Nothing Special   »   [go: up one dir, main page]

CN107358347A - Equipment cluster health state evaluation method based on industrial big data - Google Patents

Equipment cluster health state evaluation method based on industrial big data Download PDF

Info

Publication number
CN107358347A
CN107358347A CN201710543984.1A CN201710543984A CN107358347A CN 107358347 A CN107358347 A CN 107358347A CN 201710543984 A CN201710543984 A CN 201710543984A CN 107358347 A CN107358347 A CN 107358347A
Authority
CN
China
Prior art keywords
equipment
mrow
cluster
data
state
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.)
Pending
Application number
CN201710543984.1A
Other languages
Chinese (zh)
Inventor
孔宪光
王继虎
常建涛
刘尧
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.)
Xidian University
Original Assignee
Xidian University
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 Xidian University filed Critical Xidian University
Priority to CN201710543984.1A priority Critical patent/CN107358347A/en
Publication of CN107358347A publication Critical patent/CN107358347A/en
Pending legal-status Critical Current

Links

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q10/00—Administration; Management
    • G06Q10/06—Resources, workflows, human or project management; Enterprise or organisation planning; Enterprise or organisation modelling
    • G06Q10/063—Operations research, analysis or management
    • G06Q10/0639—Performance analysis of employees; Performance analysis of enterprise or organisation operations
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
    • G06F16/90—Details of database functions independent of the retrieved data types
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F30/00—Computer-aided design [CAD]
    • G06F30/20—Design optimisation, verification or simulation

Landscapes

  • Engineering & Computer Science (AREA)
  • Business, Economics & Management (AREA)
  • Theoretical Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Human Resources & Organizations (AREA)
  • General Physics & Mathematics (AREA)
  • Entrepreneurship & Innovation (AREA)
  • General Engineering & Computer Science (AREA)
  • Strategic Management (AREA)
  • Databases & Information Systems (AREA)
  • Economics (AREA)
  • Educational Administration (AREA)
  • Development Economics (AREA)
  • Tourism & Hospitality (AREA)
  • General Business, Economics & Management (AREA)
  • Quality & Reliability (AREA)
  • Operations Research (AREA)
  • Marketing (AREA)
  • Data Mining & Analysis (AREA)
  • Game Theory and Decision Science (AREA)
  • Computer Hardware Design (AREA)
  • Evolutionary Computation (AREA)
  • Geometry (AREA)
  • Management, Administration, Business Operations System, And Electronic Commerce (AREA)

Abstract

The invention discloses a kind of equipment cluster health state evaluation method based on industrial big data, solves the problems, such as the health state evaluation for equipping cluster.It is on active service full the big data environment in cycle based on equipment cluster, structure entity status information Slice administrative model is to data prediction；Equipment cluster is divided with operating mode Similarity-Based Clustering Method, the mirror image model mutually mapped with entity health status is established with homing method, obtains the health status quantitative model of different equipment clusters；To the equipment state of health data fusion of different clusters and restructuring, the health degree of different equipments is obtained, and is fitted equipment health status degenerated curve, prediction equipment residual life.The present invention uses big data cluster modeling method, can not only carry out effective otherness assessment to equipment cluster health status, and cluster modeling can reduce model redundancy, simplified model internal structure.Cluster equipment normal operation is ensured, excavates equipment use value to greatest extent.

