JP7646392B2 - Valve operation abnormality detection system, valve operation abnormality detection method, and program - Google Patents

Description

This disclosure relates to a valve operation abnormality detection system, a valve operation abnormality detection method, and a program.

Conventionally, in various processes, there is known a technique for monitoring process quantities to be managed and judging whether an abnormality has occurred in the equipment based on the monitored process quantities. For example, Patent Document 1 describes a management method in which, in semiconductor substrate processing, a band range is set in advance for monitor data such as temperature, which is defined by a reference value and upper and lower limits relative to the reference value, and if the number of times the monitor data exceeds the band range is equal to or exceeds a threshold value, it is judged that an abnormality has occurred.

JP 2012-216697 A

In plant facilities such as integrated coal gasification combined cycle (IGCC) plants, multiple control valves are installed in various devices, pipes, storage sections, etc. Such control valves can suffer from problems such as valve wear and valve body sticking, so it is necessary to monitor whether the control valves are opening and closing normally.

Here, when the control valve to be monitored opens and closes in a specific pattern, it is possible to detect abnormalities in the opening and closing of the valve by setting in advance a band range of upper and lower limit values for a reference value of the opening in the time series data of the opening of the control valve, as in the method described in Patent Document 1. However, even when the control valve opens and closes in a specific pattern, the time series data of the opening of the control valve may fluctuate along the time axis in response to fluctuations in parameters such as the processing load of the plant equipment. In this way, when fluctuations along the time axis are allowed in the pattern of the opening and closing of the control valve, normal operation may be detected as abnormal unless a relatively large band range for the opening is set. On the other hand, setting a large band range for the opening may result in a decrease in the sensitivity of abnormality detection.

The present disclosure has been made in consideration of the above, and aims to more accurately detect abnormalities in the opening and closing operation of a control valve.

In order to solve the above-mentioned problems and achieve the object, the present disclosure includes a condition setting unit that sets an allowable time for a valve opening to reach a specified opening degree, which is a time range allowed for a control valve that is controlled to open and close according to a predetermined opening and closing pattern, and an abnormality determination unit that determines that the control valve is normal if the time it takes for the valve opening to reach the specified opening degree from a closed state is within the allowable time range, and determines that the control valve is abnormal if the time it takes for the valve opening to reach the specified opening degree is outside the allowable time range.

In order to solve the above-mentioned problems and achieve the objective, the present disclosure sets an allowable time for a control valve that is controlled to open and close according to a predetermined opening and closing pattern, which is a time range allowed for the valve opening to change from a closed state to a specified opening, and determines that the control valve is normal if the time it takes for the valve opening to change from a closed state to the specified opening is within the allowable time, and determines that the control valve is abnormal if the time it takes for the valve opening to change to the specified opening is outside the allowable time range.

In order to solve the above-mentioned problems and achieve the objective, the present disclosure causes a computer to set an allowable time for a control valve that is controlled to open and close according to a predetermined opening and closing pattern, which is an allowable time range for the valve opening to change from a closed state to a specified opening, and to determine that the control valve is normal if the time it takes for the valve opening to change from a closed state to the specified opening is within the allowable time range, and to determine that the control valve is abnormal if the time it takes for the valve opening to change to the specified opening is outside the allowable time range.

The valve operation abnormality detection system, valve operation abnormality detection method, and program disclosed herein have the effect of enabling more accurate detection of abnormalities in the opening and closing operation of a control valve.

FIG. 1 is a schematic diagram illustrating an example of a valve operation abnormality detection system according to a first embodiment. FIG. 2 is an explanatory diagram showing a storage hopper equipped with a control valve that is a target for abnormality detection. FIG. 3 is an explanatory diagram showing an example of an opening and closing operation pattern of the pressure reducing valve. FIG. 4 is an explanatory diagram showing an example in which a plurality of cycles of an opening/closing operation pattern of a pressure reducing valve are plotted side by side. FIG. 5 is an explanatory diagram showing an example in which a plurality of normal and abnormal cases are plotted side by side for the valve opening degree in one cycle of the opening and closing operation pattern of the pressure reducing valve. FIG. 6 is a flowchart illustrating an example of a valve operation abnormality detection method according to the first embodiment. FIG. 7 is a schematic diagram illustrating an example of a valve operation abnormality detection system according to the second embodiment. FIG. 8 is a flowchart showing an example of a valve operation abnormality detection method according to the second embodiment. FIG. 9 is an explanatory diagram showing an example of a map that specifies the allowable arrival time according to the processing load of the IGCC.

