CN102870057B - Plant diagnosis device, diagnosis method, and diagnosis program - Google Patents

Plant diagnosis device, diagnosis method, and diagnosis program Download PDF

Info

Publication number
CN102870057B
CN102870057B CN201080065911.2A CN201080065911A CN102870057B CN 102870057 B CN102870057 B CN 102870057B CN 201080065911 A CN201080065911 A CN 201080065911A CN 102870057 B CN102870057 B CN 102870057B
Authority
CN
China
Prior art keywords
classification
plant equipment
operating condition
judged
change
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Active
Application number
CN201080065911.2A
Other languages
Chinese (zh)
Other versions
CN102870057A (en
Inventor
关合孝朗
江口彻
楠见尚弘
深井雅之
清水悟
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Hitachi Ltd
Original Assignee
Hitachi Ltd
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Hitachi Ltd filed Critical Hitachi Ltd
Publication of CN102870057A publication Critical patent/CN102870057A/en
Application granted granted Critical
Publication of CN102870057B publication Critical patent/CN102870057B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Classifications

    • G—PHYSICS
    • G05—CONTROLLING; REGULATING
    • G05B—CONTROL OR REGULATING SYSTEMS IN GENERAL; FUNCTIONAL ELEMENTS OF SUCH SYSTEMS; MONITORING OR TESTING ARRANGEMENTS FOR SUCH SYSTEMS OR ELEMENTS
    • G05B23/00—Testing or monitoring of control systems or parts thereof
    • G05B23/02—Electric testing or monitoring
    • G05B23/0205—Electric testing or monitoring by means of a monitoring system capable of detecting and responding to faults
    • G05B23/0218—Electric testing or monitoring by means of a monitoring system capable of detecting and responding to faults characterised by the fault detection method dealing with either existing or incipient faults
    • G05B23/0224—Process history based detection method, e.g. whereby history implies the availability of large amounts of data
    • G05B23/0227—Qualitative history assessment, whereby the type of data acted upon, e.g. waveforms, images or patterns, is not relevant, e.g. rule based assessment; if-then decisions
    • G05B23/0229—Qualitative history assessment, whereby the type of data acted upon, e.g. waveforms, images or patterns, is not relevant, e.g. rule based assessment; if-then decisions knowledge based, e.g. expert systems; genetic algorithms
    • G—PHYSICS
    • G05—CONTROLLING; REGULATING
    • G05B—CONTROL OR REGULATING SYSTEMS IN GENERAL; FUNCTIONAL ELEMENTS OF SUCH SYSTEMS; MONITORING OR TESTING ARRANGEMENTS FOR SUCH SYSTEMS OR ELEMENTS
    • G05B23/00—Testing or monitoring of control systems or parts thereof
    • G05B23/02—Electric testing or monitoring
    • G05B23/0205—Electric testing or monitoring by means of a monitoring system capable of detecting and responding to faults
    • G05B23/0259—Electric testing or monitoring by means of a monitoring system capable of detecting and responding to faults characterized by the response to fault detection
    • G05B23/0283—Predictive maintenance, e.g. involving the monitoring of a system and, based on the monitoring results, taking decisions on the maintenance schedule of the monitored system; Estimating remaining useful life [RUL]

Landscapes

  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Automation & Control Theory (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Evolutionary Biology (AREA)
  • Testing And Monitoring For Control Systems (AREA)

Abstract

Disclosed is a plant diagnosis device which detects abnormality in the subject being diagnosed from the measurement signal of the plant being diagnosed and reports the abnormality. Specifically disclosed is a plant diagnosis device which is configured of: a state variation detection unit which categorizes and stores the measurement signals of the plant being diagnosed and determines that the state has varied, when the measurement signal does not belong to the categories obtained by the categorization; an abnormality determination unit which determines that abnormality has occurred in the subject being diagnosed, using a signal from the state variation detection unit as a part of inputted signals; an alarm generation means which reports the output from the abnormality determination unit to the outside; and an erroneous report inhibition unit which detects events other than the abnormality in the plant being diagnosed and inhibits the output from the abnormality determination unit.

