram 52. In addition to classifying the combination of measured values based on the magnitude relationship between the two types of measured values as described above, the abnormality symptom diagnostic device 1 classifies the combination of measured values based on the difference between the two types of measured values. It is also good.

  In the existing art, such classification is performed for each of the parallel steps. For convenience of explanation, FIG. 4 to FIG. 7 are skipped and the process temporarily proceeds to FIG. The upper left view 81, the left middle view 82, and the lower left view 83 of FIG. 8 illustrate the relationship between the two types of measurement values in the upper view 51 of FIG. 3 and the four categories in the lower view 52. Among them, the upper left diagram 81 corresponds to the process PR1 in FIG. 2, the middle left diagram 82 corresponds to the process PR2 in FIG. 2, and the lower left diagram 83 corresponds to the process PR3 in FIG.

  The horizontal axis of the coordinate plane of the upper left FIG. 81 of FIG. 8 is the flow rate FA of the raw material A, and the vertical axis is the flow rate FB of the raw material B. On the coordinate plane, four spheres (clusters) indicating category 1, category 2, category 3 and category 4 are drawn. The "sphere" in the present embodiment also includes a two-dimensional circle. Among them, each of the spheres indicating category 1, category 2 and category 3 includes many points (not shown) indicating measurement values "(FA, FB)" in a normal period. The sphere indicating category 4 includes a point “◎” indicating the measured value in the diagnosis period. The user (maintenance person of the plant) of the abnormality symptom diagnostic device 1 sees the upper left diagram 81 and knows that an unknown state occurs in the process PR1. The user naturally wants to know the state of the process PR2 and the process PR3. However, in the existing technology, in order to do so, the screen has to be switched for each parallel process. It is convenient if there is a right diagram 84 in which all processes can be listed. Description returns to FIG.

(Real time and virtual time)
The real time and the virtual time will be described along FIG. The actual time is an actual time when the measurement value is measured by a sensor or the like. In the present embodiment, four types of measurement values are measured simultaneously in three parallel steps at a certain actual time. The virtual time is a virtual time on information processing, which is used for the purpose of simultaneously diagnosing parallel processes by the abnormality predictor diagnostic apparatus 1 of the present embodiment. The actual time is represented, for example, by "t + one digit" such as "t1", and the virtual time is represented by "t + two digits" such as "t11".

  Now, the flow rate of the raw material A of the process PR1 actually acquired at a certain time point t1 of the diagnosis period is represented as "F1A (t1)". Here, "t1" is an actual time. Similarly, the flow rate of the raw material B of the process PR1 actually acquired at time t1 is represented as "F1B (t1)". The pressure of the reactor of step PR1 actually acquired at time t1 is represented as "P1 (t1)". The temperature at the outlet of the reactor of step PR1 actually acquired at time t1 is represented as "T1 (t1)".

  Vectors “(F1A (t1), F1B (t1), P1 (t1), T1 (t1))” having these components as components are called measurement value vectors. Similarly, the measurement value vectors “(F2A (t1), F2B (t1), P2 (t1), T2 (t1))” of the process PR2 at time t1 are defined. Furthermore, the measurement value vectors “(F3A (t1), F3B (t1), P3 (t1), T3 (t1))” of the process PR3 at the time t1 are also defined. Measurement value vectors are similarly defined for real times other than t1. The horizontal axis of the upper diagram of FIG. 4 is the actual time. Each of dashed rectangles in the upper portion of FIG. 4 is defined as a measurement value vector.

  The abnormality symptom diagnostic device 1 of the present embodiment replaces the actual time "t1" included in the measurement value vector 63 of the process PR1 in the upper diagram of FIG. 4 with the virtual time "t11". Similarly, abnormality predictor diagnostic apparatus 1 replaces actual time “t1” included in measured value vector 64 in process PR2 with virtual time “t12”, and substitutes actual time “t1” included in measured value vector 65 in process PR3. Replace with virtual time "t13". The abnormality symptom diagnostic device 1 performs the same processing for the actual times "t2" and "t3".