Below, embodiments of the valve operation abnormality detection system, valve operation abnormality detection method, and program according to the present disclosure are described in detail with reference to the drawings. Note that the present invention is not limited to these embodiments.

[First embodiment]
1 is a schematic diagram showing an example of a valve operation abnormality detection system according to a first embodiment. The valve operation abnormality detection system 100 is a system for detecting an abnormality in the opening and closing operation of a target control valve.

In this embodiment, a control valve that is the subject of abnormality detection and a facility equipped with the control valve will be described. FIG. 2 is an explanatory diagram showing a storage hopper equipped with a control valve that is the subject of abnormality detection. The storage hopper 1 is a storage tank that stores coal (pulverized coal) that is used as a fuel in the form of powder in an integrated coal gasification combined cycle power generation facility (hereinafter referred to as "IGCC"). Note that the powder is not limited to coal. Furthermore, the storage hopper 1 is not limited to one that stores powder, and may also store gas or liquid. Furthermore, the storage hopper 1 may be one that is used in plant facilities other than IGCC.

As shown in the figure, the storage hopper 1 is connected to a supply line L1, an outlet line L2, a pressurized line L3, and a depressurized line L4. The supply line L1 is a line for supplying powder to the inside of the storage hopper 1, and is provided with a supply valve 2 as a control valve for adjusting the supply. The outlet line L2 is a line for sending the powder to the next process, and is provided with an outlet valve 3 as a control valve for adjusting the sending. The pressurized line L3 and the depressurized line L4 are lines for changing the internal pressure by introducing and releasing a carrier gas (such as nitrogen) for transporting the powder into and from the storage hopper 1. As control valves for introducing and releasing the carrier air, the pressurized line L3 is provided with a pressurized valve 4, and the depressurized line L4 is provided with a depressurized valve 5.

Control valves installed in the lines connected to the storage hopper 1 that stores powder tend to malfunction, such as becoming clogged with powder (solid), more easily than those connected to gas or liquid storage tanks. For this reason, it is desirable to appropriately monitor these control valves to see whether they are opening and closing normally. In this embodiment, abnormality detection is described for the pressure reducing valve 5 as one of the control valves. Note that abnormality detection may also be performed on the supply valve 2, outlet valve 3, or pressurizing valve 4, in addition to the pressure reducing valve 5.

Figure 3 is an explanatory diagram showing an example of an opening and closing operation pattern of a pressure reducing valve. As shown in the figure, in this embodiment, the pressure reducing valve 5 is set to repeat the same opening and closing operation at regular time intervals. That is, after opening from 0% to 100%, the valve is closed from 100% to 0% at regular time intervals. In this way, this embodiment targets abnormality detection of a control valve controlled by a predetermined specific opening and closing operation pattern. Note that the pattern shown in Figure 3 is only an example, and the range of the valve opening may be determined according to the usage mode of the target control valve. In addition, the time interval between patterns does not need to be constant, and may be determined according to the usage mode of the target control valve.

Figure 4 is an explanatory diagram showing an example in which multiple cycles of the opening and closing operation pattern of the pressure reducing valve are plotted side by side. That is, in FIG. 4, the start points of each cycle (time 0:00 in FIG. 4) are aligned and overlapped in a chronological order in the opening and closing control of the pressure reducing valve 5. The solid line in FIG. 4 shows the behavior of the valve opening of the pressure reducing valve 5 when the IGCC is operated at the rated processing load. On the other hand, the dashed line and the dashed line in FIG. 4 show the behavior of the valve opening of the pressure reducing valve 5 when the IGCC is operated at a processing load lower than the rated processing load. The pressure reducing valve 5 opens and closes according to an opening command value input from a control device (not shown), but in reality, as shown in FIG. 4, the behavior of the valve opening does not completely match, and a certain degree of variation occurs. Also, as shown by the solid line, dashed line, and dashed line in the figure, when the IGCC is operated at a processing load lower than the rated processing load, the behavior of the valve opening of the pressure reducing valve 5 is delayed compared to when the IGCC is operated at the rated processing load. This is because when the IGCC is operated at a processing load lower than the rated processing load, the amount of powder (pulverized coal) sent from the storage hopper 1 to the next process per unit time decreases. Therefore, in the pressure reducing valve 5 of this embodiment, the behavior of one cycle of the valve opening varies along the time axis depending on the processing load of the IGCC. In this case, if the normality or abnormality of the pressure reducing valve 5 is determined based on whether the valve opening falls within a band range with upper and lower limit values, the band range must be large, which may reduce the accuracy of abnormality detection. Below, a configuration and method for detecting abnormalities in the pressure reducing valve 5 whose behavior of one cycle of the valve opening varies along the time axis in this way will be described.