Description

Equipment cluster health state evaluation method based on industrial big data
Technical field
The invention belongs to field of engineering technology, is related to the modeling of equipment health status and assesses, and is specifically that one kind is based on cluster The equipment cluster health state evaluation method of modeling, the otherness quantitative evaluation available for equipment cluster health status.
Background technology
Equipment health state evaluation is primarily referred to as according to the data of the sensor of installation measurement, the data of manual measurement, gone through History data, experimental data etc. are analyzed, and are considered the influence of the factors such as the use of equipment, environment, maintenance, are commented using various Estimation algorithm establishes model, and the health status of equipment is assessed according to specified evaluation index system, clearly equips healthy shape A kind of technology of state.The residual life of the correct health status evaluation equipped and Accurate Prediction equipment is for effectively avoiding Stopping production accident is shut down, ensures equipment safety operation, ensure normal orderly production and increase economic efficiency have huge meaning.
With the rapid development of national equipment manufacture in recent years, Large Complex Equipment shows clustering, scale Development trend.So that underground engineering construction equips shield machine as an example, current national shield industry has more than 1000 shield machine, same daily When the shield machine that works up to more than 300, equipment cluster is very huge.Equipment in these clusters is with similar in structure Property, however, because its working environment is complicated, be often distributed in different geographic areas so that its under different operating modes, or even Because its different health status of service condition also has very big difference under identical operating mode.Therefore, separate unit is equipped under either simplex condition The mode of modelling evaluation far can not meet Large-Scale Equipment group variation character state real-time performance evaluation and predictive maintenance It is required that the health state evaluation method towards cluster equipment is necessary.
Filled for cluster health state evaluation problem, tradition caused by equipment configuration complexity, substantial amounts based on separate unit The health evaluating method of standby data analysis can not meet the real time health state estimation and performance prediction need of Large-Scale Equipment cluster Ask.And the sensor and data collecting system being equipped with Large Complex Equipment cluster can gather increasing equipment clothes The data such as state parameter, operating condition, service condition and the ambient parameter in labour stage, provided for equipment cluster health state evaluation Reliable big data environment.Therefore a kind of equipment cluster health state evaluation method based on industrial big data is proposed, next pair Equip health state evaluation and prediction.
The content of the invention
The purpose of the present invention aims to solve the problem that the otherness quantitative evaluation problem of large construction cluster equipment, for prior art not Foot, propose a kind of equipment cluster health state evaluation method based on industrial big data.
The present invention is a kind of equipment cluster health state evaluation method based on big data cluster modeling, it is characterised in that Cluster health state evaluation method is on active service and big data environment is equipped caused by the cycle entirely based on equipment cluster, includes the operation of equipment Ambient parameter, plant maintenance and maintenance record and performance during state parameter, the floor data of equipment operation, equipment use Class data, data above is from equipment sensor, manual record and subsequent statistical；Equip cluster health state evaluation include just like Lower step：
Step 1：Storage and analysis that big data platform is used for magnanimity equipment data are built, the big data platform includes Hadoop/Spark platforms and Storm platforms, Hadoop/Spark platforms are used for storage organization and semi-structured data, Storm platforms are used to store dynamic mass data；
Step 2：With entity state Slice management method to equipment big data carry out data prediction, according to dress The degree of correlation of standby status level is classified to equipment big data, and data are broadly divided into two classes：The first kind be fault data and Maintenance log data, the second class are state parameter data；According to inhomogeneity data, different entity state Slices is selected Management method, for primary sources, directly the reference node as equipment state renewal；It is right for secondary sources The continuous multivariable service data for equipping cluster carries out the monitoring of abnormality, is extracted using duty parameter subsequence extracting method Equipping in cluster running has the state feature of significant changes, in this, as the change node of state renewal；
Step 3：Based on Time series analysis method, operating mode automatic cluster is carried out to the operational factor of equipment, according to operating mode Least unit of the cluster that similitude clusters to obtain as cluster, equipment collection being called, equipment collection is internal to use identical modeling method, It is convenient to carry out united analysis；
Step 4：Using regression modeling method, each equipment cluster shape is established to different operating mode clusters according to historical data The mirror image model of state expression, carry out equipping the quantization of cluster health status, obtain each equipping cluster health status；
Step 5：Carry out data fusion to recombinate to obtain the health status of single equipment, and obtained using polynomial interpolation Its health status decline curve, equipment predicting residual useful life is carried out using the decline curve, provides prediction result.
The present invention is on active service big data environment caused by the cycle entirely based on equipment cluster, is entered with big data cluster modeling method Row data analysis and modeling, propose again to a whole set of flow of cluster analysis from data prediction to data modeling, so as to collection The health status of group's equipment carries out quantitative evaluation.The inventive method has cluster modeling analysis ability, overcomes conventional modeling method The defects of either simplex condition, single device can only be analyzed, magnanimity caused by cluster, multi-source, multiple types data are made full use of, and by big The storage of data platform and analysis ability, accurately and efficiently cluster health status is assessed.
The present invention compared with prior art, has advantages below：
(1) this analysis method is covered from data prediction, a whole set of flow of data modeling to cluster analysis, method master To include three core procedures：Entity status information Slice management, establishes physical system mirror image model and cluster state quantifies Model, generation equipment residual life curve, it is respectively used to solve data prediction, number in equipment cluster big data cluster modeling The problem of according to three aspects of modeling and cluster analysis.
(2) the inventive method using big data platform is used for that equipment big data is stored and calculated, and improves storage And treatment effeciency, ensure the real-time and high efficiency of analysis, provided the foundation point for equipment health status real-time update and assessment Analyse framework.
(3) the inventive method is directed to existing company-data feature, splits data into two classes：The first kind is fault data and dimension Maintenance record data is repaiied, can be directly as the reference node of equipment state renewal；Second class be state parameter data, it is necessary to The reference node updated using duty parameter subsequence extracting method as equipment state；The data classification method is by fault data With maintenance log data as a kind of update method with reference to node, the renewal of equipment cluster can be efficiently established with reference to section Point.
(4) present invention proposes a kind of automatic clustering method based on operating mode, when can avoid the selection of K-means methods initial value The too near situation of cluster initial point distance being likely to occur, can not only automatically determine the number of preliminary examination cluster point, and can improve The efficiency of initial data set division.
(5) cluster health state evaluation uses regression modeling method, models to obtain respectively respectively for different operating mode clusters The mirror image model of cluster.The coupling between model is reduced, substantially increases model accuracy rate.
Brief description of the drawings
Fig. 1 is the overall framework figure of the inventive method；
Fig. 2 is the date storage method schematic diagram of the inventive method；
Fig. 3 is entity state Slice management method flow chart in the inventive method；
Fig. 4 is operating mode automatic clustering method schematic diagram in the inventive method；
Fig. 5 is big data platform building flow chart in the inventive method；
Fig. 6 is big data platform assembly figure in the inventive method；
Fig. 7 is embodiment operating mode cluster result figure；
Fig. 8 is embodiment engine sensor parameter amplitude tendency chart；
Fig. 9 is No. 1 engine health status degenerative process result of calculation；
Embodiment
Embodiment 1：
Traditional analysis can only be directed to either simplex condition, single device carries out health status modeling analysis, with equipment scale Constantly expand, just demonstrate undoubtedly the drawbacks of conventional method for analyzing and modeling efficiency and accuracy, do not possess device clusters modeling ability, And then influence the safe efficient operation of cluster device.
The problem of present invention exists for above-mentioned analysis method, deploy research and discussion, introduce cluster modeling technology, propose A kind of equipment cluster health state evaluation method based on industrial big data, referring to Fig. 1, cluster health state evaluation method is based on Equipment cluster is on active service entirely equips big data environment caused by the cycle, include the operating mode of the running state parameter of equipment, equipment operation Ambient parameter, plant maintenance and maintenance record and performance class data during data, equipment use, data above is from equipment Sensor, manual record and subsequent statistical.Present invention equipment cluster health state evaluation includes having the following steps：
Step 1：Storage and analysis that big data platform is used for magnanimity equipment data, reference picture 2 are built, the big data is put down Platform includes Hadoop/Spark platforms and Storm platforms, and Hadoop/Spark platforms are used for storage organizationization and semi-structured number According to, such as：Design drawing data, job schedule information, make an inspection tour record form etc.；Storm platforms for stream data storage and Analysis provides good solution, for storing dynamic mass data, such as：Equip operational factor, equipment state parameter, Equip duty parameter etc..
Step 2：With entity state Slice management method to equipment big data carry out data prediction, according to dress The degree of correlation of standby status level is classified to equipment big data, and data are broadly divided into two classes：The first kind be fault data and Maintenance log data, the second class are state parameter data.According to inhomogeneity data, different entity state Slices is selected Management method, for primary sources, directly the reference node as equipment state renewal.It is right for secondary sources The continuous multivariable service data for equipping cluster carries out the monitoring of abnormality, is extracted using duty parameter subsequence extracting method Equipping in cluster running has the state feature of significant changes, the change node as state renewal.