  Thereafter, the abnormality symptom diagnosis device 1 arranges the measurement value vectors, in which the actual time is replaced with the virtual time, in time series in the lower diagram 62 of FIG. 4. The horizontal axis of the lower diagram 62 of FIG. 4 is a virtual time. At this time, it is assumed that “t1 = t11 <t12 <t13 <t2 = t21 <t22 <t23 <t3 = t31 <t32 <t33” holds. In the lower diagram 62 of FIG. 4, the length between each of t11, t12, t13, t21, t22, t23, t31, t32 and t33 is arbitrary and it is not necessary to equal each other, but it is equally spaced. Desirable (details below).

(Measurement value information)
The measurement value information 31 will be described with reference to FIG. In the measurement value information 31, in association with the process ID stored in the process ID column 101, the measurement value ID is in the measurement value ID column 102, and the type of measurement value is in the measurement value type column 103. The actual time is stored in the measurement value column 105, and the measurement value is stored in the measurement value column 105.
The process ID of the process ID column 101 is an identifier that uniquely identifies a parallel process.
The measurement value ID in the measurement value ID column 102 is an identifier that uniquely identifies the measurement value. The measurement value ID of this embodiment is such that it can be immediately known which kind of measurement value of which process at which real time it is such as “F1A (t1)”.

The type of measurement value in the measurement value type column 103 is any one of “raw material A flow rate”, “raw material B flow rate”, “reactor pressure” and “reactor outlet temperature”. Naturally, the type of measurement depends on the content of the plant. Those listed here are just an example.
The real time in the real time column 104 is the above-mentioned real time. The flow rate is defined as volume or mass per unit time. That is, to measure the flow rate, a certain time width (for example, 5 seconds) is required. The actual time here indicates the time of the start of such a time width.
The measurement values in the measurement value column 105 are the measurement values themselves measured by a sensor or the like. “#” Indicates different values in an abbreviated manner. In addition, it is preferable that the abnormal sign diagnostic apparatus 1 normalizes the measured value which the same kind of sensor which straddles a several parallel process measured, for example in the range of 0-1. The reason is that the measurement values of the same position of the same type of device may differ in the range of values that can be taken in each parallel process.

For example, the following can be understood from FIG.
There are three parallel processes: PR1, PR2 and PR3.
・ 4 kinds of measurement value in each process such as F1A (t1), F1B (t1), P1 (t1) and T1 (t1) every one minute from 10:00 on October 10, 2017 Are measured at the same time.

(Measurement value information after sorting)
The rearranged measurement value information 31b will be described with reference to FIG. In the rearranged measurement value information 31b, in association with the process ID stored in the process ID column 111, the measurement value ID is virtual in the measurement value ID column 112, and the type of measurement value is virtual in the measurement value type column 113. The virtual time is stored in the time column 114, the measured value is stored in the measured value column 115, and the original measured value ID is stored in the original measured value ID column 116.

  The process ID of the process ID column 111 is the same as the process ID of FIG. 5. However, here, the process ID of all the records (rows) is “VPR1” indicating “virtual process”. The term "virtual process" indicates that the process PR1, the process PR2, and the process PR3 in which the same type of apparatus performs the same process are collectively regarded as one process.

The measurement value ID in the measurement value ID column 112 is the same as the measurement value ID in FIG. 5. However, the measurement value ID here is such as “F1A (t11)”, and it is possible to immediately know which kind of measurement value of which process at which virtual time.
The type of measurement value of the measurement value type column 113 is any one of “raw material A flow rate”, “raw material B flow rate”, “reactor pressure” and “reactor outlet temperature”.

The virtual time in the virtual time column 114 is the virtual time described above. The virtual time is, as described above, a virtual time for causing the abnormal sign diagnostic apparatus 1 to process information, and the abnormal sign diagnostic apparatus 1 regards that the measured value is acquired at the virtual time.
The measured values in the measured value column 115 are the same as the measured values in FIG.
The original measurement value ID in the original measurement value ID column 116 is the measurement value ID of the record in FIG. 5 to which the record corresponds (the measurement value ID before the actual time is replaced with the virtual time).

(Recognition of parallel processes)
The abnormality symptom diagnostic device 1 reads the type (raw material A flow rate, raw material B flow rate, reactor pressure and reactor outlet temperature) of the measured value of the measured value information 31 (FIG. 5). Then, as a result of recognizing that these are the same across the processes PR1, PR2 and PR3, the abnormal sign diagnostic device 1 recognizes that these are parallel processes, and gives them the process ID “VPR1” of the virtual process. Give. In addition, the abnormality prognostic diagnostic device 1 may recognize the parallel process by accepting that the user inputs, for example, a plurality of measurement value IDs via the input device 12. In FIG. 6, a double line is drawn at the boundary of the actual time (t1, t2 and t3) included in the original measurement value ID. This is solely for the ease of understanding of the description, and has nothing to do with the processing content of the abnormality symptom diagnostic device 1.

