CN104781741A - Process monitoring/diagnosis device and process monitoring/diagnosis program - Google Patents

Process monitoring/diagnosis device and process monitoring/diagnosis program Download PDF

Info

Publication number
CN104781741A
CN104781741A CN201380058447.8A CN201380058447A CN104781741A CN 104781741 A CN104781741 A CN 104781741A CN 201380058447 A CN201380058447 A CN 201380058447A CN 104781741 A CN104781741 A CN 104781741A
Authority
CN
China
Prior art keywords
mentioned
data
definition
daily mode
monitoring diagnostic
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.)
Granted
Application number
CN201380058447.8A
Other languages
Chinese (zh)
Other versions
CN104781741B (en
Inventor
山中理
平冈由纪夫
吉泽直人
杉野寿治
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.)
Toshiba Corp
Original Assignee
Toshiba Corp
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 Toshiba Corp filed Critical Toshiba Corp
Publication of CN104781741A publication Critical patent/CN104781741A/en
Application granted granted Critical
Publication of CN104781741B publication Critical patent/CN104781741B/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
    • G05B23/0235—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 based on a comparison with predetermined threshold or range, e.g. "classical methods", carried out during normal operation; threshold adaptation or choice; when or how to compare with the threshold
    • 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
    • G05B23/0208—Electric testing or monitoring by means of a monitoring system capable of detecting and responding to faults characterized by the configuration of the monitoring system
    • 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
    • G05B23/0275—Fault isolation and identification, e.g. classify fault; estimate cause or root of failure

Landscapes

  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Engineering & Computer Science (AREA)
  • Automation & Control Theory (AREA)
  • Testing And Monitoring For Control Systems (AREA)

Abstract

A process monitoring/diagnosis device equipped with a classification unit, a model formulation unit having an identification unit and a fault detection data definition unit, and a monitoring/diagnosis unit having an extraction unit and a fault level assessment unit. The classification unit classifies process data or time-series data on the basis of attribute information. The identification unit creates regular pattern data and irregular pattern data from the classified time-series data. The fault detection data definition unit creates fault diagnosis models on the basis of the irregular pattern data. The extraction unit extracts degrees of deviation with respect to the regular pattern data of the classified process data. The fault level assessment unit applies the degrees of deviation to the fault diagnosis models, calculates fault detection data and a threshold value, and assesses a fault level on the basis of whether the fault detection data exceeds the threshold value.