(Valve operation abnormality detection system)
Returning to the explanation of Fig. 1, the valve operation abnormality detection system 100 is a system for detecting abnormalities in the opening and closing operation of a target control valve. The valve operation abnormality detection system 100 is a computer, and includes a storage unit 10 and a control unit 20. Note that the valve operation abnormality detection system 100 may also include an interface for inputting data, such as a mouse or a keyboard, and a display device, such as a display.

The storage unit 10 is a memory that stores various information such as the calculation contents and programs of the control unit 20, and includes at least one of a main storage device such as a RAM (Random Access Memory) or a ROM (Read Only Memory), and an external storage device such as a HDD (Hard Disk Drive).

The control unit 20 is a calculation device, i.e., a CPU (Central Processing Unit). As shown in FIG. 1, the control unit 20 includes a data acquisition unit 21, a data shaping unit 22, a feature extraction unit 23, a condition setting unit 24, and an abnormality determination unit 25. The control unit 20 reads out and executes a program (software) from the storage unit 10 to realize the data acquisition unit 21, the data shaping unit 22, the feature extraction unit 23, the condition setting unit 24, and the abnormality determination unit 25, and executes various processes.

(Data Acquisition Section)
The data acquisition unit 21 acquires a plurality of data on the valve opening of the control valve to be detected as an anomaly, that is, the pressure reducing valve 5. Here, the valve opening is not limited to a value obtained by directly measuring the actual valve opening of the pressure reducing valve 5, but includes a value calculated based on other process quantities reflecting the opening state of the pressure reducing valve 5. The other process quantities are, for example, but not limited to, the flow rate of powder flowing through the pressure reducing valve 5. Furthermore, the data on the valve opening may directly use other process quantities (for example, flow rate). In any case, it is sufficient that the value reflects the behavior of the opening and closing operation of the pressure reducing valve 5 as shown in FIG. 4. The same applies to all the cases where data on the valve opening is handled below. The data on the valve opening is input from a control device (not shown) that controls the opening and closing of the pressure reducing valve 5. The data acquisition unit 21 outputs the acquired data on the valve opening to the data shaping unit 22.

(Data Formatting Department)
The data shaping unit 22 is a data shaping unit that creates time series data D (see FIG. 4) based on data on the past valve opening of the pressure reducing valve 5. The time series data D is data obtained by cutting out data on the valve opening of the pressure reducing valve 5 acquired by the data acquisition unit 21 within a predetermined time range. In this embodiment, the time series data D is data on the behavior of one cycle of the valve opening indicated by the solid line, dashed line, or dashed line in FIG. 4. When cutting out the time series data D, the change in the target data itself may be captured, or the behavior of one cycle of the valve opening may be captured based on another signal (for example, an on/off signal of the pressure reducing valve 5). The data shaping unit 22 outputs the created time series data D to the feature extraction unit 23.

(Feature Extraction Unit)
The feature extraction unit 23 extracts, based on the time series data D, the time from when the valve opening degree becomes the specified opening degree V (see FIG. 4 ) from the closed state as a feature amount. The specified opening degree V is the valve opening degree of the pressure reducing valve 5 set in advance by the user, and is stored in the storage unit 10. As described above, the data on the valve opening degree may be a value that reflects the behavior of the opening and closing operation of the pressure reducing valve 5, and the same is true for the specified opening degree V, and other process quantities (e.g., flow rate) may be used. In this embodiment, four specified opening degrees V1, V2, V3, and V4 are set. When the opening degree when the pressure reducing valve 5 is fully opened is 100%, the specified opening degrees V1, V2, V3, and V4 are set to V1=20%, V2=40%, V3=60%, and V4=80%. The number and value of the specified opening degrees V may be set arbitrarily. In this embodiment, the number of specified opening degrees V is reduced to reduce the calculation processing load compared to the case where the specified opening degrees V are set finely and in large numbers. The feature extraction unit 23 extracts the actual time it takes for the valve opening to reach each specified opening V from 0% for time-series data D indicated by the solid line, dashed line, and dashed dotted line in Fig. 4, for example. The feature extraction unit 23 outputs the extracted feature to the condition setting unit 24 and the abnormality determination unit 25.