Description

The diagnostic device of plant equipment, diagnostic method and diagnostic routine
Technical field
The present invention relates to the diagnostic device of plant equipment, diagnostic method and diagnostic routine.
Background technology
The diagnostic device of plant equipment, when plant equipment creates abnormal transition events, accident etc., detects the generation of this exception, accident based on the measurement data from plant equipment.
Patent Document 1 discloses the diagnostic device adopting adaptive resonance theory (Adaptive Resonance Theory:ART).The diagnostic device of ART is adopted to have function multidimensional data being categorized into classification according to its similar degree.
In the technology of patent documentation 1, first use ART that measurement data time normal is categorized into multiple classification (normal category).Secondly, use ART that current measurement data is categorized into classification.When this measurement data cannot be categorized into normal category, generate new classification (new classification).The generation of new classification means that the state of plant equipment there occurs change.Therefore, utilize the generation of new classification to judge abnormal generation, be diagnosed as exception when the generation rate of new classification has exceeded threshold value.
At first technical literature
Patent documentation
Patent documentation 1: Japanese Unexamined Patent Publication 2005-165375 publication
Summary of the invention
(inventing problem to be solved)
In patent documentation 1, based on " current measurement data normal category cannot be categorized into when, generate new classification when namely the state of plant equipment there occurs change " this prerequisite, and be judged as exception.
But the state of plant equipment not only just changes when creating extremely, also can change when timeliness deterioration (aging degradation), operating condition change.Further, now also new classification can be produced.Thus, the exception creating new classification, namely create plant equipment now can not be thought.
Here, the operating condition environmental baseline (atmospheric temperature, humidity etc.), the amount (generated energy that plant equipment generates) that determined by the operation of operating personnel etc. that refer to the place set by plant equipment and instrument characteristics do not have the outside environmental elements of direct relation.Further, have nothing to do with the presence or absence that exception occurs, if operating condition changes, the value of various measurement data also changes.
Although timeliness is deteriorated, operating condition change is not abnormal, exception can be diagnosed as in the method for patent documentation 1.That is, normal condition can be diagnosed as exception, result becomes the reason producing wrong report.
The object of the invention is to by difference is deteriorated as the timeliness of the essential factor of state change, operating condition changes, the abnormal generation rate reducing wrong report.
(for solving the means of problem)
The diagnostic device of a kind of plant equipment of the present invention, detect the exception of diagnosis object plant equipment according to the measuring-signal of diagnosis object plant equipment and report exception, the diagnostic device of described plant equipment comprises: state change test section, the measuring-signal of diagnosis object plant equipment is categorized into classification to store by it, is judged to be that state there occurs change when not belonging to classified classification; Abnormality determination unit, the signal from state change test section is judged to create exception in diagnosis object plant equipment as the part inputted by it; Generate Alert block, its by the output report of this abnormality determination unit to outside; With wrong report blocking portion, the event beyond the exception of its detection diagnosis object plant equipment, stops the output of abnormality determination unit.
In addition, preferred wrong report blocking portion comprises timeliness degradation portion, this timeliness degradation portion uses the measuring-signal of diagnosis object plant equipment to obtain weight coefficient, be judged to be the scope having departed from timeliness deterioration when the variable quantity of weight coefficient has exceeded maximal value, timeliness degradation portion stops the output of abnormality determination unit when not departing from the scope of timeliness deterioration.
In addition, preferred wrong report blocking portion comprises operating condition change test section, the operating condition Data classification of diagnosis object plant equipment is become classification to store by this operating condition change test section, be judged to be that operating condition does not change when not producing new classification, operating condition change test section stops the output of abnormality determination unit when operating condition there occurs change.
In addition, preferably the outside environmental elements of direct relation is not had to form as the operating condition data of the input of operating condition change test section by the characteristic with the apparatus forming diagnosis object plant equipment.
The diagnostic device of a kind of plant equipment of the present invention, detect the exception of diagnosis object plant equipment according to the measuring-signal of diagnosis object plant equipment and report exception, the diagnostic device of described plant equipment comprises: state change test section, the measuring-signal of diagnosis object plant equipment is categorized into classification to store by it, is judged to be that state there occurs change when not belonging to classified classification; Timeliness degradation portion, it uses the measuring-signal of diagnosis object plant equipment to obtain weight coefficient, is judged to be the scope having departed from timeliness deterioration when the variable quantity of weight coefficient has exceeded maximal value; Operating condition change test section, the operating condition Data classification of diagnosis object plant equipment is become classification to store by it, is judged to be that operating condition does not change when not producing new classification; Abnormality determination unit, it there occurs change in the state that is judged to be by state change test section and is judged to have departed from the scope of timeliness deterioration by timeliness degradation portion and when being judged to be that operating condition does not change by operating condition change test section, is judged to be that diagnosis object creates exception; And Generate Alert block, its by the output report of abnormality determination unit to outside.
The diagnostic device of a kind of plant equipment of the present invention, possesses: measuring-signal database, and it preserves the measuring-signal of diagnosis object plant equipment, process data extracting unit, it extracts the diagnostic signal in order to diagnose the state of diagnosis object plant equipment to use from measuring-signal database, reference signal data storehouse, it preserves diagnostic signal, taxon, the Data classification preserved in reference signal data storehouse is become classification by it, classification results database, classification is preserved as normal category by it, diagnosis unit, it uses the information of the up-to-date diagnostic signal extracted by process data extracting unit and the normal category of preserving in classification results database, and the state of diagnosis diagnosis object plant equipment belongs to normally, exception, operating condition change, which kind of situation of timeliness deterioration, diagnostic result database, it preserves the diagnostic result of diagnosis unit, with the unit exporting the information of preserving in diagnostic result database to image display device, wherein, diagnosis unit possesses abnormality determination unit, state change test section, timeliness degradation portion and operating condition change test section, state change test section possesses following function: be judged to be that state there occurs change when the up-to-date diagnostic signal extracted by process Data extracting section does not belong to the normal category of preserving in classification results database, timeliness degradation portion possesses following function: time variations amplitude during derivation normal condition and the relation between weight coefficient variable quantity, the scope having departed from timeliness deterioration is judged to be when the maximal value of the variable quantity of the weight coefficient when having exceeded normal condition in diagnostic procedure, operating condition change test section possesses following function: when being judged to be that operating condition does not change by by when not producing new classification when becoming classification with the operating condition Data classification that instrument characteristics does not have the outside environmental elements of direct relation to form, in abnormality determination unit, be judged to be that state there occurs change by state change test section, and be judged to be by timeliness degradation portion the scope having departed from timeliness deterioration, and when not being judged to be that operating condition does not change by operating condition change test section, be judged to be exception.
The diagnostic method of a kind of plant equipment of the present invention, detect the exception of diagnosis object plant equipment according to the measuring-signal of diagnosis object plant equipment and report exception, wherein the measuring-signal of diagnosis object plant equipment is categorized into classification to store, be judged to be that state there occurs change when not belonging to classified classification, the measuring-signal of diagnosis object is used to obtain weight coefficient, the scope having departed from timeliness deterioration is judged to be when the variable quantity of weight coefficient has exceeded maximal value, the operating condition Data classification of diagnosis object is being become classification to store, be judged to be that operating condition does not change when not producing new classification, now, be judged to create exception in diagnosis object plant equipment, and report is to outside.
The diagnostic routine of a kind of plant equipment of the present invention, detect the exception of diagnosis object plant equipment according to the measuring-signal of diagnosis object plant equipment and report exception, the diagnostic routine of described plant equipment comprises: state change detecting step, the measuring-signal of diagnosis object being categorized into classification to store, being judged to be that state there occurs change when not belonging to classified classification; Abnormality juding step, is judged to create in diagnosis object exception using the signal from state change detecting step as the part inputted; Alarm generating step, by the output report of abnormality juding step to outside; Stop step with wrong report, the event beyond the exception of detection diagnosis object, stops the output of abnormality determination unit.
In addition, preferred wrong report stops step to comprise timeliness degradation step, in this timeliness degradation step, use the measuring-signal of diagnosis object plant equipment to obtain weight coefficient, the scope having departed from timeliness deterioration is judged to be when the variable quantity of weight coefficient has exceeded maximal value, in this timeliness degradation step, stop the output of abnormality juding step when not departing from the scope of timeliness deterioration.
In addition, preferred wrong report stops step to comprise operating condition change detecting step, in this operating condition change detecting step, the operating condition Data classification of diagnosis object plant equipment is become classification to store, be judged to be that operating condition does not change when not producing new classification, operating condition in operating condition change detecting step, stops the output of abnormality juding step when there occurs change.
In addition, preferably the outside environmental elements of direct relation is not had to form as the operating condition data of the input of operating condition change detecting step by the characteristic with the apparatus forming diagnosis object.
