CN108829983B - Equipment residual life prediction method based on multi-hidden-state fractional Brownian motion - Google Patents

Equipment residual life prediction method based on multi-hidden-state fractional Brownian motion Download PDF

Info

Publication number
CN108829983B
CN108829983B CN201810643898.2A CN201810643898A CN108829983B CN 108829983 B CN108829983 B CN 108829983B CN 201810643898 A CN201810643898 A CN 201810643898A CN 108829983 B CN108829983 B CN 108829983B
Authority
CN
China
Prior art keywords
state
hidden
brownian motion
nonlinear
function
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Active
Application number
CN201810643898.2A
Other languages
Chinese (zh)
Other versions
CN108829983A (en
Inventor
苗强
张恒
张新
王剑宇
莫贞凌
刘慧宇
曾小飞
王俊峰
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.)
Sichuan University
Original Assignee
Sichuan 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 Sichuan University filed Critical Sichuan University
Priority to CN201810643898.2A priority Critical patent/CN108829983B/en
Publication of CN108829983A publication Critical patent/CN108829983A/en
Application granted granted Critical
Publication of CN108829983B publication Critical patent/CN108829983B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Classifications

    • 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
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F2119/00—Details relating to the type or aim of the analysis or the optimisation
    • G06F2119/04—Ageing analysis or optimisation against ageing

Landscapes

  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • Computer Hardware Design (AREA)
  • Evolutionary Computation (AREA)
  • Geometry (AREA)
  • General Engineering & Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Management, Administration, Business Operations System, And Electronic Commerce (AREA)

Abstract

The invention relates to the field of electromechanical equipment service life prediction, discloses an equipment residual service life prediction method based on multi-hidden-state fractional Brownian motion, and solves the problem that the traditional equipment residual service life prediction method based on fractional Brownian motion only considers the current observation value and is low in equipment service life prediction precision. Firstly, selecting a nonlinear function according to the service life degradation trend of equipment, determining a nonlinear fractional Brownian motion model, and taking parameters in the nonlinear function as a hidden state; then converting the nonlinear fractional Brown motion model into a nonlinear Brown motion model; carrying out curve fitting on the training data to obtain an initial value of the mean value of the hidden state; iteratively updating the mean value and the variance of the hidden state to obtain a distribution function of the hidden state; deducing the posterior probability density distribution of the first impact time; and finally, predicting the service life by using the posterior probability density distribution of the first impact time. The method is suitable for predicting the residual effective life of the electromechanical equipment.