Description

Operation monitoring diagnostic device and operation monitoring diagnostic program
Technical field
The operation monitoring diagnostic program that embodiments of the present invention relate to operation monitoring diagnostic device and use in this operation monitoring diagnostic device, this operation monitoring diagnostic device is diagnosed the exception produced in the working procedure systemses such as lower water treatment operation, drainage sunk well operation, sludge digestion operation, water purification operation, water supply and distribution operation, chemical process and iron steel operation.
Background technology
In the factory of the working procedure systemses such as the water treatments such as lower water treatment operation, sludge digestion operation, water purification operation and water supply and distribution operation/water transport operation, petrochemistry operation, iron steel operation and semiconductor fabrication sequence, be provided with multiple on-line sensors that multiple working procedure states is measured.The operation data such as the measurement by sensor group set in working procedure systems obtains by operation monitoring arrangement (SCADA:Supervisory Control And Data Acquisition) usually, such as flow, temperature, water quality and/or operational ton, are converted to the time series datas such as such as trend chart.Plant manager (keeper) and/or running personnel (operating personnel), by monitoring time series data, grasp the state of operation thus, and the running carrying out operation is changed, controlled.
The time series data of each operation data is equipped with the upper lower limit value being called as the management limit etc. usually.When operation monitoring arrangement is when the value of time series data exceedes this management limit or lower than this management limit, alarm of transmitting messages.Plant manager and/or running personnel carry out confirmation and the re-examine of factory's utilization based on this alarm.The running management of transmitting messages based on this alarm is the basic of factory's utilization.
In the management of more advanced plant operation, not only require reply when simple operation is astable, also require on the basis of the define objective performance reaching operation, realize energy-conservation cost-saving utilization.At this, define objective refers to, such as, if lower water treatment operation, then corresponds to observing of discharge water quality restriction.In addition, if clean water treatment operation, then the concentration of residual chlorine in water purification be below set upper limit situation and/or, the situation etc. that there is not with Cryptosporidium the pathogenic microbes being representative corresponds to define objective.In addition, if chemical process and iron steel operation, then the situation quality (such as purity and intensity etc.) of product (petroleum refining product and iron steel) being maintained specialized range corresponds to define objective.When want to use the running management of factory with make to realize on the basis of the define objective performance reaching operation energy-conservation cost-saving time, focus on, the state of the operation relevant to target capabilities is monitored to the state avoiding sinking into not reach define objective, detect rapidly the state change and/or abnormality that hinder and reach define objective, and take some countermeasures in advance.Further, focus on, remain kilter by target capabilities and energy-conservation cost-saving relevant working procedure states, detect rapidly and depart from such working procedure states change from kilter possibly.
But, change and the abnormal method diagnosed as to the state of operation, the known multivariate statistics operation that is called as monitors the method for (MSPC:Multi-Variate Statistical Process Control), the method uses " the multivariate statistics analytic method " that utilize in the field of petrochemistry operation and Tie Gang factory always.As the method the most often utilized in MSPC, known principal component analysis (PCA) (PCA:Principal Component Analysis) and latent variable sciagraphy/partial least square method (PLS:Projection to Latent Structure/Partial Least Square).
In MSPC, use the multivariate analysis such as PCA and/or PLS, mainly reach the first object that the abnormal sign of factory is detected and the second object that the process variable becoming abnormal factors is inferred.For the first object, by utilizing the relevant information of multiple process variable, the slight abnormal sign that cannot be detected by a variable is detected.For the second object, at abnormality detection data (such as, the T of Q statistical magnitude and Hotelling that basis is synthesized by multiple process variable 2statistic) detect abnormal after, utilize and represent that each operation data are for the contribution amount of the contribution degree of these abnormality detection data, infer the process variable of the candidate becoming abnormal factors.So, if use MSPC, then with the supervision using the simple management limit to carry out to independently process variable in the past (in the supervision of production line etc., sometimes contrast with MSPC and be called SPC:Statistical Process Control (statistical process control)) compared with, the monitoring diagnostic for advanced person more useful plant manager and/or running personnel can be carried out.
As described above, if use MSPC, then can reach the first object that the abnormal sign of factory is detected and the second object that the process variable becoming abnormal factors is inferred, but when MSPC is introduced be used as plant supervisory system to realize in SCADA, hearing the reflection at more scene as follows: by means of only the information obtained when reaching first and second object, enough information can not be provided for user and plant manager and/or running personnel.Its reason is, plant manager and/or running personnel want to make anything finally should be taked when exception to take action such judgement, but infers by means of only the detection of abnormal sign and factor thereof, is difficult to take an immediate action when exception.
Prior art document
Patent documentation
Patent documentation 1: Japanese Unexamined Patent Publication 8-241121 publication
Patent documentation 2: Japanese Unexamined Patent Publication 2004-303007 publication
Patent documentation 3: Japanese Unexamined Patent Publication 2007-65883 publication
Non-patent literature
Non-patent literature 1:http: //tech.chase-dream.com/spc.html
Non-patent literature 2:C.Rosen, " Monitoring Wastewater Treatment Systems ", Lic.Thesis, Dept.of Industrial Electrical Engineering and Automation, LundUniversity, Lund, Sweden, 1998.
Non-patent literature 3:Mia Hubert, Peter J.Rousseeuw, Karlien V, " ROBPCA:aNew Approach to Robust Principal Component Analysis (2005) ", Technometrics.
Non-patent literature 4:C Croux, A Ruiz-Gazen, " High breakdown estimators forprincipal components:the projection-pursuit approach revisited ", Journal ofMultivariate Analysis.
Non-patent literature 5:K.-R.Muller, S.Mika, G.Ratsch, K.Tsuda, and B.Scholkopf, " An introduction to kernel-based learning algorithms ", IEEE Trans.Pattern Anal.Machine Intell., 12 (2): 181-201, March, 2001.
Non-patent literature 6:B.Scholkopf, A.J.Smola, and K.-R.Muller, " Nonlinearcomponent analysis as a kernel eigenvalue problem ", Neural Computation, 10 (5): 1299-1319,1998.
Summary of the invention
The problem that invention will solve
As described above, in the device in the past introducing MSPC, detection and the factor thereof that can carry out the abnormal sign of factory are inferred, but by means of only these information, plant manager and/or running personnel are difficult to judge to take anything to take action when exception.
Therefore, the object of the invention is to, operation monitoring diagnostic program operation monitoring diagnostic device being provided and using in the apparatus, the judgement policy anything should being taked to take action when this operation monitoring diagnostic device can be provided in exception to plant manager and/or running personnel.
For solving the means of problem
According to embodiment, a kind of operation monitoring diagnostic device, obtains the operation data measured to by the sensor arranged in object operation, possesses: division; Model construction portion, has identification part and abnormality detection data definition part; And monitoring diagnostic portion, there is extraction unit and exception level judging part.The attribute information that above-mentioned division keeps the environment residing for above-mentioned object operation and classifies, classifies to above-mentioned operation data and multiple time series datas of being made up of the operation data in the past of the multiple periods preset based on above-mentioned attribute information.Above-mentioned identification part by robustly inferring that the above-mentioned time series data typical value being separately classified as multiple classification makes the daily mode data of each above-mentioned classification, and makes non-daily mode data based on the daily mode data of each above-mentioned classification.Above-mentioned abnormality detection data definition part is each based on above-mentioned multiple non-daily mode data, makes the first definition for calculating abnormality detection data and the second definition for the threshold value that calculates above-mentioned abnormality detection data.Said extracted portion extracts the deviation degree of the above-mentioned operation data be classified relative to above-mentioned multiple daily mode data.Above-mentioned exception level judging part comes to calculate abnormality detection data according to each above-mentioned classification by above-mentioned multiple deviation degree being applied to respectively above-mentioned first definition, come according to each above-mentioned classification calculated threshold, by judging according to each above-mentioned classification whether above-mentioned abnormality detection data exceed above-mentioned threshold value and judge exception level by above-mentioned multiple deviation degree being applied to respectively above-mentioned second definition.
Accompanying drawing explanation
Fig. 1 is the figure of the functional structure representing the surveillance that the operation monitoring diagnostic device of the first embodiment possesses.
Fig. 2 represents that the division shown in Fig. 1 carries out the figure of sorted time series data.
Fig. 3 is the figure of process flow diagram when representing that the exception level of monitoring diagnostic portion to operation data shown in Fig. 1 judges.
Fig. 4 is the figure of the judged result of the exception level representing the exception level judging part shown in Fig. 1.
Fig. 5 is the figure representing the contribution amount calculated based on element paritng portion as shown in Figure 1 and the contribution amount curve made.
Fig. 6 is the figure representing the contribution amount calculated based on element paritng portion as shown in Figure 1 and the contribution amount curve made.
Fig. 7 is the block diagram of the functional structure of the operation monitoring diagnostic device representing the second embodiment.
Fig. 8 is the figure representing the matrix that the normal data register shown in Fig. 7 is registered.
Embodiment
Below, with reference to accompanying drawing, embodiment is described.
(the first embodiment)
Fig. 1 is the structural drawing of the example of the functional structure representing the surveillance that the operation monitoring diagnostic device 30 of the first embodiment possesses.In surveillance shown in Fig. 1, in order to make enforcement idea clearer and more definite, by using except the lower water height treatment process for the purpose of denitrification and phosphorus is as monitored object.But the monitored object of surveillance is not limited to lower water height treatment process.
Surveillance shown in Fig. 1 possesses lower water height treatment process 10, Data Collection storage unit 20, operation monitoring diagnostic device 30 and user interface 40.
Lower water height treatment process 10 possesses initial settling basin 101, anaerobism groove 102, anaerobic groove 103, aerobic groove 104 and final sedimentation tank 105.
In addition, lower water height treatment process 10 possesses draw-off pump 1111, fan blower 1121, ebullator 1131, loopback sludge pump 1141 and draw-off pump 1151 as actuator.Draw-off pump 1111 extracts the excess sludge of initial settling basin 101.Fan blower 1121 supplies oxygen supply to aerobic groove 104.Ebullator 1131 makes the water exported from aerobic groove 104 circulate to anaerobic groove 103.Loopback sludge pump 1141 by the excess sludge of final sedimentation tank 105 to anaerobism groove 102 loopback.Draw-off pump 1151 extracts the excess sludge of final sedimentation tank 105.Various actuator 1111 ~ 1151 carries out action with the cycle of regulation.
In addition, actuator 1111 ~ 1151 possesses operational ton sensor respectively.Namely, draw-off pump 1111 possesses extraction flow sensor 1112, and fan blower 1121 possesses air supply flow sensor 1122, and ebullator 1131 possesses circular flow sensor 1132, loopback sludge pump 1141 possesses loop back traffic sensor 1142, and draw-off pump 1151 possesses extraction flow sensor 1152.
Extract the flow of flow sensor 1112 to the mud extracted by draw-off pump 1111 to measure.The flow of air supply flow sensor 1122 to the oxygen supplied by fan blower 1121 measures.The flow of circular flow sensor 1132 to the water circulated by ebullator 1131 measures.The flow of loop back traffic sensor 1142 to the mud by the loopback of loopback sludge pump measures.Extract the flow of flow sensor 1152 to the mud extracted by draw-off pump 1151 to measure.Various operational ton sensor 1112 ~ 1152 measured operation data with the cycle of regulation.
In addition, in lower water height treatment process 10, possess as operation sensor: rain sensor 121; Quantity sensor 122 is flowed into the lower water that the lower water yield of inflow measures; To the inflow TN sensor 123 that the total nitrogen content flowed into contained by lower water measures; To the inflow TP sensor 124 that the full phosphorus amount flowed into contained by lower water measures; The inflow UV sensor 1251 that the organism amount flowed into contained by lower water is measured or inflow COD sensor 1252; To the anaerobism groove ORP sensor 126 that the ORP of anaerobism groove 102 measures; To the anaerobism groove pH sensor 127 that the pH of anaerobism groove 102 measures; To the anaerobic groove ORP sensor 128 that the ORP of anaerobic groove 103 measures; To the anaerobic groove pH sensor 129 that the pH of anaerobic groove 103 measures; To the phosphoric acid sensor 1210 that the phosphoric acid concentration of aerobic groove 104 measures; To the DO sensor 1211 that the oxyty of aerobic groove 104 measures; To the ammoniacal sensor 1212 that the ammonia density of aerobic groove 104 measures; To the MLSS sensor 1213 that active mud content measures at least 1 groove of anaerobism groove 102, anaerobic groove 103 and aerobic groove 104; To the cooling-water temperature sensor 1214 that water temperature measures at least 1 groove of anaerobism groove 102, anaerobic groove 103 and aerobic groove 104; To the excess sludge SS sensor 1215 that the solid concentration of the sludge quantity extracted from final sedimentation tank 105 measures; To the discharge SS sensor 1216 that the SS concentration of the discharge water discharged from final sedimentation tank 105 measures; To the sludge interface sensor 1217 that the sludge interface level of final sedimentation tank 105 measures; To the lower water discharge capacity sensor 1218 that the lower water yield of discharge measures; To the discharge TN sensor 1219 that the total nitrogen content contained by the lower water of discharge measures; To the discharge TP sensor 1220 that the full phosphorus amount contained by the lower water of discharge measures; And the discharge UV sensor 12211 that the organism amount contained by the lower water of discharge measured or discharge COD sensor 12212.Various operation sensor 121 ~ 12212 measured operation data with the cycle of regulation.In addition, figure 1 illustrates following situation: be provided with the inflow UV sensor 1251 flowing into UV sensor 1251 and flow among COD sensor 1252, and be provided with the discharge UV sensor 12211 among discharge UV sensor 12211 and discharge COD sensor 12212.
Data Collection storage unit 20 is collected from various operational ton sensor 1112 ~ 1152 with the operation data that obtain of cycle of regulation and the operation data that obtain with the cycle of regulation from various operation sensor 121 ~ 12212.The operation data collected export to operation monitoring diagnostic device 30 by Data Collection storage unit 20.In addition, the operation data collected are that time series data is preserved according to the format conversion preset by Data Collection storage unit 20.Preserved time series data, according to the request from operation monitoring diagnostic device 30, exports to operation monitoring diagnostic device 30 by Data Collection storage unit 20.
Operation monitoring diagnostic device 30 such as comprises CPU (Central Processing Unit) and ROM (Read Only Memory) and RAM (Random Access Memory) etc. store the storage area etc. performing program or the data processed for CPU.Operation monitoring diagnostic device 30, by making CPU executive routine, realizes the function in division 31, model construction portion 32 and monitoring diagnostic portion 33 thus.