The diagnostic routine of a kind of plant equipment of the present invention, detect the exception of diagnosis object plant equipment according to the measuring-signal of diagnosis object plant equipment and report exception, the diagnostic routine of described plant equipment comprises: state change detecting step, the measuring-signal of diagnosis object plant equipment being categorized into classification to store, being judged to be that state there occurs change when not belonging to classified classification; Timeliness degradation step, uses the measuring-signal of diagnosis object plant equipment to obtain weight coefficient, is judged to be the scope having departed from timeliness deterioration when the variable quantity of weight coefficient has exceeded maximal value; Operating condition change detecting step, becoming classification to store by the operating condition Data classification of diagnosis object plant equipment, being judged to be that operating condition does not change when not producing new classification; Abnormality juding step, there occurs change in the state that is judged to be by state change detecting step and is judged to have departed from the scope of timeliness deterioration by timeliness degradation step and is judged to be judged to be that diagnosis object creates exception when operating condition does not change by operating condition change detecting step; With alarm generating step, by the output report of abnormality juding step to outside.
(invention effect)
By changing timeliness deterioration, operating condition, extremely distinguish, thus enough generation rates subtracting reduction wrong report.
In addition, belong to timeliness deterioration, operating condition change, which kind of abnormal situation by the reason providing data trend to change to operating personnel, thus the maintenance plan contributing to plant equipment is developed programs.
Accompanying drawing explanation
Fig. 1 is the block diagram representing diagnostic device of the present invention.
Fig. 2 (a) is the figure of the process flow diagram represented under normal condition mode of learning.
Fig. 2 (b) is the figure of the process flow diagram represented under diagnostic mode.
Fig. 3 (a) is the figure representing the diagnosis making 2 Modal action in each sampling period.
Fig. 3 (b) represents the figure carrying out normal condition mode of learning within specified time limit, carry out the diagnosis of diagnostic mode within each cycle.
Fig. 3 (c) is the figure representing the example that the moment set by operating personnel carries out.
Fig. 4 (a) is the block diagram representing data pre-processing device and ART module.
Fig. 4 (b) is the block diagram of the F0 layer representing ART module.
Fig. 4 (c) is the block diagram of the F1 layer representing ART module.
The curve map of the result of Fig. 5 (a) category classification that has been the carrying out of the embodiment representing Fig. 4.
Fig. 5 (b) is the curve map of the time dependent example of measuring-signal illustrated when occurring abnormal.
Fig. 6 (a) is the picture setting diagnostic data items.
Fig. 6 (b) is the block diagram of the taxon 400 of key diagram 1.
Fig. 7 (a) is the figure of the picture of the form representing the data of preserving in measuring-signal database.
Fig. 7 (b) is the figure of the picture of the form representing the data of preserving in reference signal data storehouse.
Fig. 7 (c) is the figure of the picture of the form representing the data of preserving in classification results database.
Fig. 8 (a) is the figure of the formation representing abnormality determination unit 510.
Fig. 8 (b) is the figure of the form of the data demonstrating diagnostic result database on picture.
Fig. 8 (c) is the picture indication example of the diagnostic result demonstrating diagnostic result database in time series.
Fig. 9 (a) is the figure of the action that timeliness degradation portion is described.
Fig. 9 (b) is the figure that the determinating reference that timeliness deterioration scope confirms is described.
Figure 10 (a) is the figure of the action of the 1st embodiment that operating condition change test section is described.
Figure 10 (b) is the figure of the action of the 2nd embodiment that operating condition change test section is described.
Figure 10 (c) is the figure of the action of the 3rd embodiment that operating condition change test section is described.
Figure 11 is the block diagram representing gas turbo-generator tool equipment.
Figure 12 is the figure of the result of the action example represented when applying diagnostic device of the present invention to gas turbo-generator tool equipment.
Figure 13 is the figure of the operation condition summarising each portion of diagnosis unit.
Embodiment
Below, use accompanying drawing that the present invention is described.
Embodiment
Fig. 1 is the block diagram representing diagnostic device of the present invention.In detail in this figure, the state of plant equipment 100 is diagnosed by diagnostic device 200.
In diagnostic device 200, possess as arithmetic unit: process data extracting unit 300, taxon 400, diagnosis unit 500, Generate Alert block 600.In addition, in diagnostic device 200, possess as database: measuring-signal database 230, reference signal data storehouse 240, classification results database 250, diagnostic result database 260.In addition, in detail in this figure, database is slightly designated as DB.Database is the inscape of diagnostic device 200 referred in this, and the information recorded in each database is electronic, is commonly called e-file (electronic data).
In addition, in diagnostic device 200, possess as the interface with outside: outer input interface 210 and outside output interface 220.And, via outer input interface 210, the external input signal 2 that the measuring-signal 1 comprising the value of the various quantity of states measuring plant equipment 100 and the operation of the external input device 910 comprising keyboard 920 and mouse 930 that possessed by running management room 900 are generated is inputed to diagnostic device 200.In addition, via outside output interface 220, image display information 14 is exported to the image display device 940 (image displaying part) in running management room 900 from diagnostic device 200.
In addition, in the present embodiment, process data extracting unit 300, taxon 400, diagnosis unit 500, Generate Alert block 600, measuring-signal database 230, reference signal data storehouse 240, classification results database 250, diagnostic result database 260 are all in the inside of diagnostic device 200, but also a part for these parts can be configured at the outside of diagnostic device 200, only data be communicated.In addition, set plant equipment as diagnosis object as 1 in the present embodiment, but also can diagnose multiple stage plant equipment by 1 diagnostic device 200.
Below, the action of diagnostic device 200 is described.First, the measuring-signal 3 be transfused to via outer input interface 210 is saved in measuring-signal database 230.
In process data extracting unit 300, among the measuring-signal 5 preserved measuring-signal database 230, extract diagnostic diagnostic signal 6, be saved in reference signal data storehouse 240.In reference signal data storehouse 240, preserve the measuring-signal that operating personnel are judged to be normal period.In addition, in process data extracting unit 300, data items is extracted in the setting based on operating personnel.Fig. 6 is used to describe this detailed content later.
Reference signal 7 is categorized into classification by taxon 400.Classification results 8 is kept in classification results database 250.In addition, Fig. 4 is used to describe the contents processing of taxon 400 later.
In diagnosis unit 500, the classification results 9 processing up-to-date diagnostic signal 6 and preserve in classification results database 250, diagnoses the state of plant equipment.Diagnosis unit 500 is changed test section 540 formed by abnormality determination unit 510, state change test section 520, timeliness degradation portion 530, operating condition.
State change test section 520 judges whether the state of plant equipment 100 there occurs change.From the state change result of determination 15 that state change test section 520 exports, being " 1 " when the state of being judged to be there occurs change, is " 0 " when being judged to be that state does not change.
In state change test section 520, the up-to-date diagnostic signal 6 relatively extracted by process data extracting unit 300 and the classification results 9 preserved in classification results database 250, when belonging to the classification that classification results 9 comprises, diagnostic signal 6 is categorized into this classification.On the other hand, compare the up-to-date diagnostic signal 6 and classification results 9 that are extracted by process data extracting unit 300, when not belonging to the classification that classification results 9 comprises, producing new classification and (being labeled as new classification below.)。Fig. 4 is used to describe the detailed contents processing of state change test section 520 later.
In timeliness degradation portion 530, judge that the state change of plant equipment is whether in the scope of timeliness deterioration.The timeliness deterioration judging result 16 exported from timeliness degradation portion 530, is " 1 ", is " 0 " when in the scope of timeliness deterioration when the state change of plant equipment has departed from the scope of timeliness deterioration.Fig. 9 is used to describe the detailed contents processing in timeliness degradation portion 530 later.
In operating condition change test section 540, judge whether the operating condition of plant equipment there occurs change.From the operating condition change result of determination 17 that operating condition change test section 540 exports, be " 1 " when operating condition does not change, be " 0 " when operating condition there occurs change.Figure 10 is used to describe the detailed contents processing of operating condition change test section 540 later.
In abnormality determination unit 510, using state change result of determination 15, timeliness deterioration judging result 16 and operating condition change result of determination 17 judges whether plant equipment creates exception.Fig. 8 is used to describe the detailed contents processing of abnormality determination unit 510 later.
Comprise in the diagnostic result 10 exported from diagnosis unit 500: the result of determination in abnormity diagnosis portion 510, state change result of determination 15, timeliness deterioration judging result 16, operating condition change result of determination 17.Diagnostic result 10 is kept in diagnostic result database 260.
In Generate Alert block 600, the measuring-signal 4 being used in the diagnostic result 11 preserved in diagnostic result database 260 and the up-to-date moment of preserving in measuring-signal database 230 determines whether to produce alarm.
Generate Alert block 600 has the determinating reference (condition 1 and condition 2) of following 2 kinds of generation alarms, and these determinating reference combination in any is determined whether producing alarm.Here, the combination in any that alarm produces refers to: produce alarm when following condition 1 and these two conditions of following condition 2 are all set up, one of them condition among following condition 1 and following condition 2 produces alarm etc. when setting up.
Condition 1: the measuring-signal 4 in up-to-date moment has departed from the scope (threshold value) of regulation.
By abnormality determination unit 510, condition 2: within specified time limit, is judged to be that abnormal ratio exceedes certain value (threshold value).
In addition, the threshold value in condition 1 and condition 2 is the value set by operating personnel.
When having carried out producing the judgement of alarm in Generate Alert block 600, alarm signal 13 has been sent to outside output interface 220 by Generate Alert block 600.Alarm signal 13 is transformed into image display information 14 via outside output interface 220, is shown in image display device 940.
In the present embodiment, use image display device 940 and operator to contact alarm, but be not limited thereto, also first can produce alarm song to the generation causing operator to note alarm.In addition, also can image display device 940 be made to show image display information 14 after generation alarm song or at generation alarm song simultaneously.
In addition, also the diagnostic result 12 (result of determination in abnormity diagnosis portion 510, state change result of determination 15, timeliness deterioration judging result 16, operating condition change result of determination 17) preserved in diagnostic result database 260 can be shown in image display device 940 via outside output interface 220.
In addition, the diagnostic device information 50 of preserving in measuring-signal database 230, reference signal data storehouse 240, classification results database 250 and diagnostic result database 260 can be shown in image display device 940.In addition, also can as required and use revise these information based on the external input signal 2 of the operation of external input device 910.
The invention is characterized in diagnosis unit 500, possess abnormality determination unit 510, timeliness degradation portion 530 and operating condition change test section 540, thus timeliness degradation, operating condition change detection can not be judged to be exception.