Step 3：Based on Time series analysis method, operating mode automatic cluster is carried out to the operational factor of equipment, according to operating mode Least unit of the cluster that similitude clusters to obtain as cluster, equipment collection being called, equipment collection is internal to use identical modeling method, It is convenient to carry out united analysis.
Step 4：Using regression modeling method, each equipment cluster shape is established to different operating mode clusters according to historical data The mirror image model of state expression, carry out equipping the quantization of cluster health status, obtain each equipping cluster health status.
Step 5：Carry out data fusion to recombinate to obtain the health status of single equipment, and obtained using polynomial interpolation Its health status decline curve, equipment predicting residual useful life is carried out using the decline curve, provides prediction result.
The present invention using big data platform is used for that equipment big data is stored and calculated, and improves storage and processing effect Rate, ensure the real-time and high efficiency of analysis, provided the foundation analysis framework for equipment health status real-time update and assessment.
Embodiment 2：
Equipment cluster health state evaluation method based on industrial big data is formed with embodiment 1, reference picture 3, with reality Body state Slice management method carries out data prediction to data, comprises the following specific steps that：
Step 2.1 is classified according to the degree of correlation horizontal with equipment state to equipment big data, according to data type The reference node of equipment state renewal is determined, data are broadly divided into two classes：The first kind is fault data and maintenance log number According to the reference node directly updated as equipment state.Second class is state parameter data, it is necessary to using duty parameter The reference node that sequential extraction procedures method updates as equipment state.
Step 2.2 cluster equipment data are mainly secondary sources, therefore the processing method of secondary sources is specifically wrapped The similitude detection of time series abnormal state detection and time series is included, the detection that abnormal nodes are carried out to time series data can It was found that frequently go out in its association between different time sections series modality, the i.e. time series of reflection equipment cluster running status Existing changing pattern and the changing pattern seldom occurred, here it is the abnormal patterns of equipment running status.In time series frequently The changing pattern of appearance and the changing pattern seldom occurred, the changing pattern frequently occurred in time series, such as normal operation shape State rule or cycle sexual abnormality etc., such as changing pattern seldom occurred, chance failure.Similarity system design is carried out to time series It can be found that the similitude and otherness of current time sequence and historical time sequence, so as to the time point to be differed greatly according to it Carry out piecewise analysis and modeling, conventional method such as auto-correlation and cross-correlation analysis, curve similarity detection etc., if detecting These "abnormal" or the node of " dramatically different ", can be as equipment state with reference to node.The step results seek to obtain Equipment state is with reference to node.
The reference node that step 2.3 obtains according to step 2.2, the history data set for equipping cluster is divided into different operating modes Under discrete state sequence.
Step 2.4 is that every section obtained above discrete status switch adds corresponding state tag, and with the corresponding time The service data and floor data of point are associated storage, facilitate data to be managed collectively.
Step 2.5 is last, when the operating mode for equipping cluster changes, the last state data of automatic record equipment, and Stick new state tag to be stored, obtain the training sample set of mirror image model.The present invention passes through Slice management method pair Cluster equipment big data carries out data prediction, obtains entity state renewal with reference to node, and then obtain the training of mirror image model Sample set.
The inventive method is directed to existing company-data feature, splits data into two classes：The first kind is fault data and maintenance Record data is maintained, can be directly as the reference node of equipment state renewal；Second class is state parameter data, it is necessary to adopt The reference node updated by the use of duty parameter subsequence extracting method as equipment state；The data classification method by fault data and Maintenance log data can efficiently establish the renewal of equipment cluster with reference to section as a kind of update method with reference to node Point.
Embodiment 3：
Equipment cluster health state evaluation method based on industrial big data is formed with embodiment 1-2, reference picture 4, for Complicated multi-state service data, operating mode automatic cluster is carried out to the operational parameter data of equipment.Clustered according to operating mode similitude Least unit of the cluster arrived as cluster, it is called equipment collection.Equipment collection is internal can to use identical modeling method, convenient progress United analysis.Wherein operating mode automatic clustering method specifically comprises the following steps：
Step 3.1 carries out operating mode automatic cluster to the operational parameter data of equipment, given first coefficient of determination θ, 0 ＜ θ ＜ 1, give and judge distance DC, and appoint and take a sample as first cluster centre Z1, such as take Z1=x1。
Step 3.2 finds new cluster centre, until finding all cluster centres of the sample：
Step 3.2a calculates other all sample x on the basis of first cluster centre1To Z1Distance Di1, i=1, 2,3 ..., n, distance calculation formula are：
If step 3.2b Dk1=max { Di1, i=1, wherein 2,3 ..., n, k are { Di1In distance value maximum element Lower label, take xkAs second cluster centre Z2。
Step 3.2c calculates all samples to cluster centre Z2Distance Di2, i=1,2,3 ..., n.
If step 3.2d Di=max { min (Di1,Di2), i=1, wherein 2 ..., n, DiFor distance Di1With distance Di2In Each maximum in minimum value, and Di＞ θ D12, D12It is Z1And Z2Between distance, then take xiFor the 3rd cluster centre Z3。
If step 3.2e Z3In the presence of if Dj=max { min (Di1,Di2,Di3), i=1, wherein 2 ..., n, DjFor distance Di1, distance Di2With distance Di3In each maximum in minimum value, and Dj＞ θ D12, then the 4th polymerization site is established Z4, by that analogy, until minimax distance is not more than θ D12When, terminate to find the calculating of polymerization site, obtain all equipments The cluster centre of multi-state supplemental characteristic.
Step 3.3 is according to closest principle all sample xi, i=1,2,3 ..., n, belong to closest gather Class center, obtain cluster result cluster.
If step 3.4 cluster result intra-cluster distance d, which is less than, gives set a distance Dc, terminate iterative process, obtain equipping multi-state Supplemental characteristic automatic cluster result cluster；If cluster result intra-cluster distance d, which is not less than, gives set a distance Dc, then gravity treatment θ and first are poly- Conjunction center Z1, return to step 3.2 and find new cluster center again, carry out next round iteration.
When automatic clustering method of the present invention can avoid the K-means methods initial value from choosing the cluster initial point that is likely to occur away from From too near situation, it can not only automatically determine the number of preliminary examination cluster point, and can improve the efficiency of initial data set division. In addition automatic operating cluster is also based on following method：Spectrum Conversion, wavelet transformation, piecewise polynomial are approached.
Embodiment 4：
Equipment cluster health state evaluation method based on industrial big data is formed with embodiment 1-3, with regression modeling Method models to obtain the mirror image model of each cluster respectively for different operating mode clusters, and specific steps include：
If there is a continuous reaction in health status of the step 4.1 under specific operation (such as operating mode 1) at different moments Variable H, the quantizating index for the health status that cluster is equipped under the operating mode is represented, its codomain is [0,1], by H as equipment cluster Health degree under the operating mode, logically return and establish quantitative model, wherein health degree H is exactly the health status for equipping cluster It is the probability of normal condition,
H=P (Y=1 | X)
So equip cluster unhealthy probability be
Wherein, Y represents equipment state, and wherein Y=0 represents that equipment is in malfunction, and Y=1 represents equipment in complete Health status, X (x1,x2,...,xi,...xn) it is state parameter variable.
Then definition equipment health status is that the ratio of complete health and failure is
Odd is referred to as ratio occurs.
Step 4.2 assumes the state parameter variable X (x of logarithm ratio and reflection equipment1,x2,...,xi,...xn) between A kind of linear relationship be present, i.e.,
Wherein W={ w1,w2,...,wnIt is model parameter, n is state parameter variable X (x1,x2,...,xi,...xn) Number, above formula both sides fetching number are i.e. available
Further solve the assessment models for producing equipment health degree
Thus establish the Logistic Regression mirror image models of equipment health state evaluation.
Equipment health status is assessed using Logistic Regression methods in the inventive method, there is standard True rate is high, the features such as easily implementation.
Embodiment 5：
Equipment cluster health state evaluation method based on industrial big data is formed with embodiment 1-4, to different operating mode collection Group's equipment carries out data fusion restructuring, obtains the health status value of single equipment, and can obtain it using polynomial interpolation Health status decline curve, equipment predicting residual useful life is carried out using the decline curve, is comprised the following steps that：
Step 5.1 extracts the health status value of same equipment and corresponding temporal information under different operating mode clusters.
Step 5.2 is ranked up to the equipment health status value of each model according to time order and function.
Degenerative process equation of the step 5.3 using the health status that polynomial interpolation calculating is each equipped with the time.
Step 5.4 is obtained equipping health status decline curve, equipped using the decline curve according to polynomial equation Predicting residual useful life.
The health degree equipped in the present invention also referred to as equips health status.Data fusion restructuring in the inventive method, leads to Cross corresponding temporal information to be ranked up the equipment health status value of each model according to time order and function, and carried out with multinomial Interpolation fitting, the historical data of health status value can be equipped by cluster, obtain equipping predicting residual useful life curve, it is effectively right Equipment residual life is predicted.
A more detailed example is given below, the present invention is further described
Embodiment 6：
