CN109656236A - A kind of industrial data failure prediction method based on cyclic forecast neural network - Google Patents

A kind of industrial data failure prediction method based on cyclic forecast neural network Download PDF

Info

Publication number
CN109656236A
CN109656236A CN201910078393.0A CN201910078393A CN109656236A CN 109656236 A CN109656236 A CN 109656236A CN 201910078393 A CN201910078393 A CN 201910078393A CN 109656236 A CN109656236 A CN 109656236A
Authority
CN
China
Prior art keywords
neural network
prediction
data
failure
network model
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
CN201910078393.0A
Other languages
Chinese (zh)
Inventor
耿祖琨
张卫山
任鹏程
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.)
China University of Petroleum East China
Original Assignee
China University of Petroleum East China
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 China University of Petroleum East China filed Critical China University of Petroleum East China
Priority to CN201910078393.0A priority Critical patent/CN109656236A/en
Publication of CN109656236A publication Critical patent/CN109656236A/en
Pending legal-status Critical Current

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
    • G05B23/0243—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 model based detection method, e.g. first-principles knowledge model
    • 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
    • G05B2219/00—Program-control systems
    • G05B2219/20—Pc systems
    • G05B2219/24—Pc safety
    • G05B2219/24065—Real time diagnostics

Landscapes

  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Engineering & Computer Science (AREA)
  • Automation & Control Theory (AREA)
  • Management, Administration, Business Operations System, And Electronic Commerce (AREA)
  • Testing And Monitoring For Control Systems (AREA)

Abstract

The invention proposes a kind of industrial data failure prediction methods based on cyclic forecast neural network, include the following steps: that (1) design pattern explores branch's ring, equipment running status mode is established, for storing and indicating the various operating statuses and sequential relationship of equipment.(2) design cycle property prediction neural network is used under cloud environment, according to the probability of the next state of the time series forecasting of current a cycle and its failure.(3) by the selection of dynamic node, data antithesis in operational process under cloud environment, iteration excavates nonevent new failure, data experiment mechanism is established, to realize failure predication under real-time cloud environment.A kind of industrial data failure prediction method based on cyclic forecast neural network, industrial data failure predication and neural network are combined, and are based on algorithm precision of prediction while cyclic forecast neural network model improves algorithm implementation rate by design.