That is, although state change test section 520 detects the exception of plant equipment energetically, but originally should not be judged to be abnormal timeliness deterioration, operating condition change owing to comprising in the content of this detection, thus change and they are carried out comprehensive judgement via abnormality determination unit 510 by the deterioration of detection timeliness, operating condition, thus preventing wrong report.This just means, timeliness degradation portion 530 and operating condition change test section 540 can form the wrong report blocking portion for state change test section 520.
In state change test section 520, new classification generates when current measurement data cannot be categorized into normal category, when namely the state of plant equipment there occurs change.
But, due to the state of plant equipment be not only occur abnormal when change, also can change when timeliness deterioration, operating condition change, if thus state change test section 520 action, produce new classification.Here, the environmental baseline (atmospheric temperature, humidity etc.) that operating condition refers to the place set by plant equipment, the generated energy etc. exported from plant equipment, have nothing to do with the presence or absence that exception occurs, if operating condition changes, then the value of various measurement data also changes.
Although timeliness is deteriorated, operating condition change is not abnormal, only judging they to be diagnosed as exception in abnormal method with the generation of new classification.This becomes the reason producing wrong report.
In the present invention, by possessing abnormality determination unit 510, timeliness degradation portion 530, operating condition change test section 540 in diagnosis unit 500, thus because the timeliness having distinguished the essential factor changed as state is deteriorated, operating condition changes, exception, therefore the generation rate of wrong report can be reduced.
In addition, use the diagnostic device not possessing abnormality determination unit 510 of the present invention, timeliness degradation portion 530, operating condition change test section 540, in order to distinguish timeliness deterioration, operating condition change, extremely then need long-standing service data.That is, when making the action of normal condition mode of learning, the data evaluating all operating conditions and timeliness deterioration are needed.Therefore, from plant equipment running to start diagnosis during elongated.But the diagnostic device of the application of the invention, even if also can distinguish exception, timeliness deterioration, operating condition change in the stage that the diagnostic running of energy is few, thus can shorten during importing needed for diagnostic device.
Fig. 2 is the process flow diagram of the elemental motion of the diagnostic device 200 representing Fig. 1.Fig. 2 (a) is the figure representing normal condition mode of learning, Fig. 2 (b) is the figure representing diagnostic mode.
Below, the inscape described in Fig. 1 is used to be described.
Diagnostic device 200 has these 2 elemental motions of diagnostic mode based on the information of preserving in reference signal data storehouse 240, Data classification time normal being become the normal condition mode of learning of classification and the state of diagnosis plant equipment 100.
In Fig. 2 (a), perform normal condition mode of learning by carrying out step 1000 and 1010 successively.
First, in step S1000, make process data extracting unit 300 action, among the measuring-signal 5 of measuring-signal database 230, extract diagnostic signal 6.Diagnostic signal 6 is kept in reference signal data storehouse 240.The data of preserving in reference signal data storehouse 240 are the data that the operating condition of plant equipment 100 is judged to be normal period by operating personnel (operator).In addition, Fig. 6 is used to describe the data items preserved in reference signal data storehouse 240 later.
Then, in step S1010, make taxon 400 action, the reference signal 7 of preserving in reference signal data storehouse 240 is classified, classification results 8 is saved in classification results database 250.
Like this, under the normal condition mode of learning of Fig. 2 (a), perform the process of the process data extracting unit 300 of the row in the left side employed in the diagnostic device 200 of Fig. 1, measuring-signal database 230, reference signal data storehouse 240, taxon 400, classification results database 250.
Under the diagnostic mode shown in Fig. 2 (b), by carrying out step S1100, S1110 and S1120 to perform successively.
First, in step S1100, via outer input interface 210, the measuring-signal 1 from plant equipment 100 is taken in diagnostic device 200, and measuring-signal 3 is saved in measuring-signal database 230.
Then, make process data extracting unit 300 action, among measuring-signal database 230, extract measuring-signal 5, and the diagnostic signal 6 up-to-date moment is sent to diagnosis unit 500.
In step S1110, make diagnosis unit 500 action.In diagnosis unit 500, make operational part action according to the order of state change test section 520, timeliness degradation portion 530, operating condition change test section 540, abnormality determination unit 510.The diagnostic result 10 exported from diagnosis unit 500 is sent and is saved in diagnostic result database 260.The diagnostic result 12 exported from diagnostic result database 260 is transformed into image display information 14 via outside output interface 220, and exports image display device 940 to.
In step S1120, make Generate Alert block 600 action, determine whether to produce alarm.When producing alarm, the alarm signal 13 exported by Generate Alert block 600 is transformed into image display information 14 via outside output interface 220, and exports image display device 940 to.Thus, the operating personnel (operator) to plant equipment 100 notify alarm.
Like this, under the diagnostic mode of Fig. 2 (b), perform the diagnosis unit 500 (state change test section 520, timeliness degradation portion 530, operating condition change test section 540, abnormality determination unit 510) of the row on the right side employed in the diagnostic device 200 of Fig. 1, the process of outside output interface 220, image display device 940, Generate Alert block 600.In addition, its prerequisite yes employ outer input interface 210, measuring-signal database 230, process data extracting unit 300.
Fig. 3 illustrates the figure performing the process flow diagram (Fig. 2 (a)) of normal condition mode of learning of diagnostic device 200 and the sequential of the process flow diagram (Fig. 2 (b)) of diagnostic mode.
Diagnostic device 200 obtains measuring-signal 1 from plant equipment 100 in each sampling period.
In Fig. 3 (a), in each sampling period, normal condition mode of learning and these two Modal action of diagnostic mode are made to diagnose.
In addition, in Fig. 3 (b), make the action of normal condition mode of learning in during the setting of each regulation, only make diagnostic mode action to diagnose within each sampling period.
And in Fig. 3 (c), operating personnel implement to set the operation between the learning period, between diagnostic period, and make normal condition mode of learning and diagnostic mode action in this moment.
How no matter adopt, diagnostic mode can be performed within each sampling period, diagnose the state of plant equipment onlinely.
Below, be described successively by this order the contents processing under the contents processing of normal condition mode of learning and diagnostic mode particularly, its prerequisite is under which pattern, all perform the process inputted process variable being categorized into classification.Under normal condition mode of learning, taxon 400 is equivalent to the unit performing this process, and state change test section 520 is equivalent to the unit performing this process in the diagnostic mode.Thus, before entering the action specification under each pattern, the viewpoint of category classification method is described with the block diagram of Fig. 4 as public awareness.
Below, describe the situation to taxon 400 and state change test section 520 application self-adapting resonance theory (Adaptive Resonance Theory:ART), but also can adopt other clustering methods such as vector quantization.
As shown in Fig. 4 (a), taxon 400 and state change test section 520 are made up of data pre-processing device 710 and ART module 720.And ART module 720 possesses F0 layer 721, F1 layer 722, F2 layer 723, storer 724 and chooser system 725, and they be combined with each other.In addition, F0 layer 721, F1 layer 722 are such as formed as shown in Fig. 4 (b), Fig. 4 (c).
In this Fig. 4 (a), first, service data is transformed into the input data of ART module 720 by data pre-processing device 710.Specifically, following (1) (2) formula is performed.Below, this program (step) is described.
First, by each measure the item, calculate maximal value and minimum value.The maximal value that use calculates and minimum value carry out standardization to data.Here, for the process variable xi of plant equipment, standardized method is described.
The data number of xi is N number of, and the n-th measured value is set to xi (n).In addition, the maximal value in N number of data and minimum value are set to Max_i, Min_i respectively, then following (1) formula of the data Nxi (n) after standardization represents.
[mathematical expression 1]
Nxi(n)＝α+(1-α)×(xi(n)-Min_i)/(Max_i-Min_i)…(1)
At this, α (0≤α < 0.5) is constant, and according to above-mentioned (1) formula, data are standardized as the scope of [α, 1-α].
Then, the complement of the data after normalized, joins in input data.The complement CNxi (n) of standardized data Nxi (n) calculates by following (2) formula.
[mathematical expression 2]
CNxi(n)＝1-Nxi(n) …(2)
In data pre-processing device 710, above-mentioned (1) (2) formula is performed to multiple input data, using comprising the data of the complement CNxi (n) of standardized data Nxi (n) and the standardized data obtained as its result as input data Ii (n), input to ART module 720.Above step is included in the service data of carrying out in data pre-processing device 710 in the input data transformation process of ART module 720.
In ART module 720, data Ii (n) will be inputted and be categorized into multiple classification.For this reason, ART module 720 possesses F0 layer 721, F1 layer 722, F2 layer 723, storer 724 and chooser system 725, and they be combined with each other.F1 layer 722 and F2 layer 723 combine via weight coefficient.Weight coefficient represents the prototype (prototype) of the classification that input data are classified.Here, other typical value of prototype representation class.
Then, the algorithm of ART module 720 is described.
Algorithm summary when have input from input data to ART module 720 processes shown in 1 ~ process 5 described as follows.
Process 1: in Fig. 4 (b), by representing the F0 layer 721 of contents processing, carries out standardization by input vector, removes noise.
Process 2: by the comparison of the input data and weight coefficient that input to F1 layer 722, select the candidate of suitable classification according to the contents processing of Fig. 4 (c).
Process 3: the appropriateness evaluating the classification selected by chooser system 725 according to the ratio with parameter ρ.If be judged as suitably, then input data and be classified in this classification, enter process 4.On the other hand, if be not judged as suitably, then this classification is reset, and selects the candidate (re-treatment 2) of suitable classification among other classifications.If increase the value of parameter ρ, the classification of classification attenuates, if reduce the value of ρ, classifies thicker.This parameter ρ is called warning (vigilance) parameter.
Process 4: if known classifications whole in process 2 is reset, then judge that input data belong to new classification, generate the new weight coefficient of the prototype representing new classification.
Process 5: if input data are classified into classification J, then corresponding with classification J weight coefficient WJ (new), uses weight coefficient WJ (old) in the past and input data p (or the data gone out by input data fork) to be updated according to following (3) formula.
[mathematical expression 3]
WJ(new)＝Kw·p+(1-Kw)·WJ(old) …(3)
Here, Kw is Study rate parameter (0 < Kw < 1), is the value determining degree input vector being reflected to new weight coefficient.
The feature of the data classification algorithm of ART module 720 is above-mentioned process 4.
In process 4, when have input from the classification results database 250 of Fig. 1 record the style of (preservation) different input data, recorded style can not be changed and record new style.Thus, the style learnt of can recording over records new design.
Like this, if give as the service data that gives of input data in advance, then ART module 720 learns the style that gives.Therefore, in the ART module 720 that study is complete, have input new input data, then can judge with which style in past to be close according to above-mentioned algorithm.In addition, if the style for not lived through in the past, then new classification is classified into.
