CN107710089B

CN107710089B - Plant equipment diagnosis device and plant equipment diagnosis method

Info

Publication number: CN107710089B
Application number: CN201680036343.0A
Authority: CN
Inventors: 关合孝朗; 林喜治; 前田达矢; 定江和贵; 村上正博; 深井雅之
Original assignee: Hitachi Ltd
Current assignee: Hitachi Ltd
Priority date: 2015-06-22
Filing date: 2016-05-25
Publication date: 2020-07-10
Anticipated expiration: 2036-05-25
Also published as: CN107710089A; JP6523815B2; JP2017010232A; WO2016208315A1

Abstract

When an abnormality is detected, it is judged whether or not the abnormality should be dealt with depending on experience of plant operation, but it is preferably judged based on a risk (loss prediction amount) in a case where the abnormality is left alone. In order to solve the above-described problem, the present invention provides a plant equipment diagnosis device including a plurality of diagnosis units for diagnosing an abnormal state of plant equipment, the plant equipment diagnosis device including a comprehensive diagnosis unit for obtaining accuracy of detection of the state abnormality of each of the plurality of diagnosis units based on measurement signal data relating to a state of the plant equipment and equipment management information data relating to a past state abnormality, and evaluating a loss prediction amount based on the accuracy and a loss amount associated with the state abnormality.

Description

Plant equipment diagnosis device and plant equipment diagnosis method

Technical Field

The present invention relates to a Plant equipment diagnostic apparatus and a Plant equipment diagnostic method for diagnosing a state abnormality of a Plant.

Background

When an abnormal transient phenomenon or accident occurs in a plant, a plant diagnosis apparatus detects the occurrence of the abnormal phenomenon or accident based on measurement data from the plant.

Patent document 1 discloses a diagnostic device using Adaptive Resonance Theory (ART) which is one of clustering techniques. ART is a theory of classifying multidimensional data into different categories according to their similarity.

In the above-described technique, measurement data in a normal state is first classified into a plurality of categories (normal categories) by ART. The current measurement data is then input to the ART and classified into different categories. When the measurement data cannot be classified into the normal category, a new category (new category) is generated. The occurrence of a new category means that the state of the plant equipment has changed. Therefore, the occurrence of an abnormality is determined by the occurrence of the new category, and the abnormality is diagnosed when the occurrence rate of the new category exceeds a threshold value.

[ Prior art documents ]

[ patent document ]

Patent document 1: japanese patent laid-open publication No. 2005-165375

Disclosure of Invention

Problems to be solved by the invention

In the clustering technique, there is a parameter that determines the size of a cluster (cluster), which is the size of a category (category) in ART. This parameter is referred to as the resolution parameter. Generally, when a certain data is classified into different groups, the number of groups is reduced when the resolution is set coarsely, and the number of groups is increased when the resolution is set finely.

When the clustering technique is used in abnormality diagnosis, the magnitude of change in the trend of data occurring in a new category differs depending on the resolution coarseness and the resolution fineness. When a new category is generated in the case where the resolution is coarse, the data trend is greatly changed as compared with the normal, and therefore the accuracy of the machine abnormality is high. On the other hand, when the resolution is fine, a change in a weak tendency such as measurement noise may be detected, and therefore the accuracy of the machine abnormality is low. Thus, when the parameter setting values for specifying the size of the group are different, the accuracy of occurrence of an abnormality at the time of abnormality detection is different.

In general, since abnormality detection performance differs depending on the diagnostic method, the accuracy of detection of an abnormality at the time of abnormality detection differs.

In addition, at the time of abnormality detection, it is appropriate to stop plant equipment and perform maintenance and repair to avoid machine failure, but costs for maintenance and repair and a loss of opportunity during the stop of plant equipment occur. Therefore, if it is a slight failure, the operation is sometimes continued until a periodic inspection. On the other hand, the consequences of leaving an anomaly out can lead to machine failure, damage, and loss will be greater than for maintenance and repair.

In the current situation, when an abnormality is detected, it is determined whether the abnormality should be dealt with based on experience of plant operation, but it is also preferably determined based on a risk (loss prediction amount) when the abnormality cannot be dealt with.