(Condition Setting Section)
The condition setting unit 24 performs statistical processing on the multiple feature amounts to set an allowable arrival time T (see FIG. 4 ). The allowable arrival time T is a time range allowed for the valve opening to reach a specified opening V from a closed state, and is a time that serves as a criterion for abnormality determination in the abnormality determination unit 25. As shown in FIG. 1 , the condition setting unit 24 has a machine learning unit 24A.

The machine learning unit 24A accumulates features from time to time by unsupervised machine learning, performs statistical processing on the accumulated multiple features, and sets the arrival allowable time T. Note that unsupervised machine learning here means setting the arrival allowable time T using all accumulated data. Supervised machine learning will be described in the second embodiment. The machine learning unit 24A accumulates features extracted by the feature extraction unit 23 during the operation of the IGCC, and performs statistical processing. For the statistical processing, for example, a well-known statistical method may be used, such as setting the standard deviation of the accumulated features to a range of ±3σ from the average value of the features, where σ is the standard deviation of the accumulated features. The machine learning unit 24A sets the arrival allowable time T for all the specified openings V1, V2, V3, and V4. In the example shown in FIG. 4, the arrival allowable time T1 corresponding to the specified opening V1, the arrival allowable time T2 corresponding to the specified opening V2, the arrival allowable time T3 corresponding to the specified opening V3, and the arrival allowable time T4 corresponding to the specified opening V4 are set. The condition setting unit 24 outputs the allowable arrival time T set as described above to the abnormality determination unit 25.

The learning mode by the machine learning unit 24A can be turned on and off based on a user instruction. When the user turns on the learning mode, the machine learning unit 24A accumulates feature amounts from time to time through machine learning, and updates the arrival allowable time T based on the accumulated feature amounts. On the other hand, when the user turns off the learning mode, the accumulation of feature amounts and the update of the arrival allowable time T are temporarily stopped, and the value set up to that point is fixed as the arrival allowable time T.

(Abnormality Determination Unit)
The abnormality determination unit 25 determines whether the control valve to be detected as an abnormality, i.e., the pressure reducing valve 5, is operating normally based on the characteristic amount and the allowable arrival time T. Specifically, the abnormality determination unit 25 acquires the characteristic amount extracted by the characteristic amount extraction unit 23, i.e., the time required for the valve opening degree of the currently controlled pressure reducing valve 5 to change from the closed state to the specified opening degree V. Then, the abnormality determination unit 25 determines whether the characteristic amount of the currently controlled pressure reducing valve 5, i.e., the time required for the valve opening degree of the currently controlled pressure reducing valve 5 to change from the closed state to the specified opening degree V (V1, V2, V3, V4), is within the range of the corresponding allowable arrival time T (T1, T2, T3, T4). If the characteristic amount is within the range of each allowable arrival time T, the abnormality determination unit 25 determines that the pressure reducing valve 5 is normal, and if the characteristic amount is outside the range of each allowable arrival time T, the abnormality determination unit 25 determines that the pressure reducing valve 5 is abnormal.

With reference to FIG. 5, the abnormality detection judgment by the abnormality judgment unit 25 will be described in more detail. FIG. 5 is an explanatory diagram showing an example in which normal and abnormal examples of the valve opening of one cycle of the opening and closing operation pattern of the pressure reducing valve are plotted side by side. The solid line in the figure shows an example of the behavior of the valve opening when the pressure reducing valve 5 is normal. On the other hand, the dotted and dashed lines in the figure show an example of the behavior of the valve opening when the pressure reducing valve 5 is abnormal. In the case of the dotted line, some kind of catch occurs in the pressure reducing valve 5, and the opening operation takes longer than in the normal case. In addition, in the case of the dashed line, the opening operation of the pressure reducing valve 5 stops midway. In the normal example shown by the solid line in FIG. 5, the time to reach the specified opening V1, V2, V3, and V4 is within the range of each of the allowable reaching times T1, T2, T3, and T4, so the abnormality judgment unit 25 judges that the pressure reducing valve 5 is normal. On the other hand, in the abnormal example shown by the dotted line in Fig. 5, the time it takes to reach the specified opening degree V2 falls outside the range of the reachable time T2, so the abnormality determination unit 25 determines that an abnormality has occurred in the pressure reducing valve 5. Also, in the abnormal example shown by the dashed line in Fig. 5, the valve opening degree has not reached the specified opening degrees V2, V3, and V4 even after the reachable times T2, T3, and T4 have elapsed, so the abnormality determination unit 25 determines that an abnormality has occurred in the pressure reducing valve 5.