Specifically, above whole process 1 ~ process 5 is performed by following step.
First, Fig. 4 (b) is the block diagram of the formation representing F0 layer 721.F0 layer 721 is made up of function blocks 71,72,73,74, manages in functional block 71,72,73,74 performing following (4) (5) (6) (7) formula respectively throughout, obtains standardization input vector.
A succession of process in the F0 layer 721 of Fig. 4 (b) is: will give the input data Ii standardization again of function blocks 71 in each moment, generates the final standardization input vector Ui exporting F1 layer 721 and chooser system 725 from function blocks 74 to.In addition, in following (4) (5) (6) (7) formula, " i " represents project data number, and " 0 " represents F1 layer.
First, according to (4) formula, calculate Wi0 according to input data Ii.Here, a is constant.
[mathematical expression 4]
Wi0＝Ii+aUi0 …(4)
Then, the Xi0 after using (5) formula to calculate the Wi0 standardization making (4) formula.Here, W0 be by Wi0 norm after value.
[mathematical expression 5]
Xi0＝Wi0/W0 …(5)
Then, (6) formula is used to calculate the Vi0 to eliminate noise from Xi0 after.Wherein, θ is the constant for removing noise.According to the calculating of (6) formula, because small value becomes 0, so the noise of input data is removed.
[mathematical expression 6]
Vi0=f (Xi0)=Xi0 (as Xi0 >=θ)
=0 (in situation other than the above) ... (6)
Finally, (7) formula is used to obtain standardization input vector Ui0.Wherein, V0 be by Vi0 norm after value.The Ui0 finally obtained is the input of F1 layer.
[mathematical expression 7]
Ui0＝Vi0/V0 …(7)
Fig. 4 (c) is the block diagram of the formation representing F1 layer 722.In F1 layer 722, the Ui0 obtained by (7) formula is kept in the mode of short-term storage, calculate the Pi finally inputing to F2 layer 723.
F2 layer is made up of function blocks 75,76,77,78,79,80, manages in functional block 75,76,77,78,79,80 performing following (8) (9) (10) (11) (12) (13) formula respectively throughout.This calculating formula is concluded and is expressed as (8) ~ (13) formula.Wherein, a, b are constant, and f () serves as reasons the function shown in (6) formula, and Tj is the grade of fit calculated by F2 layer 722.In addition, the denominator of (9) (11) (12) formula is the value after norm.
[mathematical expression 8]
Wi＝Ui0+aUi …(8)
[mathematical expression 9]
Xi＝Wi/W …(9)
[mathematical expression 10]
Vi＝f(Xi)+bf(Qi)…(10)
[mathematical expression 11]
Ui＝Vi/V …(11)
[mathematical expression 12]
Qi＝Pi/P …(12)
[mathematical expression 13]
Pi＝Ui+∑g(yi)Zij
Wherein, g (yi)=d (Tj=max (Tj))
=0 (situation other than the above) ... (13)
Utilize the algorithm of the ART module 720 of Fig. 4 that the process variable of input plant equipment is carried out the results are shown in Fig. 5 of category classification.Fig. 5 (a) is the curve map of an example of presentation class result.2 projects demonstrated in measurement data as an example in this figure, mark with 2 dimension curve figure.The longitudinal axis and the measurement data of transverse axis to each project are standardized and are represented.
Measurement data is divided into multiple classification 750 (circle shown in Fig. 5 (a)) by the ART module 720 of Fig. 4 (a).
In detail in this figure, the measurement data for 2 projects represents with the curve map of 2 dimensions, but is not limited thereto, and also the coordinate of multidimensional can be used to carry out making of classification for measurement data more than 3 projects.
The classification results of Fig. 5 (a) obtains in taxon 400 under the normal condition mode of learning of the Fig. 2 (a) illustrated before, and this result is accumulated in classification results database 250.In addition, obtain in state change test section 520 under the diagnostic mode of Fig. 2 (b).
Time dependent example when Fig. 5 (b) illustrates that the measuring-signal 1 got from plant equipment 100 changes according to the generation of exception.Transverse axis gets the time, and the longitudinal axis gets the generation ratio (generation frequency) of measuring-signal, class number and new classification.Data D1 and D2 corresponds respectively to project A and B.
In detail in this figure, initial items A and B is roughly stabilized in certain value, reduces at the tight earlier data D1 of moment t1 (project A), then increases the tight subsequent data D2 (project B) of moment t1.Then, data D2 (project B) reduces, and final data D1 (project A) and data D2 (project B) increases.
In addition, before due in t1, the class number be classified is 1 ~ 4, classification, normal category when this is benchmark.In contrast, after have passed through moment t1, the class number of project A and B is 5 ~ 7, becomes and represents that the state of plant equipment there occurs the new classification of change.
Be accompanied by this, increase from the generation ratio (generation frequency) of firm new classification through after moment t1, until exceed threshold value and the state that is diagnosed as there occurs change.Here, the moving average that the generation ratio of new classification is used in the number of the new classification produced in specified time limit calculates.
Such state change changes test section 520 by state and detects under the diagnostic mode of Fig. 2 (b).In state change test section 520, the state that is diagnosed as when the moving average of the number of the new classification produced within specified time limit can be selected to have exceeded threshold value there occurs the method for change or there occurs change and one of them method among the method that the mode do not changed for being diagnosed as state when normal category is diagnosed each sample according to the state that is diagnosed as when produced classification is new classification.
Fig. 6 (a) be for set by the data items that extracts of process data extracting unit 300, the picture of the image display device 940 of Fig. 1.Diagnostic data items is decided by the object (wanting the content of the exception detected) of diagnosis.
This picture 940 can scroll up square in length and breadth and append label (showing the label of group 1 and group 2 in legend).On picture, process variable (data items) A, B, C, D etc. of plant equipment show together with its process number PID.In addition, show maximum, the minimum value of each process variable (data items) etc., the object (wanting the content of the exception detected) etc. of the relevant or diagnosis that operating personnel judge between each process variable selects to want the combination of the process variable dividing into groups to manage.In the label of the group 1 of figure, represent that process variable (data items) A, C, D have relation each other and navigate to the group that monitor and select.
Fig. 6 (b) is the block diagram of the formation of presentation class unit 400, is configured with data pre-processing device and ART module that Fig. 4 (a) illustrated for the process variable in the group set by Fig. 6 (a).That is, in taxon 400, with the quantity of group correspondingly, perform a succession of process in data pretreating device and ART module.
Fig. 7 (a), Fig. 7 (b), Fig. 7 (c) illustrate the form of the data of preserving in the measuring-signal database 230 of Fig. 1, reference signal data storehouse 240 and classification results database 250 respectively.Also these figure can be thought the display frame of the image display device 940 of Fig. 1.In addition, in picture 940, in length and breadth direction scroll bar and carried out the point of label display as required, be identical in Fig. 6 (a) or other pictures.
As shown in Fig. 7 (a), in measuring-signal database 230, preserve the value of the multiple data items (project A, B, C etc.) measured by plant equipment 100 by each sampling period (moment of the longitudinal axis).In display frame 940, on the longitudinal axis, temporally sequence shows the value of these data items (project A, B, C etc.), by use can cross shifting scroll box 56a and 56b thus can the data of roll display wide region.
In addition, when the reference signal data storehouse 240 of Fig. 7 (b), by selecting mark 57a, 57b of the data sheet representing benchmark 1 ~ 2, thus only can gather the project of display by benchmark classification.
In the process data extracting unit 300 of Fig. 1, among measuring-signal database 230, be extracted in the data group used the diagnosis of plant equipment 100.Such as, the state that the data group when data group of display reference signal is 2 (" benchmark 1 ", " benchmark 2 ") and have selected " benchmark 1 " in Fig. 7 (b) is made up of project A, project C and project D.The data corresponding with the data items of the group 1 set by Fig. 6 (a) are benchmark 1, and the data corresponding with the data items of group 2 are benchmark 2.
Like this, in measuring-signal database 230, as shown in Fig. 7 (a), the measured value of total data project is preserved in seasonal effect in time series mode as 1 data group, and the measured value of the data items extracted by process data extracting unit 300 as shown in Fig. 7 (b) in reference signal data storehouse 240 is preserved in seasonal effect in time series mode as multiple data group.
And then Fig. 7 (c) is the display frame of the form representing the data of preserving in the classification results database 250 of Fig. 1.In Fig. 7 (c) left side display moment and the relation between the data of now inscribing and the class number of classification, the relation between Fig. 7 (c) right side Display Category numbering and weight coefficient.Like this, in classification results database 250, save the classification results of each data group that reference signal data storehouse 240 is preserved.
Fig. 8 (a) is the block diagram forming abnormality determination unit 510.In abnormality determination unit 510, the operating condition change result of determination 17 in the state change result of determination 15 in using state change test section 520, the timeliness deterioration judging result 16 in timeliness degradation portion 530, operating condition change test section 540 judges whether plant equipment creates exception.
As Figure 13 conclude, state change result of determination 15 when the state of being judged to be there occurs change for " 1 ", when being judged to be that state does not change is " 0 ".Timeliness deterioration judging result 16 is " 1 " when the state change of plant equipment has departed from the scope of timeliness deterioration, is " 0 " when in the scope of timeliness deterioration.Operating condition change result of determination 17 is " 1 " when operating condition does not change, be then " 0 " when operating condition there occurs change.
The result obtained with door 514 these digital signals being inputed to Fig. 8 (a) becomes abnormality juding result 10.That is, to change and the change of the state of plant equipment has departed from the scope of timeliness deterioration and when operating condition do not change, abnormality juding result 10 has been set to " 1 ", is judged to create exception in the state of plant equipment.Under condition in addition, abnormality juding result 10 becomes " 0 ".
Fig. 8 (b) is the figure of the form demonstrating the data of preserving in diagnostic result database 260 on picture 940.State change result of determination, operating condition change result of determination, timeliness deterioration judging result, abnormality juding result are preserved by seasonal effect in time series mode respectively, and are shown on picture.In addition, in diagnostic result database 260, also save as make state change the moment shown in the result of test section 520 action and Fig. 7 (c) of obtaining and this time the data of inscribing and the class number of classification between relation and relation between class number and weight coefficient.
Fig. 8 (c) be by image display device 940 with seasonal effect in time series mode show in diagnostic result database 260 preserve diagnostic result 12 when picture indication example.Can as shown in Fig. 8 (b) result of numerically displaying time sequence, and can with the result of the mode displaying time sequence of curve map as shown in Fig. 8 (c).In addition, not only show " 0 ", " 1 ", also can show the moving average of certain specified time limit.In Fig. 8 (c), curve 551 represents the moving average of state change result of determination, curve 552 represents the moving average of operating condition change result of determination, curve 553 represents the moving average of timeliness deterioration judging result, and curve 554 represents the moving average of abnormality juding result.
Fig. 9 (a) is the accompanying drawing of the action that timeliness degradation portion 530 is described.As shown in the curve map of Fig. 9 (a) upper left, when timeliness deterioration, measured value changes gradually.That is, if horizontal axis representing time, the longitudinal axis represents the size of process variable, then when timeliness changes, project A, B show the trend of slowly change.On the other hand, as shown in the curve map of upper right, measured value changes with departing from the scope of timeliness deterioration sometimes, is now the situation representing that generation exception or operating condition change mostly.In addition, about the curve map of upper right, be also horizontal axis representing time, the longitudinal axis represents the size of process variable.