By comparing FIGS. 5 and 6, for example, the following can be understood.
The real time “t1” for the process PR1 is replaced with the virtual time “t11”. "T1" is "10:00:00 on October 10, 2017". "T11" is also "10:00:00 on October 10, 2017". The abnormality prediction device 1 selects any one of the parallel processes, and makes the virtual time coincide with the real time for the process.

The real time "t1" for the process PR2 is replaced with the virtual time "t12". While "t1" is "10:00 10:00 on Oct. 10, 2017", "t12" is "10:00 on Oct. 10, 2017". The virtual time is 20 seconds behind the real time. This delay will be described later.

The real time "t1" for the process PR3 is replaced with the virtual time "t13". While “t1” is “10:00 10:00 on October 10, 2017”, “t13” is “10:00:00 on October 10, 2017”. The virtual time is 40 seconds behind the real time. This delay will also be described later.

(Conversion time difference)
The difference between the virtual time and the real time will be described, leaving the explanation of FIGS. 5 and 6 once. The difference obtained by subtracting the real time from the virtual time is called "conversion time difference". In the above-mentioned example, the abnormal sign diagnostic apparatus 1 sets the conversion time difference of each process as “(PR1, PR2, PR3) = (0 seconds, 20 seconds, 40 seconds)”. Preferably, the conversion time difference of any one step is "0 seconds". Moreover, it is preferable that the conversion time difference between the other two steps is not the same, and furthermore, "(0 seconds, 10 seconds, 20 seconds)", "(0 seconds, 20 seconds, 40 seconds)" It is more desirable that the intervals be even.

  However, if, for example, the interval between the real time t1 and the next real time t2 is 60 seconds, and the conversion time difference becomes “(0 seconds, 40 seconds, 80 seconds)”, the virtual times of all steps are actually It can not be fit between time t1 and the next actual time t2. Therefore, in general, the abnormal sign diagnostic apparatus 1 converts the conversion time difference into “(PR1, PR2, PR3,...) = (0, p / q, 2 p / q, 3 p / q,. It is most preferable to Where p is the real time interval and q is the number of parallel steps. When p = 60 seconds and q = 3, the most preferable conversion time difference is "(PR1, PR2, PR3) = (0 seconds, 20 seconds, 40 seconds)" as described above. This conversion time difference is called "standard time difference". The above-mentioned delay is the result of the abnormal sign diagnostic device 1 applying the standard time difference.

Returning to the description of FIGS. 5 and 6, for example, the following can be understood.
The real time “t2” for the process PR1 is replaced with the virtual time “t21”. "T2" is "10:01:00 on October 10, 2017". "T21" is also "10:01:00 on October 10, 2017".

The real time “t2” for the process PR2 is replaced with the virtual time “t22”. While "t2" is "10:01:00 on October 10, 2017", "t22" is "10:01 on October 10, 2017". The virtual time is 20 seconds behind the real time. This delay is a result of the abnormal sign diagnostic device 1 applying the standard time difference.

The real time "t2" for the process PR3 is replaced with the virtual time "t23". While "t2" is "10:01:00 on October 10, 2017", "t23" is "10:01 on October 10, 2017". The virtual time is 40 seconds behind the real time. This delay is also the result of the abnormal sign diagnostic device 1 applying the standard time difference.

The real time "t3" for the process PR1 is replaced with the virtual time "t31". “T3” is “10:02:00 on October 10, 2017”. "T31" is also "10:02:00 on October 10, 2017".

The real time "t3" for the process PR2 is replaced with the virtual time "t32". While "t3" is "10:02 10 seconds on October 10, 2017", "t32" is "10:02 10 seconds on October 10, 2017". The virtual time is 20 seconds behind the real time. This delay is a result of the abnormal sign diagnostic device 1 applying the standard time difference.