Means for solving the problems

In order to solve the above-described problems, the present invention provides a plant equipment diagnosis apparatus including a plurality of diagnosis units for diagnosing a state abnormality of plant equipment, the plant equipment diagnosis apparatus including a comprehensive diagnosis unit for obtaining accuracy of detection of the state abnormality of each of the plurality of diagnosis units based on measurement signal data relating to a state of the plant equipment and equipment management information data relating to a past state abnormality, and estimating a predicted loss amount based on the accuracy and a loss amount associated with the state abnormality.

ADVANTAGEOUS EFFECTS OF INVENTION

The loss prediction amount is obtained at the time of abnormality detection, and information useful for determining whether or not to deal with the detected abnormality can be provided.

Drawings

Fig. 1 is a block diagram illustrating a diagnostic apparatus as a 1 st embodiment of the present invention.

Fig. 2 is a flowchart illustrating the operation of the integrated diagnostic unit in the evaluation mode and the diagnostic mode of the diagnostic apparatus.

Fig. 3 is a diagram illustrating the timing of operating the evaluation mode and the diagnosis mode.

Fig. 4 is a diagram illustrating a case where data stored in the measurement signal database and the device management information database.

Fig. 5 is a diagram illustrating a case where data is stored in the diagnosis result database.

Fig. 6 is an explanatory diagram of the adaptive resonance theory.

Fig. 7 is a diagram showing the result of classifying measurement signals into different categories.

Fig. 8 is a diagram illustrating a relationship between the size of classification and detection timing, accuracy, and loss prediction amount.

Fig. 9 is a diagram illustrating a change with time of the detection result and the loss prediction amount of each diagnosis unit.

Fig. 10 is a diagram illustrating a correction method of accuracy.

Fig. 11 is a diagram showing an example of a screen clearly shown in the image display apparatus.

Fig. 12 is a diagram showing an example of a screen clearly shown in the image display apparatus.

Fig. 13 is a diagram illustrating model diagnosis.

Fig. 14 is a diagram illustrating an effect obtained by combining the clustering technique diagnosis and the model diagnosis.

Fig. 15 is a diagram for explaining an example in which the diagnostic device of the present invention is applied to a thermal power plant.

Description of the reference numerals

1: measuring a signal; 2: an external input signal; 3: measuring a signal; 4: a device management information signal; 5: measuring a signal; 6: a device management information signal; 7: diagnostic result database information; 8: diagnostic result database information; 9, measuring a signal; 10, diagnosis result; synthesizing the diagnosis result signal; synthesizing the diagnosis result signal; 100, factory equipment; 200, a diagnostic device; 210, a data input interface; 220, a data output interface; 300, measuring a signal database; a device management information database 310; 320, a diagnosis result database; 400, a comprehensive diagnosis unit; 500, a diagnostic unit; 800: an image display device; 900 external input device; 910, a keyboard; 920 mouse

Detailed Description

Hereinafter, a diagnostic apparatus suitable for carrying out the present invention will be described with reference to the drawings. The following description is merely an example, and the present invention itself is not intended to be limited to the following specific description.

Example 1

Fig. 1 is a block diagram illustrating a diagnostic apparatus as a 1 st embodiment of the present invention. The diagnostic apparatus 200 is connected to the plant 100, the image display apparatus 800, and the external input apparatus 900, and monitors and diagnoses the plant 100. The diagnostic apparatus 200 is configured by connecting, by wire or wirelessly, a communication Unit that performs communication between the devices or apparatuses, a computer and a computer server (CPU), a memory, various databases DB, and the like. The external input device 900 is configured by a keyboard switch, a pointing device such as a mouse, a touch panel, an audio pointing device, and the like, and the image display device 800 is configured by a liquid crystal display or the like.

The diagnostic device 200 includes an integrated diagnostic unit 400 and a diagnostic unit 500 as arithmetic devices. The diagnostic unit 500 is provided in plural and the number thereof can be arbitrarily set. The diagnostic apparatus 200 includes a measurement signal database 300 as a database, a device management information database 310, and a diagnostic result database 320. In addition, the database is abbreviated as DB in fig. 1.

The measurement signal database 300, the device management information database 310, and the diagnosis result database 320 store electronic information, and store the information in a form generally called an electronic file (electronic data).

The diagnostic apparatus 200 has an external input interface 210 and an external output interface 220 as interfaces with the outside.