In this way, by setting the standard used for detecting an abnormality in the pressure reducing valve 5 to the allowable time T specified in the time axis direction, it becomes easier to determine that the pressure reducing valve 5 is operating normally even if the behavior of the opening and closing operation of the pressure reducing valve 5 fluctuates in the time axis direction, for example, as in the example of low processing load shown by the dashed line and the dashed-dotted line in Figure 4. Furthermore, in the abnormality example shown by the dotted line and dashed line in Figure 5, the behavior of the valve opening does not fall within the range of the allowable time T at some point, and an abnormality in the pressure reducing valve 5 can be detected.

(Method for detecting valve operation abnormality)
Next, a valve operation abnormality detection method according to the first embodiment will be described with reference to Fig. 6. Fig. 6 is a flowchart showing an example of the valve operation abnormality detection method according to the first embodiment. The process shown in Fig. 6 is repeatedly executed by the control unit 20 at predetermined time intervals.

In step S1, the control unit 20 acquires data on the current valve opening of the pressure reducing valve 5 by the data acquisition unit 21. The valve opening data is input from a control device (not shown) that controls the opening and closing of the pressure reducing valve 5. As described above, the valve opening data is not limited to a value obtained by directly measuring the actual valve opening, but may be any value that reflects the behavior of the opening and closing operation of the pressure reducing valve 5. The data acquisition unit 21 outputs the acquired valve opening data to the data shaping unit 22.

In step S2, the control unit 20 creates time series data D using the data shaping unit 22. As described above, the time series data D is data on one cycle of the valve opening behavior indicated by the solid line, dashed line, and dashed dotted line in FIG. 4. The data shaping unit 22 outputs the created time series data D to the feature extraction unit 23.

In step S3, the control unit 20 causes the feature extraction unit 23 to extract, as a feature, the time it takes for the valve opening to change from the closed state to the specified opening V. In this embodiment, the specified opening V is set to V1 = 20%, V2 = 40%, V3 = 60%, and V4 = 80%, as shown in FIG. 4. The feature extraction unit 23 extracts, from the time-series data D, the actual time it takes for the valve opening to change from 0% to each specified opening V as a feature, and outputs the extracted feature to the condition setting unit 24 and the abnormality determination unit 25.

In step S4, the control unit 20 determines whether the learning mode is on. The learning mode is switched on and off based on the user's instruction. When the control unit 20 determines that the learning mode is on (Yes in step S4), in step S5, the control unit 20 sets the reachable time T based on the feature amount by unsupervised machine learning. Specifically, in the machine learning unit 24A of the condition setting unit 24, a plurality of feature amounts extracted in step S3 are accumulated, and the accumulated plurality of feature amounts are subjected to a known statistical process to set the reachable time T. The machine learning unit 24A sets the corresponding reachable times T1, T2, T3, and T4 for all the specified openings V1, V2, V3, and V4. Note that since the feature amounts are not sufficiently accumulated immediately after the start of this routine, it is sufficient to accumulate the feature amounts in advance using data on past valve openings, and perform statistical processing to set the reachable time T in advance. The condition setting unit 24 outputs the reachable time T set as described above to the abnormality determination unit 25, and proceeds to step S6.

On the other hand, if the control unit 20 determines that the learning mode is off (No in step S4), it proceeds to step S6 without updating the reachable time T by machine learning. That is, if this routine has been executed for a sufficient time and the reachable time T has been sufficiently updated by machine learning, the learning mode may be turned off at the user's discretion. In this case, the reachable time T that was set up to that point is set as a fixed value. Also, as described above, the learning mode may be turned off at the user's discretion even if multiple pieces of past valve opening data have been obtained in advance, multiple feature amounts have been accumulated, and the reachable time T has been set in advance by statistical processing.

In step S6, the control unit 20 determines, by the abnormality determination unit 25, whether the valve opening of the pressure reducing valve 5 reaches the specified opening V within the range of the allowable arrival time T. In other words, the abnormality determination unit 25 determines whether the feature amount extracted in step S3 is within the range of the allowable arrival time T. If the feature amount is within the range of each of the allowable arrival times T1, T2, T3, and T4 (Yes in step S6), the abnormality determination unit 25 determines that the pressure reducing valve 5 is normal (step S7). On the other hand, if the feature amount is outside the range of each of the allowable arrival times T1, T2, T3, and T4 (No in step S6), the abnormality determination unit 25 determines that the pressure reducing valve 5 is abnormal (step S8).