The real time "t3" for the process PR3 is replaced with the virtual time "t33". While "t3" is "10:02:00 on October 10, 2017", "t33" is "10:02:40 on October 10, 2017". The virtual time is 40 seconds behind the real time. This delay is also the result of the abnormal sign diagnostic device 1 applying the standard time difference.

(Effect of virtual time and virtual process introduction)
The effects of virtual time and virtual process introduction will be described with reference to FIGS. 7 and 8. Now, it is assumed that the abnormal sign diagnostic device 1 reads the measurement value information 31 of FIG. 5 according to the existing technology. Then, the abnormality prognostic diagnostic device 1 recognizes that there are three processes and the same four types of time series measurement values exist for each process. That is, the abnormality symptom diagnosis device 1 recognizes that the information as shown in the upper left view 71, the middle left view 72, and the lower left view 73 of FIG. 7 is present. The upper left FIG. 81, the left middle FIG. 82, and the lower left of FIG. 8 show the results of diagnosis of the measurement values (clustering using adaptive resonance theory) in the diagnosis period based on the existence of such information as the abnormality predictor diagnostic device 1. 83. Note that these figures show, for the purpose of illustration only, the simplest two-dimensional plane in a multi-dimensional space. The horizontal axis is the flow rate FA of the raw material A, and the vertical axis is the flow rate FB of the raw material B (the same applies to FIG. 84 to the right).

  The position and radius of the sphere indicating category 1 in the upper left FIG. 81 is not limited to be the same as the position and radius of the sphere indicating category 1 in the left middle FIG. It is not necessarily the same as the radius. The same is true for categories 2-4. That is, even though it appears that all of the measured value ““ ”of process PR1, the measured value“ ○ ”of process PR2 and the measured value“ ● ”of process PR3 are included in the same category, they are the same abnormality (or It is unknown whether they belong to the same unknown state).

  On the other hand, it is assumed that the abnormality prognostic diagnostic device 1 reads the post-sorted measurement value information 31b of FIG. Then, the abnormality prognostic diagnostic device 1 recognizes that there is only one process and that four types of time series measurement values exist for the one process. That is, the abnormality symptom diagnosis device 1 recognizes that the information as shown in the right diagram 74 of FIG. 7 exists. The virtual process is a process recognized by the abnormality predictor diagnostic apparatus 1 in this manner.

  The right view 74 of FIG. 7 is at first glance similar to the upper left view 71, the middle left view 72 and the lower left view 73. However, each of the upper left FIG. 71, the left middle FIG. 72, and the lower left FIG. 73 has measurement values of only one process, whereas the right FIG. 74 has measurement values of three processes. That is, for example, the right diagram 74 has all of the measurement values at the times indicated by the dashed line in the upper left diagram 71, the one-dot chain line in the left middle diagram, and the two-dot chain line in the lower left diagram 73 (data amount of time axis is Tripled). Note that “*” in the right diagram 74 collectively represents “1”, “2” and “3”.

  The right figure 84 of FIG. 8 shows the result of the diagnosis of the measured value in the diagnosis period by the abnormality predictor diagnostic device 1 on the premise of the presence of such information. In the right diagram 84, the measured value "◎" of the process PR1 in the diagnosis period, the measured value "○" of the process PR2, and the measured value "●" of the process PR3 are drawn in the same space (plane). Here, these three points belong to the sphere indicating category 4. The user knows that an unknown state has occurred in three parallel steps. The sphere indicating category 1 in the right diagram 84 of FIG. 8 includes all the spheres indicating the category 1 in the upper left diagram 81, the middle left diagram 82, and the lower left diagram 83 of FIG. The same is true for categories 2-4.

(Clustering using adaptive resonance theory)
Clustering using adaptive resonance theory will be described along FIG. The abnormality symptom diagnostic device 1 draws points representing multidimensional measurement values in a multidimensional space. A virtual time is associated with the measurement value, and the number of drawn points corresponds to the number of virtual times. The two-dimensional plane of FIG. 9 is also an example of the simplest explanatory purpose in the multidimensional space, having the flow rate FA of the raw material A on the horizontal axis and the flow rate FB of the raw material B on the vertical axis. The abnormality symptom diagnostic device 1 classifies a plurality of points indicating measurement values acquired in normal time into category 1, category 2 and category 3. These points are any of the measurement value “◎” acquired in the process PR1, the measurement value “○” acquired in the process PR2, and the measurement value “●” acquired in the process PR3.