Equipment cluster health state evaluation method based on industrial big data is formed with embodiment 1-5, reference picture 1, this hair It is bright to be used to solve the problems, such as to equip cluster health state evaluation., may simultaneously in view of following with the continuous expansion of cluster scale The magnanimity equipment cluster big data application scenarios occurred, specially propose a kind of equipment cluster health based on industrial big data State evaluating method.Based on big data parallel processing technique, equipment cluster mass data is obtained by Data Collection, then utilized Slice management method pre-processes to data, obtains equipping abnormality renewal with reference to node, and then be based on time series Analysis method carries out operating mode automatic cluster to the operational factor of equipment, and utilizes regression modeling method, according to historical data to not The mirror image model of each equipment cluster state expression is established with operating mode cluster, equipment cluster health status is quantified, finally carried out Data fusion, and equipment health status decline curve is obtained using polynomial interpolation, it is surplus to carry out equipment using the decline curve Remaining life prediction.
According to method of the present invention, the cluster health state evaluation process based on industrial big data is realized, specific bag Include and have the following steps：
(1) the hardware node topological structure in industrial big data platform is determined, number of nodes is carried out according to data volume size It is determined that its scale can be extended and shrink to provide the data storage service of different scales demand.
The industrial big data platform network node topology structure of the present embodiment is as shown in table 1 below, and table 1 is given comprising five The topological structure of server node, one of node is as host node NameNode, and four additional node is as back end DataNode, in order to ensure the robustness of platform, backup node SecondaryNameNode is deployed in non-master On NameNode.
The network node topological structure of table 1
Node Ip Attribute Remarks
Master.Hadoop 192.168.137.2 NameNode
Slave1.Hadoop 192.168.137.3 DataNode SecondaryNameNode
Slave2.Hadoop 192.168.137.4 DataNode
Slave3.Hadoop 192.168.137.5 DateNode
Slave4.Hadoop 192.168.137.6 DateNode
Each network node configuration in big data platform is as shown in table 2 below, and the configuration of four node computers is identical.
The node concrete configuration of table 2
Attribute Parameter
Memory size 1G
Hard drive space 20G
CPU number 1
Linux system version CentOS-6.3
Jdk versions jdk-6u32
Hadoop versions Hadoop-2.6.0
Spark versions Spark-1.4.0
Storm versions Storm-1.0.2
Reference picture 5, platform specific builds flow, comprising the installation of installation Linux system, meshed network configuration, ssh without password Login configurations, installation java environment, configuration Hadoop, configuration Spark, configuration Storm, and the platform put up progress is initial Change operation, carry out data storage and prepare with analysis.
Reference picture 6, after putting up industrial big data platform, console module includes HDFS, Hadoop Yarn, Hadoop The component such as MapReduce, Spark, Spark Streaming, Storm, the Hadoop/Spark platforms in the big data platform For storage organization and semi-structured data, Storm platforms are used to store dynamic mass data.
(2) related data is gathered, is mainly included：The state parameter of equipment operation, the floor data of equipment operation, equipment make The maintaining record and performance class data of ambient parameter, equipment during, each data concrete meaning are as follows：
The state parameter of equipment operation：Refer mainly to from sensor, such as：Perceive sensor and the control of the indexs such as vibration, temperature What is obtained in device processed is capable of the data of consersion unit operating condition and health status, that is, traditional monitoring data, The sample frequency of such data is often very high, and the variable of collection is also most numerous and diverse, comprising information metric density it is also maximum.
The floor data of equipment operation：Refer to the power equipped, rotating speed, the amount of feeding, moment of torsion, pressure etc. and determine that armament-related work is appointed Business and the setup parameter of condition, such data mostly come from equipment control device, and the floor data of equipment is for analytical equipment Running state parameter is of great significance, because only that equipment status parameter is compared and divided at the same conditions Analysis can just reflect the change of equipment health status and performance.
Ambient parameter during equipment use：Refer to the environmental information for being possible to influence equipment performance and running status, Such as temperature, wind speed, state of weather, such as when ship rides the sea, the environmental data such as unrestrained height, ocean current, wind speed, wind direction for There is highly important effect the economy danger of analysis ship, and collection ambient parameter information can help us to more fully understand that equipment is transported Row rule affected by environment, helps us performance change will be made a distinction caused by equipment state and environmental change.
The maintaining record of equipment：Record is changed in scheduled maintenance inspection and maintenance in the whole life span of equipment.This The reference that a little data can update as equipment state, mutually compares with the status data of equipment, can both be used as equipment state More new node carry out the health forecast model of more new equipment, the health status of equipment status parameter reflection can also be utilized to safeguard Front and rear changes to judge the validity of maintenance and maintenance work, and the statistical analysis of a large amount of such historical datas obtains equipment key The MTBF (Mean Time Between Failure, MTBF) of part, using this can as reliability lifted and The foundation that design is improved and service parts planning is formulated, these data can generally access acquisition from the systems such as ERP, EAM, BOM.
Performance class data：To the related performance of equipment operation and the index class number judged equipment running status According to, include energy consumption, the quality of production, machining accuracy etc. for manufacturing equipment, then can be according to the mesh of its work for other equipments Mark formulates corresponding performance indicators.The central role of performance indicators is to aid in us and understands performance state residing for current equipment, Help our labels to health, inferior health or failure in the data post of different time sections.
(3) caused a large amount of structural datas such as history static data and performance class data in cluster running will be equipped Stored using line data storage storehouse SQL server, design drawing, material consumption statistics and maintenance log etc. are non-structural Change and semi-structured static data using Hadoop distributed file systems HDFS storage, equip cluster real-time running data, The stream data such as Condition Monitoring Data and real-time working condition data is stored using Storm.
(4) data caused by equipment cluster are pre-processed using entity state Slice management method, specifically included Following steps：
Step 4a. extracts relevant with equipment state in equipment cluster initial data from unified equipment big data environment Data are divided by a large amount of failures and maintenance data and state parameter data according to the degree of correlation horizontal with equipment state Class, the reference node of equipment state renewal is determined according to data type, and data are broadly divided into two classes：The first kind be fault data and Maintenance log data；Second class is state parameter data；, can be directly as equipment state more to primary sources New reference node, mutually compareed with the operational parameter data of equipment, establish the mirror image model of equipment state；For he second-class number According to by the node for obtaining equipping cluster state renewal to continuous state argument sequence significantly abnormal detection or similarity system design (i.e. the time points of equipment state generation significant changes).
The reference node that step 4b. obtains according to above-mentioned steps, the history data set for equipping cluster is divided into different works Discrete state sequence under condition.
Step 4c. is that every section obtained above discrete status switch adds corresponding state tag, and with the corresponding time The service data and floor data of point are associated storage, facilitate data to be managed collectively.
Step 4d. is last, when the operating mode for equipping cluster changes, the last state data of automatic record equipment, and Stick new state tag to be stored, obtain the training sample set of mirror image model.The present invention passes through Slice management method pair Cluster equipment big data carries out data prediction, obtains entity state renewal with reference to node, and then obtain the training of mirror image model Sample set.
(5) automatic cluster is carried out to equipment company-data using operating mode similitude automatic clustering method, specifically included as follows Step：
Step 5a. carries out operating mode automatic cluster to the operational parameter data of equipment, given first coefficient of determination θ, 0 ＜ θ ＜ 1, give and judge distance DC, and appoint and take a sample as first cluster centre Z1, such as take Z1=x1。
Step 5b. finds new cluster centre, until finding all cluster centres of the sample：
Step 5ba. calculates other all sample x on the basis of first cluster centre1To Z1Distance Di1, i=1, 2,3 ..., n, distance calculation formula are：
If step 5bb. Dk1=max { Di1, i=1, wherein 2,3 ..., n, k are { Di1In distance value maximum element Lower label, take xkAs second cluster centre Z2。
Step 5bc. calculates all samples to cluster centre Z2Distance Di2, i=1,2,3 ..., n.
If step 5bd. Di=max { min (Di1,Di2), i=1, wherein 2 ..., n, DiFor distance Di1With distance Di2In Each maximum in minimum value, and Di＞ θ D12, D12It is Z1And Z2Between distance, then take xiFor the 3rd cluster centre Z3。
If step 5be. Z3In the presence of if Dj=max { min (Di1,Di2,Di3), i=1, wherein 2 ..., n, DjFor distance Di1, distance Di2With distance Di3In each maximum in minimum value, and Dj＞ θ D12, then the 4th polymerization site is established Z4, by that analogy, until minimax distance is not more than θ D12When, terminate to find the calculating of polymerization site, obtain all equipments The cluster centre of multi-state supplemental characteristic.
Step 5c. is according to closest principle all sample xi, i=1,2,3 ..., n, belong to closest gather Class center, obtain cluster result cluster.
If step 5d. cluster result intra-cluster distances d, which is less than, gives set a distance Dc, terminate iterative process, obtain equipping multi-state Supplemental characteristic automatic cluster result cluster；If cluster result intra-cluster distance d, which is not less than, gives set a distance Dc, then gravity treatment θ and first are poly- Conjunction center Z1, return to step 5b and find new cluster center again, carry out next round iteration.
(6) specific steps of equipment cluster health status mirror image model are established including as follows with regression modeling method：
If there is a continuous reaction in health status of the step 6a. under specific operation (such as operating mode 1) at different moments Variable H, the quantizating index for the health status that cluster is equipped under the operating mode is represented, its codomain is [0,1], by H as equipment cluster Health degree under the operating mode, logically return and establish quantitative model, wherein health degree H is exactly the health status for equipping cluster It is the probability of normal condition,