Division 31 possesses attribute supply unit 311 and determined property portion 312.
Moon information when attribute supply unit 311 keeps operation data to be collected, season (spring (March ~ May), summer (June ~ August), autumn (September ~ November), winter (Dec ~ February)) information, climatic information, temperature or the attribute information such as water temperature information and operation mode information.
The time series data in the past during data Collection and conservation portion 20 request of 312 pairs, determined property portion presets.The attribute information that determined property portion 312 use attribute supply unit 311 keeps, classifies to the time series data supplied from Data Collection storage unit 20.Attribute information is the relevant information of the attribute that is easily classified with the external condition residing for operation, and attribute such as has month, season, weather, temperature, water temperature and operation mode etc.Such as, time series data, based on attribute information, is categorized as the data of each weathers such as data monthly, the data in per four seasons, fine, cloudy, rain or snow, temperature or water temperature etc. is divided into the data etc. of each temperature of multiple grade or the data of water temperature and the dispar each running of operation mode by determined property portion 312.Determined property portion 312 exports being categorized as time series data of all categories to model construction portion 32.
In addition, when the diagnosis of operation data being carried out current time (online) by monitoring diagnostic portion 33, the online operation data collected by Data Collection storage unit 20 are classified by attribute information that determined property portion 312 use attribute supply unit 311 keeps as described above.Determined property portion 312 exports being categorized as operation data of all categories to monitoring diagnostic portion 33.
Model construction portion 32 possesses identification part 321, abnormality detection data definition part 322 and contribution amount definition part 323.Model construction portion 32, by identification part 321, makes multiple daily mode data and makes non-daily mode data from the beginning based on each of these daily mode datas.In addition, model construction portion 32, by abnormality detection data definition part 322, defines becoming the abnormality detection data etc. whether producing abnormal index based on non-daily mode data.In addition, model construction portion 32 is by contribution amount definition part 323, and define contribution amount, this contribution amount represents that the process variable to operational ton sensor and operation sensor represent makes much contributions to produced exception.
Identification part 321 is categorized as time series data of all categories based on by division 31, makes daily mode data and the non-daily mode data based on this daily mode data.Daily mode data is corresponding to the object wanting confirmation trend, be manufactured with multiple according to the definition relevant with calendar, the phase of such as time series data in the same time, this day unit daily mode data, and/or, what day (is such as all Monday, or is all Saturday etc.) identical of time series data and mutually in the same time, i.e. the daily mode data etc. of what day unit.
Such as, when making the daily mode data of day unit, on the basis of the information removing of identification part 321 when in advance process actions being become abnormal based on climatic information and what day information etc., based on the phase measurement value in the same time of the time series data received, calculate and robustly infer in each moment the typical value.At this, robustly infer that the typical value is mean value, pruning mean value, intermediate value and the mode value etc. of phase measurement value in the same time.When time series data may comprise exceptional value, mean value is not representatively worth, and intermediate value etc., the value lower relative to exceptional value sensitivity are representatively worth.In addition, when making the daily mode data of what day unit, identification part 321 for time series data phase in the same time and what day measurement value identical carries out process similar to the above.The process that identification part 321 also can proceed to calculate typical value and make daily mode data.
In addition, robustly infer the method for typical value, also known other are several.Such as, if use the method such as HL (Hodges-Lehmann) supposition, MCD (Minimum Covariance Determinant), bootstrap or subsample method, even if when being more complex mixed into abnormal data, also can inferring and typical typical value.
The measurement value that identification part 321 uses when calculating the daily mode data of making among time series data, deviate from which kind of degree relative to the value in the daily mode data produced, using result of calculation as non-daily mode data.
The quantity of the daily mode data of definition is set to m by identification part 321, when the quantity of the classification of being classified by division 31 is l, makes m × l daily mode data and m × l non-daily mode data by above-mentioned process.
In addition, the making of the daily mode data carried out based on time series data and non-daily mode data, such as, also can use the digital filter such as low-pass filter or discrete wavelet conversion to implement.But, if supposition has been mixed into abnormal data, then uses known various methods in Robust Statistics robustly to infer typical value, the daily pattern that reliability is higher can have been made.
Abnormality detection data definition part 322 accepts the non-daily mode data produced by identification part 321.Now, definition number m of non-daily mode data and l the classification of being classified by division 31 and daily mode data is provided as l × m data collection accordingly.
In addition, in the following description, by contained by the data set of non-daily mode data, for the past during presetting time series data deviate from value, be recited as < Xk >, wherein, k=1,2 ..., l × m.< Xk > represents the matrix that deviate from value, in a column direction represent sequential relative with non-day regular data of the operation data measured by operational ton sensor and operation sensor in the row direction.Namely, < Xk > be, s capable for r row key element < Xk > (r, s), store the matrix that deviate from value relative with the non-day regular data calculated according to r the sensor measurement value of moment s.If the quantity of operational ton sensor and operation sensor is set to n, the data number of sequential is set to Q, then < Xk > comprises Q × n data.In addition, following, need < Xk >, k=1,2 ..., the abnormality detection data of l × m define.But difference is only, the non-daily mode data inputted provides from different classes of.The define method of abnormality detection data is one, therefore as long as no the worry confused, then < Xk > is simply recited as < X >, omits subscript k.
Abnormality detection data definition part 322, by various methods such as the parsing of < X > using multivariate or rote learnings, generates abnormality detection data.To the method that < X > applies, be the method often used as operation diagnostic techniques, be called MSPC (process management of multivariate statistics).In MSPC, usual principal component analysis (PCA) (PCA) or latent variable sciagraphy (PLS) is utilized to generate the T being called as Q statistical magnitude and Hotelling 2the abnormality detection data of statistic.
Below, concrete calculating formula when using PCA to generate abnormality detection data is recorded.If use PCA, then < X > is decomposed as described below.In addition, in (1) formula, < X > is recorded with the X of thick word.
[several 1]
x = &Sigma; i = 1 n t i &times; P i T = T a P a T = &Sigma; i = 1 p t i &times; P i T + &Sigma; i = p 1 n t i &times; P i T = TP T + E - - - ( 1 )
At this, < T a> (the thick word T in (1) formula a) ∈ R q × n, be the matrix being called as score matrix be made up of q sample and n number of principal components.< P a> (the thick word P in (1) formula a) ∈ R n × n, be represent n the matrix being called as loading matrix forming the relation between variable and n major component.< T > (the thick word T in (1) formula) ∈ R q × p, be the < T intercepted by p < < n major component athe part matrix of >, is commonly called score matrix.Equally, < P > (the thick word P in (1) formula) ∈ R n × p, be represent that the major component intercepted by p < < n is relative to the < P of the relation of n variable athe part matrix (also referred to as submatrix) of >, is commonly called loading matrix.In addition, < E > (the thick word E in (1) formula) ∈ R q × n, be the error matrix be made up of q sample and n variable, represent error when having intercepted major component by p < < n.
Below, by < T a> and < T >, < P a> and < P > distinguishes clearly, by < T a> and < P a> is called score matrix and loading matrix, and < T > and < P > is called main score matrix and Main Load matrix.
Use these matrixes, define following Q statistical magnitude and T as abnormality detection data 2statistic.
[several 2]
Q(x(t))＝x T(t)(I-PP T)x(T) (2)
[several 3]
T 2(x(t))＝x T(t)P T∧ -1Px(t) (3)
At this, < Λ > (the thick word Λ in (3) formula) is the matrix that the dispersion of major component will usually be had as diagonal angle, and the meaning refers to carries out normalization to dispersion.In addition, < I > (the thick word I in (2) formula) is the unit matrix of appropriate size.In addition, x (t) is t the key element of matrix L EssT.LTssT.LT X >.In addition, when carrying out anomaly monitoring diagnosis in monitoring diagnostic portion 33, the operation data substituting into online measurement to this x (t) calculated relative to daily the deviating from value of mode data.
Abnormality detection with data definition part 322 set as by (2), (3) formula definition Q statistical magnitude and T 2statistic identifies the threshold value of abnormal and normal judgment standard.Threshold value change to state and/or the detection of abnormal sign relevant significantly, therefore its establishing method is more important.But, owing to haveing nothing to do with the technological thought of present embodiment, therefore only record typical establishing method.
When supposing that time series data relative to the past during presetting is without any prior information, setting method by default, can use the statistics confidence limit angle value relevant to Q statistical magnitude and the T with Hotelling 2the statistics confidence limit angle value (with reference to non-patent literature 2) that statistic is relevant.The statistics confidence limit angle value relevant to Q statistical magnitude and the T with Hotelling 2the statistics confidence limit angle value that statistic is relevant, can write as described below.
[several 4]
h 0：＝1-2Θ 1Θ 3/3Θ 2 2
Θ i：＝λ p+1 i+λ p+2 i+.........+λ n i
At this, p is the quantity of variable residual in model.C askew (example: when a=0.01, the c of the standard deviation of standard normal distribution when be the limit of fiducial interval being 1-a a=2.53, when a=0.05, c a=1.96).In addition, λ idiagonal angle key element (that is, the Θ of < Λ > ibe each composition contained by error term i power and.)。
[several 5]
T 2limit＝p(q-1)/(q-p)F(p，q-p，a) (5)
At this, p is the quantity of variable residual in model, and q is the quantity of entire variable.F (p, q-p, a) be degree of freedom be (p, q-p), F distribution when cofidence limit being set to a.In addition, the situation being set to a=0.01 or 0.05 is more.
Abnormality detection data definition part 322 defines (1) ~ (5) formula according to l × m each of non-daily mode data.
In addition, when being assumed in the data containing a large amount of outliers etc., abnormality detection data definition part 322 such as also can use the various robust PCA algorithms considering the robustness for outlier of non-patent literature 3 and non-patent literature 4 etc.Or, also can be carried out expanding and using as robust PLS.
Further, when be assumed to have between data stronger non-linear be correlated with such, abnormality detection data definition part 322 such as also can use the core PCA etc. described in non-patent literature 5 and non-patent literature 6 etc. to consider the PCA of non-linear.Or, also can be carried out expanding and using as core PLS.
Further, when there is the problem of both sides of non-linear and outlier, abnormality detection data definition part 322 also can use method robust PCA and core PCA are combined with.
In addition, as the technology similar with MSPC, the field mouth method etc. that the field that also can be used in quality engineering uses, uses mahalanobis distance to generate abnormality detection data.
Contribution amount definition part 323 sets the process variable, the definition relative to the contribution amount of the abnormality detection data defined by (2) formula and (3) formula that represent operational ton sensor and operation sensor.The define method of contribution amount also exists multiple, but such as can define as described below.
[several 6]
Q cont ( n , t ) = x T ( t , n ) F ( : , n ) T F ( : , n ) x ( t , n ) F = ( I - PP T ) - - - ( 6 )
[several 7]
T 2cont(n，t)＝x T(t)P T∧ -1P(：，n)x(t，n) (7)
At this, the n meaning refers to the n-th process variable, and t is the variable represented sometime.(6) formula and the definition shown in (7) formula, defines according to l × m each of non-daily mode data.By using (6) formula and (7) formula, contribution amount definition part 323 can provide to monitoring diagnostic portion 33 can make the definition of which kind of degree contribution by calculation process variable respectively to the value of abnormality detection data.
In addition, if as long as contribution amount definition part 323 has the abnormality detection data being transfused to and being calculated based on the operation data be supplied to by monitoring diagnostic portion 33, which process variable can become the possibility of abnormal factor is high to sort and the framework exported to, can be arbitrary framework.
As described above, model construction portion 32 is about definition when making the classification in the moon or season etc. and daily mode data, defining (1) ~ (7) formula respectively, building the multiple abnormity diagnosis models for evaluating the deviation degree of online operation data and daily mode data thus.
Monitoring diagnostic portion 33 possesses extraction unit 331, exception level judging part 332, element paritng portion 333 and anomalous identification portion 334.Monitoring diagnostic portion 33 is extracted the non-daily data pattern data in online operation data by extraction unit 331.In addition, monitoring diagnostic portion 33, by exception level judging part 332, uses the abnormality detection data calculated based on the non-daily data pattern extracted, judges the exception level of produced exception.In addition, monitoring diagnostic portion 33, by element paritng portion 333, calculates the contribution amount of each process variable to produced exception.In addition, monitoring diagnostic portion 33 by anomalous identification portion 334, based on the contribution amount calculated, identify produced exception be caused by the exception of operational ton sensor and operation sensor or caused by the exception of operation.
Extraction unit 331 accepts to be categorized as online operation data of all categories by division 31.Extraction unit 331 pairs of model construction portions 32 ask the daily mode data relevant with the classification of accepted operation data.Extraction unit 331 obtains the difference of daily mode data and the operation data supplied from division 31 provided from identification part 321 according to request, the non-daily mode data thus in abstraction process data.
The non-daily mode data extracted by extraction unit 331 is substituted into l × m (2) formula and (3) formula that are defined by abnormality detection data definition part 322 by exception level judging part 332, calculates Q statistical magnitude and T 2statistic.
Exception level judging part 332 judges the Q statistical magnitude that calculates for multiple abnormity diagnosis model and T 2whether statistic exceedes in each abnormity diagnosis model based on the threshold value that (4) formula and (5) formula calculate.Multiple abnormity diagnosis model is the abnormity diagnosis model being suitable for evaluating the exception level of online operation data, is preset.Such as, these abnormity diagnosis models belong to the adjacent classification among the classification of identical type, based on identical definition daily mode data and make.In addition, online operation data obtain situation, be the situation belonging to identical category with the some abnormity diagnosis models in multiple abnormity diagnosis model.
Exception level judging part 332 is based on Q statistical magnitude and T 2the comparative result of statistic and threshold value, decides whether create abnormal exception level in the online operation data of expression.Such as, exception level judging part 332 first to judge for operation data between the Q statistical magnitude that calculates of the highest abnormity diagnosis model of adaptability and T 2whether statistic exceedes the threshold value calculated for this abnormity diagnosis model.Grade of fit between abnormity diagnosis model is predefined, and the abnormity diagnosis model that adaptability is the highest such as refers to that the situation that obtains with operation data belongs to the abnormity diagnosis model of identical category.At Q statistical magnitude and T 2when statistic is less than threshold value, exception level judging part 332 is judged as that operation data are normal, and exception level is 0%.In addition, at this, evaluate the value of exception level according to number percent, but be not limited to evaluate according to number percent.