Description

Equipment residual life prediction method based on multi-hidden-state fractional Brownian motion
Technical Field
The invention relates to the field of life prediction of electromechanical equipment, in particular to a method for predicting the residual life of equipment based on multi-hidden-state fractional Brownian motion.
Background
With the rapid development of modern technology and the continuous increase of functional requirements, a large number of electromechanical devices gradually show the trend of complication, integration and intellectualization, and the trend urgently needs the improvement of the health management capability and reliability of the electromechanical devices. The electromechanical device has inevitable performance degradation during operation. When the performance of the equipment is degraded to the extent that the equipment is not enough to complete the function of the equipment, the equipment can be shut down or even break down, and huge economic loss or even casualties are caused. Accurate prediction of the remaining useful life of the equipment can provide a correct and effective maintenance strategy, playing an important role in avoiding these serious safety incidents and economic losses. Therefore, the prediction of the remaining useful life of the equipment has become a hot research point in the fields of system failure prediction and health management. Lithium batteries are used as power sources for many electromechanical devices, and whether they can provide the power required by the devices has a significant impact on the safe operation of the electromechanical devices. Therefore, it is also necessary to predict the remaining life of the lithium battery.
The current methods for predicting the remaining useful life of equipment are mainly divided into two categories: one is a regression model-based method, which relies mainly on a state space model consisting of a state equation and a measurement equation and regression analysis, for example, a method combining an empirical degradation function obtained from lithium battery degradation data and particle filtering; the second type is a method based on a random process model, which mainly takes the equipment degradation process as a random process, then represents the degradation process through the random process, such as a gamma process, a Markov process, a wiener process and the like, and further analytically obtains a posterior distribution function of the first impact time. Because the probability distribution function of the first impact time is inverse Gaussian distribution, the linear Brownian motion model is widely concerned in the field of residual effective life prediction. However, in engineering practice, the equipment degradation process mostly presents nonlinear characteristics, resulting in limited use of the linear brownian motion model. Then, the distribution function of the first impact time of the nonlinear brownian motion model is approximated through space-time conversion, so that the nonlinear brownian motion model is widely applied. However, the brownian motion model assumes that the increments are independent throughout the degradation process, which is inconsistent with the degradation process of some devices, such as lithium batteries, bearings, etc. There is a correlation between the increments in the degradation process of these devices, that is, there is a long dependency on the degradation process. The nonlinear fractional brownian motion model is introduced to predict the remaining useful life of the equipment with long dependency on the degradation process, so that the problem of relevant increment is effectively solved, but the residual life prediction method based on the nonlinear fractional brownian motion model in the prior literature takes parameters in the model as constants, so that the problem that only the current observed value is considered exists.
Disclosure of Invention
The technical problem to be solved by the invention is as follows: the method for predicting the residual life of the equipment based on the multi-hidden-state fractional Brownian motion solves the problem that the traditional method for predicting the residual life of the equipment based on the fractional Brownian motion only considers the current observed value and is low in equipment life prediction precision.
In order to solve the problems, the invention adopts the technical scheme that: the method for predicting the residual life of the equipment based on the multi-hidden-state fractional Brownian motion comprises the following steps:
step 1: selecting a non-linear function according to the service life degradation trend of the equipment, determining a non-linear fractional Brownian motion model of the non-linear function, and taking parameters in the non-linear function as non-observable state variables, wherein the non-observable state variables are hidden states;
step 2: converting the nonlinear fractional Brown motion model into a nonlinear Brown motion model;
and step 3: selecting training data, performing curve fitting on the training data based on the nonlinear function selected in the step 1, and obtaining an initial value of a hidden state mean value through a fitting curve;
and 4, step 4: iteratively updating the mean value and the variance of the hidden state by using the historical degradation data of the equipment to obtain a distribution function of the hidden state;
and 5: deducing posterior probability density distribution of the first impact time by combining the distribution function of the hidden state;
step 6: and predicting the residual effective life of the equipment by using the posterior probability density distribution of the first impact time.
Further, the nonlinear fractional brownian motion model determined in step 1 is as follows:
Figure BDA0001703043690000021
wherein X (t) is the state of the device at time t; x (0) is an initial state; μ (τ; θ) is a non-linear function; tau is an integral variable in the nonlinear function, and theta is a parameter vector in the nonlinear function; sigmaHIs the drift coefficient; b isH(t) is a fractional Brownian motion function with a Herster index of H.
Further, in order to better fit the curve to the training data in step 3, the non-linear function selected in step 1 is:
μ(τ；θ)＝a·b·exp(bτ)+c·d·exp(dτ)
where a, b, c, and d are parameters in a nonlinear function, and θ ═ a, b, c, and d.
Further, step 2 converts the nonlinear fractional brownian motion model into a nonlinear brownian motion model through a weak convergence theory.
Further, step 3 selects the Battery Data Set test Data provided by the central station for predicting the excellent failure of the American aviation and space administration as the training Data.
Further, in order to update the mean and variance of the hidden state in a better iteration manner, step 4 updates the mean and variance of the hidden state in an iteration manner by using a tasteless particle filter method.
Further, the distribution function of the hidden state obtained in step 4 is:
Figure BDA0001703043690000031
wherein m iskMean value, P, representing the hidden state at time kkRepresenting the variance of the hidden state at time k, X0:kRepresenting the observed value of the particle, theta, at times 0 to kkRepresenting an unobservable state variable at time k.
Further, the posterior probability density distribution function of the first impact time derived in step 5 is:
Figure BDA0001703043690000032
wherein:
Figure BDA0001703043690000033
lkfor remaining useful life, omegathIs a capacitance threshold value, NsAs to the number of the particles,
Figure BDA0001703043690000034
the weight of the ith particle at time k,
Figure BDA0001703043690000035
σ (t) is a time-varying coefficient generated when the fractional brownian motion model is converted into the brownian motion model, and σ (t) is a value of a hidden state of the ith particle at the time k.
The invention has the beneficial effects that: according to the method, parameters of a nonlinear function in a nonlinear fractional Brown motion model are taken as hidden variables, so that the model is more flexible, and a tasteless particle filtering method is adopted, historical data are used for iteratively updating the mean value and the variance of a plurality of hidden states, so that a distribution function of the hidden states is obtained; meanwhile, the sum of the double exponential functions is used as a nonlinear function in the nonlinear fractional Brownian motion model to approximately obtain a posterior distribution function of the first impact time, finally, the residual effective service life of the lithium battery is predicted, and the service life prediction precision of the equipment is improved.
Drawings
FIG. 1 is a flow chart of the present invention;
FIGS. 2 a-2 c are graphs comparing the predicted results of method 4 with methods 1-3, respectively, at 40,45,50,55,60,65, and 70 cycles;
FIG. 3 is a graph of the relative error of the predicted results of the four methods at 40,45,50,55,60,65, and 70 cycles.
In the figure: L1-L4 represent the curves of the probability density function of the predicted remaining useful life of methods 1-4, respectively; a represents the true remaining life value; relative error curves for the predicted results of methods 1 to 4, M1-M4, respectively.
Detailed Description
The method for predicting the residual life of equipment based on the multi-hidden-state fractional Brownian motion, disclosed by the invention, as shown in figure 1, comprises the following steps of:
step 1: and selecting a non-linear function according to the service life degradation trend of the equipment, determining a non-linear fractional Brownian motion model of the non-linear function, and taking parameters in the non-linear function as non-observable state variables, wherein the non-observable state variables are hidden states. This step may select μ (τ; θ) ═ a · b · exp (b τ) + c · d · exp (d τ) as the nonlinear function, where a, b, c, and d are parameters in the nonlinear function, and θ ═ a, b, c, d](ii) a And the nonlinear fractional brownian motion model determined in the step 1 is as follows:
Figure BDA0001703043690000041
wherein X (t) is the state of the device at time t; x (0) is an initial state, and is generally 0; μ (τ; θ) is a non-linear function; tau is an integral variable in the nonlinear function, and theta is a parameter vector in the nonlinear function; sigmaHIs the drift coefficient; b isH(t) is fractional Brownian motion.
Step 2: and converting the nonlinear fractional Brownian motion model into a nonlinear Brownian motion model. This step can convert the nonlinear fractional brownian motion model to a nonlinear brownian motion model by weak convergence theory.
And step 3: and (3) selecting training data, carrying out curve fitting on the training data based on the nonlinear function selected in the step (1), and obtaining an initial value of the hidden state mean value through a fitting curve. In the step, the Battery Data Set test Data provided by the Excellent failure prediction center of the American aerospace agency can be selected as training Data.
And 4, step 4: and iteratively updating the mean value and the variance of the hidden state by using the historical degradation data of the equipment to obtain a distribution function of the hidden state. This step can iteratively update the mean and variance of the hidden states using a method of unscented particle filtering.
And 5: and (4) combining the distribution function of the hidden state to deduce the posterior probability density distribution of the first impact time.