H=P (Y=1 | X)
So equip cluster unhealthy probability be
Wherein, Y represents equipment state, and wherein Y=0 represents that equipment is in malfunction, and Y=1 represents equipment in complete Health status, X (x1,x2,...,xi,...xn) it is state parameter variable.
Then definition equipment health status is that the ratio of complete health and failure is
Odd is referred to as ratio occurs.
Step 6b. assumes the state parameter variable X (x of logarithm ratio and reflection equipment1,x2,...,xi,...xn) between A kind of linear relationship be present, i.e.,
Wherein W={ w1,w2,...,wnIt is model parameter, n is state parameter variable X (x1,x2,...,xi,...xn) Number, above formula both sides fetching number are i.e. available
Further solve the assessment models for producing equipment health degree
Thus establish the Logistic Regression mirror image models of equipment health state evaluation.
When step 6c. learns to above-mentioned regression model, for given training set T, T={ (x1,y1),(x2, y2),...,(xN,yN), wherein training input sample xi∈Rn, RnRepresent n dimension sets of real numbers, training output sample yi∈ { 0,1 }, N Represent sample total number in training set T, i is specimen number, i=1,2,3 ..., n.
Using Maximum Likelihood Estimation Method estimation model parameter W={ w1,w2,...,wn, so as to obtain Logistic Regression mirror image models.
If P (Y=1 | X)=π (x), P (Y=1 | X)=π (x), then its likelihood function is
Log-likelihood function is
To L (w) maximizings, w estimation is obtained.
Assuming that w Maximum-likelihood estimation isThe Logistic Regression mirror image models for so learning to obtain are
Because the state parameter variable X of equipment is the function of time, therefore health degree is also what is changed over time, thus Realize the quantization to equipping health status.
Step 6d. Y={ 1,2,3 ..., K } in the case of training sample Y value set is more classified variables, K is Classification number, then Logistic Regression mirror image models are
X(x1,x2,...,xi,...xn) state parameter variable, W={ w1,w2,...,wnIt is model parameter.
Multimode label, i.e., the cluster health status modeling of more classified variables and assessment, its method for parameter estimation and binomial Logistic Regression method of estimation is similar.
(7) after the health status value equipped under different operating modes is calculated based on above-mentioned model, line number is entered to state of health data The health status degenerative process of each equipment is obtained according to fusion restructuring, specific steps include as follows：
Step 7a. extracts the health status value of same equipment and corresponding temporal information under different operating mode clusters.
Step 7b. is ranked up to the equipment health status value of each model according to time order and function.
Degenerative process equations of the step 7c. using the health status that polynomial regression calculating is each equipped with the time.
Step 7d. is obtained equipping health status decline curve, equipped using the decline curve according to polynomial equation Predicting residual useful life, the health state evaluation of the equipment cluster of industrial big data is realized with this.
The inventive method becomes by carrying out health state evaluation to equipment cluster, and with fitting of a polynomial equipment health status Change trend, equipment performance decline curve is obtained, residual life of equipment efficiently predicted with this.
With reference to emulation and experimental data, checking explanation is carried out to the technique effect of the present invention
Embodiment 7：
Equipment cluster health state evaluation method based on industrial big data is formed with embodiment 1-6, Binding experiment data The technique effect of the inventive method is described as follows：
The present embodiment uses certain company caused data set during the subway shield tunnel construction of various regions, and the data set includes 26 Row variable (the wherein the 3rd~5 row are floor datas, and 6~26 be state parameter data), data are all from each sensor.Tie below Accompanying drawing is closed to be explained in detail the application principle of the present invention.
Step 7.1 entity state Slice management
Because each Cycle of initial data is the state more new node that significant changes occur for engine health status, because This eliminates the process of entity state Slice management during processing and analysis, and directly data can be carried out with related pre- place Reason and feature extraction can establish the mirror image model of engine condition.
Step 7.2 engine cluster operating mode similitude automatic cluster
Floor data is clustered according to the above-mentioned clustering method based on minimax distance, obtains cluster result such as Shown in Fig. 7.After operating mode automatic cluster, operating mode is divided into six classes, as shown in table 1 per accounting of the class sample number in totality.It is poly- Class result shares six classes, will cluster operating mode mark of the acquired results according to operating mode 1, operating mode 2, operating mode 3, operating mode 4, operating mode 5 and operating mode 6 Label model respectively to the engine cluster in corresponding operating mode respectively.
The Different categories of samples distribution situation of table 3 counts
Classification Operating mode 1 Operating mode 2 Operating mode 3 Operating mode 4 Operating mode 5 Operating mode 6
Number of samples 6882 11571 6954 6881 6771 6859
Number of samples accounting 14.99% 25.20% 15.14% 14.99% 14.75% 14.94%
Step 7.3 determines engine health degree H
Sensor magnitudes trend under different operating modes is contrasted to obtain as drawn a conclusion：
(1) four classes as shown in Figure 8 be can be divided mainly into and become for any operating mode, the change shape of sensing data amplitude Gesture；
(2) under every kind of operating mode, in early stage in engine life stage (Cycle ＜ 50) sensing data highly dense, Cheng Lian Continuous shape distribution, in lifetime stage latter stage (Cycle ＞ 300), there is significant discretization trend in sensing data point；
(3) if assuming in Cycle ＜ 5, engine is in 1 state of health status, during Cycle ＞ 300 at engine In 0 state of health status, then the time intrinsic motivation performance between 0 ＜ Cycle ＜ 300 is in catagen phase.
No. 2, No. 3, No. 4, No. 7, No. 11, No. 12, No. 15, No. 20 and No. 21 sensing datas are chosen according to above-mentioned conclusion to make To characterize the healthy variable of engine health status.As Cycle ＜ 50, health degree H is 1；As Cycle ＞ 300, health degree H is 0.The sensing data with both labels under different operating modes is extracted to be modeled as shown in 2~table of table 7.
The modeling data of the operating mode 1 of table 4 extraction
The modeling data of the operating mode 2 of table 5 extraction
The modeling data of the operating mode 3 of table 6 extraction
The modeling data of the operating mode 4 of table 7 extraction
The modeling data of the operating mode 5 of table 8 extraction
The modeling data of the operating mode 6 of table 9 extraction
The healthy quantitative model modeling of step 7.4
Establish health status Logistic Regression mirror image model of the engine under operating mode n
The model is exactly health status quantitative model, W={ w1,w2,...,wnIt is to require parameter.
The modeling data obtained using step 7.3 is calculated, and it is α=0.05 to take confidence level, is tied as shown in table 10 Fruit, its every evaluation index parameter are as shown in table 11.
Mirror image model parameter list under 10 different operating modes of table
Model parameter Operating mode 1 Operating mode 2 Operating mode 3 Operating mode 4 Operating mode 5 Operating mode 6
w0 37.464 124.547 10.235 80.790 -19.654 94.930
w1 -0.078 -0.133 -0.055 -0.113 -0.056 -0.092
w2 -0.006 -0.009 -0.006 -0.013 0 -0.012
w3 -0.014 -0.015 -0.008 -0.014 -0.012 -0.017
w4 -0.012 -0.017 0.030 0.029 0.047 0.062
w5 -0.134 -0.323 -0.085 -0.240 -0.046 -0.143
w6 0.110 -0.019 0.046 0.103 0.160 -0.069
w7 -0.002 -0.274 0.026 -0.392 -0.133 -0.154
w8 0.165 0.017 0.129 0.273 0.098 0.132
w9 0.234 0.204 0.250 -0.060 0.405 0.024
The different operating mode drag evaluation indexes of table 11
Step 7.5 data fusion recombinates
Data shown in table 12 are the state of health data of each engine, and fusion restructuring is carried out to it and is obtained 1 shown in table 13 Number engine state of health data.Using polynomial regression, data shown in table 13 are entered with row interpolation recurrence, obtains different engines Health status change with time trend, be called and do health status degenerated curve.Fig. 9 is must wait until according to the method described above 1 The health status degenerated curve of number engine.The residual life of equipment is predicted using health status degenerated curve.
Health value table under 12 No. 1 each operating modes of engine of table
No. 1 engine health status table after the fusion restructuring of table 13
Engine model Time/Cycle Health status H (t)
1 1 1.164407
1 2 1.206333
1 3 0.980242
1 4 1.003762
… … …
1 219 0.504856
1 220 0.190922
1 222 0.260011
1 223 0.25399
The inventive method is equipped based on magnanimity cluster, and the health status that can be correctly equipped using cluster modeling method is commented The residual life of valency and Accurate Prediction equipment is for effectively avoiding shutting down stopping production accident, guarantee equipment safety operation, ensuring just Chang Youxu production and the use value of raising equipment, are that enterprise safety operation and raising enterprise are reasonable, are efficiently carried using equipment For important leverage.
In summary, a kind of equipment cluster health state evaluation based on industrial big data cluster modeling disclosed by the invention Method, the big data environment in cycle of being on active service entirely based on equipment cluster, first with entity status information Slice management method pair Data are pre-processed；Then equipment cluster is divided with operating mode Similarity-Based Clustering Method, forms different cluster knots Fruit cluster；Secondly the mirror image model mutually mapped with entity health status is established with Logistic Regression homing methods, Obtain the health status quantitative model of different equipment clusters；Finally the equipment health status of different clusters is carried out data fusion with Restructuring, obtains the health status value of different equipments, and uses polynomial interpolation, and equipment health status trend is fitted, Obtain equipping health status degenerated curve, for predicting equipment residual life.This analysis method is covered from data prediction, number According to a whole set of flow of modeling to cluster analysis, method mainly includes three core procedures：Entity status information Slice management, Physical system mirror image model and cluster state quantitative model are established, generation equipment residual life curve, is respectively used to solve equipment The problem of in terms of data prediction, data modeling and cluster analysis three in cluster big data cluster modeling.
The present invention has used big data cluster modeling method, can not only carry out effective difference to equipment cluster health status Property assess, and cluster modeling can reduce model redundancy, simplified model internal structure.
The foregoing is merely illustrative of the preferred embodiments of the present invention, is not intended to limit the invention, all essences in the present invention All any modification, equivalent and improvement made within refreshing and principle etc., should be included in the scope of the protection.