  Also in the diagnosis period, the abnormality symptom diagnostic device 1 draws points (◎, 及 び, and 示 す) indicating measurement values in a multidimensional space. These points may belong to any of category 1, category 2 or category 3 and may not belong to any of them. In the example of FIG. 9, as the virtual time passes, ◎, ○, and ● are repeatedly drawn in this order. These points belong to category 4 which was unknown in the normal period. The user is aware of the occurrence of a state in which the flow rate of the raw material A and the flow rate of the raw material B are both high in the three parallel processes.

(Abnormality and contribution)
The degree of abnormality and the degree of contribution will be described along FIG. The vertical axis, the horizontal axis, and the categories 1 to 4 in FIG. 10 are the same as in FIG. 9. If the measured value 85 85 in the diagnosis period belongs to the category 4 which is unknown in the normal period, the user wants to know the following.
-From what known category the measurement value is abnormal to what extent.
Of the multiple types of measurement values (flow rate of raw material A, flow rate of raw material B, etc.), what is the one that contributes most to the abnormality?

Therefore, the abnormality predictor diagnostic apparatus 1 first performs the following processing to calculate the degree of abnormality.
(1) The abnormality symptom diagnostic device 1 specifies a representative point カ テ ゴ リ 86a of category 1, a representative point カ テ ゴ リ 86b of category 2, and a representative point 86 86c of category 3. Here, the representative point (representative value) is, for example, a median value (maximum value-minimum value), an average value (gravity center) or the like of measurement values belonging to each category.
(2) The abnormality symptom diagnosis device 1 calculates the distance between each of the specified representative points と and the measurement value 85 85, and sets these as candidates for the degree of abnormality. The number of anomaly degree candidates matches the number of known categories.
(3) The abnormality symptom diagnosis device 1 sets the smallest one of the calculated abnormality degree candidates as the “abnormality degree”. In the example of FIG. 10, the distance between the measured value 85 85 and the representative point カ テ ゴ リ 86 c of the category 3 is the abnormality degree 87.

Subsequently, the abnormality symptom diagnosis device 1 performs the following processing to calculate the degree of contribution.
(4) The abnormality symptom diagnostic device 1 decomposes the degree of abnormality 87 into a component in the horizontal axis direction and a component in the vertical axis direction. Here, “decompose into components in the axial direction” means finding the length of projection of the degree of abnormality 87 on each axis.
(5) The abnormality symptom diagnosis device 1 sets the length 88 a as a contribution of the flow rate of the raw material A, and sets the length 88 b as a contribution of the flow rate of the raw material B.

  In general, the degree of anomaly in multi-dimensional space can be projected to any coordinate axis in multi-dimensional space. That is, the number of contributions corresponds to the order of the multidimensional space. As described above, the measurement values are represented as measurement value vectors each having the value of each axis of the multidimensional space as a component. Therefore, it can be said that the abnormality degree is the magnitude of a vector (subtraction result vector) as a result of subtracting the position vector of the representative point カ テ ゴ リ of the category from the measurement value vector. Furthermore, it can be said that the degree of contribution is an arbitrary component of the subtraction result vector itself.

(Processing procedure)
The processing procedure will be described with reference to FIG. 12 and 13 will be appropriately referred to during the explanation. As a premise of starting the processing procedure, the auxiliary storage device 15 stores measurement value information 31 (FIG. 5) for a normal period and a diagnosis period.
In step S201, the pre-processing unit 21 of the abnormality symptom diagnosis device 1 learns a category at the normal time. Specifically, first, the preprocessing unit 21 reads the measurement value information 31 from the auxiliary storage device 15.

Second, the preprocessing unit 21 receives an input of a normal period through the input device 12 by the user. The normal period is a period during which the plant is known to be normal. If a period in which a certain process is known to be abnormal is partially included in a period in which a normal period is desired, the user inputs a period excluding that period as a normal period. In order to simplify the following description, it is assumed that there is no period excluded here.
Thirdly, the preprocessing unit 21 extracts a record corresponding to the normal period received in the “second” of the step 201 from the measurement value information 31 read out in the “first” of the step S201.