Step 6: and predicting the residual effective life of the lithium battery by using the posterior probability density distribution of the first impact time.
According to the method, parameters of a nonlinear function in a nonlinear fractional Brown motion model are taken as hidden variables, so that the model is more flexible, and a tasteless particle filtering method is adopted, historical data are used for iteratively updating the mean value and the variance of a plurality of hidden states, so that a distribution function of the hidden states is obtained; meanwhile, the sum of the double exponential functions is used as a nonlinear function in a nonlinear fractional Brownian motion model to approximately obtain a posterior distribution function of the first impact time, and finally, the residual effective life of the lithium battery is predicted.
Examples
The embodiment takes a lithium ion battery as an example, and provides a device residual life prediction method based on multi-hidden-state fractional Brownian motion, which comprises the following specific steps:
step 1: and selecting a non-linear function according to the service life degradation trend of the equipment, determining a non-linear fractional Brownian motion model of the non-linear function, and taking parameters in the non-linear function as non-observable state variables, wherein the non-observable state variables are hidden states.
The step takes the empirical degradation function of the lithium battery as a nonlinear function in a nonlinear fractional brownian motion model. The nonlinear function chosen was:
μ(τ；θ)＝a·b·exp(bτ)+c·d·exp(dτ)
wherein tau is a variable in a nonlinear function, and a, b, c and d are parameters in the nonlinear function;
the nonlinear fractional brownian motion model is as follows:
X(t)＝X(0)+a·exp(bt)+c·exp(dt)+σHBH(t)
x (t) is a function of state, in this case the capacity of the lithium battery; x (0) is an initial value, set here to 0; b isH(t) is the fraction of the hurst index HBrownian motion; sigmaHIs the drift coefficient.
In this example, θ ═ a, b, c, d is set to the hidden state.
Step 2: and converting the nonlinear fractional Brownian motion model into a nonlinear Brownian motion model by a weak convergence theory.
Because the above-described degeneration process is neither a markov process nor a half-yoke process, and therefore a deterministic yoke for the first attack time is difficult to derive, embodiments may utilize weak convergence theory to approximately convert a nonlinear fractional brownian motion model to a nonlinear brownian motion model, which may be expressed as:
Figure BDA0001703043690000051
b (t) is standard Brownian motion; σ (t) is a time-varying coefficient, which can be expressed as:
Figure BDA0001703043690000052
cHas a normalization constant, it can be expressed as:
Figure BDA0001703043690000053
and step 3: and selecting training data, carrying out curve fitting on the training data based on a nonlinear function, and obtaining an initial value of the hidden state mean value through a fitting curve.
In the step, the given nonlinear function is used, Battery Data Set test Data provided by the central station for predicting the excellent fault of the American aerospace agency is used as training Data, Battery Data B0006, Battery Data B0007 and Battery Data B0018 are used as training Data, Battery Data B0005 is used for predicting the service life, and the Battery Data B0006, Battery Data B0007 and Battery Data B0018 are subjected to curve fitting by using a matlab curve fitting tool box.
And then averaging the parameter values of the obtained fitting curve to obtain an initial value of the hidden state average value.
And 4, step 4: and (3) iteratively updating the mean value and the variance of the hidden state by using historical degradation data by using a tasteless particle filtering method to obtain a distribution function of the hidden state.
When θ is considered as a hidden state, it is no longer a constant, in this example, it is assumed that θ follows a multivariate gaussian distribution of a 4-dimensional mean vector m and a 4 × 4-dimensional variance matrix P. The state space model can be described as:
θk＝θk-1+ν
xk＝xk-1+φ(k)-φ(k-1)+σ(tk)B(tk)-σ(tk-1)B(tk-1)+n
wherein:
φ(k)＝ak·exp(bktk)+ck·exp(dktk)
v is a state error and follows multivariate Gaussian distribution of a 4-dimensional 0-mean vector and a 4 multiplied by 4-dimensional variance matrix Q; n is a measurement error and follows multivariate Gaussian distribution of a zero mean vector and a variance matrix R; by using historical data, the state space model is solved, and the mean value and the variance of the hidden state can be estimated in a posteriori mode. The unscented particle filter is the prior art, and the process of estimating the mean and variance of the hidden state by the unscented particle filter of the embodiment is as follows:
(1) initializing;
Figure BDA0001703043690000061
Figure BDA0001703043690000062
(2) generating a proposed distribution using unscented kalman filtering;
Figure BDA0001703043690000063
Figure BDA0001703043690000064
Figure BDA0001703043690000065
Figure BDA0001703043690000071
Figure BDA0001703043690000072
Figure BDA0001703043690000073
Figure BDA0001703043690000074
Figure BDA0001703043690000075
Figure BDA0001703043690000076
Figure BDA0001703043690000077
Figure BDA0001703043690000078
as described above in detail with reference to the drawings,
Figure BDA0001703043690000079
is the selected particle; λ is a parameter of the tasteless transform;
Figure BDA00017030436900000710
and
Figure BDA00017030436900000711
the weight coefficients are respectively the first-order statistical characteristic and the second-order statistical characteristic; m isk|k-1、Pk|k-1And Zk|k-1One-step prediction of state quantity, variance and measurement value, respectively; kkIs the filter gain;
Figure BDA00017030436900000712