Claims (5)

  1. A kind of 1. equipment cluster health state evaluation method based on big data cluster modeling, it is characterised in that cluster health shape State appraisal procedure is on active service and big data environment is equipped caused by the cycle entirely based on equipment cluster, including the running state parameter of equipment, The floor data of equipment operation, the ambient parameter during equipment use, plant maintenance and maintenance record and performance class data, with Upper data are from equipment sensor, manual record and subsequent statistical；Equipment cluster health state evaluation includes having the following steps：
    Step 1：Storage and analysis that big data platform is used for magnanimity equipment data are built, the big data platform includes Hadoop/ Spark platforms and Storm platforms, Hadoop/Spark platforms are used for storage organization and semi-structured data, and Storm platforms are used In storage dynamic mass data；
    Step 2：Data prediction is carried out to equipment big data with entity state Slice management method, according to equipping shape The horizontal degree of correlation of state is classified to equipment big data, and data are broadly divided into two classes：The first kind is fault data and maintenance Record data is maintained, the second class is state parameter data；According to inhomogeneity data, different entity state Slice management is selected Method, for primary sources, directly the reference node as equipment state renewal；For secondary sources, to equipment The continuous multivariable service data of cluster carries out the monitoring of abnormality, is extracted and equipped using duty parameter subsequence extracting method There is the state feature of significant changes, in this, as the change node of state renewal in cluster running；
    Step 3：Based on Time series analysis method, operating mode automatic cluster is carried out to the operational factor of equipment, it is similar according to operating mode Property least unit as cluster of the cluster that clusters to obtain, be called equipment collection, equipment collection is internal use identical modeling method, conveniently Carry out united analysis；
    Step 4：Using regression modeling method, each equipment cluster state table is established to different operating mode clusters according to historical data The mirror image model reached, carry out equipping the quantization of cluster health status, obtain each equipping cluster health status；
    Step 5：Carry out data fusion to recombinate to obtain the health status of single equipment, and obtain it using polynomial interpolation and be good for Health state decays curve, equipment predicting residual useful life is carried out using the decline curve, provides prediction result.
  2. 2. as claimed in claim 1 based on industrial big data equipment cluster health state evaluation method, it is characterised in that step Utilization entity state Slice management described in two carries out data prediction to data, specifically comprises the following steps：
    Step 2.1 is classified according to the degree of correlation horizontal with equipment state to equipment big data, is determined according to data type The reference node of equipment state renewal；
    Step 2.2 is according to obtained reference node, the discrete shape history data set for equipping cluster being divided under different operating modes State sequence；
    Step 2.3 is that obtained every section of discrete status switch adds corresponding state tag, and with the operation at corresponding time point Data and floor data are associated storage；
    Step 2.4 is when the operating mode for equipping cluster changes, the last state data of automatic record equipment, and sticks new shape State label is stored, and obtains the training set sample set of mirror image model.
  3. 3. as claimed in claim 1 based on industrial big data equipment cluster health state evaluation method, it is characterised in that step Operating mode automatic clustering method described in three specifically comprises the following steps：
    Step 3.1 carries out operating mode automatic cluster to the operational parameter data of equipment, coefficient of determination θ given first, 0 ＜ θ ＜ 1, gives It is fixed to judge distance DC, and appoint and take a sample as first cluster centre Z1, such as take Z1=x1；
    Step 3.2 finds new cluster centre, until finding all cluster centres of the sample：
    Step 3.2a calculates other all sample x on the basis of first cluster centre1To Z1Distance Di1, i=1,2, 3 ..., n, distance calculation formula are
    <mrow> <msub> <mi>D</mi> <mrow> <mi>i</mi> <mn>1</mn> </mrow> </msub> <mo>=</mo> <mo>|</mo> <mo>|</mo> <msub> <mi>x</mi> <mi>i</mi> </msub> <mo>-</mo> <msub> <mi>Z</mi> <mn>1</mn> </msub> <mo>|</mo> <mo>|</mo> <mo>=</mo> <msqrt> <mrow> <munderover> <mo>&amp;Sigma;</mo> <mrow> <mi>i</mi> <mo>=</mo> <mn>1</mn> </mrow> <mi>d</mi> </munderover> <msup> <mrow> <mo>|</mo> <mrow> <msub> <mi>x</mi> <mi>i</mi> </msub> <mo>-</mo> <msub> <mi>Z</mi> <mn>1</mn> </msub> </mrow> <mo>|</mo> </mrow> <mn>2</mn> </msup> </mrow> </msqrt> <mo>,</mo> <msub> <mi>D</mi> <mrow> <mi>i</mi> <mn>2</mn> </mrow> </msub> <mo>=</mo> <mo>|</mo> <mo>|</mo> <msub> <mi>x</mi> <mi>i</mi> </msub> <mo>-</mo> <msub> <mi>Z</mi> <mn>2</mn> </msub> <mo>|</mo> <mo>|</mo> </mrow>
    If step 3.2b Dk1=max { Di1, i=1, wherein 2,3 ..., n, k are { Di1In the maximum element of distance value subscript Number, take xkAs second cluster centre Z2；
    Step 3.2c calculates all samples to cluster centre Z2Distance Di2, i=1,2,3 ..., n；
    If step 3.2d Di=max { min (Di1,Di2), i=1, wherein 2 ..., n, DiFor distance Di1With distance Di2In each away from From the maximum in minimum value, and Di＞ θ D12, D12It is Z1And Z2Between distance, then take xiFor the 3rd cluster centre Z3；
    If step 3.2e Z3In the presence of if Dj=max { min (Di1,Di2,Di3), i=1, wherein 2 ..., n, DjFor distance Di1、 Distance Di2With distance Di3In each maximum in minimum value, and Dj＞ θ D12, then the 4th polymerization site Z is established4, By that analogy, until minimax distance is not more than θ D12When, terminate to find the calculating of polymerization site, it is more to obtain all equipments The cluster centre of duty parameter data；
    Step 3.3 is according to closest principle all sample xi, i=1,2,3 ..., n, belong in closest cluster The heart, obtain cluster result cluster；
    If step 3.4 cluster result intra-cluster distance d, which is less than, gives set a distance Dc, terminate iterative process, obtain equipping multi-state parameter number According to automatic cluster result cluster；If cluster result intra-cluster distance d, which is not less than, gives set a distance Dc, then gravity treatment θ and first polymerization site Z1, return to step 3.2 and find new cluster center again, carry out next round iteration.
  4. 4. as claimed in claim 1 based on industrial big data equipment cluster health state evaluation method, it is characterised in that step Four specific steps that equipment cluster health status quantization mirror image model is obtained with regression modeling method include：
    If health status of the step 4.1 under specific operation has a continuous response variable H at different moments, the work is represented The quantizating index of the health status of cluster is equipped under condition, its codomain is [0,1], and H is strong under the operating mode as equipment cluster Kang Du, logically return and establish quantitative model, wherein health degree H is exactly that to equip the health status of cluster be the general of normal condition Rate,
    H=P (Y=1 | X)
    So equip cluster unhealthy probability be
    <mrow> <mover> <mi>H</mi> <mo>&amp;OverBar;</mo> </mover> <mo>=</mo> <mn>1</mn> <mo>-</mo> <mi>H</mi> <mo>=</mo> <mi>P</mi> <mrow> <mo>(</mo> <mi>Y</mi> <mo>=</mo> <mn>0</mn> <mo>|</mo> <mi>X</mi> <mo>)</mo> </mrow> </mrow>
    Wherein, Y represents equipment state, and wherein Y=0 represents that equipment is in malfunction, and Y=1 represents equipment in health completely State, X (x1,x2,...,xi,...xn) it is state parameter variable,
    Then definition equipment health status is that the ratio of complete health and failure is
    <mrow> <mi>o</mi> <mi>d</mi> <mi>d</mi> <mo>=</mo> <mfrac> <mrow> <mi>P</mi> <mrow> <mo>(</mo> <mi>Y</mi> <mo>=</mo> <mn>1</mn> <mo>|</mo> <mi>X</mi> <mo>)</mo> </mrow> </mrow> <mrow> <mn>1</mn> <mo>-</mo> <mi>P</mi> <mrow> <mo>(</mo> <mi>Y</mi> <mo>=</mo> <mn>1</mn> <mo>|</mo> <mi>X</mi> <mo>)</mo> </mrow> </mrow> </mfrac> <mo>=</mo> <mfrac> <mi>H</mi> <mrow> <mn>1</mn> <mo>-</mo> <mi>H</mi> </mrow> </mfrac> </mrow>
    Odd is referred to as ratio occurs；
    Step 4.2 assumes the state parameter variable X (x of logarithm ratio and reflection equipment1,x2,...,xi,...xn) between exist A kind of linear relationship, i.e.,
    <mrow> <mi>l</mi> <mi>o</mi> <mi>g</mi> <mrow> <mo>(</mo> <mi>o</mi> <mi>d</mi> <mi>d</mi> <mo>)</mo> </mrow> <mo>=</mo> <mi>l</mi> <mi>o</mi> <mi>g</mi> <mrow> <mo>(</mo> <mfrac> <mrow> <mi>P</mi> <mrow> <mo>(</mo> <mi>Y</mi> <mo>=</mo> <mn>1</mn> <mo>|</mo> <mi>X</mi> <mo>)</mo> </mrow> </mrow> <mrow> <mn>1</mn> <mo>-</mo> <mi>P</mi> <mrow> <mo>(</mo> <mi>Y</mi> <mo>=</mo> <mn>1</mn> <mo>|</mo> <mi>X</mi> <mo>)</mo> </mrow> </mrow> </mfrac> <mo>)</mo> </mrow> <mo>=</mo> <mi>l</mi> <mi>o</mi> <mi>g</mi> <mrow> <mo>(</mo> <mfrac> <mi>H</mi> <mrow> <mn>1</mn> <mo>-</mo> <mi>H</mi> </mrow> </mfrac> <mo>)</mo> </mrow> <mo>=</mo> <msup> <mi>W</mi> <mi>T</mi> </msup> <mi>X</mi> </mrow>
    Wherein W={ w1,w2,...,wnIt is model parameter, n is state parameter variable X (x1,x2,...,xi,...xn) number, Above formula both sides fetching number is i.e. available
    <mrow> <mfrac> <mrow> <mi>P</mi> <mrow> <mo>(</mo> <mi>Y</mi> <mo>=</mo> <mn>1</mn> <mo>|</mo> <mi>X</mi> <mo>)</mo> </mrow> </mrow> <mrow> <mn>1</mn> <mo>-</mo> <mi>P</mi> <mrow> <mo>(</mo> <mi>Y</mi> <mo>=</mo> <mn>1</mn> <mo>|</mo> <mi>X</mi> <mo>)</mo> </mrow> </mrow> </mfrac> <mo>=</mo> <mfrac> <mi>H</mi> <mrow> <mn>1</mn> <mo>-</mo> <mi>H</mi> </mrow> </mfrac> <mo>=</mo> <msup> <mi>e</mi> <mrow> <msup> <mi>W</mi> <mi>T</mi> </msup> <mi>X</mi> </mrow> </msup> </mrow>
    Further solve the assessment models for producing equipment health degree
    <mrow> <mi>H</mi> <mo>=</mo> <mfrac> <msup> <mi>e</mi> <mrow> <msup> <mi>W</mi> <mi>T</mi> </msup> <mi>X</mi> </mrow> </msup> <mrow> <mn>1</mn> <mo>+</mo> <msup> <mi>e</mi> <mrow> <msup> <mi>W</mi> <mi>T</mi> </msup> <mi>X</mi> </mrow> </msup> </mrow> </mfrac> </mrow>
    Thus establish the Logistic Regression mirror image models of equipment health state evaluation.
  5. 5. as claimed in claim 1 based on industrial big data equipment cluster health state evaluation method, it is characterised in that step Five described equipped to different operating mode clusters carry out data fusion restructuring, obtain the health status value of single equipment, and can utilize Polynomial interpolation obtains its health status decline curve, and equipment predicting residual useful life is carried out using the decline curve, specific step It is rapid as follows：
    Step 5.1 extracts the health status value of same equipment and corresponding temporal information under different operating mode clusters；
    Step 5.2 is ranked up to the equipment health status value of each model according to time order and function；
    Degenerative process equation of the step 5.3 using the health status that polynomial interpolation calculating is each equipped with the time；
    Step 5.4 obtains equipping health status decline curve, it is remaining to carry out equipment using the decline curve according to polynomial equation Life prediction.
CN201710543984.1A 2017-07-05 2017-07-05 Equipment cluster health state evaluation method based on industrial big data Pending CN107358347A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201710543984.1A CN107358347A (en) 2017-07-05 2017-07-05 Equipment cluster health state evaluation method based on industrial big data