Then, the measurement signal 1 obtained by measuring various state quantities as the operation states of the plant equipment 100 and the external input signal 2 created by the operation of the keyboard 910 and the mouse 920 provided in the external input device 900 are taken into the diagnostic device 200 via the external input interface 210. Further, the integrated diagnosis result signal 12 is output to the image display apparatus 800 via the external output interface 220.

In the diagnostic apparatus 200 shown in fig. 1, a detection signal 1 obtained by measuring various state quantities of the plant 100 is taken in via the external input interface 210. The measurement signal 3 taken into the diagnostic apparatus 200 is stored in the measurement signal database 300. Plant equipment management information such as failure information and maintenance information generated in the plant equipment 100 is received in the diagnostic apparatus 200 by the external input signal 2 generated by the operation of the keyboard 910 and the mouse 920. The device management information signal 4 taken into the diagnostic apparatus 200 is stored in the device management information database 310.

The diagnostic apparatus 200 has two processing modes, an evaluation mode and a diagnostic mode. The flow of the evaluation mode and the diagnosis mode and the actions of the integrated diagnosis unit 400 and the diagnosis unit 500 will be described with reference to fig. 1 and 2.

In the diagnostic apparatus 200 of the present embodiment, the comprehensive diagnostic unit 400, the diagnostic unit 500, the measurement signal database 300, the device management information database 310, and the diagnostic result database 320 are disposed inside the diagnostic apparatus 200, but some of these devices may be disposed outside the diagnostic apparatus 200 and only data may be communicated between the devices.

Further, the information stored in the database provided in the diagnostic apparatus 200 can be all the information displayed on the image display apparatus 100, and the information can be modified by the external input signal 1 generated by operating the external input apparatus 900.

In the present embodiment, the external input device 900 is constituted by a keyboard and a mouse, but may be a device for inputting data such as a microphone for inputting voice or a touch panel.

It is needless to say that the embodiment of the present invention can be implemented as a diagnostic method or an information providing service that provides information obtained by operating the diagnostic apparatus 200.

Fig. 2 is a flowchart illustrating the operation of the comprehensive diagnosis unit 400 in the evaluation mode and the diagnosis mode of the diagnosis device 200.

Fig. 2 (a) is a flowchart of the evaluation mode.

First, in step 2000, the integrated diagnostic unit 400 extracts the measurement signal 5 stored in the measurement signal database 300 for a predetermined period.

In step 2010, the comprehensive diagnostic unit 400 sends the measurement signal 9 to the diagnostic unit 500. The diagnosis unit 500 processes the measurement signal 9 and diagnoses the state of the plant device 100, and transmits a diagnosis result 10 to the comprehensive diagnosis unit 400. The integrated diagnostic unit 400 aggregates the received diagnostic results 10, and transmits and stores the diagnostic result database information 8 to the diagnostic result database 320.

In step 2020, the integrated diagnostic unit 400 extracts the device management information signal 6 stored in the device management information database 310.

In step 2030, the detection result of each diagnostic unit of the diagnostic result database information 7 stored in the diagnostic result database 320 is compared with the device management information signal 6 extracted in step 2020, and the accuracy and the average lead time are calculated. Here, the accuracy is obtained by dividing the number of failures by the number of detections. The average lead time is a time obtained by subtracting the time detected by the corresponding diagnostic unit from the time detected by the threshold determination, and indicates how long it is detected. The accuracy and average lead time for each diagnostic unit found in step 2030 are stored in the diagnostic result database 320.

In step 2040, the integrated diagnostic unit 400 extracts the diagnostic result database information 7 stored in the diagnostic result database 320, and sends it to the external output interface 220 as the integrated diagnostic result signal 11. The integrated diagnosis result signal 12 is transmitted to the image display apparatus 800, and is displayed in the image display apparatus 800.

Fig. 2 (b) is a flowchart illustrating the operation in the diagnosis mode.

In step 2100, the integrated diagnostic unit 400 extracts the operation data 5 stored in the measurement signal database during the period of diagnosis.

In step 2110, the integrated diagnostic unit 400 transmits the measurement signal 9 to the diagnostic unit 500. The diagnosis unit 500 processes the measurement signal 9 and diagnoses the state of the plant device 100, and transmits a diagnosis result 10 to the comprehensive diagnosis unit 400. The integrated diagnostic unit 400 aggregates the received diagnostic results 10, and transmits and stores the diagnostic result database information 8 to the diagnostic result database 320.