and
Figure BDA00017030436900000713
respectively the final mean value and the variance of the tasteless Kalman filtering; omegakIs the particle weight.
(3) Importance sampling and weight calculation;
taking N from the proposed distribution obtained abovesA particle, wherein the particle obeys
Figure BDA00017030436900000714
The calculation weight formula is as follows:
Figure BDA00017030436900000715
(4) resampling;
if the number of valid particles is below the threshold, resampling is performed, and N is re-adopted from the current set of particlessAnd (4) particles. The threshold calculation formula is as follows:
Figure BDA00017030436900000716
(5) outputting;
the hidden state mean value calculation formula is as follows:
Figure BDA0001703043690000081
and if k is less than or equal to the latest observation time T, making k equal to k +1, and returning to the step (2), otherwise, outputting a prediction result.
After obtaining the mean and the variance, the distribution function of the hidden state can be obtained by substituting the standard normal distribution function, and the distribution function of the hidden state can be expressed as:
Figure BDA0001703043690000082
in the formula, mkMean value, P, representing the hidden state at time kkRepresenting the variance of the hidden state at time k, X0:kRepresenting the observed value of the particle, theta, at times 0 to kkRepresenting an unobservable state variable at time k.
And 5: and (4) combining the distribution function of the hidden state to deduce the posterior probability density distribution of the first impact time.
Under the multi-hidden state nonlinear fractional brownian motion model, the posterior probability density distribution function of the first impact time can be expressed as:
Figure BDA0001703043690000083
approximating the integral calculation described above with the particles and weights generated in step 4, the following results may be obtained:
Figure BDA0001703043690000084
wherein:
Figure BDA0001703043690000085
Figure BDA0001703043690000086
lkfor remaining useful life, omegathIs the threshold value of the capacity, i.e. the battery capacity is below omegathRegarded as invalid, NsAs to the number of the particles,
Figure BDA0001703043690000091
the weight of the ith particle at time k,
Figure BDA0001703043690000092
σ (t) is a time-varying coefficient generated when the fractional brownian motion model is converted into the brownian motion model, and σ (t) is a value of a hidden state of the ith particle at the time k.
Step 6: and predicting the residual effective life of the lithium battery by using the posterior probability density distribution of the first impact time.
Given the life threshold, the posterior distribution of the first impact time is calculated by the formula obtained in step 6. The result of predicting the remaining effective life of the B0005 battery by using the method in 70 cycles is shown in FIG. 3, and it can be seen that in 70 cycles, the prediction model can well track the degradation trend of the capacity, and the peak value of the posterior distribution function of the first impact time obtained by prediction is very close to the real remaining life, so that the method provided by the invention can obtain very good precision in predicting the remaining effective life of the lithium battery.
To verify the superiority of the method presented in the examples, the method presented in the examples was compared with three other methods, four of which are described below:
the method comprises the following steps: adopting a non-linear fractional Brownian motion model without a hidden state, wherein the non-linear function mu (tau; theta) is selected from the non-linear functions a.b.exp (b tau) + c.d.exp (d tau), and an iterative updating method is odorless particle filtering;
the method 2 comprises the following steps: adopting a multi-hidden state nonlinear fractional Brownian motion model, wherein the nonlinear function mu (tau; theta) is selected from the provided nonlinear functions a.b.exp (b tau) + c.d.exp (d tau), and an iterative updating method is particle filtering;
the method 3 comprises the following steps: a nonlinear fractional Brownian motion model with multiple hidden states is adopted, wherein a nonlinear function mu (tau; theta) is a nonlinear function a.b.exp (b tau) commonly used in the prior literature, and an iterative updating method is odorless particle filtering;
the method 4 comprises the following steps: adopting a multi-hidden state nonlinear fractional Brownian motion model, wherein the nonlinear function mu (tau; theta) is selected from the proposed nonlinear functions a.b.exp (b tau) + c.d.exp (d tau), and an iterative updating method is odorless particle filtering, namely the method proposed by the embodiment;
in addition, in order to verify the robustness of the method, the embodiment respectively predicts the remaining effective life at different time points of the lithium batteries 40,45,50,55,60,65,70 cycles and the like. The predicted result pairs of the graphs comparing the predicted results of method 4 with those of methods 1-3, respectively, at different prediction times are shown in fig. 2 a-2 c. Fig. 3 shows the prediction error of the four methods at different times. As can be seen from fig. 3, the embodiment takes the parameter of the non-linear function as a hidden state, which is an improvement on the non-linear fractional brownian motion model; a better result than particle filtering can also be obtained by adopting tasteless particle filtering to update the mean value and the variance of the hidden state; compared with the nonlinear function in the prior literature, the nonlinear function can better track the degradation trend of the lithium battery. Overall, the embodiments can significantly improve the prediction accuracy.