Fourth, the preprocessing unit 21 rearranges the records extracted in the “third” of step S201 according to the above-described method, and generates post-rearranged measurement value information 31b (FIG. 6).
Fifth, the preprocessing unit 21 draws points representing measurement values in a multi-dimensional space based on the post-rearranged measurement value information 31 b created in the “fourth” of step S 201, and forms a plurality of spheres (clusters) Create At this stage, the preprocessing unit 21 is drawing category 1, category 2 and category 3 of FIG. 9 in the multidimensional space.

In step S202, the preprocessing unit 21 acquires a measurement value of a diagnosis period. Specifically, first, the preprocessing unit 21 receives an input of a diagnosis period via the input device 12 by the user. The diagnosis period is a period during which it is not known whether each step is normal, and is usually a period following the normal period.
Second, the preprocessing unit 21 extracts a record corresponding to the diagnosis period received in the “first” of step 202 from the measurement value information 31 read out in the “first” of step S201.

  In step S203, the preprocessing unit 21 rearranges the measurement values of the diagnosis period. Specifically, the preprocessing unit 21 rearranges the records extracted in the “second” of step S202 according to the above-described method, and creates post-rearranged measurement value information 31 b (FIG. 6). At this time, the preprocessing unit 21 associates a plurality of different virtual times with those having the same actual time among the measurement values acquired from each of the parallel processes (see FIG. 4). . In FIG. 4, for example, virtual times t11, t12 and t13 are included in a predetermined time width (time width between real times t1 and t2).

In step S204, the classification unit 22 of the abnormal symptom diagnostic device 1 classifies the measurement value of the diagnosis period. Specifically, first, the classification unit 22 draws a point indicating the measurement value in the multi-dimensional space based on the post-reorder measurement value information 31 b created in step S203. Then, the classification unit 22 extracts the “post-processing target point” from the drawn points according to any one of the following criteria.
<Criteria 1> Among the drawn points, points not belonging to any of the known category 1, category 2 and category 3 are taken as post-processing target points.
<Reference 2> Let all points drawn be post-processing target points.

  Second, the classification unit 22 displays the multidimensional space on the output device 13. FIG. 9 is an example in which the multidimensional space is a two-dimensional plane. At this time, it is desirable that the classification unit 22 also display the anteroposterior relation of the process to which the post-processing object point belongs and the virtual time together. In the example of FIG. 9, the classification unit 22 changes the shape of the point for each process ((, 及 び, and ●), and draws an arrow from the point where the virtual time is early to the point where the virtual time is late.

  In step S205, the post-processing unit 23 of the abnormality symptom diagnosis device 1 calculates the degree of abnormality and the degree of contribution. Specifically, the post-processing unit 23 calculates the degree of abnormality and the degree of contribution for each post-processing object point by the method described above with reference to FIG.

In step S206, the post-processing unit 23 displays the abnormality degree / contribution degree information 32 (FIG. 12). Specifically, the post-processing unit 23 displays the abnormality degree / contribution degree information 32 on the output device 13 for each post-processing target point.
The actual time and / or the virtual time are displayed in the measurement time field 91a.
The category column 91b displays the number of the category to which the post-processing object point belongs. When the criteria 1 are followed, the numbers here are numbers other than 1, 2 and 3.
In the normal / abnormal column 91 c, “normal” or “abnormal” is displayed. When the criterion 1 is followed, "abnormal" is always displayed here.

The type of measurement value (the largest of the components of the subtraction result vector) having the largest degree of contribution to the degree of abnormality is displayed in the abnormality cause column 91 d. For example, when “F1A” is displayed in the column, it is understood that an abnormality has occurred in the process PR1 and it is highly possible that a change in the flow rate of the raw material A from the past example is the cause of the abnormality. .
An abnormality degree is displayed in the abnormality degree column 91e. The degree of anomaly is the square root of the sum of squares of the components of the subtraction result vector that differ in unit. Thus, the degree of anomaly does not have its own unit. "[-]" Indicates this.

In the contribution degree ranking column 92a, the order counted from the one with the largest contribution degree is displayed. When the number of dimensions of the multidimensional space (the number of types of measurement values) is n, the order is 1, 2, 3,.
The type of measurement value indicated by the axis onto which the degree of abnormality is projected is displayed in the measurement point column 92b.
The degree of contribution is displayed in the degree of contribution column 92c.
The measured value column 92 d shows the measured value itself.
For example, when there are ten post-processing target points, the post-processing unit 23 displays the abnormality degree / contribution degree information 32 of FIG. 12 ten times in order of the virtual time.