Description

A kind of industrial data failure prediction method based on cyclic forecast neural network
Technical field
The present invention relates to neural network, failure predication, big data cloud computing environments, and in particular to a kind of based on periodically The industrial data failure prediction method of prediction neural network.
Background technique
As the extensive use of cloud computing technology is transmitted to cloud after industrial equipment operational management uses field data collection End carries out Data Analysis Services, this has become a general mode.One urgent problems is having for a large amount of historical datas How the real-time processing that effect excavates with dynamic operation data organically blends, to realize data intelligence, solves to go out in equipment operation Existing problem.Obviously, this needs a kind of industrial data failure prediction method based on cyclic forecast neural network, supports big rule The Intelligent treatment of mould time series data.
The difficult point of fault diagnosis and prediction is: in the extensive time series data being collected into actual scene, fault data is logical Chang Feichang is rare, and following failure may be the failure that there is no crossing in the past, it is difficult to realize effective accident analysis and prediction. Various scene restriction bring data it is incomplete be current big data treatment research a common problem.If only from reality Border data are analyzed, and the issuable all scenes of real system failure institute can not be theoretically verified, to also be difficult to send out It digs all modes that failure generates and effectively carries out failure predication.
This method can be more accurate using a kind of industrial data failure prediction method based on cyclic forecast neural network Failure predication.This method improves the speed of algorithm training by establishing periodical neural network prediction model；Worked as by basis The probability of the next state of the time series forecasting in previous period and its failure reduces prediction data exception error rate and makes Algorithm stability greatly promotes；
Summary of the invention
To solve shortcoming and defect in the prior art, the invention proposes a kind of based on cyclic forecast neural network Industrial data failure prediction method carries out industrial data failure predication by cyclic forecast neural network model, accurate to carry out Industrial equipment failure predication and promotion predictablity rate.
The technical solution of the present invention is as follows:
A. design pattern explores branch's ring, equipment running status mode is established, for storing and indicating the various fortune of equipment Row state and sequential relationship.
B. design cycle property prediction neural network, under cloud environment, the time series according to current a cycle to be pre- Survey next state and the probability of its failure.
C. by the selection of dynamic node, data antithesis in operational process under cloud environment, iteration is excavated nonevent new Failure establishes data experiment mechanism, to realize failure predication under real-time cloud environment.
Beneficial effects of the present invention:
(1) building is based on cyclic forecast neural network model, to improve data forecast quality；
(2) this method improves model training speed by the network structure of Neural memory model；
(3) by the network model of fusion mean square deviation and cross entropy design, prediction accuracy is improved.
Detailed description of the invention
In order to more clearly explain the embodiment of the invention or the technical proposal in the existing technology, to embodiment or will show below There is attached drawing needed in technical description to be briefly described, it should be apparent that, the accompanying drawings in the following description is only this Some embodiments of invention for those of ordinary skill in the art without creative efforts, can be with It obtains other drawings based on these drawings.
Fig. 1 is a kind of neuron mould of the industrial data failure prediction method based on cyclic forecast neural network of the present invention Type；
Specific embodiment
Following will be combined with the drawings in the embodiments of the present invention, and technical solution in the embodiment of the present invention carries out clear, complete Site preparation description, it is clear that described embodiments are only a part of the embodiments of the present invention, instead of all the embodiments.It is based on Embodiment in the present invention, it is obtained by those of ordinary skill in the art without making creative efforts every other Embodiment shall fall within the protection scope of the present invention.
As shown in Figure 1, in a kind of industrial data failure prediction method based on cyclic forecast neural network model tool Body design is described in detail:
Cyclic forecast neural network model is based primarily upon GRU, i.e. gate recursive unit.For the spy of periodic data Point redesigns cyclic forecast neural network model, it is preferred that emphasis is design neuron connection type and loss function.
Neural network output module is designed as two, an output next step state Vi+1, an output is for next step shape The assessment result P of statei+1。
Q (s_i)=(Vi+1,Pi+1) ⑿
The s_i of cyclic forecast neural network model input is the past data mode for being pushed forward a cycle of current data node Variation track, the V of outputi+1To select next back end and selecting the probability of malfunction P after iti+1。
The structure of cyclic forecast neural network model is as shown in Figure 1.Output valveFor laststate valueAnd candidate ValueWithThis update ratio is superimposed to obtain.Determine current input value and laststate input value in current candidate In shared scale.Calculation formula are as follows:
HereinIt is equivalent to received state value ht-1With current state value xtOne it is cumulative and, WzAnd UzIt is Corresponding weight.
The calculation formula of current candidate are as follows:
In this rtResetting door is represented, ⊙ represents multiselect state.Work as rtWhen equal to 0, all state values front input are whole Forget about there will not be any influence to the value newly inputted.rtCalculation formula are as follows:
Its loss function is designed as the function of comprehensive assessment expectation and cross entropy, ydFor prediction result,For expectation as a result, pdFor prediction probability,For expected probability, λ is empirical parameter, controls mean square deviation weight shared in the loss function:
α is L2 regular terms parameter, and w is the relevant parameter in neural network, and the solution of L2 regularization is all smaller, anti-interference kinetic energy Power is strong, is conducive to the performance for improving cyclic forecast neural network model.
A kind of industrial data failure prediction method based on cyclic forecast neural network of the invention, by nerve net Network, gate recursive unit and periodic data feature combine, and this method is able to achieve pre- according to the time series of current a cycle Next state and the probability of its failure are surveyed, the deficiency for industrial data use of information is avoided.Square by merging Difference and cross entropy effectively improve computational efficiency, while increasing the accuracy of prediction algorithm；
The foregoing is merely illustrative of the preferred embodiments of the present invention, is not intended to limit the invention, all in essence of the invention Within mind and principle, any modification, equivalent replacement, improvement and so on be should all be included in the protection scope of the present invention.

Claims (1)

1. a kind of industrial data failure prediction method based on cyclic forecast neural network, which is characterized in that using periodically Prediction neural network model method, and industrial mass data is combined, carry out equipment fault prediction, comprising the following steps:
Cyclic forecast neural network model is based primarily upon GRU, i.e. gate recursive unit.The characteristics of for periodic data, weight New design cycle property prediction neural network model, it is preferred that emphasis is design neuron connection type and loss function.
Neural network output module is designed as two, an output next step state Vi+1, an output is for next step state Assessment result Pi+1。
Q (s_i)=(Vi+1, Pi+1) (1)
The s_i of cyclic forecast neural network model input is that current data node changes toward the data mode for being pushed forward a cycle Track, the V of outputi+1To select next back end and selecting the probability of malfunction P after iti+1。
The structure of cyclic forecast neural network model is as shown in Figure 1.Output valveFor laststate valueAnd candidate value WithThis update ratio is superimposed to obtain.Determine current input value and laststate input value institute in current candidate The scale accounted for.Calculation formula are as follows:
HereinIt is equivalent to received state value ht-1With current state value xtOne it is cumulative and, WzAnd UzIt is corresponding Weight.
The calculation formula of current candidate are as follows:
In this rtResetting door is represented, ⊙ represents multiselect state.Work as rtWhen equal to 0, all state values that front inputs all are forgotten Fall, there will not be any influence to the value newly inputted.rtCalculation formula are as follows:
Its loss function is designed as the function of comprehensive assessment expectation and cross entropy, ydFor prediction result,It is expected as a result, pdIt is pre- Probability is surveyed,For expected probability, λ is empirical parameter, controls mean square deviation weight shared in the loss function:
α is L2 regular terms parameter, and w is the relevant parameter in neural network, and the solution of L2 regularization is all smaller, Ability of Resisting Disturbance By force, be conducive to improve the performance of cyclic forecast neural network model.
CN201910078393.0A 2019-01-28 2019-01-28 A kind of industrial data failure prediction method based on cyclic forecast neural network Pending CN109656236A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201910078393.0A CN109656236A (en) 2019-01-28 2019-01-28 A kind of industrial data failure prediction method based on cyclic forecast neural network