Claims (5)

1. The method for predicting the residual life of the equipment based on the multi-hidden-state fractional Brownian motion is characterized by comprising the following steps of:
step 1: selecting a non-linear function according to the service life degradation trend of the equipment, determining a non-linear fractional Brownian motion model of the non-linear function, and taking parameters in the non-linear function as non-observable state variables, wherein the non-observable state variables are hidden states;
step 2: converting the nonlinear fractional Brown motion model into a nonlinear Brown motion model;
and step 3: selecting training data, performing curve fitting on the training data based on the nonlinear function selected in the step 1, and obtaining an initial value of a hidden state mean value through a fitting curve;
and 4, step 4: iteratively updating the mean value and the variance of the hidden state by using the historical degradation data of the equipment and adopting a tasteless particle filtering method to obtain a distribution function of the hidden state;
the distribution function of the hidden states is:
Figure FDA0002934084720000011
wherein m iskMean value, P, representing the hidden state at time kkRepresenting the variance of the hidden state at time k, X0:kRepresenting the observed value of the particle, theta, at times 0 to kkA state variable representing unobservable at time k;
and 5: and deducing the posterior probability density distribution of the first impact time by combining the distribution function of the hidden state, wherein the deduced posterior probability density distribution function of the first impact time is as follows:
Figure FDA0002934084720000012
wherein:
Figure FDA0002934084720000013
lkfor remaining useful life, omegathIs a capacitance threshold value, NsAs to the number of the particles,
Figure FDA0002934084720000014
the weight of the ith particle at time k,
Figure FDA0002934084720000015
σ (t) is a time-varying coefficient generated when the fractional brownian motion model is converted into the brownian motion model, and is the value of the hidden state of the ith particle at the moment k;
step 6: and predicting the residual effective life of the equipment by using the posterior probability density distribution of the first impact time.
2. The method for predicting the remaining life of a device based on multi-hidden-state fractional brownian motion according to claim 1, wherein the non-linear fractional brownian motion model determined in step 1 is as follows:
Figure FDA0002934084720000016
wherein X (t) is the state of the device at time t; x (0) is an initial state; μ (τ; θ) is a non-linear function; tau is an integral variable in the nonlinear function, and theta is a parameter vector in the nonlinear function; sigmaHIs the drift coefficient; b isH(t) is a fractional Brownian motion function with a Herster index of H.
3. The method for predicting the remaining life of equipment based on the multi-hidden-state fractional brownian motion according to claim 2, wherein the non-linear function selected in the step 1 is as follows:
μ(τ；θ)＝a·b·exp(bτ)+c·d·exp(dτ)
where a, b, c, and d are parameters in a nonlinear function, and θ ═ a, b, c, and d.
4. The method for predicting the remaining life of a device based on multi-hidden-state fractional brownian motion according to claim 1, wherein the step 2 converts the nonlinear fractional brownian motion model into the nonlinear brownian motion model by a weak convergence theory.
5. The method for predicting the residual life of equipment based on the multi-hidden-state fractional brownian motion as claimed in claim 1, wherein step 3 selects Battery Data Set test Data provided by the central prominent failure prediction of the U.S. aviation and space administration as training Data.
CN201810643898.2A 2018-06-21 2018-06-21 Equipment residual life prediction method based on multi-hidden-state fractional Brownian motion Active CN108829983B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201810643898.2A CN108829983B (en) 2018-06-21 2018-06-21 Equipment residual life prediction method based on multi-hidden-state fractional Brownian motion

