CN110456199B - Residual life prediction method of multi-sensor system - Google Patents

Residual life prediction method of multi-sensor system Download PDF

Info

Publication number
CN110456199B
CN110456199B CN201910750989.0A CN201910750989A CN110456199B CN 110456199 B CN110456199 B CN 110456199B CN 201910750989 A CN201910750989 A CN 201910750989A CN 110456199 B CN110456199 B CN 110456199B
Authority
CN
China
Prior art keywords
sensor system
calculating
data
health
improved
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
CN201910750989.0A
Other languages
Chinese (zh)
Other versions
CN110456199A (en
Inventor
苗强
张恒
罗冲
莫贞凌
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 CN201910750989.0A priority Critical patent/CN110456199B/en
Publication of CN110456199A publication Critical patent/CN110456199A/en
Application granted granted Critical
Publication of CN110456199B publication Critical patent/CN110456199B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Classifications

    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01R—MEASURING ELECTRIC VARIABLES; MEASURING MAGNETIC VARIABLES
    • G01R31/00—Arrangements for testing electric properties; Arrangements for locating electric faults; Arrangements for electrical testing characterised by what is being tested not provided for elsewhere

Landscapes

  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Management, Administration, Business Operations System, And Electronic Commerce (AREA)
  • Measuring And Recording Apparatus For Diagnosis (AREA)

Abstract

The invention relates to the technical field of electromechanical equipment health management, and discloses a residual life prediction method of a multi-sensor system, which comprises the following specific implementation steps of: s1 calculating improved second order permutation entropy of the sensor data; s2: calculating and comparing improved second-order permutation entropy values of the sensor data according to the calculation method of S1; s3: an improved second order permutation entropy value is selected and the data of the sensor is used for data fusion. The method for predicting the residual life of the multi-sensor system provides an improved monotonicity trend of second-order permutation entropy index measurement data, and fuses data of a plurality of sensors through a data fusion method to construct a health index with a good degradation trend and a smaller threshold variance to represent the health state of equipment, and then estimates and predicts the health state of the equipment by taking the data of the health index as an observed value of odorless particle filtering, so that the residual life of the multi-sensor system is predicted finally.