  In step S207, the post-processing unit 23 displays the diagnostic result information 33 (FIG. 13). Specifically, first, the post-processing unit 23 displays the diagnostic result information 33 on the output device 13. The diagnosis result information 33 is information in which the degree-of-abnormality / contribution information 32 displayed for each post-processing object point in step S206 is combined into one.

The category column 93a, the normal / abnormal column 93b, and the abnormality cause column 93d of the diagnosis result information 33 correspond to the column of the same name in the abnormality degree / contribution degree information 32 of FIG.
“Learning” or “diagnosis” is displayed in the learning / diagnosis column 93 c. "Learning" indicates that the category of the record has been created for a normal (learning) period. "Diagnostic" indicates that the category of the record has been created for the diagnostic period.
In the decider column 93e, a user name (or an identifier) that confirms the content of the record is displayed.
In the finalized date column 93f, the date when the content of the record is confirmed is displayed.

  The record of the diagnosis result information 33 is divided into an upper part and a lower part with the double line 95 as a boundary. The upper part records 94a to 94c are for the user to confirm the category created in the normal (learning) period. In the normal / abnormal column 93b, "normal" is always displayed. In the abnormality cause column 93 d, “−” (no data) is always displayed. The post-processing unit 23 initially displays the determiner column 93 e and the determination date column 93 f of these records as blanks. The user visually recognizes these and confirms that there are three normal categories, and then inputs his / her identifier in the determiner column 93e of these records via the input device 12, and the determination date column 93f Enter the fixed date in.

  Each of the lower records 96a to 96d corresponds to each post-processing target point in the diagnosis period. First, the post-processing unit 23 displays the determiner column 93 e and the determination date column 93 f of these records as blanks, and displays “diagnosis” in the learning / diagnosis column 93 c. When the criterion 2 is followed, the post-processing unit 23 displays “normal” or “abnormal” in the normal / abnormal column 93 b of these records.

  If the post-processing unit 23 displays "abnormal" in the normal / abnormal column 93b, the post-processing unit 23 displays numbers other than "1", "2" and "3" in the category column 93a of the record. To indicate the type of measurement with the highest contribution. When the post-processing unit 23 displays "normal" in the normal / abnormal column 93b, the post-processing unit 23 displays "1", "2" or "3" in the category column 93a of the record, and displays "normal" in the abnormality cause column 93d. Display "-".

  After visually recognizing the records 96a to 96d, the user inputs his / her identifier into the determined person column 93e of these records via the input device 12, and inputs the determined date in the determined date column 93f. The user can change the data displayed in the normal / abnormal column 93 b and the abnormal cause column 93 d as needed, as in the following example.

<Example 1: Record 96b>
The post-processing unit 23 displays “normal” in the normal / abnormal column 93 b. However, the user once wanted to put the diagnosis result on hold. At this time, the post-processing unit 23 accepts that the user changes “normal” to “pending”.

<Example 2: Record 96d>
The post-processing unit 23 displays “abnormal” in the normal / abnormal column 93 b. However, the user has determined that the diagnosis result is not caused by the failure of the virtual process but is caused by a change in the environment in which the virtual process is arranged. At this time, the post-processing unit 23 accepts that the user changes “normal” to “new” and deletes the data displayed in the abnormality cause column 93 d. "New" here has a meaning of "normal" of a new category.

  In addition to the example 1 and the example 2, when the measured value is not included in any of the known categories, the post-processing unit 23 may first display “careful” in the normal / abnormal column 93 b. Then, the user may change the "careful" to "new", "normal" or "abnormal".

The explanation of the screen display has been lengthened, but it returns to the processing procedure. Second, the post-processing unit 23 accepts that the user performs an arbitrary operation (for example, pressing of an “end” button not shown) via the input device 12.
Thirdly, the post-processing unit 23 stores the diagnostic result information 33 after receiving the input of the identifier of the decider and the like and the other changes in the auxiliary storage device 15, and ends the processing procedure.