At Q statistical magnitude and T 2when statistic exceedes threshold value, exception level judging part 332 is judged as that operation data are for abnormal, to operation data setting V%.In addition, V% is the arbitrary numerical value can evaluated exception level.Next, exception level judging part 332 to judge for operation data between the Q statistical magnitude that calculates of the high abnormity diagnosis model of adaptability second and T 2whether statistic exceedes the threshold value calculated for this abnormity diagnosis model.The abnormity diagnosis model that adaptability second is high, such as, refer to the abnormity diagnosis model of the classification adjacent with the classification belonging to operation data.At Q statistical magnitude and T 2when statistic is less than threshold value, set V% is judged as exception level by exception level judging part 332.At Q statistical magnitude and T 2when statistic exceedes threshold value, exception level judging part 332 is judged as operation data exception, set V% is increased and is set to V ' %.
Exception level judging part 332 is at Q statistical magnitude and T 2when statistic exceedes threshold value, becoming Q statistical magnitude and T 2till statistic is less than threshold value, abnormity diagnosis model is switched, repeatedly carries out Q statistical magnitude and T 2statistic compares with threshold value.In addition, exception level judging part 332 all implements Q statistical magnitude and T for the multiple abnormity diagnosis models preset 2statistic compares with threshold value, at Q statistical magnitude and T 2when statistic exceedes threshold value, be judged as that operation data are 100% exception.
Element paritng portion 333 is not when the exception level judged by exception level judging part 332 is 0%, namely, when operation data are abnormal, (6) formula and (7) formula is used to calculate Q statistical magnitude to being calculated by exception level judging part 332 and T 2the contribution amount of statistic.Now, element paritng portion 333 to use for online operation data between the definition of the highest abnormity diagnosis model of adaptability.Element paritng portion 333 makes contribution amount curve based on the contribution amount calculated, and the contribution amount curve produced is shown in user interface 40.Thus, the user of user interface 40 can infer the process variable becoming and produce abnormal factor.
Anomalous identification portion 334 based on the process variable being speculated as abnormal factor by element paritng portion 333, identify abnormal kind be caused by the exception of operational ton sensor and operation sensor or caused by the exception of operation.Recognition result is shown in user interface 40 by anomalous identification portion 334.
Such as, in contribution amount curve, when the data of only some process variable illustrate outstanding exceptional value, for the possibility of the exception of sensor is higher.Particularly, under there is relevant environment to multiple operation measurement information, when the data of a process variable illustrate the exceptional value detected, for the possibility of sensor abnormality is high.Therefore, anomalous identification portion 334, when the data of only some process variable illustrate outstanding exceptional value, is speculated as sensor abnormality.
On the contrary, when the contribution amount of multiple process variable illustrates exceptional value, for the possibility of the exception of operation is higher.Therefore, anomalous identification portion 334, when the contribution amount of multiple process variable illustrates exceptional value, is speculated as operation abnormal.But, even if be the state of operation exception, also there is the state of certain sensor failure.Even if in this case, the contribution amount of the process variable broken down illustrates that the situation of relatively extreme exceptional value is more, therefore in most cases can identify sensor abnormality and operation exception.
As carrying out this knowledge method for distinguishing quantitatively, method as described below can be considered.
The data be made up of the contribution amount of each process variable are considered as a data set by anomalous identification portion 334, carry out the abnormity diagnosis for these data.When inferring abnormal factors by contribution amount curve, owing to detecting exception, therefore must premised on the situation being wherein mixed into certain abnormal data.Owing to can not calculate exception by common average or dispersion (standard deviation) for contribution amount data set, therefore anomalous identification portion 334 utilizes and replaces the robust of average and dispersion to infer method.Such as, can replace average, and use intermediate value and prune average.In addition, standard deviation can be replaced, and the standard deviation utilizing median absolute deviation (MAD) and prune.Anomalous identification portion 334, when employing intermediate value and median absolute deviation, uses the contribution amount X of certain process variable to calculate
K=(X-intermediate value)/median absolute deviation,
To the threshold k max of value setting such as 5 ~ 10 degree of this K.When the process variable that the value of K exceedes this Kmax be extracted to only one, anomalous identification portion 334 is judged as sensor abnormality, when also not extracting or extract multiple for one, is judged as that operation is abnormal.
In addition, as additive method, also method as following can be adopted.Between contribution amount and statistic, there is the summation this character consistent with statistic of the contribution amount of each process variable.That is, if (6) formula is obtained summation for each process variable, (2) formula is become, or, if (7) formula is obtained summation for each process variable, become (3) formula.Utilize this character, can adopt with the following method: when the contribution amount of maximum abnormal factors candidate exceedes the setting value of regulation relative to the ratio of the statistic ((2) formula or (3) formula) the abnormal moment being detected, be judged as sensor abnormality.Such as, 0.75 (75%) is set as setting value, when abnormality detection, during contribution amount ((6) formula or (7) formula) the ÷ abnormality detection of the maximum abnormal factors candidate of t, the value of the statistic ((2) formula or (3) formula) of t exceedes setting value 0.75, anomalous identification portion 334 is judged as sensor abnormality.
The measurement information of the sensor of identical type contains multiple in process variable, the precision of this identification can be made to improve further.Such as, when being provided with the DO sensor 1211 of multiple respectively different principle, assuming that the position that some differences, flow direction are different in place is provided with the situation etc. of multiple DO sensor 1211.At this, contribution amount that is significantly high in the contribution amount of the only measurement value of certain DO sensor, DO sensor beyond it is not high, the DO sensor failure that strong doubt contribution amount is high.In this case, sensor abnormality can be judged as.
In addition, once be judged as sensor abnormality, then be judged as abnormal sensor maintained and recover normal before, continue to detect abnormal all the time.Therefore, surveillance is difficult to proceed abnormity diagnosis after sensor recovers normally.In order to should situation, anomalous identification portion 334 is for by the operation data being judged as abnormal sensor and measuring, before notifying that the reparation of this sensor terminates, input the data in the object moment of the subject sensor in the daily mode data defined by identification part 321.Or, anomalous identification portion 334 to the phase in nearest past in the same time, what day the several sample of Data Mining identical, the intermediate value of the sample obtained etc. is inputted by being judged as the operation data that abnormal sensor measures.Thus, even if anomalous identification portion 334 also can proceed abnormity diagnosis after being diagnosed as sensor fault.In addition, even if when multiple sensor failure, only otherwise breaking down in the same time mutually, anomalous identification portion 334 just can identification sensor fault and operation exception.
Next, the action of the operation monitoring diagnostic device 30 formed as described above is specifically described.
First, the situation of model construction portion 32 build exception diagnostic model is described.In the following description, the situation for the determined property portion 312 of division 31, the time series data supplied from Data Collection storage unit 20 being categorized as the data in June is described.Fig. 2 is the figure of the example representing the time series data that the data in June are classified.According to Fig. 2, sorted time series data comprises the data of 1 hour unit with crossing over multiple date-time.Such as, the data of moment 0:00 are made up of X0=[5 4 66 35 111 7] this data set.
Identification part 321 makes the daily mode data of day unit based on the time series data shown in Fig. 2.Such as, when calculating the typical value of moment 0:00, according to X0=[5 4663 5,111 7], the typical value being predicted as moment 0:00 becomes the value near about 5,6.If it is namely average to get to X0 the most straightforward procedure calculating typical value, then the mean value of X0 becomes 18.375.This value is different significantly from predicted typical value.Its reason can be thought, the 8th data are exceptional value.Therefore, if not this simple averaging operation but calculate intermediate value, then 5.5 are become, close to the typical value predicted.In addition, such as, if calculate the mode value of X0, then become 6, this value is also close to the typical value predicted.In addition, if remove maximal value and minimum value and be averaged in pruning average (pruning average), then become 5.5, this value is also close to the typical value predicted.Identification part 321, by calculating typical value as described above in each moment, makes the daily mode data in June.
Which kind of degree identification part 321 deviates from by calculating the measurement value used when making daily mode data with the daily mode data produced, and makes non-daily mode data.Such as, when having carried out calculating to the typical value of daily mode data with intermediate value, the 1st data in X0=[5 4663 5,111 7] and daily mode data deviate from for 5-5.5=-0.5, deviating from for 111-5.5=105.5 of the 7th data and daily mode data.
Abnormality detection data definition part 322, with reference to the non-daily mode data produced by identification part 321, makes the definition of the Q statistical magnitude shown in (2) formula and the T shown in (3) formula 2the definition of statistic.In addition, abnormality detection data definition part 322 calculates the statistics confidence limit angle value relevant with Q statistical magnitude based on (4) formula, calculates and T based on (5) formula 2the statistics confidence limit angle value that statistic is relevant.
Contribution amount definition part 323, as shown in (6) formula and (7) formula, sets the process variable that measured by operational ton sensor and operation sensor to the Q statistical magnitude defined by (2) formula and (3) formula and T 2the definition of the contribution amount of statistic.
Above, describe model construction portion 32 builds the abnormity diagnosis model in June situation based on the daily mode data in June, but model construction portion 32 is in order to evaluate the exception level of online operation data, and the situation that supposition builds the abnormity diagnosis model of the every month in January to Dec.That is, model construction portion 32 makes the daily mode data of the every month in January to Dec, and makes the definition based on (2) ~ (5) formula of these daily mode datas.
Then, the situation that monitoring diagnostic portion 33 diagnoses online operation data is described.Fig. 3 is the figure of process flow diagram when representing that the exception level of the 33 pairs of operation data in monitoring diagnostic portion judges.In addition, in the following description, for the data of the operation data supplied from Data Collection storage unit 20 collected by the moment 0:00 of xx day in June, be categorized as the situation in June by determined property portion 312 and be described.
Which kind of degree is extraction unit 331 deviate from by calculating the operation data of being classified by division 31 from the daily mode data in the June of being produced by identification part 321, makes non-daily mode data (step S31) thus.Such as, the value of collected operation data is 8, and when the identification part 321 daily mode data that used intermediate value to make, the value of non-daily mode data is calculated as 8-5.5=2.5.
The non-daily mode data that exception level judging part 332 will be made by step S31, substitutes into (2) formula and (3) formula that the abnormity diagnosis model for June defines, calculates Q statistical magnitude and T 2statistic.The non-daily mode data that exception level judging part 332 will be made by step S31, substitutes into (4) formula and (5) formula that the abnormity diagnosis model for June defines, calculates threshold value and the T of Q statistical magnitude 2the threshold value (step S32) of statistic.Exception level judging part 332 is by the Q statistical magnitude that calculated by step S32 and T 2statistic, with the threshold value of the Q statistical magnitude calculated by step S32 and T 2the threshold value of statistic compares (step S33), judges Q statistical magnitude and T 2whether statistic is less than threshold value (step S34).At Q statistical magnitude and T 2under statistic is less than the situation (step S34 is) of threshold value, exception level judging part 332 be judged as exception level be 0%, namely operation data be normal (step S35), process is terminated.
At Q statistical magnitude and T 2under statistic exceedes the situation (step S34's is no) of threshold value, exception level judging part 332 judges that whether as the abnormity diagnosis model of comparison other be last abnormity diagnosis model (step S36) in the multiple abnormity diagnosis models preset.In addition, the classification belonging to operation data is June, and therefore last abnormity diagnosis model is herein the abnormity diagnosis model in Dec.The abnormity diagnosis model becoming comparison other in step S34 is the abnormity diagnosis model in June (step S36's is no), and therefore exception level judging part 332 is set to 10% (step S37) by tentative for exception level.
Then, monitoring diagnostic portion 33, based on the classification abnormity diagnosis model in May adjacent with the abnormity diagnosis model in June and the abnormity diagnosis model in July, carries out the diagnosis of operation data.That is, extraction unit 331 by the abnormity diagnosis model that becomes comparison other from June increment to May and July (step S38).Extraction unit 331 is calculated the operation data of being classified by division 31 and deviates from which kind of degree from the daily mode data in the May of being produced by identification part 321 and July, makes the non-daily mode data (step S39) in May and July thus.
The non-daily mode data in the May of being produced by step S39 and July is substituted into (2) formula for the abnormity diagnosis model definition in May and July and (3) formula by exception level judging part 332 respectively, calculates Q statistical magnitude and the T in May and July 2statistic (step S310).Exception level judging part 332 is by the Q statistical magnitude that calculated by step S310 and T 2statistic, to compare (step S311) with the threshold value calculated by abnormality detection data definition part 322 in the abnormity diagnosis model in May and July, process is shifted to step S34.At the Q statistical magnitude relevant with the abnormity diagnosis model in May or July and T 2under statistic is less than the situation (step S34 is) of threshold value, exception level judging part 332 is judged as that operation data are suitable for the abnormity diagnosis model in May or July, and exception level is judged as YES 10% (the step S35) set tentatively, process is terminated.At the Q statistical magnitude relevant with the abnormity diagnosis model both sides in May and July and T 2under statistic exceedes the situation (step S34's is no) of threshold value, whether the abnormity diagnosis model that exception level judging part 332 judges as comparison other is the abnormity diagnosis model (step S36) in Dec.The abnormity diagnosis model becoming comparison other in step S34 is the abnormity diagnosis model in May and July (step S36's is no), and therefore exception level judging part 332 is set as 30% (step S37) by tentative for exception level.