Description

Residual life prediction method of multi-sensor system
Technical Field
The invention relates to the technical field of electromechanical equipment health management, in particular to a residual life prediction method of a multi-sensor system.
Background
With the rapid development of modern technology and the continuous improvement of functional requirements, the complexity, comprehensiveness and intelligence of a large number of electromechanical devices are continuously improved, so that the electromechanical devices become a large-scale complex multi-sensing system, and at the same time, the reliability and safe operation of these devices are becoming more and more important, however, the complex operating conditions and the harsh operating environment, which lead to the inevitable performance degradation of the electromechanical devices during operation, when the performance of the electromechanical device degrades such that the device is insufficient to perform its function, it can result in equipment downtime or even failure, and huge economic loss and even casualties are brought, as a key technology of prediction and health management, the residual life prediction can effectively predict the potential degradation trend of the system and estimate the failure time of the electromechanical equipment, and has important significance and practical value for ensuring the safe service and reliable operation of the electromechanical equipment.
At present, most scholars research on the prediction of the residual life of the electromechanical equipment and adopt different methods and technologies to predict the residual life of the electromechanical equipment, and some scholars integrate the advantages of particle filters and bond diagram modeling to realize RUL prediction and obtain higher precision and accurate confidence limits; some scholars establish an improved autoregressive model and predict RUL by using a particle filtering method, so that the prediction accuracy is improved, but the methods only use data of one sensor in electromechanical equipment to predict the residual life, so that the performance degradation mechanism of the electromechanical equipment is clear, and single sensor data can describe the degradation process of the electromechanical equipment, so that the methods are effective.
Disclosure of Invention
The invention provides a method for predicting the residual service life of a multi-sensor system, which has the advantage of accurately predicting the residual service life of electromechanical equipment and solves the problems in the background technology.
The invention provides the following technical scheme: a method for predicting the residual life of a multi-sensor system comprises the following specific implementation steps:
s1 calculating improved second order permutation entropy of the sensor data;
s2: calculating and comparing improved second-order permutation entropy values of the sensor data according to the calculation method of S1;
s3: selecting an improved second-order permutation entropy value, using data of the sensor for data fusion, and constructing a health index capable of representing the health state of the multi-sensor system;
s4: and the health index is used as an observed value of a particle filtering method, so that the health state of the multi-sensor system is predicted.
As a preferable aspect of the method for predicting remaining life of a multi-sensor system of the present invention, wherein: the calculating of the improved second order permutation entropy of the sensor data comprises the steps of:
s1.1: the ith sensor data xiMapping to a 2-dimensional phase space for reconstruction, and obtaining a phase space matrix as shown in the following:
Figure GDA0002934962680000021
a is a reconstructed phase space matrix, B is a submatrix in A, and n is the length of sensor data;
s1.2: for each state vector of the above reconstructed phase space matrix, a symbol sequence is obtained:
S(l)＝[m1,m2],l＝1,2,…,n*(n-1)/2
where l is the state vector index, [ m ]1,m2]There are two possibilities pi for permutation according to the magnitude of the state vector values1＝[0,1]And pi2＝[1,0]When two numerical values in the state vector are equal, take [0, 1%]Comparing the magnitude of the numerical value in each state vector to obtain the arrangement relation of each state vector;
s1.3: calculating the corresponding permutation pi of all state vectors1,π2The probability of (d);
s1.4: a second order rank entropy value of the sensor data improvement is calculated.
As a preferable aspect of the method for predicting remaining life of a multi-sensor system of the present invention, wherein: calculating the corresponding arrangement pi of all state vectors1,π2The calculation formula used for the probability is as follows:
Figure GDA0002934962680000031
Figure GDA0002934962680000032
wherein # represents the number.
As a preferable aspect of the method for predicting remaining life of a multi-sensor system of the present invention, wherein: the calculation formula for calculating the improved second order entropy value of the sensor data is as follows:
S2＝abs(p(π1)log2[p(π1)]-p(π2)log2[p(π2)])
where abs () is the absolute value.
As a preferable aspect of the method for predicting remaining life of a multi-sensor system of the present invention, wherein: the method for constructing the health index capable of representing the health state of the multi-sensor system specifically comprises the following steps:
s3.1: constructing a health index at the time t;