In step 2120, the presence or absence of abnormality detection is evaluated, and if there is a diagnostic unit in which abnormality is detected, the routine proceeds to step 2130, and if not, the routine proceeds to step 2160.

In step 2130, the integrated diagnostic unit 400 extracts the diagnostic result database information 7 stored in the diagnostic result database 320, and grasps the information on the accuracy of the diagnostic unit in which the abnormality is detected in step 2120.

In step 2140, the integrated diagnostic unit 400 extracts the device management information 6 stored in the device management information database 310, and grasps the amount of loss due to the failure.

In step 2150, the integrated diagnosis unit 400 calculates a loss prediction amount based on the accuracy extracted in step 2130 and the loss amount extracted in step 2140. The loss prediction amount may be obtained by various methods such as multiplying the accuracy by the loss amount, or by evaluating the loss amount using a predetermined parameter.

In step 2160, the detection results of the diagnostic units are displayed on the image display apparatus 800, and when there is a diagnostic unit in which an abnormality is detected, the image display apparatus 800 also displays the predicted loss amount calculated in step 2150.

As described above, in the diagnostic apparatus 200 according to the present invention, when the diagnostic means 500 detects an abnormality, the loss prediction amount is displayed, whereby information useful for determining whether or not to process the detected abnormality can be provided.

In the method shown in fig. 3 (a), after the operation data is accumulated for a certain period, the evaluation mode is operated once, and the diagnosis mode is operated at a certain cycle.

In the method shown in fig. 3 (b), the evaluation mode is operated at regular intervals, and after the accuracy and average lead time data stored in the diagnostic result database 320 is updated, the diagnostic mode is operated.

In the method shown in fig. 3 (c), the evaluation mode is operated when an instruction is issued from the user. The evaluation mode is executed at an arbitrary timing, and the accuracy and average lead time are updated to make the diagnosis mode act.

In addition, the timing of operating the evaluation mode and the diagnosis mode may be arbitrarily set in addition to the timing set forth in the present embodiment.

Fig. 4 is a diagram illustrating a case where data stored in the measurement signal database 300 and the device management information database 310.

As shown in fig. 4 a, the measurement signal database 300 stores the value of the measurement signal 1 (in the figure, the data item A, B, C is described) as the operation data measured for the plant 100 for each sampling period (time on the vertical axis).

By using the

scroll boxes

302 and 303 which are vertically and horizontally movable on the display screen 301, a wide range of data can be displayed in a scrolling manner.

As shown in fig. 4 (b), the device management information database 310 stores failure information such as failure contents, countermeasure costs, lead time required to avoid a failure, the number of times of stoppage due to a failure, and the amount of opportunity loss due to stoppage of plant equipment.

As shown in fig. 4 (c), the device management information database 310 stores maintenance information such as maintenance contents, maintenance costs, the number of days required for maintenance, and the amount of opportunity loss due to maintenance.

Fig. 5 is a diagram illustrating a case where data is stored in the diagnosis result database 320.

As shown in fig. 5 a, the diagnostic result database 320 stores the detection results of the diagnostic units for each sampling period (time of the vertical axis) (the diagnostic units A, B, C are shown in the figure).

By using the scroll frames 312 and 313 that are vertically and horizontally movable on the display screen 311, a wide range of data can be displayed in a scrolling manner.

The diagnostic result database 320 stores the detection results in the respective diagnostic units, and stores the detection results by replacing the diagnostic results with digital information, for example, as 1 in the case of the abnormality determination and 0 in the case of the normal determination.

As shown in (b) of fig. 5, the accuracy and the average lead time calculated in the evaluation mode are saved for each diagnostic unit in the diagnostic result database.

Fig. 6 depicts a case of applying Adaptive Resonance Theory (ART) as an embodiment of the diagnosis unit 500. In addition, other clustering methods such as vector quantization and support vector machines can also be used.

As shown in fig. 6 (a), the data sorting function is composed of a data preprocessing device 610 and an ART module 620. The data preprocessing unit 610 converts the operation data into input data of the ART module 620.