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201710543984.1A CN107358347A (en) 2017-07-05 2017-07-05 Equipment cluster health state evaluation method based on industrial big data

Publications (1)

Publication Number Publication Date
CN107358347A true CN107358347A (en) 2017-11-17

Family

ID=60293127

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201710543984.1A Pending CN107358347A (en) 2017-07-05 2017-07-05 Equipment cluster health state evaluation method based on industrial big data

Country Status (1)

Country Link
CN (1) CN107358347A (en)

Cited By (36)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN108035838A (en) * 2017-12-07 2018-05-15 武汉四创自动控制技术有限责任公司 Full factory's Turbine Governor System health state evaluation and optimization method
CN108197647A (en) * 2017-12-28 2018-06-22 中南大学 A kind of Fast Speed Clustering of automobile starter durable test data
CN108491965A (en) * 2018-03-14 2018-09-04 广东省智能制造研究所 Trend prediction method, device, electronic equipment and the storage medium of stamping equipment
CN108520332A (en) * 2018-03-12 2018-09-11 北京天润新能投资有限公司 A kind of wind farm device asset management system and method
CN108804710A (en) * 2018-06-25 2018-11-13 浪潮软件集团有限公司 Method and device for refining label through model tool based on business rule
CN108874640A (en) * 2018-05-07 2018-11-23 北京京东尚科信息技术有限公司 A kind of appraisal procedure and device of clustering performance
CN108898254A (en) * 2018-07-03 2018-11-27 柳州市木子科技有限公司 A kind of production line spare part spoilage forecasting system based on LR
CN108959498A (en) * 2018-06-26 2018-12-07 北京航空航天大学 A kind of big data processing platform and its design method for health monitoring
CN109253826A (en) * 2018-08-01 2019-01-22 西安交通大学 A kind of calorimeter method for predicting residual useful life based on the fusion of more degeneration sample datas
CN109359100A (en) * 2018-09-28 2019-02-19 北京赛博贝斯数据科技有限责任公司 The visual modeling method of PB grades of historical datas and online data calculated in real time
CN109711664A (en) * 2018-11-16 2019-05-03 国网江苏省电力有限公司盐城供电分公司 A kind of power transmission and transforming equipment health evaluation system based on big data
CN109740809A (en) * 2018-12-27 2019-05-10 北京五维星宇科技有限公司 The equipment cluster method for predicting residual useful life and terminal to be developed based on single dress service life
CN110146317A (en) * 2019-04-29 2019-08-20 北京和欣运达科技有限公司 A kind of diagnostic method and device of building electromechanical equipment health status
WO2019174142A1 (en) * 2018-03-14 2019-09-19 山东科技大学 Multi-mode degradation process modelling and remaining service life prediction method
CN110647122A (en) * 2019-09-30 2020-01-03 佛山科学技术学院 Automatic production line fault prediction method and device based on multiple sensors
CN110750526A (en) * 2019-10-17 2020-02-04 北京五维星宇科技有限公司 Method and apparatus for generating single package use plan, and storage medium
CN111160616A (en) * 2019-12-05 2020-05-15 广东工业大学 Kitchen electrical equipment predictive maintenance system and method based on edge cloud cooperation
CN111680806A (en) * 2020-06-11 2020-09-18 河南财政金融学院 Building equipment management system
CN112084166A (en) * 2019-06-13 2020-12-15 上海杰之能软件科技有限公司 Sample data establishment method, data model training method, device and terminal
CN112258689A (en) * 2020-10-26 2021-01-22 上海船舶研究设计院（中国船舶工业集团公司第六0四研究院） Ship data processing method and device and ship data quality management platform
CN112613646A (en) * 2020-12-08 2021-04-06 上海交通大学烟台信息技术研究院 Equipment state prediction method and system based on multi-dimensional data fusion
CN112633617A (en) * 2019-09-24 2021-04-09 北京国双科技有限公司 Maintenance strategy generation method and device, computer equipment and readable storage medium
CN112651605A (en) * 2020-12-10 2021-04-13 河钢乐亭钢铁有限公司 Equipment monitoring and state analysis system
CN112764788A (en) * 2021-01-19 2021-05-07 南京大学 Software dynamic update hot patch synthesis method based on program source code slice recombination
CN113168544A (en) * 2018-12-19 2021-07-23 西门子股份公司 Method and system for providing services for complex industrial systems
CN113255795A (en) * 2021-06-02 2021-08-13 杭州安脉盛智能技术有限公司 Equipment state monitoring method based on multi-index cluster analysis
CN113357138A (en) * 2021-07-21 2021-09-07 中国人民解放军陆军工程大学 Method and device for predicting remaining service life of hydraulic pump and terminal equipment
CN113438110A (en) * 2021-06-23 2021-09-24 曙光信息产业(北京)有限公司 Cluster performance evaluation method, device, equipment and storage medium
CN113673532A (en) * 2021-10-21 2021-11-19 北京科技大学 Optimization method and device of quantization model
CN113743461A (en) * 2021-07-29 2021-12-03 中国人民解放军军事科学院国防科技创新研究院 Unmanned aerial vehicle cluster health degree assessment method and device
CN114298616A (en) * 2022-03-11 2022-04-08 西南石油大学 Equipment health state evaluation method and device and computer equipment
CN114491052A (en) * 2020-10-27 2022-05-13 中移物联网有限公司 System health degree determination method, device, equipment and readable storage medium
CN115169650A (en) * 2022-06-20 2022-10-11 四川观想科技股份有限公司 Equipment health prediction method for big data analysis
CN115774652A (en) * 2023-02-13 2023-03-10 浪潮通用软件有限公司 Cluster control equipment health monitoring method, equipment and medium based on clustering algorithm
CN115795922A (en) * 2023-02-06 2023-03-14 浪潮通用软件有限公司 Equipment operation management method, system and medium based on equipment health degree
CN116881673A (en) * 2023-09-06 2023-10-13 山东济矿鲁能煤电股份有限公司阳城煤矿 Shield tunneling machine operation and maintenance method based on big data analysis

Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN102682180A (en) * 2011-03-07 2012-09-19 上海宝信软件股份有限公司 Evaluation method for performance degradation of rotary mechanical equipment
CN104820716A (en) * 2015-05-21 2015-08-05 中国人民解放军海军工程大学 Equipment reliability evaluation method based on data mining
CN106651188A (en) * 2016-12-27 2017-05-10 贵州电网有限责任公司贵阳供电局 Electric transmission and transformation device multi-source state assessment data processing method and application thereof

Patent Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN102682180A (en) * 2011-03-07 2012-09-19 上海宝信软件股份有限公司 Evaluation method for performance degradation of rotary mechanical equipment
CN104820716A (en) * 2015-05-21 2015-08-05 中国人民解放军海军工程大学 Equipment reliability evaluation method based on data mining
CN106651188A (en) * 2016-12-27 2017-05-10 贵州电网有限责任公司贵阳供电局 Electric transmission and transformation device multi-source state assessment data processing method and application thereof

Non-Patent Citations (4)

* Cited by examiner, † Cited by third party
Title
DAVID SIEGEL等: "Wind turbine performance assessment using multi-regime modeling approach", 《RENEWABLE ENERGY》 *
KONG XIANGUANG等: "A novel approach for product makespan prediction in production life cycle", 《INTERNATIONAL JOURNAL OF ADVANCED MANUFACTURING TECHNOLOGY》 *
孔宪光等: "工业大数据环境下的混合故障诊断模型研究", 《2015年全国机械行业可靠性技术学术交流会暨第五届可靠性工程分会第二次全体委员大会论文集》 *
施浩等: "基于多工况自动聚类的飞机发动机健康状态评估", 《机电一体化》 *