s3.2: solving the dual-target optimization problem through a genetic algorithm to obtain an optimal weight vector omega*And calculating the integral monotonous trend and the fault threshold variance of the health index.
As a preferable aspect of the method for predicting remaining life of a multi-sensor system of the present invention, wherein: the construction formula for constructing the health index at the time t is as follows:
Ht＝x·,tω t＝0,1,2,…,ni
where ω is a weight vector, x·,t∈R1×SIs a matrix of measurements collected from the S sensors;
the calculation formula for calculating the overall monotonous trend of the health index is as follows:
Figure GDA0002934962680000041
s.t.ω′1S＝1
wherein S2ω(d) Is an improved second order permutation entropy value of the health index determined by the weight vector omega,
Figure GDA0002934962680000042
is the fault threshold variance of the health index determined by the weight vector ω,1S∈RS×1Is a vector with all values of 1;
the fault threshold variance calculation formula is as follows:
Figure GDA0002934962680000043
wherein Y ∈ RM×SIs a matrix of fault values, M being a training unitThe number of the cells.
As a preferable aspect of the method for predicting remaining life of a multi-sensor system of the present invention, wherein: the particle filtering method specifically comprises the following steps:
s4.1: constructing a state space model of the multi-sensing system:
θk＝f(θk-1,υk)
Hk＝h(θk,νk)
wherein θ is a state vector, f is a nonlinear function, upsilon is a process variance, H is a health index, H is a nonlinear observation function, and v is a measurement variance;
s4.2: generating an important density function of the particle filter by using unscented Kalman filtering;
s4.21: setting k to 0, extracting N from the prior distributionsParticles and determining the weight to be 1/NsThen, initializing parameters:
Figure GDA0002934962680000044
Figure GDA0002934962680000045
augmented initialization state vector and covariance matrix:
Figure GDA0002934962680000046
Figure GDA0002934962680000051
s4.22: gaussian point and weight calculation:
Figure GDA0002934962680000052
s4.23: and (3) time updating:
Figure GDA0002934962680000053
Figure GDA0002934962680000054
Figure GDA0002934962680000055
Figure GDA0002934962680000056
Figure GDA0002934962680000057
s4.24: and (3) updating the measurement value:
Figure GDA0002934962680000058
Figure GDA0002934962680000059
Figure GDA00029349626800000510
Figure GDA00029349626800000511
Figure GDA00029349626800000512
s4.3: calculating the sampling particle and updating the particle weight:
Figure GDA00029349626800000513
Figure GDA0002934962680000061
s4.4: if the number of significant particles is below a given threshold, resampling is performed:
Figure GDA0002934962680000062
Figure GDA0002934962680000063
s4.5: and (3) state estimation:
Figure GDA0002934962680000064
if k is greater than T, ending the prediction;
wherein
Figure GDA0002934962680000065
For selected particles, alpha and beta are parameters of the tasteless transform, Wi (m)And Wi (c)The weight coefficients, θ, of the first-order statistical characteristic and the second-order statistical characteristic, respectivelyk|k-1、Pk|k-1And Hk|k-1One-step prediction of the state quantity, variance and measured value, respectively, KkIn order to filter the gain of the filter,
Figure GDA0002934962680000066
and
Figure GDA0002934962680000067
respectively the final mean value and the variance of the tasteless Kalman filtering;
s4.6: predicting the health state of the system by using a prediction model obtained by a tasteless particle filtering algorithm;
s4.7: determining the remaining life of the multi-sensing system by predicting the health state:
Figure GDA0002934962680000068
where TH is the fault threshold of the system, TRULIs the remaining life of the system and T is the current time.
As a preferable aspect of the method for predicting remaining life of a multi-sensor system of the present invention, wherein: when k ≦ T, let k ═ k +1, then calculate by:
s4.22: gaussian point and weight calculation:
Figure GDA0002934962680000069
s4.23: and (3) time updating:
Figure GDA00029349626800000610
Figure GDA0002934962680000071
Figure GDA0002934962680000072
Figure GDA0002934962680000073
Figure GDA0002934962680000074
s4.24: and (3) updating the measurement value:
Figure GDA0002934962680000075
Figure GDA0002934962680000076
Figure GDA0002934962680000077
Figure GDA0002934962680000078
Figure GDA0002934962680000079
s4.3: calculating the sampling particle and updating the particle weight:
Figure GDA00029349626800000710
Figure GDA00029349626800000711
s4.4: if the number of significant particles is below a given threshold, resampling is performed:
Figure GDA00029349626800000712
Figure GDA00029349626800000713
s4.5: and (3) state estimation:
Figure GDA00029349626800000714
if k > T, then the prediction is ended
Wherein
Figure GDA0002934962680000081
For selected particles, alpha and beta are parameters of the tasteless transform, Wi (m)And Wi (c)The weight coefficients, θ, of the first-order statistical characteristic and the second-order statistical characteristic, respectivelyk|k-1、Pk|k-1And Hk|k-1One-step prediction of the state quantity, variance and measured value, respectively, KkIn order to filter the gain of the filter,
Figure GDA0002934962680000082
and
Figure GDA0002934962680000083
respectively the final mean value and the variance of the tasteless Kalman filtering;
s4.6: predicting the health state of the system by using a prediction model obtained by a tasteless particle filtering algorithm;