The following describes steps performed by the data preprocessing unit 610 and the ART module 620.

First, in the data preprocessing device 610, data is normalized for each measurement item. Data including data nxi (n) obtained by normalizing the measurement signal and a complement cnxi (n) (1-nxi (n)) of the normalized data is input data ii (n). The input data ii (n) is input to the ART module 620.

In the ART module 620, the measurement signal 10 or the operation signal 11 as input data is classified into a plurality of categories.

The ART module 620 has an F0 layer 621, an F1 layer 622, an F2 layer 623, a memory 624, and a selection subsystem 625, which are combined with each other. The F1 layer 622 and the F2 layer 623 are combined by a weighting coefficient. The weighting coefficients represent the original type (prototype) of the class into which the input data is classified. Here, the original type represents a representative value of the category.

Next, an algorithm of the ART module 620 will be described.

When input data is input to the ART module 620, the outline of the algorithm is as the following processing 1 to processing 5.

Process 1 the input vector is normalized by the F0 layer 621 to remove noise.

Process 2. candidates in the appropriate category are selected by comparing the input data to the F1 layer 622 with the weighting coefficients.

Process 3 the adequacy of the class selected in the selection subsystem 625 is evaluated by the ratio to the parameter p. If judged to be valid, the input data is classified into the category and proceeds to process 4. On the other hand, if it is judged to be inappropriate, the category is reset and a suitable candidate for the category is selected from the other categories (process 2 is repeated). If the value of the parameter ρ is increased, classification of the class becomes finer. That is, the category size becomes smaller. Conversely, if the value of the parameter ρ is decreased, classification of the category becomes rough. The category size becomes larger. This parameter ρ is called a vigilance parameter.

And a process 4 of determining that the input data belongs to the new class when all the existing classes are reset in the process 2, and generating a new weighting coefficient indicating the original type of the new class.

In the process 5, when the input data is classified into different categories J, the weighting coefficients wj (new) corresponding to the categories J are updated by using the past weighting coefficients wj (old) and the input data p (or data derived from the input data) by the formula 1.

(formula 1)

WJ(new)＝Kw·p+(1－Kw)·WJ(old)

Here, Kw is a learning rate parameter (0< Kw <1), and is a value that determines the degree to which the input vector is reflected in the new weighting coefficient.

In addition, the operational expressions of formula 1 and the following formulas 2 to 12 are embedded in the ART module 620.

The data classification algorithm of the ART module 620 is characterized by the above-described process 4.

In the process 4, when input data different from the pattern at the time of learning is input, a new pattern can be recorded without changing the pattern of recording. Thus, a new pattern can be recorded while a past learned pattern is recorded.

Thus, when the pre-supplied operation data is supplied as the input data, the ART module 620 learns the supplied pattern. Thus, when new input data is input to the learned ART module 620, it can be determined whether or not it is close to a certain pattern in the past by the above algorithm. In addition, if the pattern has not been experienced in the past, the pattern is classified into a new category.

Fig. 6 (b) shows a block diagram of the structure of the F0 layer 621. In the F0 layer 621, input data I is again input at each time_iNormalization, to make a normalized input vector for input to the F1 layer 621 and the selection subsystem 625

First, according to equation 2, based on the input data I_iComputing

Where a is a constant.

(formula 1)

Then, the pair is calculated using equation 3

Obtained by normalization

Here, | w⁰I represents w⁰Norm of (d).

(formula 2)

Then, using equation 4, the equation

For removing noise

Where θ is a constant for eliminating noise. According to the calculation of equation 4, the minute value becomes 0, and therefore, the noise of the input data is eliminated.

(formula 3)

Finally, the normalized input vector is found using equation 5

Is the input to layer F1.

(formula 4)

Fig. 6 (c) is a block diagram showing the structure of the F1 layer 622. In the F1 layer 622, the value obtained from equation 5

Keeping as short-term storage, P input to F2 level 722 is calculated_i. The calculation formulas of the F2 layers are collectively shown as formula 6 to formula 12. Where a, b are constants, f (-) is a function represented by equation 4, T_jIs the degree of fit calculated by the F2 layer 623.

(formula 5)

(formula 6)

(formula 7)

v_i＝f(x_i)+bf(q_i)

(formula 8)

(formula 9)

(formula 10)

Wherein (formula 11)

Fig. 7 is a diagram showing an example of the result of classifying measurement signals into different categories.