Then, monitoring diagnostic portion 33, based on the abnormity diagnosis model in classification April adjacent with the abnormity diagnosis model in May and the abnormity diagnosis model in the abnormity diagnosis model in classification and July adjacent August, carries out the diagnosis of operation data.Monitoring diagnostic portion 33 after implementing step S38 ~ step S311, in step S34, at the Q statistical magnitude relevant with the abnormity diagnosis model in April or August and T 2under statistic is less than the situation (step S34 is) of threshold value, exception level judging part 332 is judged as that operation data are suitable for the abnormity diagnosis model in April or August, and exception level is judged as YES 30% (the step S35) set tentatively, process is terminated.At the Q statistical magnitude relevant with the abnormity diagnosis model both sides in April and August and T 2under statistic exceedes the situation (step S34's is no) of threshold value, whether the abnormity diagnosis model that exception level judging part 332 judges as comparison other is the abnormity diagnosis model (step S36) in Dec.The abnormity diagnosis model becoming comparison other in step S34 is the abnormity diagnosis model in April and August (step S36's is no), and therefore exception level judging part 332 is set as 50% (step S37) by tentative for exception level.
Exception level judging part 332 carries out the process of step S34 ~ step S311 repeatedly.Q statistical magnitude relevant with the abnormity diagnosis model in Dec in step S34 and T 2under statistic exceedes the situation no and step S36 of the step S34 (be) of threshold value, exception level is set as 100% (step S312) by exception level judging part 332, and process is shifted to step S35.
Fig. 4 is the figure of the judged result representing exception level judging part 332 pairs of exception level.The transverse axis meaning of Fig. 4 refers to the classification of the operation data becoming diagnosis object.In the present note, using the operation data collected by the moment 0:00 of xx day in June as diagnosis object, the data of the row represented by oblique line in Fig. 4 are therefore conceived to.The longitudinal axis meaning of Fig. 4 refers to the abnormity diagnosis model of each moon.Numeric representation described in each table is not suitable for the exception level when model represented by the longitudinal axis.Judged result is exported by user interface 40.
In addition, in Fig. 3 and Fig. 4, show as an example when operation data be not suitable for January ~ all abnormity diagnosis model in Dec monitoring diagnostic portion 33 exception level of operation data is judged as 100% situation, if but were not suitable for several abnormity diagnosis model, then certainly also exception level could are set as 100%.
Then, illustrate that the kind of monitoring diagnostic portion 33 to exception is situation that is that caused by sensor abnormality or that identified by abnormal the carrying out caused of operation.
When the operation data of element paritng portion 333 collected by the moment 0:00 being supplied to xx day in June, (6) formula and (7) formula is used to calculate the contribution amount of the abnormity diagnosis model to June.Fig. 5 and Fig. 6 is the contribution amount curve made based on the contribution amount calculated by element paritng portion 333.The transverse axis of Fig. 5 and Fig. 6 is the process variable to MSPC input, and the longitudinal axis is to Q statistical magnitude or T 2the contribution amount of statistic.
Anomalous identification portion 334, when only the contribution amount of some process variable exceedes threshold k max as shown in Figure 5, is speculated as sensor abnormality.In addition, anomalous identification portion 334, when the contribution amount of multiple process variable as shown in Figure 6 exceedes threshold k max, is speculated as operation abnormal.
As described above, the typical value that the identification part 321 of present embodiment uses the time series data according to the past and robustly infers, makes the daily mode data of multiple kind.Identification part 321 makes the non-daily mode data of multiple kind based on the daily mode data of the multiple kinds produced.Abnormality detection data definition part 322 makes multiple abnormity diagnosis model based on the non-daily mode data of the multiple kinds produced.Extraction unit 331 calculates the operation data that supply from division 31 relative to the deviation degree of daily mode data, makes the non-daily mode data of diagnosis thus.The non-daily mode data of diagnosis is applied to the multiple abnormity diagnosis models built by model construction portion 32 by exception level judging part 332, makes abnormality detection data thus.Whether the abnormality detection data that exception level judging part 332 is produced according to each judgement of multiple abnormity diagnosis model illustrate exceptional value, determine the exception level of operation data thus.Thus, operation monitoring diagnostic device 30 can not only carry out judgement that is abnormal or normally this 2 values, can also provide abnormal degree.In the surveillance of carrying out based on MSPC in the past, only can carry out the detection of abnormal sign and being separated of variable factors, on the other hand, operation monitoring diagnostic device 30 according to the present embodiment, can produce engineering (engineering) hardly and mechanically provide the information of the detailed levels of the abnormality of factory.
In addition, the identification part 321 of present embodiment, based on the phase measurement value in the same time of accepted time series data, uses pruning mean value, intermediate value, mode value etc. in each moment and robustly calculates typical value.Then, identification part 321 utilizes typical value to make daily mode data, and makes non-daily mode data based on the daily mode data produced.So, identification part 321, by implementing fairly simple statistical treatment to time series data, automatically can identify daily pattern and non-daily pattern.Therefore, the user of surveillance, in abnormity diagnosis, easily can identify the exceptional value departed from from daily mode data or paroxysmal exceptional value.
In addition, the identification part 321 of present embodiment, by carrying out constantly calculating the process that typical value makes daily mode data, automatically can learn the slow change of daily mode data.Thereby, it is possible to easily carry out the maintenance of operation monitoring diagnostic device.
In addition, average, intermediate value or more each method HL supposition of the Robust Statistics of height, MCD, bootstrap or subsample method etc. are pruned by utilizing in calculating in typical value in the identification part 321 of present embodiment, even if thus when containing a large amount of exceptional value, also can robustly synthesize daily pattern.
In addition, each contribution amount to exception of contribution amount definition part 323 to multiple process variable that multiple abnormity diagnosis model contains of present embodiment defines.The definition of the non-daily mode data of diagnosis to contribution amount inputs by element paritng portion 333, calculates each contribution amount to exception of multiple process variable.Thus, operation monitoring diagnostic device 30 can be provided for the information to producing abnormal factor and inferring.
In addition, the anomalous identification portion 334 of present embodiment exists multiple when exceeding the contribution amount of threshold value in the contribution amount calculated by element paritng portion 333, be speculated as operation abnormal, when the contribution amount exceeding threshold value is one, be speculated as sensor abnormality.Thus, operation monitoring diagnostic device 30 can be provided for the factor of exception is the information that operation exception or sensor abnormality are inferred.Particularly, when being provided with the redundant system of the sensor of identical type in the factory, can the accurately abnormal and operation exception of identification sensor further.Therefore, anomalous identification portion 334 identification sensor fault and operation extremely, when sensor fault, can urge the reply such as the maintenance of sensor and/or the replacing of sensor clearly, when operation exception, the abnormality of operation promptly can be tackled.
Therefore, operation monitoring diagnostic device 30 according to the present embodiment, the judgement policy anything should being taked to take action when can be provided in exception to plant manager and/or running personnel.Thus, the action when user of surveillance easily takes exception.
In addition, the abnormality diagnostic a series of action supposition of present embodiment realizes in the operation monitoring diagnostic device 30 of surveillance, but is not limited thereto.Such as, also can be, make the function of operation monitoring diagnostic device 30, i.e. division 31, the function in model construction portion 32 and monitoring diagnostic portion 33 realize on Cloud Server, exception level is judged, abnormal factors infer and sensor operation anomalous identification information remote provide to the user needed.
Such as, monitoring diagnostic portion 33, to the mobile terminal of the office of sensor manufacturer or the maintainer of sensor manufacturer that provide operational ton sensor 1112 ~ the 1152 and operation sensor 121 ~ 12212 carrying out in the factory measuring, provides that exception level judges, abnormal factors is inferred and the information of sensor operation anomalous identification.Then, sensor manufacturer judges based on exception level, abnormal factors is inferred and the information of sensor operation anomalous identification, judges that sensor is the need of maintenance, can realize the raising of maintenance efficiency thus.
In addition, if exception level judges by the monitoring diagnostic portion 33 realized on Cloud Server, abnormal factors is inferred and the information of sensor operation anomalous identification sends to the mobile terminal being in long-range factory operating personnel, so when confirm not to be sensor fault but certain exception of operation, factory operating personnel just can carry out fast processing to abnormal.So, by utilizing the modes such as cloud, remotely information can be provided to multiple user.
(the second embodiment)
Fig. 7 is the block diagram of the example of the functional structure of the operation monitoring diagnostic device 50 representing the second embodiment.Operation monitoring diagnostic device 50 shown in Fig. 7 possesses division 31, anomaly monitoring diagnostics division 33, model construction portion 51 and normal data register 52.In addition, in the figure 7, identical symbol is given for the part common with Fig. 1.
Normal data register 52 accepts to be categorized as time series data of all categories by division 31.Normal data register 52 is registered accepted time series data, according to the request from model construction portion 51, is exported by registered time series data to model construction portion 51.
In addition, normal data register 52 is judged as that exception level is 0% in the judged result of exception level judging part 332, be namely judged as being normal situation as the data set of diagnosis object under, can think in the data of this data centralization not containing exception, therefore this data set be registered.
As the register method of normal data, arbitrary method can be used.Such as, normal data register 52 matrix (table) for preparing that there is process variable in the row direction, there is time (interval between diagnosis) in a column direction.Then, under being judged as that the data set of diagnosis object is normal situation, data set writes to prepared matrix by normal data register 52.In addition, when being judged as that exception level is not 0%, data set does not such as write-99999 to matrix by normal data register 52.Fig. 8 represents the example of the matrix that normal data register 52 records.Data beyond-99999 are registered normal data.Then, the normal data registered when being endowed the triggering of the condition preset in starting in zero month zero day etc. according to preassigned specified period, or when user has transferred, is exported by model construction portion 51.
Model construction portion 51 possesses identification part 511, abnormality detection data definition part 512 and contribution amount definition part 513.
Normal data is read in from normal data register 52 in identification part 511, and adds read in normal data to the matrix L EssT.LTssT.LT X > defined abnormality detection data.Or the information obsolescence that the old times in matrix L EssT.LTssT.LT X > carve by identification part 511, replaces and is taken into normal data.
Abnormality detection data definition part 512 uses matrix L EssT.LTssT.LT X > to upgrade (1) formula ~ (5) formula, redefines abnormality detection data.
Contribution amount definition part 513 uses matrix L EssT.LTssT.LT X > to upgrade (6) formula ~ (7) formula, redefines contribution amount.
In addition, identification part 511 also can use normal data to upgrade daily mode data as required.But, due to various methods generate pattern from the time series data in past that pattern recognition device uses robust to infer originally, therefore also can not use this normal data and proceed to upgrade with the cycle of regulation.
As described above, the normal data register 52 of the second embodiment, under being judged as that by exception level judging part 332 operation data are normal situation, is registered normal data.Then, model construction portion 51, by data registered, upgrades abnormity diagnosis model.Thus, operation monitoring diagnostic device 50 only can extract normal data while automatically upgrade abnormity diagnosis model from abnormality diagnostic result.
Also there is the known method such as ecad MSPC in the renewal of abnormity diagnosis model, but in ecad MSPC, automatically not to upgrade abnormity diagnosis model with the cycle of regulation and carry out normal, the abnormal judgement of data.Therefore, when containing abnormal data, abnormity diagnosis model can be adapted to abnormal data sometimes, and diagnostic accuracy can worsen.In the operation monitoring diagnostic device 50 of the second embodiment, only extract and be judged as that normal data upgrade abnormity diagnosis model by initial abnormity diagnosis model, therefore, it is possible to automatically improve abnormality diagnostic performance, and constantly can be suitable for the characteristic of the factory changed lentamente.Therefore, the maintenance of the abnormity diagnosis model becoming bottleneck in the realization of diagnostic system is not needed.
In addition, the abnormality diagnostic a series of action of present embodiment is assumed to and realizes in the operation monitoring diagnostic device 50 of surveillance, but is not limited thereto.Such as, also can be, make the function of operation monitoring diagnostic device 50, i.e. division 31, model construction portion 51, monitoring diagnostic portion 33 and normal data register 52 function realize on Cloud Server, exception level is judged, abnormal factors infer and sensor operation anomalous identification information remote provide to the user needed.
In addition, the program performed by operation monitoring diagnostic device 30,50 and operation monitoring diagnostic program, also can be recorded in the recording medium of embodied on computer readable.
Several embodiment of the present invention is illustrated, but these embodiments are point out as an example, are not intended to limit scope of invention.These new embodiments can be implemented in other various modes, can carry out various omission, displacement, change in the scope of purport not departing from invention.These embodiments and distortion thereof are contained in scope of invention and purport, and be contained in claims record invention and equivalent scope in.
The explanation of symbol
10 ... lower water height treatment process, 101 ... initial settling basin, 102 ... anaerobism groove, 103 ... anaerobic groove, 104 ... aerobic groove, 105 ... final sedimentation tank, 1111 ... draw-off pump, 1112 ... extract flow sensor, 1121 ... fan blower, 1122 ... air supply flow sensor, 1131 ... ebullator, 1132 ... circular flow sensor, 1141 ... loopback sludge pump, 1142 ... loop back traffic sensor, 1151 ... draw-off pump, 1152 ... extract flow sensor, 121 ... rain sensor, 122 ... lower water flows into quantity sensor, 123 ... flow into TN sensor, 124 ... flow into TP sensor, 1251 ... flow into UV sensor, 1252 ... flow into COD sensor, 126 ... anaerobism groove ORP sensor, 127 ... anaerobism groove pH sensor, 128 ... anaerobic groove ORP sensor, 129 ... anaerobic groove pH sensor, 1210 ... phosphoric acid sensor, 1211 ... DO sensor, 1212 ... ammoniacal sensor, 1213 ... MLSS sensor, 1214 ... cooling-water temperature sensor, 1215 ... excess sludge SS sensor, 1216 ... discharge SS sensor, 1217 ... sludge interface sensor, 1218 ... lower water discharge capacity sensor, 1219 ... discharge TN sensor, 1220 ... discharge TP sensor, 12211 ... discharge UV sensor, 12212 ... discharge COD sensor, 20 ... Data Collection storage unit, 30, 50 ... operation monitoring diagnostic device, 31 ... division, 311 ... attribute supply unit, 312 ... determined property portion, 32, 51 ... model construction portion, 321, 511 ... identification part, 322, 512 ... abnormality detection data definition part, 323, 513 ... contribution amount definition part, 33 ... monitoring diagnostic portion, 331 ... extraction unit, 332 ... exception level judging part, 333 ... element paritng portion, 334 ... anomalous identification portion, 40 ... user interface, 52 ... normal data register.