In addition, the identification of adjoint above-mentioned time variations also can be judged with the variation delta z of the weight coefficient of the variable quantity of measured value, classification.Relevant owing to existing between these indexs, so timeliness deterioration can be judged with Δ z.The curve map of the bottom of Fig. 9 (a) is all that transverse axis gets project A, and the longitudinal axis gets project B, and describes the variation delta z of the weight coefficient of the classification between the different moment.In timeliness change in left side, the variation delta z of the weight coefficient of classification is showed little, and when depart from the scope of timeliness deterioration at measured value changing, the variation delta z of the weight coefficient of classification is showed greatly.
In addition, even if when normal condition, along with the timeliness deterioration of plant equipment, the trend of data also can change.The intensity of variation of data trend during pre-recorded normal condition, if within the scope of this, be judged to be in the scope of timeliness deterioration.
Fig. 9 (b) is the figure of the determinating reference being illustrated as the scope of timeliness deterioration or the scope of disengaging timeliness deterioration.In the figure, transverse axis gets time variations amplitude, the heavy index variation amount of longitudinal axis weighting, the judgement line of the maximal value obtained under normal condition mode of learning as benchmark is prepared, according to actual measurement to time variations amplitude and weight coefficient variable quantity between relation whether departed from the scope that border determines whether changing in timeliness.
In the present embodiment, judge based on the result of relation between time variations amplitude when summarizing normal condition and weight coefficient variable quantity.Here, following (14) formula is used to calculate weight coefficient variable quantity relative to time variations amplitude.
[mathematical expression 14]
Max(∑(Wi(t+Δt)-Wi(t) 2) …(14)
In (14) formula, Δ t is time variations amplitude, and i is the symbol for identification data project, and wherein 1≤i≤n (n is data items sum), Wi (t) is the weight coefficient of the data items i under moment t.
In addition, in the diagnostic mode, when the maximal value of the variable quantity of the weight coefficient when having exceeded normal condition, the scope having departed from timeliness deterioration is judged to be.In addition, in the present embodiment, the maximal value of the variable quantity of weight coefficient during normal condition is judged to be threshold value, but also the mean value of the variable quantity of weight coefficient during normal condition can be set as threshold value.
Figure 10 (a) is the accompanying drawing of the 1st embodiment that operating condition change test section 540 is described.In the present embodiment, the ART process (process of data pre-processing device 710c and ART module 720d) of Fig. 4 identical with changing the process implemented in test section 520 in taxon 400, state is performed.Wherein, now to the input of ART and the process variable of non-mechanical devices but operating condition data, different in this.That is, utilize the sorted result of ART is carried out to judge whether operating condition there occurs change to the data of the group be only made up of operating condition data.As operating condition data candidate and enumerate the outside environmental elements that " generation power ", " temperature " etc. and instrument characteristics do not have direct relation.
When creating new classification when operating condition data of classifying with ART, the possibility that operating condition changes is high.On the contrary, when the operating condition data of classifying with ART, do not produce new classification, when creating new classification when category measurement data, the reason of state change is not the change of operating condition.Can judge whether operating condition there occurs change like this.
Figure 10 (b) is the figure of the 2nd embodiment that operating condition change test section 540 is described.
Here, also ART is used.In ART, as shown in Figure 10 (b), the longitudinal axis and transverse axis represent project A and project B respectively, and Data classification during benchmark is become several normal category.In this action, when creating new classification, calculating the similar degree become between the data of object and normal category, extracting most similar (similar degree is maximum) normal category.
Similar degree S such as uses following (15) formula to calculate, and extracts (similar degree is minimum) normal category that S is minimum.
[mathematical expression 15]
Sj＝∑(Di-Wij) 2 …(15)
Here, i is the symbol for identification data project, wherein 1≤i≤n (n is data items sum).In addition, j is for other symbol of recognition category, wherein 1≤j≤m (m is the sum of normal category).And Sj is similar degree, Di is the value of the data items i of the data becoming object, and Wij is the weight coefficient of the data items i in classification j.
Then, such as use following (16) formula to calculate the contribution degree Ci of each data items.
[mathematical expression 16]
Ci＝|Di-Wij| …(16)
The data items higher due to contribution degree is more separated with normal category, thus can be described as the data items becoming the reason producing new classification.When this data items is operating condition data, the reason that state changes is determined to be operating condition and there occurs change.
In the example shown in Figure 10 (b), observe for the position on the two-dimensional coordinate of normal category, new classification coordinate there occurs the change of what degree on the longitudinal axis, transverse axis.The change (contribution degree) of transverse axis is larger than the change of the longitudinal axis in the figure, it can thus be appreciated that the contribution degree of project A is large.
Figure 10 (c) is the figure of the 3rd embodiment that operating condition change test section 540 is described.In the present embodiment, the function illustrated in Figure 10 (a), Figure 10 (b) is combined with.
Result 542 after using Figure 10 (a) to judge and the contribution degree result 543 after using Figure 10 (b) to judge are input to operating condition change detection unit 544.In operating condition change detection unit 544, implement the digital processing of AND or OR, export operating condition change result of determination 545.
Below, action when diagnostic device 200 of the present invention being applied to thermal power generation machinery equipment is described.Figure 11 is the block diagram representing thermal power generation machinery equipment.
In detail in this figure, thermal power generation machinery equipment 100 comprises gas turbo-generator 110, control device 120 and data sending device 130.Gas turbo-generator 110 comprises generator 111, compressor 112, burner 113 and turbine 114.
When generating electricity, the air being sucked into compressor 112 being carried out compression and is used as pressurized air, this pressurized air is delivered to burner 113, carrying out mixing after-combustion with fuel.Use the gases at high pressure produced by burning that turbine 114 is rotated, and generated electricity by generator 111.
In control device 120, control the output of gas turbo-generator 110 according to electricity needs.In addition, the service data 102 that measured by the sensor (not shown) being arranged at gas turbo-generator 110 of control device 120 is as input data.Service data 102 is the quantity of states such as suction temperature, fuel input amount, turbine exhaust gas temperature, turbine rotation number, generator power output, turboshaft vibration, measures within each sampling period.In addition, also the weather informations such as atmospheric temperature are measured.
In control device 120, these service datas 102 are used to calculate control signal 101 for controlling gas turbo-generator 110.
The measuring-signal 1 comprising the service data 102 measured by control device 120 and the control signal 101 calculated by control device 120 is sent to diagnostic device 200 by signal data dispensing device 130.
Figure 12 is the accompanying drawing that to illustrate with the thermal power generation described in fig. 11 machinery equipment be object and result after using diagnostic device 200 to diagnose.
Figure 12 be data items as the operating condition data of group 1 and select generator to export a, atmospheric temperature b and as the data items of group 2 result that have selected when generator exports a, atmospheric temperature b, fuel flow rate c.
At moment T 0-T 1between, as normal data, for learning normal condition under mode of learning.Even if in normal state, due to timeliness change, efficiency reduces, so the fuel flow rate c obtained needed for identical data a adds.
From moment T 1work the diagnosis have come to using diagnostic device 200.At moment T 1-T 2although between be normal condition, due to timeliness change impact and cause fuel flow rate c to increase.Its result, because the value of fuel flow rate c changes, causes producing new classification when making state change test section 520 action.Therefore, through moment T 1afterwards, state change result of determination d moves closer to 1.In addition, in timeliness degradation portion 530, because in the scope when increase ratio of fuel flow rate c is in study, so be judged to be in the scope of timeliness deterioration.Above, be judged to exception not to occur in abnormality determination unit 510.
At moment T 2-T 3between, the temporary rising of atmospheric temperature b.If because atmospheric temperature b rises, efficiency reduces, then in order to the fuel flow rate c obtained needed for identical output a increases.Its result, produces new classification when making state change test section 520 action.In addition, in timeliness degradation portion 530, be judged to be that the increase ratio of fuel flow rate c has departed from scope when learning.In operating condition change test section 540, because atmospheric temperature b there occurs change, be thus judged to be that operating condition f changes.Based on more than, be judged to exception not to occur in abnormality determination unit 510.
After atmospheric temperature reduces, at moment T 4occur abnormal, concomitantly fuel flow rate c increases with it.Increase ratio due to fuel flow rate c departs from the timeliness deterioration scope of e and operating condition f does not also change, and is thus judged to create exception in abnormality determination unit 510.
When only using state changes the result enforcement abnormality juding of test section 520, at moment T 1can diagnose to produce later and give birth to exception.This is wrong report.The timeliness degradation portion 530 described in the application of the invention, operating condition change test section 540, thus moment T can be judged to be 1-T 4scope be normal, can rate of false alarm be cut down.
Like this, the diagnostic device 200 of the application of the invention, obtains the effect that can lower rate of false alarm.In addition, belong to timeliness deterioration, operating condition change, which kind of abnormal situation by the reason providing data trend to change to operating personnel, thus the maintenance plan contributing to plant equipment is developed programs.
In addition, although possessed abnormality determination unit 510, state change test section 520, timeliness degradation portion 530, operating condition change test section 540 in diagnosis unit 500 of the present invention, but also only can possess a wherein side of timeliness degradation portion 530 or operating condition change test section 540, for detecting wherein a kind of situation of the change of timeliness deterioration or operating condition.
In addition, can form having carried out illustrating the item also illustrated in the description above as diagnostic device, using computing machine to it can be used as diagnostic method to implement, or a series of program is embodied in program.
Utilizability in industry
The present invention can be widely used in various plant equipment etc. as the diagnostic device of plant equipment etc., diagnostic method, diagnostic routine.
Symbol description
100: plant equipment
200: diagnostic device
210: outer input interface
220: outside output interface
230: measuring-signal database
240: reference signal data storehouse
250: classification results database
260: diagnostic result database
300: process data extracting unit
400: taxon
500: diagnosis unit
510: abnormality determination unit
520: state change test section
530: timeliness degradation portion
540: operating condition change test section
600: Generate Alert block
900: running management room
910: external input device
920: keyboard
930: mouse
940: image display device