Fig. 7 (a) is a diagram showing an example of a classification result of classifying the measurement signal 1 of the plant device 100 into different classes.

Fig. 7 (a) shows two items in the measurement signal as an example, and is marked with a two-dimensional graph. The vertical axis and the horizontal axis are normalized to show the measurement signals of the respective items.

The measurement signal is divided into a plurality of categories 630 (circles shown in (c) of fig. 4) by the ART module 620 of (a) of fig. 3. One circle corresponds to one category.

In the present embodiment, the measurement signals are classified into four categories. Class number 1 is a group with a large value for item a and a small value for item B, class number 2 is a group with a small value for both item a and item B, class number 3 is a group with a small value for item a and a large value for item B, and class number 4 is a group with a large value for both item a and item B.

Fig. 7 (b) is a diagram illustrating a result of classifying the measurement signals 1 acquired from the plant 100 into different categories. The horizontal axis is time and the vertical axis is the measurement signal, class number.

As shown in FIG. 7 (b), data in the normal period before the start of diagnosis is classified into different categories 1 to 3. The data of the first half after the start of monitoring is classified into category 2 and the same category as the model data. In this case, since the trends of the data are the same, it is determined that the state has not changed. On the other hand, the data of the second half of the period in which monitoring is started is classified into a category 4, and is classified into a category different from the model data. Since the trends of the data are different, it is determined that the state of the plant equipment has changed.

Thus, the diagnostic technique to which the clustering technique is applied has a feature of detecting a trend change of data.

Fig. 8 is a diagram illustrating a relationship between the size of the category and the detection timing, accuracy, loss prediction amount.

As shown in fig. 8 (a), when the resolution-determining parameter ρ is set large and the category size is reduced, even a slight change is detected. Can be detected in advance. In contrast, since a minute change in measurement noise or the like is detected, the accuracy is lowered.

On the other hand, when the parameter ρ is set small and the class size is increased, a new class appears when the deviation from the normal state is large.

Apart from the normal state, the accuracy of being an abnormality becomes high. On the other hand, the timing of detection becomes late.

Thus, the accuracy becomes high as the classification size becomes large. Since the loss prediction amount is increased with high accuracy, the classification size and the loss prediction amount are exponentially functional as shown in fig. 8 (b).

In step 2030 of fig. 2, integrated diagnostic section 400 analyzes the past data, finds the relationship of (b) of fig. 8, stores the relationship in diagnostic result database 320, and may display the relationship of (b) of fig. 8 on image display device 800 in step 2040.

Fig. 9 is a diagram illustrating the change with time of the detection result and the loss prediction amount of each diagnosis unit.

The diagnostic unit A, B, C is composed of 3 ARTs of different category sizes. At time 2200, diagnostic unit a detects, at time 2210 diagnostic unit B detects, and at time 2220 diagnostic unit C detects. The loss amount (1000 ten thousand yen in this embodiment) is multiplied by the maximum accuracy value of the unit to be detected, and the loss prediction amount is calculated.

Thus, during the period of time 2200-. As time elapses, the change in the measurement value becomes large, an abnormality is detected by the diagnosis unit with high accuracy, and the loss prediction amount also increases.

As described above, according to the diagnostic apparatus 200 of the present invention, it is possible to obtain information for determining whether or not to process an abnormality based on the loss prediction amount at each time.

Fig. 10 is a diagram illustrating a correction method of accuracy.

The possibility of damage varies depending on the degree and content of the failure.

For example, with respect to faults related to machine damage, tripping, improved accuracy and improved correction of loss predictions; regarding minor failures that have not been noticed until the periodic inspection, the accuracy is reduced and a reduction correction is made to the loss prediction amount.

In this way, the accuracy is corrected according to the degree of influence of the contents of the failure, and the loss prediction amount can be estimated more accurately.

Fig. 11 is a diagram illustrating an example of a screen displayed in the image display apparatus 800.

Fig. 11 (a) is a diagram illustrating an example of a screen displayed in the image display apparatus 800 when the diagnosis mode is performed. The diagnosis unit and the loss prediction amount in which the abnormality is detected are displayed on a screen. Thus, the loss prediction amount is displayed on the screen display device, and information for determining whether or not to perform processing can be provided.