s4.7: determining the remaining life of the multi-sensing system by predicting the health state:
Figure GDA0002934962680000084
where TH is the fault threshold of the system, TRULIs the remaining life of the system and T is the current time.
As a preferable aspect of the method for predicting remaining life of a multi-sensor system of the present invention, wherein: firstly, calculating the mean value and covariance of each particle by using the unscented Kalman filter in a sampling stage; and secondly using the mean and covariance to guide sampling.
The invention has the following beneficial effects:
the method for predicting the residual life of the multi-sensor system provides an improved monotonicity trend of second-order permutation entropy index measurement data, and fuses data of a plurality of sensors through a data fusion method to construct a health index with a good degradation trend and a smaller threshold variance to represent the health state of equipment, and then estimates and predicts the health state of the equipment by taking the data of the health index as an observed value of odorless particle filtering, so that the residual life of the multi-sensor system is predicted finally.
Drawings
FIG. 1 is a block diagram of the method of the present invention.
Detailed Description
The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
Example 1
Referring to fig. 1, a method for predicting the remaining life of a multi-sensor system mainly includes representing a health index of an electromechanical device by a data fusion method and predicting the health state of the electromechanical device by a particle filtering method, the data fusion method constructs the health index having a good degradation trend and a smaller threshold variance by fusing data of a plurality of sensors to represent the health state of the device, and when the health index is constructed by the data fusion method, the health index is often allowed to have a larger improved second-order entropy, so that the health index has a good overall monotonic trend, thereby improving the accuracy of predicting the remaining life of the electromechanical device, and the specific implementation steps are as follows:
s1, calculating an improved second-order permutation entropy of sensor data, and constructing a health index by the improved second-order permutation entropy and a data fusion method to have a good integral monotonous trend and a smaller fault threshold error so that the residual life prediction precision is higher;
s1.1: the ith sensor data xiMapping to a 2-dimensional phase space for reconstruction, and obtaining a phase space matrix as shown in the following:
Figure GDA0002934962680000091
a is a reconstructed phase space matrix, B is a submatrix in A, and n is the length of sensor data;
s1.2: for each state vector of the above reconstructed phase space matrix, a symbol sequence is obtained:
S(l)＝[m1,m2],l＝1,2,…,n*(n-1)/2
where l is the state vector index, [ m ]1,m2]There are two possibilities pi for permutation according to the magnitude of the state vector values1＝[0,1]And pi2＝[1,0]When two numerical values in the state vector are equal, take [0, 1%]Comparing the magnitude of the numerical value in each state vector to obtain the arrangement relation of each state vector;
s1.3: calculating the corresponding permutation pi of all state vectors1,π2The probability of (c) is as follows:
Figure GDA0002934962680000101
Figure GDA0002934962680000102
wherein # represents the number;
s1.4: calculating an improved second-order permutation entropy value of the sensor data, wherein the improved second-order permutation entropy calculation formula is as follows:
S2＝abs(p(π1)log2[p(π1)]-p(π2)log2[p(π2)])
wherein abs () is the absolute value;
s2: calculating and comparing improved second-order permutation entropy values of the sensor data according to the calculation method of S1;
s3: selecting improved sensor data with maximum second-order arrangement entropy and good integral monotony trend for data fusion, constructing a health index capable of representing the health state of the multi-sensor system,
s3.1: and (3) constructing a health index at the time t, wherein a health index construction formula is as follows:
Ht＝x·,tω t＝0,1,2,...,ni
where ω is a weight vector, x·,t∈R1×SIs a matrix of measurements collected from the S sensors;
s3.2: solving the above two-target optimization problem by genetic algorithm to obtain the optimal weight vector omega*Ensuring that the health indicator has good overall monotonous trend and smaller fault threshold variance:
Figure GDA0002934962680000103
s.t.ω′1S＝1
wherein S2ω(d) Is an improved second order permutation entropy value of the health index determined by the weight vector omega,
Figure GDA0002934962680000111
is the fault threshold variance of the health index determined by the weight vector ω,1S∈RS×1The vector with the value of 1 is adopted, and the fault threshold variance calculation formula is as follows:
Figure GDA0002934962680000112
wherein Y ∈ RM×SIs a matrix of fault values, M is the number of training units;
s4: the health index is used as an observed value of a particle filter method, the health state prediction of a multi-sensor system is realized, and the particle filter can be widely applied to the prediction of the residual service life of equipment because the particle filter can better solve the state estimation problem of a nonlinear and non-Gaussian system.