Claims (16)

1. an operation monitoring diagnostic device, obtains the operation data measured to by the sensor arranged in object operation, wherein,
Possess: division; Model construction portion, has identification part and abnormality detection data definition part; And monitoring diagnostic portion, there is extraction unit and exception level judging part,
The attribute information that above-mentioned division keeps the environment residing for above-mentioned object operation and classifies, and based on above-mentioned attribute information, above-mentioned operation data and multiple time series datas of being made up of the operation data in the past of the multiple periods preset are classified,
Above-mentioned identification part by robustly inferring that the above-mentioned time series data typical value being separately classified as multiple classification makes the daily mode data of each above-mentioned classification, and makes non-daily mode data based on the daily mode data of each above-mentioned classification,
Above-mentioned abnormality detection data definition part is each based on above-mentioned multiple non-daily mode data, makes the first definition for calculating abnormality detection data and the second definition for the threshold value that calculates above-mentioned abnormality detection data,
Said extracted portion extracts the deviation degree of the above-mentioned operation data be classified relative to above-mentioned multiple daily mode data,
Above-mentioned exception level judging part comes to calculate abnormality detection data according to each above-mentioned classification by above-mentioned multiple deviation degree being applied to respectively above-mentioned first definition, come according to each above-mentioned classification calculated threshold, by judging according to each above-mentioned classification whether above-mentioned abnormality detection data exceed above-mentioned threshold value and judge exception level by above-mentioned multiple deviation degree being applied to respectively above-mentioned second definition.
2. as the operation monitoring diagnostic device that claim 1 is recorded, wherein,
Above-mentioned model construction portion also has contribution amount definition part, and this contribution amount definition part is each based on above-mentioned multiple non-daily mode data, makes for calculating three definition of each process variable for the contribution amount of above-mentioned abnormality detection data,
Above-mentioned monitoring diagnostic portion also has element paritng portion, this element paritng portion is by the deviation degree of daily mode data the highest relative to adaptability between the above-mentioned operation data be classified for the above-mentioned operation data be classified, be applied to above-mentioned 3rd definition produced for the non-daily mode data that adaptability between the above-mentioned operation data be classified is the highest, calculate the contribution amount of each above-mentioned process variable thus.
3. as the operation monitoring diagnostic device that claim 2 is recorded, wherein,
Above-mentioned monitoring diagnostic portion also possesses anomalous identification portion, this anomalous identification portion is when the process variable of the contribution amount exceeding contribution amount threshold value is 1, being speculated as sensor produces abnormal, when the process variable of the contribution amount exceeding above-mentioned contribution amount threshold value is more than 2, is speculated as above-mentioned object operation and produces abnormal.
4. as the operation monitoring diagnostic device that claim 1 is recorded, wherein,
Above-mentioned identification part makes above-mentioned daily mode data according to the definition of multiple kinds relevant with calendar, and makes the non-daily mode data based on each daily mode data defined above,
Above-mentioned abnormality detection data definition part, for each non-daily mode data produced according to each above-mentioned classification and each above-mentioned definition, makes first and second definition above-mentioned,
Said extracted portion, for each daily mode data produced according to each above-mentioned classification and each above-mentioned definition, extracts above-mentioned deviation degree,
The abnormality detection data that above-mentioned exception level judging part calculates but multiple deviation degrees that above-mentioned classification is different identical according to above-mentioned definition each and threshold value compare, and result judges exception level based on the comparison.
5. as the operation monitoring diagnostic device that claim 3 is recorded, wherein,
Above-mentioned anomalous identification portion is when the process variable of the contribution amount exceeding contribution amount threshold value is 1, replace the measurement value creating above-mentioned abnormal sensor, by based on create above-mentioned abnormal sensor normal time the value of data that measures, supply to above-mentioned extraction unit.
6. as the operation monitoring diagnostic device that claim 1 is recorded, wherein,
Also possess normal data register in above-mentioned exception level judging part, when being judged as not producing abnormal in operation data, this normal data register is registered not producing above-mentioned abnormal operation data as normal data,
Above-mentioned abnormality detection data definition part uses above-mentioned normal data to upgrade first and second definition above-mentioned.
7. as the operation monitoring diagnostic device that claim 6 is recorded, wherein,
Above-mentioned normal data is enrolled the above-mentioned time series data be classified by above-mentioned identification part, and robustly infers the typical value of the time series data after having enrolled above-mentioned normal data, upgrades thus to above-mentioned daily mode data.
8. as the operation monitoring diagnostic device that claim 2 is recorded, wherein,
Also possess normal data register in above-mentioned exception level judging part, when being judged as not producing abnormal in operation data, this normal data register is registered not producing above-mentioned abnormal operation data as normal data,
Above-mentioned abnormality detection data definition part uses above-mentioned normal data to upgrade first and second definition above-mentioned,
Above-mentioned contribution amount definition part uses above-mentioned normal data to upgrade above-mentioned 3rd definition.
9. as the operation monitoring diagnostic device that claim 8 is recorded, wherein,
Above-mentioned normal data is enrolled the above-mentioned time series data be classified by above-mentioned identification part, and robustly infers the typical value of the time series data after having enrolled above-mentioned normal data, upgrades thus to above-mentioned daily mode data.
10. as the operation monitoring diagnostic device that claim 1 is recorded, wherein,
Above-mentioned division, above-mentioned model construction portion and above-mentioned monitoring diagnostic portion realize on Cloud Server.
11. 1 kinds of operation monitoring diagnostic devices, obtain the operation data measured to by the sensor arranged in object operation, wherein,
Possess: division; Model construction portion, has identification part, abnormality detection data definition part and contribution amount definition part; And monitoring diagnostic portion, there is extraction unit and element paritng portion,
The attribute information that above-mentioned division keeps the environment residing for above-mentioned object operation and classifies, and based on above-mentioned attribute information, above-mentioned operation data and the time series data that is made up of the operation data in the past during presetting are classified,
Above-mentioned identification part by robustly inferring that the typical value of the above-mentioned time series data be classified makes daily mode data, and makes non-daily mode data based on above-mentioned daily mode data,
Above-mentioned abnormality detection data definition part, based on above-mentioned non-daily mode data, makes the first definition for calculating abnormality detection data,
Above-mentioned contribution amount definition part, based on above-mentioned non-daily mode data, makes for calculating second definition of each process variable for the contribution amount of above-mentioned abnormality detection data,
Said extracted portion extracts the deviation degree of the above-mentioned operation data be classified relative to above-mentioned daily mode data,
Above-mentioned element paritng portion is by being applied to the contribution amount that first and second definition above-mentioned calculates each above-mentioned process variable by above-mentioned deviation degree.
The 12. operation monitoring diagnostic devices recorded as claim 11, wherein,
Above-mentioned monitoring diagnostic portion also has anomalous identification portion, this anomalous identification portion is when the process variable of the contribution amount exceeding contribution amount threshold value is 1, being speculated as sensor produces abnormal, when the process variable of the contribution amount exceeding above-mentioned contribution amount threshold value is more than 2, is speculated as above-mentioned object operation and produces abnormal.
The 13. operation monitoring diagnostic devices recorded as claim 12, wherein,
Above-mentioned anomalous identification portion is when the process variable of the contribution amount exceeding contribution amount threshold value is 1, replace the measurement value creating above-mentioned abnormal sensor, by based on create above-mentioned abnormal sensor normal time the value of data that measures, supply to above-mentioned extraction unit.
The 14. operation monitoring diagnostic devices recorded as claim 11, wherein,
Above-mentioned division, above-mentioned model construction portion and above-mentioned monitoring diagnostic portion realize on Cloud Server.
15. 1 kinds of operation monitoring diagnostic programs, use, make the computer-implemented following process of above-mentioned operation monitoring diagnostic device in the operation monitoring diagnostic device obtaining in by object operation the operation data that the sensor that arranges measures:
Classification process, the attribute information of classifying based on the environment residing for above-mentioned object operation, classifies to above-mentioned operation data and multiple time series datas of being made up of the operation data in the past of the multiple periods preset;
Identifying processing, by robustly inferring that the above-mentioned time series data typical value being separately classified as multiple classification makes the daily mode data of each above-mentioned classification, and makes non-daily mode data based on the daily mode data of each above-mentioned classification;
Definition process, each based on above-mentioned multiple non-daily mode data, makes the first definition for calculating abnormality detection data and the second definition for the threshold value that calculates above-mentioned abnormality detection data;
Extraction process, extracts the deviation degree of the above-mentioned operation data be classified relative to above-mentioned multiple daily mode data; And
Judge process, come to calculate abnormality detection data according to each above-mentioned classification by above-mentioned multiple deviation degree being applied to respectively above-mentioned first definition, come according to each above-mentioned classification calculated threshold, by judging according to each above-mentioned classification whether above-mentioned abnormality detection data exceed above-mentioned threshold value and judge exception level by above-mentioned multiple deviation degree being applied to respectively above-mentioned second definition.
16. 1 kinds of operation monitoring diagnostic programs, use, make the computer-implemented following process of above-mentioned operation monitoring diagnostic device in the operation monitoring diagnostic device obtaining in by object operation the operation data that the sensor that arranges measures:
Classification process, the attribute information of classifying based on the environment residing for above-mentioned object operation, classifies to above-mentioned operation data and the time series data that is made up of the operation data in the past during presetting;
Identifying processing, by robustly inferring that the typical value of the above-mentioned time series data be classified makes daily mode data, and makes non-daily mode data based on above-mentioned daily mode data;
First definition process, based on above-mentioned non-daily mode data, makes the first definition for calculating abnormality detection data;
Second definition part, based on above-mentioned non-daily mode data, makes for calculating second definition of each process variable for the contribution amount of above-mentioned abnormality detection data;
Extraction process, extracts the deviation degree of the above-mentioned operation data be classified relative to above-mentioned daily mode data; And
Element paritng process, by being applied to the contribution amount that first and second definition above-mentioned calculates each above-mentioned process variable by above-mentioned deviation degree.
CN201380058447.8A 2012-11-09 2013-09-05 Operation monitors diagnostic device and operation monitoring diagnostic program Active CN104781741B (en)