Claims (10)

1. a diagnostic device for plant equipment, detects the exception of diagnosis object plant equipment according to the measuring-signal of diagnosis object plant equipment and reports exception, and the feature of the diagnostic device of this plant equipment is, comprising:
State change test section, the measuring-signal of described diagnosis object plant equipment is categorized into classification to store by it, is judged to be that state there occurs change when not belonging to classified classification;
Abnormality determination unit, the signal from this state change test section is judged to create exception in described diagnosis object plant equipment as the part inputted by it;
Generate Alert block, its by the output report of this abnormality determination unit to outside; With
Wrong report blocking portion, it detects the event beyond the exception of described diagnosis object plant equipment, stops the output of described abnormality determination unit,
Described wrong report blocking portion comprises timeliness degradation portion, this timeliness degradation portion uses the measuring-signal of described diagnosis object plant equipment to obtain the weight coefficient corresponding with classification, the scope having departed from timeliness deterioration is judged to be when the variable quantity of the weight coefficient corresponding with classification has exceeded maximal value
This timeliness degradation portion stops the output of described abnormality determination unit when not departing from the scope of timeliness deterioration.
2. the diagnostic device of plant equipment according to claim 1, is characterized in that,
Described wrong report blocking portion comprises operating condition change test section, the operating condition Data classification of described diagnosis object plant equipment is become classification to store by this operating condition change test section, be judged to be that operating condition does not change when not producing new classification
This operating condition change test section stops the output of described abnormality determination unit when operating condition there occurs change.
3. the diagnostic device of plant equipment according to claim 2, is characterized in that,
As the operating condition data of the input of operating condition change test section, the outside environmental elements of direct relation is not had to form by the characteristic with the apparatus forming described diagnosis object plant equipment.
4. a diagnostic device for plant equipment, detects the exception of diagnosis object according to the measuring-signal of diagnosis object plant equipment and reports exception, and the feature of the diagnostic device of described plant equipment is, comprising:
State change test section, the measuring-signal of described diagnosis object plant equipment is categorized into classification to store by it, is judged to be that state there occurs change when not belonging to classified classification;
Timeliness degradation portion, it uses the measuring-signal of diagnosis object plant equipment to obtain the weight coefficient corresponding with classification, is judged to be the scope having departed from timeliness deterioration when the variable quantity of the weight coefficient corresponding with classification has exceeded maximal value;
Operating condition change test section, the operating condition Data classification of diagnosis object plant equipment is become classification to store by it, is judged to be that operating condition does not change when not producing new classification;
Abnormality determination unit, it there occurs change in the state that is judged to be by described state change test section and is judged to have departed from the scope of timeliness deterioration by described timeliness degradation portion and when being judged to be that operating condition does not change by described operating condition change test section, is judged to be that described diagnosis object creates exception; With
Generate Alert block, its by the output report of this abnormality determination unit to outside.
5. a diagnostic device for plant equipment, possesses:
Measuring-signal database, it preserves the measuring-signal of diagnosis object plant equipment;
Process data extracting unit, it extracts the diagnostic signal in order to diagnose the state of described diagnosis object plant equipment to use among described measuring-signal database;
Reference signal data storehouse, it preserves described diagnostic signal;
Taxon, the Data classification preserved in described reference signal data storehouse is become classification by it;
Classification results database, described classification is preserved as normal category by it;
Diagnosis unit, it uses the up-to-date described diagnostic signal that extracted by described process data extracting unit and the information of described normal category of preserving in described classification results database, diagnoses that the state of described diagnosis object plant equipment belongs to normally, exception, operating condition change, which kind of situation of timeliness deterioration;
Diagnostic result database, it preserves the diagnostic result of described diagnosis unit; With
The unit of the information of preserving in described diagnostic result database is exported to image display device,
The feature of the diagnostic device of described plant equipment is,
Abnormality determination unit, state change test section, timeliness degradation portion and operating condition change test section is possessed in described diagnosis unit,
Described state change test section possesses following function: be judged to be that state there occurs change when the up-to-date described diagnostic signal extracted by described process Data extracting section does not belong to the described normal category of preserving in described classification results database,
Described timeliness degradation portion possesses following function: time variations amplitude during derivation normal condition and the relation between the weight coefficient variable quantity corresponding with classification, the scope having departed from timeliness deterioration is judged to be when the maximal value of the variable quantity of the weight coefficient corresponding with classification when having exceeded normal condition in diagnostic procedure
Described operating condition change test section possesses following function: when by by when not producing new classification when becoming classification with the operating condition Data classification that instrument characteristics does not have the outside environmental elements of direct relation to form, be judged to be that operating condition does not change
In described abnormality determination unit, there occurs change in the state that is judged to be by described state change test section and be judged to have departed from the scope of timeliness deterioration by described timeliness degradation portion and when not being judged to be that operating condition does not change by described operating condition change test section, be judged to be exception.
6. a diagnostic method for plant equipment, detects the exception of diagnosis object plant equipment according to the measuring-signal of diagnosis object plant equipment and reports exception, and the feature of the diagnostic method of described plant equipment is,
The measuring-signal of described diagnosis object plant equipment is categorized into classification to store, be judged to be that state there occurs change when not belonging to classified classification, the measuring-signal of diagnosis object plant equipment is used to obtain the weight coefficient corresponding with classification, the scope having departed from timeliness deterioration is judged to be when the variable quantity of the weight coefficient corresponding with classification has exceeded maximal value, the operating condition Data classification of diagnosis object is become classification to store, be judged to be that operating condition does not change when not producing new classification, now, be judged to create exception in described diagnosis object plant equipment, and report is to outside.
7. a diagnostic routine for plant equipment, detects the exception of diagnosis object plant equipment according to the measuring-signal of diagnosis object plant equipment and reports exception, and the feature of the diagnostic routine of described plant equipment is, comprising:
State change detecting step, being categorized into classification to store by the measuring-signal of described diagnosis object plant equipment, being judged to be that state there occurs change when not belonging to classified classification;
Abnormality juding step, is judged to create exception using the signal from this state change detecting step in described diagnosis object as the part inputted;
Alarm generating step, by the output report of this abnormality juding step to outside; With
Wrong report stops step, and the event beyond the exception detecting described diagnosis object, stops the output of described abnormality juding step,
Described wrong report stops step to comprise timeliness degradation step, in this timeliness degradation step, use the measuring-signal of described diagnosis object plant equipment to obtain the weight coefficient corresponding with classification, the scope having departed from timeliness deterioration is judged to be when the variable quantity of the weight coefficient corresponding with classification has exceeded maximal value
In this timeliness degradation step, stop the output of described abnormality juding step when not departing from the scope of timeliness deterioration.
8. the diagnostic routine of plant equipment according to claim 7, is characterized in that,
Described wrong report stops step to comprise operating condition change detecting step, in this operating condition change detecting step, the operating condition Data classification of described diagnosis object plant equipment is become classification to store, be judged to be that operating condition does not change when not producing new classification
Operating condition in this operating condition change detecting step, stops the output of described abnormality juding step when there occurs change.
9. the diagnostic routine of plant equipment according to claim 8, is characterized in that,
As the operating condition data of the input of operating condition change detecting step, the outside environmental elements of direct relation is not had to form by the characteristic with the apparatus forming described diagnosis object.
10. a diagnostic routine for plant equipment, detects the exception of diagnosis object plant equipment according to the measuring-signal of diagnosis object plant equipment and reports exception, and the feature of the diagnostic routine of described plant equipment is, comprising:
State change detecting step, being categorized into classification to store by the measuring-signal of described diagnosis object plant equipment, being judged to be that state there occurs change when not belonging to classified classification;
Timeliness degradation step, uses the measuring-signal of diagnosis object plant equipment to obtain the weight coefficient corresponding with classification, is judged to be the scope having departed from timeliness deterioration when the variable quantity of the weight coefficient corresponding with classification has exceeded maximal value;
Operating condition change detecting step, becoming classification to store by the operating condition Data classification of diagnosis object plant equipment, being judged to be that operating condition does not change when not producing new classification;
Abnormality juding step, there occurs change in the state that is judged to be by described state change detecting step and be judged to have departed from the scope of timeliness deterioration by described timeliness degradation step and when being judged to be that operating condition does not change by described operating condition change detecting step, be judged to be that described diagnosis object creates exception; With
Alarm generating step, by the output report of this abnormality juding step to outside.
CN201080065911.2A 2010-04-08 2010-04-08 Plant diagnosis device, diagnosis method, and diagnosis program Active CN102870057B (en)