According to the working principle, data of a plurality of sensors are fused by a data fusion method, a health index with a good degradation trend and a small threshold variance is constructed to represent the health state of equipment, then the health index is used as an observed value of a particle filtering method, a prediction model obtained by a particle filtering algorithm is used for predicting the health state of a system, the residual life of the electromechanical equipment is finally calculated, and the data fusion method and the particle filtering method are matched with each other for use, so that the prediction precision of the residual life of the electromechanical equipment is further improved.
Example 2
Referring to fig. 1, the difference between the present embodiment and the above embodiment is: the unscented kalman filter is used by the present embodiment to select the proposed distribution for particle filtering, since the particle filtering algorithm uses a prior distribution as the proposed density distribution function, when the coincidence degree of the prior distribution and the posterior distribution is very small, the effect of the particle filter algorithm is poor, so that when the particle filter algorithm is used for predicting the residual effective life of the electromechanical equipment, the prediction precision is correspondingly reduced, the present embodiment therefore employs unscented kalman filtering to select the proposed distribution for particle filtering, first in the sampling phase, with unscented kalman filtering to calculate for each particle its mean and covariance, secondly, the mean value and the covariance are used for guiding sampling, and the latest observation information is utilized when the mean value and the covariance are calculated by the tasteless Kalman filtering, it is closer to the posterior distribution and thus improves the accuracy of the remaining life of the computer electronics.
The method for predicting the residual life of the multi-sensor system mainly comprises the steps of representing a health index of electromechanical equipment by adopting a data fusion method and realizing the prediction of the health state of the electromechanical equipment by adopting a tasteless particle filtering method, constructing the health index with a good degradation trend and a smaller threshold variance by fusing data of a plurality of sensors by the data fusion method to represent the health state of the equipment, and generally enabling the health index to have a larger improved second-order permutation entropy when constructing the health index by the data fusion method to enable the health index to have a good integral monotonous trend, so that the prediction precision of the residual life of the electromechanical equipment is improved, and the method comprises the following specific implementation steps:
s1, calculating an improved second-order permutation entropy of sensor data, and constructing a health index by the improved second-order permutation entropy and a data fusion method to have a good integral monotonous trend and a smaller fault threshold error so that the residual life prediction precision is higher;
s1.1: the ith sensor data xiMapping to a 2-dimensional phase space for reconstruction, and obtaining a phase space matrix as shown in the following:
Figure GDA0002934962680000121
a is a reconstructed phase space matrix, B is a submatrix in A, and n is the length of sensor data;
s1.2: for each state vector of the above reconstructed phase space matrix, a symbol sequence is obtained:
S(l)＝[m1,m2],l＝1,2,...,n*(n-1)/2
where l is the state vector index, [ m ]1,m2]There are two possibilities pi for permutation according to the magnitude of the state vector values1＝[0,1]And pi2＝[1,0]When two numerical values in the state vector are equal, take [0, 1%]Comparing the magnitude of the numerical value in each state vector to obtain the arrangement relation of each state vector;
s1.3: calculating the corresponding permutation pi of all state vectors1,π2The probability of (c) is as follows:
Figure GDA0002934962680000131
Figure GDA0002934962680000132
wherein # represents the number;
s1.4: calculating an improved second-order permutation entropy value of the sensor data, wherein the improved second-order permutation entropy calculation formula is as follows:
S2＝abs(p(π1)log2[p(π1)]-p(π2)log2[p(π2)])
wherein abs () is the absolute value;
s2: calculating and comparing improved second-order permutation entropy values of the sensor data according to the calculation method of S1;
s3: selecting improved sensor data with maximum second-order arrangement entropy and good integral monotony trend for data fusion, constructing a health index capable of representing the health state of the multi-sensor system,
s3.1: and (3) constructing a health index at the time t, wherein a health index construction formula is as follows:
Ht＝x·,tω t＝0,1,2,...,ni
where ω is a weight vector, x·,t∈R1×SIs a matrix of measurements collected from the S sensors;
s3.2: solving the above two-target optimization problem by genetic algorithm to obtain the optimal weight vector omega*Ensuring that the health indicator has good overall monotonous trend and smaller fault threshold variance:
Figure GDA0002934962680000133
s.t.ω′1S＝1
wherein S2ω(d) Is an improved second order permutation entropy value of the health index determined by the weight vector omega,
Figure GDA0002934962680000134
is the fault threshold variance of the health index determined by the weight vector ω,1S∈RS×1The vector with the value of 1 is adopted, and the fault threshold variance calculation formula is as follows:
Figure GDA0002934962680000141
wherein Y ∈ RM×SIs a matrix of fault values, M is the number of training units;
s4: the health index is used as an observed value of a particle filtering method to realize the health state prediction of a multi-sensor system, the particle filtering method specifically comprises the following steps, the particle filtering is widely applied and used for predicting the residual service life of equipment because the particle filtering can better solve the state estimation problem of a nonlinear and non-Gaussian system,
s4.1: constructing a state space model of the multi-sensing system:
θk＝f(θk-1,υk)
Hk＝h(θk,νk)
wherein θ is a state vector, f is a nonlinear function, upsilon is a process variance, H is a health index, H is a nonlinear observation function, and v is a measurement variance;
s4.2: generating an important density function of the particle filter by using unscented Kalman filtering;
the particle filter algorithm adopts prior distribution as a suggested density distribution function, and when the coincidence degree of the prior distribution and the posterior distribution is very small, the particle filter effect is poor, so that the suggested distribution of the particle filter is selected by adopting tasteless Kalman filtering, particularly, in the sampling stage, the mean value and the covariance of each particle are calculated by using the tasteless Kalman filtering, and then the sampling is guided by using the mean value and the covariance, and the latest observation information is utilized when the mean value and the covariance are calculated by using the tasteless Kalman filtering, so that the particle filter algorithm is closer to the posterior distribution;
s4.21: first, k is set to 0, and N is extracted from the prior distributionsParticles and determining the weight to be 1/NsThen, initializing parameters:
Figure GDA0002934962680000142
Figure GDA0002934962680000143
augmented initialization state vector and covariance matrix:
Figure GDA0002934962680000151
Figure GDA0002934962680000152
s4.22: gaussian point and weight calculation:
Figure GDA0002934962680000153
s4.23: and (3) time updating:
Figure GDA0002934962680000154
Figure GDA0002934962680000155
Figure GDA0002934962680000156
Figure GDA0002934962680000157
Figure GDA0002934962680000158
s4.24: and (3) updating the measurement value:
Figure GDA0002934962680000159
Figure GDA00029349626800001510
Figure GDA00029349626800001511
Figure GDA00029349626800001512
Figure GDA00029349626800001513
s4.3: calculating the sampling particle and updating the particle weight:
Figure GDA00029349626800001514
Figure GDA0002934962680000161
s4.4: if the number of significant particles is below a given threshold, resampling is performed:
Figure GDA0002934962680000162
Figure GDA0002934962680000163
s4.5: and (3) state estimation:
Figure GDA0002934962680000164
if k is less than or equal to T, making k equal to k +1, and returning to step S4.22, otherwise, ending prediction;
wherein
Figure GDA0002934962680000165
For selected particles, alpha and beta are parameters of the tasteless transform, Wi (m)And Wi (c)The weight coefficients, θ, of the first-order statistical characteristic and the second-order statistical characteristic, respectivelyk|k-1、Pk|k-1And Hk|k-1One-step prediction of the state quantity, variance and measured value, respectively, KkIn order to filter the gain of the filter,
Figure GDA0002934962680000166
and
Figure GDA0002934962680000167
respectively the final mean value and the variance of the tasteless Kalman filtering;
s4.6: predicting the health state of the system by using a prediction model obtained by a tasteless particle filtering algorithm;
s4.7: determining the remaining life of the multi-sensing system by predicting the health state:
Figure GDA0002934962680000168
where TH is the fault threshold of the system, TRULIs the remaining life of the system and T is the current time.
The working principle of the embodiment is as follows: the method comprises the steps of firstly fusing data of a plurality of sensors by adopting a data fusion method, constructing a health index with good degradation trend and smaller threshold variance by adopting an improved second-order permutation entropy and data fusion method to represent the health state of equipment, then taking the health index as an observed value of an odorless particle filtering method, obtaining a prediction model by adopting an odorless particle filtering algorithm and predicting the health state of a system, and further improving the accuracy of predicting the residual life of the electromechanical equipment by mutually matching the data fusion method and the particle filtering method.
It is noted that, herein, relational terms such as first and second, and the like may be used solely to distinguish one entity or action from another entity or action without necessarily requiring or implying any actual such relationship or order between such entities or actions. Also, the terms "comprises," "comprising," or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements does not include only those elements but may include other elements not expressly listed or inherent to such process, method, article, or apparatus.
Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that changes, modifications, substitutions and alterations can be made in these embodiments without departing from the principles and spirit of the invention, the scope of which is defined in the appended claims and their equivalents.