Applications Claiming Priority (3)

Application Number Priority Date Filing Date Title
JP2012-247560 2012-11-09
JP2012247560A JP5996384B2 (en) 2012-11-09 2012-11-09 Process monitoring diagnostic device, process monitoring diagnostic program
PCT/JP2013/073998 WO2014073261A1 (en) 2012-11-09 2013-09-05 Process monitoring/diagnosis device and process monitoring/diagnosis program

Publications (2)

Publication Number Publication Date
CN104781741A true CN104781741A (en) 2015-07-15
CN104781741B CN104781741B (en) 2017-07-04

Family

ID=50684378

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201380058447.8A Active CN104781741B (en) 2012-11-09 2013-09-05 Operation monitors diagnostic device and operation monitoring diagnostic program

Country Status (3)

Country Link
JP (1) JP5996384B2 (en)
CN (1) CN104781741B (en)
WO (1) WO2014073261A1 (en)

Cited By (10)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN107179752A (en) * 2016-03-10 2017-09-19 横河电机株式会社 Process monitoring arrangement, process monitoring system, process monitoring method, process monitoring program and recording medium
CN110226140A (en) * 2017-01-25 2019-09-10 Ntn株式会社 State monitoring method and state monitoring apparatus
CN110260913A (en) * 2018-03-12 2019-09-20 株式会社理光 Diagnostic device, diagnostic method and computer readable storage medium
CN110347561A (en) * 2019-06-11 2019-10-18 平安科技（深圳）有限公司 Monitoring alarm method and terminal device
TWI705317B (en) * 2018-05-30 2020-09-21 日商橫河電機股份有限公司 Abnormality detecting apparatus, abnormality detecting method, abnormality detecting program, and recording medium
CN111695584A (en) * 2019-03-14 2020-09-22 株式会社日立制作所 Time series data monitoring system and time series data monitoring method
CN112567306A (en) * 2018-08-31 2021-03-26 东芝三菱电机产业系统株式会社 Manufacturing process monitoring device
CN112955837A (en) * 2018-10-30 2021-06-11 国立研究开发法人宇宙航空研究开发机构 Abnormality diagnosis device, abnormality diagnosis method, and program
CN114265390A (en) * 2021-12-22 2022-04-01 苏州华星光电技术有限公司 Equipment data acquisition diagnosis method and device, server and storage medium
CN114375389A (en) * 2019-09-13 2022-04-19 住友重机械工业株式会社 Diagnostic system