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201810643898.2A CN108829983B (en) 2018-06-21 2018-06-21 Equipment residual life prediction method based on multi-hidden-state fractional Brownian motion

Publications (2)

Publication Number Publication Date
CN108829983A CN108829983A (en) 2018-11-16
CN108829983B true CN108829983B (en) 2021-03-23

Family

ID=64143042

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201810643898.2A Active CN108829983B (en) 2018-06-21 2018-06-21 Equipment residual life prediction method based on multi-hidden-state fractional Brownian motion

Country Status (1)

Country Link
CN (1) CN108829983B (en)

Families Citing this family (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN110781592B (en) * 2019-10-25 2020-12-22 中国人民解放军火箭军工程大学 Method and system for predicting residual life of equipment
CN111413621A (en) * 2020-04-03 2020-07-14 重庆大学 Lithium battery aging rate estimation method based on iterative learning
CN114662226B (en) * 2022-02-24 2024-03-29 北京工业大学 Rolling bearing residual service life prediction method based on time-varying Kalman filtering

Citations (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN107403242A (en) * 2017-08-10 2017-11-28 四川大学 Based on the equipment method for predicting residual useful life for improving tasteless particle filter

Family Cites Families (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20170091358A1 (en) * 2015-09-27 2017-03-30 Caterpillar Inc. Model-based prognosis of machine health
CN106353691A (en) * 2016-10-31 2017-01-25 首都师范大学 Method for degradation modeling and life prediction of lithium battery with self-healing effect
CN107688687A (en) * 2017-07-10 2018-02-13 山东科技大学 One kind considers time-length interrelation and the probabilistic life-span prediction method of part

Patent Citations (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN107403242A (en) * 2017-08-10 2017-11-28 四川大学 Based on the equipment method for predicting residual useful life for improving tasteless particle filter

Non-Patent Citations (2)

* Cited by examiner, † Cited by third party
Title
Nonlinear-drifted Brownian motion with multiple hidden states for remaining useful life prediction of rechargeable batteries;Dong Wang等;《Mechanical Systems and Signal Processing》;20170901;第93卷;第531-544页 *
Remaining Useful Life Prediction for Degradation Processes With Long-Range Dependence;Hanwen Zhang等;《IEEE Transactions on Reliability》;20170829;第1368-1379页 *

Also Published As

Publication number Publication date
CN108829983A (en) 2018-11-16

Similar Documents

Publication Publication Date Title
Xu et al. Real-time reliability prediction for a dynamic system based on the hidden degradation process identification
CN107957562B (en) Online prediction method for residual life of lithium ion battery
Mo et al. A remaining useful life prediction approach for lithium-ion batteries using Kalman filter and an improved particle filter
Daigle et al. An efficient deterministic approach to model-based prediction uncertainty estimation
CN108829983B (en) Equipment residual life prediction method based on multi-hidden-state fractional Brownian motion
CN107480440B (en) Residual life prediction method based on two-stage random degradation modeling
CN104504296B (en) Gaussian of Mixture Hidden Markov Model and the method for predicting residual useful life of regression analysis
CN107918103B (en) Lithium ion battery residual life prediction method based on gray particle filtering
CN112415414A (en) Method for predicting remaining service life of lithium ion battery
CN103577710B (en) Aviation Power Converter faults Forecasting Methodology based on fractional order UPF
CN110398368B (en) FBM-based bearing inner ring fault residual life prediction method based on long correlation model
CN108304685A (en) A kind of non-linear degradation equipment method for predicting residual useful life and system
CN104459560A (en) Method for predicting remaining service life of lithium battery based on wavelet denoising and relevant vector machine
CN104615866B (en) A kind of life-span prediction method based on physical-statistical model
CN113239132B (en) Online out-of-tolerance identification method for voltage transformer
Yan et al. Low-cost adaptive lebesgue sampling particle filtering approach for real-time li-ion battery diagnosis and prognosis
CN110687450B (en) Lithium battery residual life prediction method based on phase space reconstruction and particle filtering
Zhao et al. Model-based fault diagnosis methods for systems with stochastic process–a survey
CN107403242B (en) Equipment residual life prediction method based on improved odorless particle filtering
CN105868557A (en) Online prediction method for remaining life of electromechanical equipment under situation of two-stage degradation
CN113791351B (en) Lithium battery life prediction method based on transfer learning and difference probability distribution
Zheng et al. Real-time estimation of battery state of charge with metabolic grey model and LabVIEW platform
Daigle et al. Advanced methods for determining prediction uncertainty in model-based prognostics with application to planetary rovers
CN111460692A (en) Equipment residual life prediction method and system considering degradation rate mutual influence
Peng et al. Collaborative online RUL prediction of multiple assets with analytically recursive Bayesian inference

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
GR01 Patent grant
GR01 Patent grant