Claims (5)

1. A method for predicting the remaining life of a multi-sensor system is characterized in that: the specific implementation steps are as follows:
s1 calculating improved second order permutation entropy of the sensor data;
s2: calculating improved second-order permutation entropy values of the sensor data according to the calculation method of S1 and comparing;
s3: selecting an improved second-order permutation entropy value, using data of the sensor for data fusion, and constructing a health index capable of representing the health state of the multi-sensor system;
s4: the health index is used as the observed value of the particle filtering method to realize the health state prediction of the multi-sensor system,
wherein: the calculating of the improved second order permutation entropy of the sensor data comprises the steps of:
s1.1: the ith sensor data xiMapping to a 2-dimensional phase space for reconstruction, and obtaining a phase space matrix as shown in the following:
Figure FDA0002934962670000011
a is a reconstructed phase space matrix, B is a submatrix in A, and n is the length of sensor data;
s1.2: for each state vector of the reconstructed phase-space matrix, a symbol sequence is obtained:
S(l)＝[m1,m2],l＝1,2,…,n*(n-1)/2
where l is the state vector index, [ m ]1,m2]There are two possibilities pi for permutation according to the magnitude of the state vector values1＝[0,1]And pi2＝[1,0]When two numerical values in the state vector are equal, take [0, 1%]Comparing the magnitude of the numerical value in each state vector to obtain the arrangement relation of each state vector;
s1.3: calculating the corresponding permutation pi of all state vectors1,π2The probability of (d);
s1.4: a second order rank entropy value of the sensor data improvement is calculated.
2. The remaining life prediction method of a multi-sensor system according to claim 1, characterized in that: calculating the corresponding arrangement pi of all state vectors1,π2The calculation formula used for the probability is as follows:
Figure FDA0002934962670000021
Figure FDA0002934962670000022
wherein # represents the number.
3. The remaining life prediction method of a multi-sensor system according to claim 2, characterized in that: the calculation formula for calculating the improved second order entropy value of the sensor data is as follows:
S2＝abs(p(π1)log2[p(π1)]-p(π2)log2[p(π2)])
where abs () is the absolute value.
4. The remaining life prediction method of a multi-sensor system according to claim 3, characterized in that: the method for constructing the health index capable of representing the health state of the multi-sensor system specifically comprises the following steps:
s3.1: constructing a health index at the time t;
s3.2: solving the dual-target optimization problem through a genetic algorithm to obtain an optimal weight vector omega*And calculating the integral monotonous trend and the fault threshold variance of the health index.
5. The remaining life prediction method of a multi-sensor system according to claim 4, characterized in that: firstly, using tasteless Kalman filtering to generate an important density function of particle filtering, and calculating the mean value and covariance of each particle by using the tasteless Kalman filtering in a sampling stage; and secondly using the mean and covariance to guide sampling.
CN201910750989.0A 2019-08-14 2019-08-14 Residual life prediction method of multi-sensor system Active CN110456199B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201910750989.0A CN110456199B (en) 2019-08-14 2019-08-14 Residual life prediction method of multi-sensor system