Fig. 11 (b) is a diagram illustrating an example of a screen displayed in the image display apparatus 800 when the evaluation mode is performed. Assuming that the faults detected earlier than the lead time are faults that can be prevented by introducing a diagnostic scheme, the sum of the loss amounts of these faults is shown as a cost advantage. The calculated cost advantage and the diagnostic plan service price are displayed, and whether to purchase the service can be judged.

Fig. 12 is a diagram illustrating an example of an image displayed in the image display apparatus 800.

Assuming that maintenance is performed when a fault is detected, it is recommended to use a diagnostic scheme that minimizes the loss forecast relative to the maintenance cost target value. If maintenance is performed based on a detection result with low detection accuracy, the loss prediction amount (risk) can be reduced, but the number of times of maintenance increases, and the maintenance cost increases.

The appropriate diagnostic scenario is output relative to the input maintenance cost target (maintenance cost target used annually). In this way, the diagnostic plan with the smallest loss prediction amount can be flexibly used as the target value of the input maintenance cost.

Example 2

In embodiment 2 of the present invention, a case where model diagnosis and clustering are applied as the diagnosis technique 500 is explained. The technique described in example 1 was applied with respect to clustering.

Fig. 13 is a diagram illustrating model diagnosis. In the model diagnosis, a device model that simulates characteristics of devices constituting the plant 100 is used. As a method of constructing a model simulating the plant 100, there are a physical model using a physical formula such as a mass conservation formula and a heat transfer formula, and a statistical model such as a neural network, and japanese patent application laid-open No. 2006-57595 is a publicly known technique.

Input/output information of the machines constituting the plant apparatus 100 is measured as a signal a and a signal B, respectively. In the machine model, a predicted value of the signal B with respect to the input of the signal a is output. In the model diagnosis technique, when an error between a model predicted value and an actually measured value of the signal B exceeds a threshold value, an abnormality is detected.

Fig. 14 is a graph illustrating the effect obtained by using both the clustering technique and the model diagnosis.

Factory equipment is connected to machine a and machine B. In the clustering technology (ART) diagnosis of the diagnosis machine B, data B and data C are used as input data to ART, and data trend change is detected. In the model diagnosis, a predicted value of data C is output with respect to data B as an input, and an abnormality is detected when an error between the predicted value and an actually measured value of data C exceeds a threshold value.

In this case, at time 2300, a failure that does not stop the plant equipment occurs in the machine a. Due to the effect of a fault occurring in the machine a, the flow rate, pressure, and temperature from the machine a to the machine B change, and the signal B changes. Between time 2300 and time 2310, machine B operates normally. At time 2310, the flow rate, pressure, and temperature of the fluid flowing through the device B change, and thus the device B malfunctions.

In this case, since a change in the signal B is detected in the ART diagnosis, the ART diagnosis detects an abnormality at the timing 2300. On the other hand, since the machine B is in a normal state, no abnormality is detected in the model diagnosis.

At timing 2310 when the failure occurs in the machine B, the model diagnosis detects an abnormality.

Thus, the ART diagnosis detects an abnormality earlier than the model diagnosis. Further, the device B does not malfunction when an abnormality is detected in ART, and malfunctions when an abnormality is detected in model diagnosis. That is, the accuracy of detection when an abnormality is detected in the model diagnosis is high, and in the diagnostic device 200 of the present invention, the loss prediction amount is calculated to be high in consideration of the accuracy.

By calculating and displaying the loss prediction amount from the diagnosis results obtained by using diagnosis units having different detection timings and accuracies, such as clustering technique and model diagnosis, it is possible to provide information useful for determining whether or not to process a detected abnormality.

Example 3

The effect of applying the diagnostic apparatus 200 of the present invention to a C/C plant will be described.