By the above processing, it is possible to determine in real time whether the behavior of the opening and closing operation of the pressure reducing valve 5 is normal or abnormal. The abnormality determination unit 25 outputs the determination result to a display unit (not shown) and causes the result to be displayed on the display unit. This allows the user to look at the display result and determine whether or not an abnormality has occurred in the pressure reducing valve 5, and if an abnormality is detected, to take appropriate action such as inspecting the pressure reducing valve 5.

As described above, the valve operation abnormality detection system 100 according to the first embodiment includes a condition setting unit 24 that sets the allowable time T, which is the time range allowed for the valve opening to reach the specified opening V from the closed state, for the pressure reducing valve 5 (control valve) that is controlled to open and close according to a predetermined opening and closing pattern, and an abnormality determination unit 25 that determines that the pressure reducing valve 5 is normal if the time it takes for the valve opening to reach the specified opening V from the closed state is within the allowable time T, and determines that the pressure reducing valve 5 is abnormal if the time it takes for the valve opening to reach the specified opening V is outside the allowable time T.

With this configuration, in the opening and closing operation of the pressure reducing valve 5, the allowable arrival time T is specified as the time range allowed for the valve opening to reach the specified opening V, and the normality and abnormality of the pressure reducing valve 5 can be determined based on whether the change in the valve opening falls within the range of the allowable arrival time T. Therefore, even if the behavior of the opening and closing operation of the pressure reducing valve 5 fluctuates in the time axis direction, the possibility of detecting normal operation of the pressure reducing valve 5 as abnormal or failing to detect abnormal operation can be reduced without setting the range of the allowable arrival time T larger than necessary. Therefore, according to the valve operation abnormality detection system 100 and valve operation abnormality detection method of this embodiment, it is possible to more accurately detect abnormalities in the opening and closing operation of the pressure reducing valve 5 (control valve).

The valve operation abnormality detection system 100 according to the first embodiment further includes a data acquisition unit 21 that acquires data on the valve opening of the pressure reducing valve 5, a data shaping unit 22 that creates time series data D by extracting the data on the valve opening within a predetermined time range, and a feature extraction unit 23 that extracts, as a feature, the time it takes for the valve opening to change from a closed state to a specified opening V from the multiple time series data D, and the condition setting unit 24 performs statistical processing on the multiple feature values to set the arrival allowable time T. With this configuration, it is possible to use multiple behavior data on the opening and closing operation of the valve opening to set the arrival allowable time T that matches the actual situation through statistical processing.

The condition setting unit 24 also updates the allowable arrival time T at any time through machine learning. With this configuration, the allowable arrival time T can be updated at any time through machine learning, improving the accuracy of anomaly detection.

In the first embodiment, the condition setting unit 24 sets the allowable arrival time T based on the time series data D, but the method of setting the allowable arrival time T is not limited to this. For example, the condition setting unit 24 may set a predetermined time as the allowable arrival time T based on the design information of the target pressure reducing valve 5 (control valve) and the IGCC, which is the equipment in which the pressure reducing valve 5 is used. In other words, the user may determine in advance the variation in the time required for the valve opening to reach the specified opening V based on design information such as the specifications of the pressure reducing valve 5 and the usage conditions in the IGCC, and set the time as the allowable arrival time T.

[Second embodiment]
7 is a schematic diagram showing an example of a valve operation abnormality detection system according to the second embodiment. A valve operation abnormality detection system 200 according to the second embodiment includes a true/false result acquisition unit 26 in addition to the components of the valve operation abnormality detection system 100 according to the first embodiment. Other components of the valve operation abnormality detection system 200 are similar to those of the valve operation abnormality detection system 100, so the same components are denoted by the same reference numerals and description thereof will be omitted.

(Acquisition of correct/incorrect results)
The correct/incorrect result acquisition unit 26 acquires the correct/incorrect result of the judgment by the abnormality judgment unit 25. Here, the correct/incorrect result of the judgment is a result of confirming whether the judgment results in steps S3 to S5 in FIG. 6 were correct. For example, when it is determined that the pressure reducing valve 5 is abnormal, the user inspects the pressure reducing valve 5 after the fact. At this time, if the pressure reducing valve 5 is actually abnormal, the correct/incorrect result is positive, and if the pressure reducing valve 5 is actually normal, the correct/incorrect result is negative. In addition, when it is determined that the pressure reducing valve 5 is normal, for example, if some abnormality occurs around the storage hopper 1 and the inspection result shows that the abnormality is caused by the abnormality of the pressure reducing valve 5, the correct/incorrect result is also negative. In this way, the user inputs the correct/incorrect result of the judgment results in steps S3 to S5 in FIG. 6 after the fact into the valve operation abnormality detection system 200. In other words, information in which the past data on the valve opening degree of the pressure reducing valve 5 is categorized into cases where the judgment results are correct and cases where the judgment results are not correct is input into the valve operation abnormality detection system 200. The correct/incorrect result acquisition unit 26 acquires the correct/incorrect result input by the user, and outputs the acquired correct/incorrect result to the machine learning unit 24 A. Note that the user may input the correct/incorrect result to a control device of the IGCC (not shown), and the correct/incorrect result acquisition unit 26 may acquire information on the correct/incorrect result from a control device of the IGCC (not shown).