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201910750989.0A CN110456199B (en) 2019-08-14 2019-08-14 Residual life prediction method of multi-sensor system

Publications (2)

Publication Number Publication Date
CN110456199A CN110456199A (en) 2019-11-15
CN110456199B true CN110456199B (en) 2021-03-23

Family

ID=68486659

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201910750989.0A Active CN110456199B (en) 2019-08-14 2019-08-14 Residual life prediction method of multi-sensor system

Country Status (1)

Country Link
CN (1) CN110456199B (en)

Families Citing this family (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN111222290B (en) * 2020-01-13 2024-04-09 浙江工业大学 Multi-parameter feature fusion-based method for predicting residual service life of large-scale equipment
CN111553590B (en) * 2020-04-27 2021-09-24 中国电子科技集团公司第十四研究所 Radar embedded health management system
CN111751508A (en) * 2020-05-12 2020-10-09 北京华科仪科技股份有限公司 Performance evaluation prediction method and system for life cycle of water quality sensor
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
CN112836381B (en) * 2021-02-19 2023-03-14 震兑工业智能科技有限公司 Multi-source information-based ship residual life prediction method and system
CN113627088B (en) * 2021-08-23 2024-04-09 上海交通大学 Machine performance degradation evaluation method and system based on gene programming and data fusion

Family Cites Families (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN104792529A (en) * 2015-04-12 2015-07-22 北京化工大学 Rolling bearing life prediction method based on state-space model
CN106845866A (en) * 2017-02-27 2017-06-13 四川大学 Equipment method for predicting residual useful life based on improved particle filter algorithm
CN109376401B (en) * 2018-09-29 2022-12-09 西安交通大学 Self-adaptive multi-source information optimization and fusion mechanical residual life prediction method
CN109948860A (en) * 2019-03-26 2019-06-28 哈工大机器人（合肥）国际创新研究院 A kind of mechanical system method for predicting residual useful life and system

Also Published As

Publication number Publication date
CN110456199A (en) 2019-11-15

Similar Documents

Publication Publication Date Title
CN110456199B (en) Residual life prediction method of multi-sensor system
CN112380098B (en) Timing sequence abnormality detection method and device, computer equipment and storage medium
CN107145645B (en) Method for predicting residual life of non-stationary degradation process with uncertain impact
CN107765347B (en) Short-term wind speed prediction method based on Gaussian process regression and particle filtering
CN111694879B (en) Multielement time sequence abnormal mode prediction method and data acquisition monitoring device
CN113568819B (en) Abnormal data detection method, device, computer readable medium and electronic equipment
Liu et al. A novel method for online health prognosis of equipment based on hidden semi-Markov model using sequential Monte Carlo methods
CN111368403A (en) Self-adaptive non-linear degradation residual life prediction method
CN112712209B (en) Reservoir warehousing flow prediction method and device, computer equipment and storage medium
CN112966432A (en) Method and device for predicting remaining effective life of lithium ion battery
CN106170708B (en) Individual electric equipment working condition estimation unit and its method
Kulikov et al. Square‐root Kalman‐like filters for estimation of stiff continuous‐time stochastic systems with ill‐conditioned measurements
CN103714045A (en) Information fusion estimation method for asynchronous multi-rate non-uniform sampled observation data
CN112949026A (en) Age and state dependence considered degradation equipment residual life prediction method
CN112668238B (en) Rainfall processing method, rainfall processing device, rainfall processing equipment and storage medium
Wei et al. Forecasting the geomagnetic activity of the Dst index using multiscale radial basis function networks
CN111007454A (en) Extended target tracking method based on cooperative target information
CN115469260A (en) Hausdorff-based current transformer abnormity identification method and system
CN111523727B (en) Method for predicting remaining life of battery by considering recovery effect based on uncertain process
CN117407771A (en) Bearing health state assessment method and device based on digital twin and related equipment
CN115730845A (en) Power system dynamic estimation method, device, equipment and readable storage medium
CN116539994A (en) Substation main equipment operation state detection method based on multi-source time sequence data
Shao et al. Remaining useful life prediction via a hybrid DBN-KF-based method: a case of subsea Christmas tree valves
Hu et al. Tracking multiple extended targets with multi‐Bernoulli filter
CN117910351A (en) Method for predicting residual life of digital-analog linkage random degradation tool

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