Cited By (51)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN108035838B (en) * 2017-12-07 2020-11-03 武汉四创自动控制技术有限责任公司 Health state evaluation and optimization method for speed regulation system of water turbine of whole plant
CN108035838A (en) * 2017-12-07 2018-05-15 武汉四创自动控制技术有限责任公司 Full factory's Turbine Governor System health state evaluation and optimization method
CN108197647B (en) * 2017-12-28 2021-08-31 中南大学 Rapid clustering method for automobile starter endurance test data
CN108197647A (en) * 2017-12-28 2018-06-22 中南大学 A kind of Fast Speed Clustering of automobile starter durable test data
CN108520332A (en) * 2018-03-12 2018-09-11 北京天润新能投资有限公司 A kind of wind farm device asset management system and method
CN108491965A (en) * 2018-03-14 2018-09-04 广东省智能制造研究所 Trend prediction method, device, electronic equipment and the storage medium of stamping equipment
WO2019174142A1 (en) * 2018-03-14 2019-09-19 山东科技大学 Multi-mode degradation process modelling and remaining service life prediction method
CN108874640A (en) * 2018-05-07 2018-11-23 北京京东尚科信息技术有限公司 A kind of appraisal procedure and device of clustering performance
CN108804710A (en) * 2018-06-25 2018-11-13 浪潮软件集团有限公司 Method and device for refining label through model tool based on business rule
CN108959498A (en) * 2018-06-26 2018-12-07 北京航空航天大学 A kind of big data processing platform and its design method for health monitoring
CN108898254A (en) * 2018-07-03 2018-11-27 柳州市木子科技有限公司 A kind of production line spare part spoilage forecasting system based on LR
CN109253826A (en) * 2018-08-01 2019-01-22 西安交通大学 A kind of calorimeter method for predicting residual useful life based on the fusion of more degeneration sample datas
CN109359100A (en) * 2018-09-28 2019-02-19 北京赛博贝斯数据科技有限责任公司 The visual modeling method of PB grades of historical datas and online data calculated in real time
CN109711664A (en) * 2018-11-16 2019-05-03 国网江苏省电力有限公司盐城供电分公司 A kind of power transmission and transforming equipment health evaluation system based on big data
CN109711664B (en) * 2018-11-16 2022-08-19 国网江苏省电力有限公司盐城供电分公司 Power transmission and transformation equipment health assessment system based on big data
CN113168544A (en) * 2018-12-19 2021-07-23 西门子股份公司 Method and system for providing services for complex industrial systems
CN109740809B (en) * 2018-12-27 2021-03-23 北京五维星宇科技有限公司 Equipment cluster residual life prediction method and terminal based on single-mounted life evolution
CN109740809A (en) * 2018-12-27 2019-05-10 北京五维星宇科技有限公司 The equipment cluster method for predicting residual useful life and terminal to be developed based on single dress service life
CN110146317A (en) * 2019-04-29 2019-08-20 北京和欣运达科技有限公司 A kind of diagnostic method and device of building electromechanical equipment health status
CN110146317B (en) * 2019-04-29 2021-11-12 北京和欣运达科技有限公司 Method and device for diagnosing health state of building electromechanical equipment
CN112084166A (en) * 2019-06-13 2020-12-15 上海杰之能软件科技有限公司 Sample data establishment method, data model training method, device and terminal
CN112633617A (en) * 2019-09-24 2021-04-09 北京国双科技有限公司 Maintenance strategy generation method and device, computer equipment and readable storage medium
CN110647122A (en) * 2019-09-30 2020-01-03 佛山科学技术学院 Automatic production line fault prediction method and device based on multiple sensors
CN110647122B (en) * 2019-09-30 2021-11-09 佛山科学技术学院 Automatic production line fault prediction method and device based on multiple sensors
CN110750526A (en) * 2019-10-17 2020-02-04 北京五维星宇科技有限公司 Method and apparatus for generating single package use plan, and storage medium
CN111160616B (en) * 2019-12-05 2021-09-07 广东工业大学 Kitchen electrical equipment predictive maintenance system and method based on edge cloud cooperation
CN111160616A (en) * 2019-12-05 2020-05-15 广东工业大学 Kitchen electrical equipment predictive maintenance system and method based on edge cloud cooperation
CN111680806A (en) * 2020-06-11 2020-09-18 河南财政金融学院 Building equipment management system
CN112258689A (en) * 2020-10-26 2021-01-22 上海船舶研究设计院（中国船舶工业集团公司第六0四研究院） Ship data processing method and device and ship data quality management platform
CN114491052A (en) * 2020-10-27 2022-05-13 中移物联网有限公司 System health degree determination method, device, equipment and readable storage medium
CN112613646A (en) * 2020-12-08 2021-04-06 上海交通大学烟台信息技术研究院 Equipment state prediction method and system based on multi-dimensional data fusion
CN112651605A (en) * 2020-12-10 2021-04-13 河钢乐亭钢铁有限公司 Equipment monitoring and state analysis system
CN112764788B (en) * 2021-01-19 2023-06-16 南京大学 Software dynamic update hot patch synthesis method based on program source code slice recombination
CN112764788A (en) * 2021-01-19 2021-05-07 南京大学 Software dynamic update hot patch synthesis method based on program source code slice recombination
CN113255795A (en) * 2021-06-02 2021-08-13 杭州安脉盛智能技术有限公司 Equipment state monitoring method based on multi-index cluster analysis
CN113438110A (en) * 2021-06-23 2021-09-24 曙光信息产业(北京)有限公司 Cluster performance evaluation method, device, equipment and storage medium
CN113357138A (en) * 2021-07-21 2021-09-07 中国人民解放军陆军工程大学 Method and device for predicting remaining service life of hydraulic pump and terminal equipment
CN113743461A (en) * 2021-07-29 2021-12-03 中国人民解放军军事科学院国防科技创新研究院 Unmanned aerial vehicle cluster health degree assessment method and device
CN113743461B (en) * 2021-07-29 2024-04-09 中国人民解放军军事科学院国防科技创新研究院 Unmanned aerial vehicle cluster health degree assessment method and device
CN113673532A (en) * 2021-10-21 2021-11-19 北京科技大学 Optimization method and device of quantization model
CN114298616A (en) * 2022-03-11 2022-04-08 西南石油大学 Equipment health state evaluation method and device and computer equipment
CN114298616B (en) * 2022-03-11 2022-05-06 西南石油大学 Equipment health state evaluation method and device and computer equipment
CN115169650A (en) * 2022-06-20 2022-10-11 四川观想科技股份有限公司 Equipment health prediction method for big data analysis
CN115169650B (en) * 2022-06-20 2022-11-22 四川观想科技股份有限公司 Equipment health prediction method for big data analysis
CN115795922A (en) * 2023-02-06 2023-03-14 浪潮通用软件有限公司 Equipment operation management method, system and medium based on equipment health degree
CN115795922B (en) * 2023-02-06 2023-04-25 浪潮通用软件有限公司 Equipment operation management method, system and medium based on equipment health degree
US20240302833A1 (en) * 2023-02-06 2024-09-12 Inspur Genersoft Co., Ltd. Device operation management method, system, and medium based on device health degree
CN115774652A (en) * 2023-02-13 2023-03-10 浪潮通用软件有限公司 Cluster control equipment health monitoring method, equipment and medium based on clustering algorithm
WO2024169123A1 (en) * 2023-02-13 2024-08-22 浪潮通用软件有限公司 Clustering-algorithm-based cluster control device health monitoring method and device, and medium
CN116881673A (en) * 2023-09-06 2023-10-13 山东济矿鲁能煤电股份有限公司阳城煤矿 Shield tunneling machine operation and maintenance method based on big data analysis
CN116881673B (en) * 2023-09-06 2023-12-19 山东济矿鲁能煤电股份有限公司阳城煤矿 Shield tunneling machine operation and maintenance method based on big data analysis

Similar Documents

Publication Publication Date Title
CN107358347A (en) Equipment cluster health state evaluation method based on industrial big data
CN108320040B (en) Acquisition terminal fault prediction method and system based on Bayesian network optimization algorithm
CN107330056B (en) Wind power plant SCADA system based on big data cloud computing platform and operation method thereof
CN115578015A (en) Sewage treatment overall process supervision method and system based on Internet of things and storage medium
CN106651188A (en) Electric transmission and transformation device multi-source state assessment data processing method and application thereof
CN102411729B (en) Wind power prediction method based on adaptive linear logic network
CN105701596A (en) Method for lean distribution network emergency maintenance and management system based on big data technology
CN109412155B (en) Power distribution network power supply capacity evaluation method based on graph calculation
CN107832876A (en) Subregion peak load Forecasting Methodology based on MapReduce frameworks
CN110297207A (en) Method for diagnosing faults, system and the electronic device of intelligent electric meter
CN115822887A (en) Performance evaluation and energy efficiency diagnosis method and system of wind turbine generator
CN109471698A (en) System and method for detecting abnormal behavior of virtual machine in cloud environment
Momtazpour et al. Analyzing invariants in cyber-physical systems using latent factor regression
CN110033102A (en) A kind of huge hydroelectric power plant has the intelligent diagnosing method and expert system of learning functionality
CN117318033B (en) Power grid data management method and system combining data twinning
CN117372201A (en) Rapid construction method of intelligent water conservancy digital twin model applied to reservoir
Peng et al. Review of key technologies and progress in industrial equipment health management
CN117111544A (en) Automatic-adaptation building internet of things monitoring method and system
CN116227676A (en) Production equipment energy consumption analysis method and device, electronic equipment and storage medium
Yan et al. An Efficient Method for Optimizing Sensors’ Layout for Accurate Measurement of Underground Ventilation Networks
Fawzy et al. Trio-V wind analyzer: a generic integral system for wind farm suitability design and power prediction using big data analytics
CN118353011B (en) Method and system for detecting panoramic dynamic topology anomaly of power distribution network based on graph calculation
Feijoo-Martinez et al. Event analysis on power communication networks with big data for maintenance forms
CN117314240A (en) Method and system for evaluating toughness of park comprehensive energy system under extreme weather condition
CN115600695A (en) Fault diagnosis method of metering equipment

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination
WD01 Invention patent application deemed withdrawn after publication

Application publication date: 20171117

WD01 Invention patent application deemed withdrawn after publication