In the second embodiment, the machine learning unit 24A sets the allowable arrival time T by supervised machine learning that reflects the correct/incorrect result input from the correct/incorrect result acquisition unit 26 in the machine learning. That is, the machine learning unit 24A obtains information on the features of each accumulated time-series data D that is categorized into cases where the judgment result is correct and cases where it is incorrect. As a result, when performing machine learning, the machine learning unit 24A deletes the features categorized into cases where the judgment result is incorrect, and then performs statistical processing on the accumulated features to set the allowable arrival time T.

Figure 8 is a flow chart showing an example of a valve operation abnormality detection method according to the second embodiment. Steps S1 to S4 in Figure 8 are similar to steps S1 to S4 in Figure 6, and steps S7 to S9 in Figure 8 are similar to steps S6 to S8 in Figure 6, so their explanations are omitted.

When the control unit 20 determines that the learning mode is on (Yes in step S4), in step S5, the correct/incorrect result acquisition unit 26 acquires the correct/incorrect result of the judgment made by the anomaly judgment unit 25. As described above, the correct/incorrect result of the judgment is input by the user. Then, in step S6, the control unit 20 sets the allowable arrival time based on the features by supervised machine learning that reflects the acquired correct/incorrect result of the judgment in machine learning. Specifically, as described above, the machine learning unit 24A deletes the features of each accumulated time-series data D that are categorized as incorrect, and then performs statistical processing on the accumulated features to set the allowable arrival time T.

As described above, in the valve operation abnormality detection system 200 according to the second embodiment, the condition setting unit 24 acquires the result of the judgment by the abnormality judgment unit 25, and sets the allowable arrival time T by reflecting the result of the judgment in the machine learning. With this configuration, the allowable arrival time T can be set more appropriately in the machine learning without taking into account the data of the feature amount in which the judgment result by the abnormality judgment unit 25 differs from the actual phenomenon.

In the first and second embodiments, the allowable arrival time T is set regardless of fluctuations in the processing load of the IGCC, but the allowable arrival time T may be set to a value according to the processing load. That is, the condition setting unit 24 may further specify the allowable arrival time T described in the first and second embodiments as a function of the processing load of the IGCC. FIG. 9 is an explanatory diagram showing an example of a map that specifies the allowable arrival time according to the processing load of the IGCC. As shown in the figure, the condition setting unit 24 sets the allowable arrival time T according to a map that specifies the relationship between the processing load and the allowable arrival time T for each specified opening V (V1, V2, V3, V4).

A more detailed explanation will be given with reference to the map of the specified opening degree V4 (=80%). The solid line in the figure indicates the average value of the feature quantity when the reachable time T is determined by statistical processing, and the dashed and dotted lines indicate the range of ±3σ relative to the average value. At this time, a function is determined in advance so that the average value and the range of ±3σ, i.e., the reachable time T, change according to the processing load. Specifically, the function is specified so that the average value and the range of ±3σ tend to decrease as the processing load increases. Also, the range of ±3σ tends to widen as the processing load increases. As a result, as shown in the figure, for example, the reachable time T when the processing load is 100% is shorter than the reachable time T when the processing load is 50%, and the range of ±3σ is wider. This is because the opening speed of the pressure reducing valve 5 becomes faster (the time to open to the specified opening degree V becomes shorter) as the processing load increases, and the time to open to the specified opening degree V is more likely to vary due to the faster opening speed. In this way, by defining the allowable arrival time T for each specified opening V using a function according to the processing load, the allowable arrival time T can be defined within a range corresponding to the processing load, and it is not necessary to set the allowable arrival time T unnecessarily long. As a result, it is possible to further improve the accuracy of anomaly detection.

In this case, when setting the allowable arrival time T in machine learning, the features may be categorized according to the processing load, and statistical processing may be performed for each categorized feature to set the allowable arrival time T according to the processing load.