(Effect of this embodiment)
The effects of the abnormal sign diagnostic apparatus of the present embodiment are as follows.
(1) The abnormality symptom diagnosis device can classify an abnormality of the same cause into the same category for a plant composed of a plurality of parallel processes.
(2) The abnormality sign diagnosis system enables maintenance personnel to efficiently diagnose a plurality of parallel processes.
(3) The abnormality sign diagnosis system facilitates maintenance personnel to take measures to restore the parallel process to the normal state.
(4) The abnormality symptom diagnosis device can quantitatively show the basis of the diagnosis result to maintenance personnel.
(5) The abnormality symptom diagnosis device can show maintenance results in a list format to maintenance personnel.

  The present invention is not limited to the embodiments described above, but includes various modifications. For example, the above-described embodiments are described in detail to explain the present invention in an easy-to-understand manner, and are not necessarily limited to those having all the described configurations. Also, part of the configuration of one embodiment can be replaced with the configuration of another embodiment, and the configuration of another embodiment can be added to the configuration of one embodiment. In addition, with respect to a part of the configuration of each embodiment, it is possible to add, delete, and replace other configurations.

Further, each of the configurations, functions, processing units, processing means, etc. described above may be realized by hardware, for example, by designing part or all of them with an integrated circuit. In addition, each configuration, function, and the like described above may be realized by software by a processor interpreting and executing a program that realizes each function. Information such as a program, a table, and a file for realizing each function can be placed in a memory, a hard disk, a recording device such as a solid state drive (SSD), or a recording medium such as an IC card, an SD card, or a DVD.
Further, control lines and information lines indicate what is considered to be necessary for the description, and not all control lines and information lines in the product are necessarily shown. In practice, almost all the configurations may be considered to be mutually connected.

DESCRIPTION OF SYMBOLS 1 abnormality precursor diagnostic apparatus 2 plant 3 measurement and control apparatus 4 network 11 central control apparatus 12 input device 13 output device 14 main memory 15 auxiliary storage device 16 communication device 21 pre-processing unit 22 classification unit 23 post-processing unit 31 measured value information 32 anomaly degree / contribution information 33 diagnostic result information

Claims (7)

Acquiring, in time series, measurement values of a plurality of parallel processes in which the same type of apparatus performs the same type of processing;
A pre-processing unit that regards measured values of the plurality of acquired parallel processes as measured values in one process;
A classification unit that classifies the measured value regarded as the measured value in one process by the pre-processing unit;
An abnormal sign diagnostic apparatus comprising: The pre-processing unit
Among the measured values acquired from each of the parallel processes, a plurality of different virtual times included in a predetermined time width are respectively associated with one having the same actual time actually acquired,
The classification unit
Considering that the measured value is obtained at the virtual time,
The abnormal sign diagnostic apparatus according to claim 1, characterized in that The pre-processing unit
Let the real time of a parallel process be the virtual time as it is,
Associating the virtual times of all parallel processes other than the parallel process in a time zone after the real time of the parallel process and before the next real time;
The abnormal sign diagnostic apparatus according to claim 2, characterized in that In a multi-dimensional space having various measurement values as axes, there is provided a post-processing unit that decomposes a distance between a representative value of past measurement values and a measurement value at a diagnosis target time into components for any axis of the multi-dimensional space. about,
The abnormal sign diagnostic apparatus according to claim 3, characterized in that The classification unit
Clustering using adaptive resonance theory,
Displaying, in the same space, time series measurement values of a plurality of parallel processes in which the same type of apparatus performs the same type of processing;
The abnormal sign diagnostic apparatus according to claim 4, characterized in that The pre-processing unit of the abnormality sign diagnosis system is
Acquiring, in time series, measurement values of a plurality of parallel processes in which the same type of apparatus performs the same type of processing;
The measured values of the plurality of parallel processes acquired as above are regarded as measured values in one process,
The classification unit of the abnormal sign diagnostic apparatus is
Classifying the measured value regarded as the measured value in one process by the pre-processing unit;
A method for diagnosing abnormality sign of abnormality prediction apparatus characterized by For the pre-processing unit of the abnormality sign diagnosis system,
Acquiring, in time series, measurement values of a plurality of parallel processes in which the same type of apparatus performs the same type of processing;
A process of regarding the measured values of the plurality of acquired parallel processes as measured values in one process is executed,
With respect to the classification unit of the abnormal sign diagnostic device,
Causing the pre-processing unit to execute a process of classifying measurement values regarded as measurement values in one process;
An abnormality precursor diagnosis program for operating an abnormality precursor diagnosis device, characterized in that

Images