Families Citing this family (25)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US9767671B2 (en) * 2014-11-05 2017-09-19 Intel Corporation System for determining sensor condition
JP6514494B2 (en) * 2014-11-28 2019-05-15 ローム株式会社 Sensor network system and its operation method
WO2016088362A1 (en) * 2014-12-05 2016-06-09 日本電気株式会社 System analyzing device, system analyzing method and storage medium
CN104731083B (en) * 2015-02-03 2017-06-20 浙江大学 A kind of industrial method for diagnosing faults and application based on self-adaptive feature extraction
JP6595245B2 (en) * 2015-07-27 2019-10-23 株式会社東芝 Monitoring system, information providing method, and computer program
DE102016008987B4 (en) 2015-07-31 2021-09-16 Fanuc Corporation Machine learning method and machine learning apparatus for learning failure conditions, and failure prediction apparatus and failure prediction system including the machine learning apparatus
JP6148316B2 (en) * 2015-07-31 2017-06-14 ファナック株式会社 Machine learning method and machine learning device for learning failure conditions, and failure prediction device and failure prediction system provided with the machine learning device
JP6140331B1 (en) 2016-04-08 2017-05-31 ファナック株式会社 Machine learning device and machine learning method for learning failure prediction of main shaft or motor driving main shaft, and failure prediction device and failure prediction system provided with machine learning device
JP6607821B2 (en) * 2016-04-12 2019-11-20 株式会社神戸製鋼所 Blast furnace sensor failure detection method and abnormal situation prediction method
JP7009438B2 (en) * 2016-07-07 2022-01-25 アスペン テクノロジー インコーポレイテッド Computer systems and methods for monitoring key performance indicators (KPIs) using time series pattern models
CN106371939B (en) * 2016-09-12 2019-03-22 山东大学 A kind of time series data method for detecting abnormality and its system
JP6888291B2 (en) * 2016-12-16 2021-06-16 富士電機株式会社 Process monitoring device, process monitoring system and program
EP3367261A1 (en) * 2017-02-28 2018-08-29 Fraunhofer-Gesellschaft zur Förderung der angewandten Forschung e.V. Method for classifying information and classification processor
WO2019049688A1 (en) * 2017-09-06 2019-03-14 日本電信電話株式会社 Abnormal sound detecting device, abnormality model learning device, abnormality detecting device, abnormal sound detecting method, abnormal sound generating device, abnormal data generating device, abnormal sound generating method, and program
JP6915693B2 (en) * 2017-10-10 2021-08-04 日本電気株式会社 System analysis method, system analyzer, and program
JP6615963B1 (en) * 2018-08-31 2019-12-04 株式会社日立パワーソリューションズ Abnormal sign diagnostic apparatus and abnormal sign diagnostic method
JP7163117B2 (en) * 2018-09-14 2022-10-31 株式会社東芝 Process monitoring support device, process monitoring support system, process monitoring support method, and process monitoring support program
JP7408911B2 (en) * 2018-10-26 2024-01-09 富士電機株式会社 Anomaly detection device and anomaly detection method
CN113330379A (en) 2019-01-21 2021-08-31 三菱电机株式会社 Abnormality cause estimation device, abnormality cause estimation method, and program
CN113678074A (en) * 2019-03-26 2021-11-19 杰富意钢铁株式会社 Process state monitoring device and process state monitoring method
JP7398270B2 (en) * 2019-12-23 2023-12-14 株式会社トヨタプロダクションエンジニアリング Transmitting device, receiving device, transmitting method, receiving method, transmitting program and receiving program
CN115380294A (en) * 2020-03-31 2022-11-22 Abb瑞士股份有限公司 Data processing for industrial machine learning
WO2021255783A1 (en) * 2020-06-15 2021-12-23 株式会社日立製作所 Automatic inspection system and wireless slave machine
CN113377078B (en) * 2021-04-20 2023-04-07 山东师范大学 Industrial process fault detection method and system based on kernel principal component regression
JP7052914B1 (en) 2021-08-05 2022-04-12 富士電機株式会社 Abnormality diagnosis system, abnormality diagnosis device, abnormality diagnosis method, and program

Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JP2004303007A (en) * 2003-03-31 2004-10-28 Mitsubishi Chemicals Corp Process monitoring method
JP2007065883A (en) * 2005-08-30 2007-03-15 Toshiba Corp Process monitoring device and its method
CN102282516A (en) * 2009-02-17 2011-12-14 株式会社日立制作所 Abnormality detecting method and abnormality detecting system
CN102460529A (en) * 2009-06-11 2012-05-16 株式会社日立制作所 Device abnormality monitoring method and system
JP2012141712A (en) * 2010-12-28 2012-07-26 Toshiba Corp Process monitoring diagnostic system

Family Cites Families (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JP3495129B2 (en) * 1995-03-03 2004-02-09 株式会社東芝 Plant abnormality detection device
JP4762088B2 (en) * 2006-08-31 2011-08-31 株式会社東芝 Process abnormality diagnosis device

Patent Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JP2004303007A (en) * 2003-03-31 2004-10-28 Mitsubishi Chemicals Corp Process monitoring method
JP2007065883A (en) * 2005-08-30 2007-03-15 Toshiba Corp Process monitoring device and its method
CN102282516A (en) * 2009-02-17 2011-12-14 株式会社日立制作所 Abnormality detecting method and abnormality detecting system
CN102460529A (en) * 2009-06-11 2012-05-16 株式会社日立制作所 Device abnormality monitoring method and system
JP2012141712A (en) * 2010-12-28 2012-07-26 Toshiba Corp Process monitoring diagnostic system

Cited By (11)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN107179752A (en) * 2016-03-10 2017-09-19 横河电机株式会社 Process monitoring arrangement, process monitoring system, process monitoring method, process monitoring program and recording medium
CN110226140A (en) * 2017-01-25 2019-09-10 Ntn株式会社 State monitoring method and state monitoring apparatus
CN110260913A (en) * 2018-03-12 2019-09-20 株式会社理光 Diagnostic device, diagnostic method and computer readable storage medium
TWI705317B (en) * 2018-05-30 2020-09-21 日商橫河電機股份有限公司 Abnormality detecting apparatus, abnormality detecting method, abnormality detecting program, and recording medium
CN112567306A (en) * 2018-08-31 2021-03-26 东芝三菱电机产业系统株式会社 Manufacturing process monitoring device
CN112955837A (en) * 2018-10-30 2021-06-11 国立研究开发法人宇宙航空研究开发机构 Abnormality diagnosis device, abnormality diagnosis method, and program
CN111695584A (en) * 2019-03-14 2020-09-22 株式会社日立制作所 Time series data monitoring system and time series data monitoring method
CN110347561A (en) * 2019-06-11 2019-10-18 平安科技（深圳）有限公司 Monitoring alarm method and terminal device
CN114375389A (en) * 2019-09-13 2022-04-19 住友重机械工业株式会社 Diagnostic system
CN114265390A (en) * 2021-12-22 2022-04-01 苏州华星光电技术有限公司 Equipment data acquisition diagnosis method and device, server and storage medium
CN114265390B (en) * 2021-12-22 2024-02-20 苏州华星光电技术有限公司 Equipment data acquisition diagnosis method, device, server and storage medium

Also Published As

Publication number Publication date
JP2014096050A (en) 2014-05-22
WO2014073261A1 (en) 2014-05-15
CN104781741B (en) 2017-07-04
JP5996384B2 (en) 2016-09-21

Similar Documents

Publication Publication Date Title
CN104781741A (en) Process monitoring/diagnosis device and process monitoring/diagnosis program
JP5284433B2 (en) Process monitoring / diagnosis / support equipment
JP4468269B2 (en) Process monitoring apparatus and method
EP0708390B1 (en) Supporting method and system for process operation
JP5259797B2 (en) Learning type process abnormality diagnosis device and operator judgment estimation result collection device
JP4762088B2 (en) Process abnormality diagnosis device
CN112863134B (en) Intelligent diagnosis system and method for rural sewage treatment facility abnormal operation
CN104793605B (en) A kind of method that utilization normal distribution judges equipment fault
JP2009070071A (en) Learning process abnormality diagnostic device and operator&#39;s judgement estimation result collecting device
CN105027010A (en) Control parameter adjustment method and control parameter adjustment system
Yel et al. Prediction of primary treatment effluent parameters by Fuzzy Inference System (FIS) approach
Liu et al. Transforming data into actionable knowledge for fault detection, diagnosis and prognosis in urban wastewater systems with AI techniques: A mini-review
CN108052092A (en) A kind of subway electromechanical equipment abnormal state detection method based on big data analysis
CN112861350A (en) Temperature overheating defect early warning method for stator winding of water-cooled steam turbine generator
Andreides et al. The prediction of WWTP influent characteristics: Good practices and challenges
CN110426996B (en) Environmental pollution monitoring method based on big data and artificial intelligence
DE102008045840B4 (en) Method for operating a process measuring point
CN113221442B (en) Method and device for constructing health assessment model of power plant equipment
Shepherd et al. Cloud-based artificial intelligence analytics to assess combined sewer overflow performance
Dixon et al. Experience with data mining for the anaerobic wastewater treatment process
CN116293454A (en) Management system, method and medium based on steam pipe network dynamic balance
Dürrenmatt Data mining and data-driven modeling approaches to support wastewater treatment plant operation
Cicceri et al. A machine learning approach for anomaly detection in environmental iot-driven wastewater purification systems
Qambar et al. Anomaly detection in wastewater treatment plants using unsupervised machine learning algorithms
Rustum et al. Fault detection in the activated sludge process using the kohonen self-organising map

Legal Events

Date Code Title Description
C06 Publication
PB01 Publication
EXSB Decision made by sipo to initiate substantive examination
SE01 Entry into force of request for substantive examination
GR01 Patent grant
GR01 Patent grant