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201910078393.0A CN109656236A (en) 2019-01-28 2019-01-28 A kind of industrial data failure prediction method based on cyclic forecast neural network

Publications (1)

Publication Number Publication Date
CN109656236A true CN109656236A (en) 2019-04-19

Family

ID=66121510

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201910078393.0A Pending CN109656236A (en) 2019-01-28 2019-01-28 A kind of industrial data failure prediction method based on cyclic forecast neural network

Country Status (1)

Country Link
CN (1) CN109656236A (en)

Cited By (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN110335168A (en) * 2019-04-22 2019-10-15 山东大学 Method and system based on GRU optimization power information acquisition terminal fault prediction model
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
CN111221318A (en) * 2019-12-11 2020-06-02 中山大学 Multi-robot state estimation method based on model predictive control algorithm
CN116403605A (en) * 2023-06-08 2023-07-07 宁德时代新能源科技股份有限公司 Equipment fault prediction method, stacker fault prediction method and related devices

Cited By (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN110335168A (en) * 2019-04-22 2019-10-15 山东大学 Method and system based on GRU optimization power information acquisition terminal fault prediction model
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
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
CN111221318A (en) * 2019-12-11 2020-06-02 中山大学 Multi-robot state estimation method based on model predictive control algorithm
CN111221318B (en) * 2019-12-11 2023-03-28 中山大学 Multi-robot state estimation method based on model predictive control algorithm
CN116403605A (en) * 2023-06-08 2023-07-07 宁德时代新能源科技股份有限公司 Equipment fault prediction method, stacker fault prediction method and related devices

Similar Documents

Publication Publication Date Title
Groshev et al. Learning generalized reactive policies using deep neural networks
CN109656236A (en) A kind of industrial data failure prediction method based on cyclic forecast neural network
CN108448610B (en) Short-term wind power prediction method based on deep learning
CN104951836A (en) Posting predication system based on nerual network technique
CN111709754B (en) User behavior feature extraction method, device, equipment and system
Severa et al. Spiking neural algorithms for markov process random walk
Xue et al. Self-adaptive particle swarm optimization-based echo state network for time series prediction
Galluppi et al. Real time on-chip implementation of dynamical systems with spiking neurons
CN113112791A (en) Traffic flow prediction method based on sliding window long-and-short term memory network
CN114004338A (en) Mixed time period mode multivariable time sequence prediction method based on neural network
CN113704075A (en) Fault log-based high-performance computing system fault prediction method
Salloom et al. Proportional integral derivative booster for neural networks-based time-series prediction: Case of water demand prediction
Karimpanal et al. Experience replay using transition sequences
Guoqiang et al. Study of RBF neural network based on PSO algorithm in nonlinear system identification
Safari et al. Quantum neural networks (QNN) application in weather prediction of smart grids
CN116843083A (en) Carbon emission prediction system and method based on hybrid neural network model
US11911902B2 (en) Method for obstacle avoidance in degraded environments of robots based on intrinsic plasticity of SNN
CN116578858A (en) Air compressor fault prediction and health degree evaluation method and system based on graphic neural network
CN112070200B (en) Harmonic group optimization method and application thereof
CN114386666A (en) Wind power plant short-term wind speed prediction method based on space-time correlation
Wan et al. Energy efficiency optimization and carbon emission reduction targets of resource-based cities based on BiLSTM-CNN-GAN model
Wang et al. Interpretable uncertainty forecasting framework for robust configuration of energy storage in a virtual power plant
Prasad et al. A Hybrid Time Series Rainfall Prediction Model Using Neural Prophet and LS TM
Rana Time series prediction of the COVID-19 outbreak in India using LSTM based deep learning models
Alharbi et al. Wind Speed and Solar Irradiance Prediction Using a Bidirectional Long Short-Term Memory Model Based on Neural Networks. Energies, 2021, 14, 6501

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
WD01 Invention patent application deemed withdrawn after publication

Application publication date: 20190419

WD01 Invention patent application deemed withdrawn after publication