In addition, although the allowable time to reach T is defined here as a function corresponding to the processing load, the allowable time to reach may also be defined as a function corresponding to some parameter that affects the behavior of the opening and closing operation of the pressure reducing valve 5 (control valve) in the time axis direction. For example, in an IGCC, the allowable time to reach T may be defined as a function corresponding to the calories consumed instead of the processing load. Also, in cases where the behavior of the opening and closing operation may be affected depending on the type of powder (here, pulverized coal) stored in the storage hopper 1, the allowable time to reach T may be changed depending on the type of powder.

REFERENCE SIGNS LIST 1 Storage hopper 2 Supply valve 3 Outlet valve 4 Pressurization valve 5 Pressure reduction valve 10 Memory unit 20 Control unit 21 Data acquisition unit 22 Data shaping unit 23 Feature extraction unit 24 Condition setting unit 24A Machine learning unit 25 Abnormality determination unit 26 Correct/incorrect result acquisition unit 100, 200 Valve operation abnormality detection system D Time series data L1 Supply line L2 Outlet line L3 Pressurization line L4 Pressure reduction line T, T1, T2, T3, T4 Allowable time to reach V, V1, V2, V3, V4 Specified opening degree

Claims (8)

a condition setting unit that sets an allowable time for a valve opening degree to reach a specified opening degree from a closed state for a control valve that is controlled to open and close in accordance with a predetermined opening and closing pattern;
an abnormality determination unit that determines that the control valve is normal when a time taken for the valve opening to reach the specified opening from a closed state is within the range of the allowable time to reach, and that the control valve is abnormal when the time taken for the valve opening to reach the specified opening is outside the range of the allowable time to reach;
Equipped with
The condition setting unit sets the specified opening degree to a plurality of opening degrees,
The abnormality determination unit is a valve operation abnormality detection system that determines whether the time taken for the valve opening to reach each of the specified opening degrees from a closed state is within each of the allowable arrival times . A data acquisition unit that acquires data on the valve opening degree of the control valve;
a data shaping unit that creates time series data by extracting the valve opening data within a predetermined time range;
a feature extraction unit that extracts, from the plurality of time-series data, a time from when the valve opening degree changes from a closed state to when the valve opening degree changes to the specified opening degree as a feature;
Further equipped with
The valve operation abnormality detection system according to claim 1 , wherein the condition setting unit sets the permissible arrival time by performing statistical processing on a plurality of the feature amounts. The valve operation abnormality detection system according to claim 2, wherein the condition setting unit updates the allowable arrival time from time to time using machine learning. The valve operation abnormality detection system according to claim 3, wherein the condition setting unit obtains a result of the judgment made by the abnormality judgment unit, and sets the allowable arrival time by reflecting the result of the judgment in the machine learning. The valve operation abnormality detection system according to claim 1, wherein the condition setting unit sets the allowable arrival time to a time that is predetermined based on design information of the control valve and the equipment in which the control valve is used. The valve operation abnormality detection system according to any one of claims 1 to 5, wherein the condition setting unit specifies the allowable arrival time using a function according to a parameter that affects the behavior of the opening and closing operation of the control valve in the time axis direction. For a control valve that is controlled to open and close according to a predetermined opening and closing pattern, a permissible time to reach a specified opening is set, which is a time range permitted for the valve opening to change from a closed state to a specified opening;
Regarding the control valve, if the time taken for the valve opening to reach the specified opening degree from the closed state is within the range of the allowable time to reach, the control valve is determined to be normal, and if the time taken for the valve opening to reach the specified opening degree is outside the range of the allowable time to reach, the control valve is determined to be abnormal ,
A method for detecting abnormalities in valve operation , comprising: setting the specified opening degree to a number of different opening degrees; and determining whether the time it takes for the valve opening degree to reach each of the specified opening degrees from a closed state is within the range of each of the allowable arrival times . For a control valve that is controlled to open and close according to a predetermined opening and closing pattern, a permissible time to reach a specified opening is set, which is a time range permitted for the valve opening to change from a closed state to a specified opening;
Regarding the control valve, if the time taken for the valve opening to reach the specified opening degree from the closed state is within the range of the allowable time to reach, the control valve is determined to be normal, and if the time taken for the valve opening to reach the specified opening degree is outside the range of the allowable time to reach, the control valve is determined to be abnormal ,
setting the specified opening degree to a plurality of different opening degrees, and determining whether the time taken for the valve opening degree to reach each of the specified opening degrees from a closed state is within the range of each of the allowable arrival times;
A program that causes a computer to do something.

Images