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
PCT/JP2010/002575 WO2011125130A1 (en) 2010-04-08 2010-04-08 Plant diagnosis device, diagnosis method, and diagnosis program

Publications (2)

Publication Number Publication Date
CN102870057A CN102870057A (en) 2013-01-09
CN102870057B true CN102870057B (en) 2015-01-28

Family

ID=44762120

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201080065911.2A Active CN102870057B (en) 2010-04-08 2010-04-08 Plant diagnosis device, diagnosis method, and diagnosis program

Country Status (3)

Country Link
JP (1) JP5544418B2 (en)
CN (1) CN102870057B (en)
WO (1) WO2011125130A1 (en)

Families Citing this family (24)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JP6116466B2 (en) * 2013-11-28 2017-04-19 株式会社日立製作所 Plant diagnostic apparatus and diagnostic method
JP6366520B2 (en) * 2015-02-02 2018-08-01 日立Ｇｅニュークリア・エナジー株式会社 Output fluctuation monitoring apparatus and method
US20190011506A1 (en) * 2016-01-20 2019-01-10 Mitsubishi Electric Corporation Malfunction detection apparatus capable of detecting actual malfunctioning device not due to abnormal input values
JP6031202B1 (en) * 2016-01-29 2016-11-24 ファナック株式会社 Cell control device for finding the cause of abnormalities in manufacturing machines
JP6609689B2 (en) 2016-03-15 2019-11-20 株式会社日立製作所 Abnormality diagnosis system
KR101876185B1 (en) 2016-08-29 2018-07-09 한국수력원자력 주식회사 Learning method on recent data considering external effect in early alarm system, and system using thereof
JP7017851B2 (en) * 2016-12-01 2022-02-09 住友重機械工業株式会社 Fault diagnosis system and processing unit
JP6793565B2 (en) * 2017-02-06 2020-12-02 三菱パワー株式会社 State analyzer, display method, and program
JP6721563B2 (en) * 2017-11-28 2020-07-15 ファナック株式会社 Numerical control device
JP7180985B2 (en) * 2018-03-01 2022-11-30 株式会社日立製作所 Diagnostic device and diagnostic method
JP6810097B2 (en) 2018-05-21 2021-01-06 ファナック株式会社 Anomaly detector
JP6777686B2 (en) * 2018-05-29 2020-10-28 ファナック株式会社 Diagnostic equipment, diagnostic methods and diagnostic programs
WO2020013252A1 (en) * 2018-07-10 2020-01-16 住友重機械工業株式会社 Display method for construction machine and assistance device for construction machine
JP7057252B2 (en) * 2018-08-08 2022-04-19 Dmg森精機株式会社 Abnormal state judgment system
WO2020044533A1 (en) * 2018-08-31 2020-03-05 東芝三菱電機産業システム株式会社 Manufacturing process monitoring device
JP7260292B2 (en) * 2018-12-04 2023-04-18 日立グローバルライフソリューションズ株式会社 Abnormality diagnosis device and abnormality diagnosis method
CN113168171B (en) * 2018-12-05 2023-09-19 三菱电机株式会社 Abnormality detection device and abnormality detection method
WO2020144747A1 (en) * 2019-01-08 2020-07-16 眞鍋造機株式会社 Abnormality detection device, crane system, and abnormality detection method
JP7316816B2 (en) * 2019-03-26 2023-07-28 株式会社日本総合研究所 Information processing device and program
CN114391093B (en) * 2019-09-17 2023-10-24 日产自动车株式会社 Abnormality determination device and abnormality determination method
JP2021189964A (en) * 2020-06-03 2021-12-13 日立グローバルライフソリューションズ株式会社 Anomaly diagnostic device and anomaly diagnostic method
WO2023188018A1 (en) * 2022-03-29 2023-10-05 三菱電機株式会社 Abnormality detection device, machine system, and abnormality detection method
WO2023199567A1 (en) * 2022-04-12 2023-10-19 三菱重工業株式会社 Data processing device and data processing method
JP7485473B1 (en) 2023-03-22 2024-05-16 東芝エネルギーシステムズ株式会社 Generator cooler performance monitoring device

Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JPH04186119A (en) * 1990-11-21 1992-07-02 Hitachi Ltd Method and device for monitoring abnormality
JPH10301619A (en) * 1997-04-23 1998-11-13 Toshiba Corp Diagnostic device for fuel battery power generation plant
JP2005165375A (en) * 2003-11-28 2005-06-23 Hitachi Ltd Diagnostic device and method for facility
CN101118694A (en) * 2007-05-18 2008-02-06 李克明 Vehicle intellectualized management system

Family Cites Families (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JPH0660280A (en) * 1992-08-10 1994-03-04 Mitsubishi Electric Corp Transformation equipment monitor
JP3745783B2 (en) * 1994-06-08 2006-02-15 株式会社アドイン研究所 Pattern classification device and state monitoring system using this device
JP4140634B2 (en) * 2006-01-26 2008-08-27 松下電工株式会社 Anomaly monitoring method and anomaly monitoring apparatus
JP4767039B2 (en) * 2006-02-28 2011-09-07 株式会社フジクラ Vibration detection sensor system and vibration detection method

Patent Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JPH04186119A (en) * 1990-11-21 1992-07-02 Hitachi Ltd Method and device for monitoring abnormality
JPH10301619A (en) * 1997-04-23 1998-11-13 Toshiba Corp Diagnostic device for fuel battery power generation plant
JP2005165375A (en) * 2003-11-28 2005-06-23 Hitachi Ltd Diagnostic device and method for facility
CN101118694A (en) * 2007-05-18 2008-02-06 李克明 Vehicle intellectualized management system

Also Published As

Publication number Publication date
CN102870057A (en) 2013-01-09
WO2011125130A1 (en) 2011-10-13
JPWO2011125130A1 (en) 2013-07-08
JP5544418B2 (en) 2014-07-09

Similar Documents

Publication Publication Date Title
CN102870057B (en) Plant diagnosis device, diagnosis method, and diagnosis program
CN102999038B (en) The diagnostic device of generating set and the diagnostic method of generating set
WO2023071217A1 (en) Multi-working-condition process industrial fault detection and diagnosis method based on deep transfer learning
CN109524139B (en) Real-time equipment performance monitoring method based on equipment working condition change
CN108460144B (en) Coal equipment fault early warning system and method based on machine learning
EP2905665B1 (en) Information processing apparatus, diagnosis method, and program
US7181365B2 (en) Diagnostic data detection and control
CN102576227B (en) Plant diagnostic equipment
KR101713985B1 (en) Method and apparatus for prediction maintenance
US8682824B2 (en) Method and device for monitoring the state of a facility
US20030225466A1 (en) Methods and apparatus for early fault detection and alert generation in a process
JP6200833B2 (en) Diagnostic equipment for plant and control equipment
CN107229017A (en) A kind of wind generating set pitch control battery abnormal failure Forecasting Methodology
EP3469434B1 (en) Automatic visual and acoustic analytics for event detection
KR102618023B1 (en) Failure prediction diagnosis system and method through pattern analysis according to failure type
TWI780434B (en) Abnormal diagnosis device and method
Lahrache et al. Anomaly detection in a cutting tool by k-means clustering and support vector machines
CN114004059B (en) Health portrait method for hydroelectric generating set
Purarjomandlangrudi et al. Application of anomaly technique in wind turbine bearing fault detection
CN111108455A (en) Data sorting device
Richter et al. Automatic Defect Detection by Classifying Aggregated Vehicular Behavior
US20230418272A1 (en) System for Determining Operating Points in a Chemical Plant
US20230046709A1 (en) Prediction apparatus, prediction method, and program
CN117687382A (en) Vehicle fault checking method, system and computer medium
Ardakani et al. A New Scheme for Monitoring and Diagnosis of Multistage Manufacturing Processes Using Product Quality Measurements

Legal Events

Date Code Title Description
C06 Publication
PB01 Publication
C10 Entry into substantive examination
SE01 Entry into force of request for substantive examination
C14 Grant of patent or utility model
GR01 Patent grant