Fig. 15 is a diagram showing a machine configuration of a C/C plant as an embodiment of the plant 1000. The gas turbine 1080 is constituted by a compressor 1010, an expander 1020, and a combustor 1030. In the gas turbine 1080, the compressor 1010 takes in air and compresses it, and then the combustor 1030 takes in compressed air and fuel and generates combustion gas, and the expander 1020 takes in the combustion gas and obtains power. The output of the gas turbine 1080 is the difference between the power output by the expander 1020 and the power used by the compressor 1010. The heat recovery boiler 1050 is provided with a heat exchanger 1060, and generates high-temperature steam using high-temperature exhaust gas from the gas turbine 1080. In the steam turbine 1070, high-temperature steam generated by the waste heat recovery boiler 1050 is sucked and power is obtained. In condenser 1090, the exhaust gas sucked into steam turbine 1070 exchanges heat with cooling water, thereby condensing steam into water. In the generator 1040, the output of the gas turbine 1080 and the steam turbine 1070 is used to generate electricity.

In the present plant, the fuel flow rate is controlled so that the exhaust gas temperature becomes a target value.

As an abnormal phenomenon occurring in the plant, a phenomenon in which holes for passing cooling air (blade surface cooling holes) of the blades of the expander 1020 become large can be exemplified. When this abnormal phenomenon occurs, the amount of cooling air increases, the exhaust gas temperature decreases, and the fuel flow rate of the combustor 1030 increases. The combustion temperature rises due to the effect of the increase in the fuel flow rate, and damages the combustor 1030. Thus, the abnormality of the expander 1020 is propagated to the combustor 1030.

When the abnormality is involved, as described in example 2, by calculating and displaying the loss prediction amount based on the diagnosis results obtained using the diagnosis units having different detection timings and accuracies, it is possible to provide information useful for determining whether or not to process the detected abnormality.

Industrial applicability of the invention

The present invention can be widely applied as a diagnostic device for plant equipment.

Claims

1. A plant equipment diagnosis device provided with a plurality of diagnosis units for diagnosing abnormality of a plant equipment, comprising:

a comprehensive diagnosis unit which obtains the accuracy of detection of the state abnormality of each of the plurality of diagnosis units based on the measurement signal data on the state of the plant and the equipment management information data on the past state abnormality, and estimates the predicted loss amount based on the accuracy and the loss amount associated with the state abnormality,

the plurality of diagnostic units each differ in timing of detection of the state abnormality, and the loss prediction amount is evaluated based on a maximum value among the accuracies of each of the plurality of diagnostic units and the loss amount accompanying the state abnormality.

2. The plant equipment diagnostic apparatus of claim 1,

the diagnostic device further includes a display unit that displays the detection result of the diagnostic unit and the loss prediction amount.

3. The plant equipment diagnostic apparatus of claim 1,

the integrated diagnostic unit determines the accuracy by dividing the number of state abnormalities in a predetermined period by the number of detection of state abnormalities by the diagnostic unit.

4. The plant equipment diagnostic apparatus of claim 1,

the equipment management information data includes failure information including failure contents, countermeasure costs, lead time required for preventing a failure, the number of days of plant equipment stoppage when a failure occurs, and the amount of opportunity loss due to the plant equipment stoppage.

5. The plant equipment diagnostic apparatus of claim 1,

the integrated diagnostic unit obtains an average lead time obtained by subtracting the time at which the occurrence of the state abnormality is detected by the diagnostic unit from the time at which the measurement signal data deviates from a predetermined threshold value set based on the device management information data and diagnosed as the state abnormality.

6. The plant equipment diagnostic apparatus of claim 1,

the diagnosis unit uses at least one of a model diagnosis using a machine model simulating characteristics of machines constituting the plant equipment or a cluster diagnosis using an adaptive resonance theory.

7. The plant equipment diagnostic apparatus of claim 1,

the plurality of diagnostic units classify the measured signal data into a plurality of categories according to similarity.

8. The plant equipment diagnostic apparatus of claim 1,

the comprehensive diagnosis unit corrects the accuracy according to the degree of influence of the state abnormality.

9. The plant equipment diagnostic apparatus of claim 2,

the display unit displays a detection result obtained by a diagnosis unit that minimizes the loss prediction amount with respect to a target value of a cost for maintaining the state abnormality.

10. A method for diagnosing plant equipment, which diagnoses abnormality of the state of the plant equipment by using a plurality of diagnostic units,

obtaining accuracy of detection of the state abnormality of each of the plurality of diagnostic units based on measurement signal data on the state of the plant equipment and equipment management information data on a past state abnormality, evaluating a loss prediction amount based on the accuracy and a loss amount associated with the state abnormality,