CN112729878A - Method for evaluating health state of CRH380 type running gear system - Google Patents

Method for evaluating health state of CRH380 type running gear system Download PDF

Info

Publication number
CN112729878A
CN112729878A CN202011187692.7A CN202011187692A CN112729878A CN 112729878 A CN112729878 A CN 112729878A CN 202011187692 A CN202011187692 A CN 202011187692A CN 112729878 A CN112729878 A CN 112729878A
Authority
CN
China
Prior art keywords
reliability
confidence
uncertainty
rule
running gear
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
CN202011187692.7A
Other languages
Chinese (zh)
Inventor
程超
王久赫
王艳
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.)
Changchun University of Technology
Original Assignee
Changchun University of Technology
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 Changchun University of Technology filed Critical Changchun University of Technology
Priority to CN202011187692.7A priority Critical patent/CN112729878A/en
Publication of CN112729878A publication Critical patent/CN112729878A/en
Pending legal-status Critical Current

Links

Images

Classifications

    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01M—TESTING STATIC OR DYNAMIC BALANCE OF MACHINES OR STRUCTURES; TESTING OF STRUCTURES OR APPARATUS, NOT OTHERWISE PROVIDED FOR
    • G01M17/00—Testing of vehicles
    • G01M17/08—Railway vehicles
    • 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
    • G06F30/27—Design optimisation, verification or simulation using machine learning, e.g. artificial intelligence, neural networks, support vector machines [SVM] or training a model
    • 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
    • G06F2119/02—Reliability analysis or reliability optimisation; Failure analysis, e.g. worst case scenario performance, failure mode and effects analysis [FMEA]

Landscapes

  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Evolutionary Computation (AREA)
  • General Physics & Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Medical Informatics (AREA)
  • Software Systems (AREA)
  • Artificial Intelligence (AREA)
  • Computer Hardware Design (AREA)
  • Geometry (AREA)
  • General Engineering & Computer Science (AREA)
  • Testing Of Devices, Machine Parts, Or Other Structures Thereof (AREA)

Abstract

The invention discloses a CRH380 type running gear system health state evaluation method based on a confidence rule base, and particularly relates to the field of reliability of large complex systems. The evaluation method comprises the following steps: collecting measurement data of a CRH380 type running gear system in different health states as training data, giving out initial parameters of a library by a domain expert, and establishing a corresponding training model; monitoring and environmental uncertainty existing in the quantitative evaluation process according to the measured data; obtaining the corrected attribute reliability based on an evidence discount method; quantifying cognitive uncertainty according to prior information, and constructing an expert reliability model; collecting measurement data of a CRH380 type running gear system under real-time working conditions as test data; and (4) integrating the quantized three-layer uncertainty factors into a confidence rule base, fitting the confidence rule base with the real health state threshold of the system, and analyzing the evaluation result of the test data. The invention does not need to know the mathematical model of the complex system and can effectively evaluate the health state of the large complex equipment.

Description

Method for evaluating health state of CRH380 type running gear system
Technical Field
The invention belongs to the field of reliability, and particularly relates to a health state method of a running gear system of a high-speed train.
Background
In the method of evaluating the health status of the running gear of the high-speed train, there can be distinguished (i) a method based on data driving; (ii) a model-based approach; (iii) a knowledge-based method. The health state evolution is known from the time sequence of the observation system based on the data driving method, and an information-rich result is provided for the health state evaluation of the high-speed train running part. However, the process of solving is not transparent, i.e., there is a lack of physical explanation for the changes to the system; it is necessary to know the appropriate failure threshold or health level, which may be difficult to achieve in some systems, especially given the limited knowledge of the system mechanics, and the limited ability to assess the health of the high speed train running gear.
Model-based methods assess the health of the system by analyzing residuals, not by setting a health level or failure threshold. In fact, the method is closely combined with a system mechanism model, and the state monitoring of the system can be conveniently realized. But the performance is often poor due to lack of accurate analytic expressions of the system; knowledge-based modeling methods are good at finding causal relationships between local faults and system anomalies, and are therefore often used in engineering. However, with the development of scientific technology and the continuous breakthrough of scientific problems in other fields, the integration level of the running part of the high-speed train is higher and higher, the complex coupling relation is more and more difficult to distinguish, and the complete and accurate knowledge of the system is difficult to obtain; furthermore, for existing assessment methods, it is crucial to assess the expected mismatch between the true health status of the system and the expected outcome. For this reason, all sources of uncertainty that affect the health status assessment must be considered:
(1) monitoring uncertainty affected by the performance of the sensor itself. For example, a decrease in monitoring performance due to microstructural differences between homologous sensors, or due to unforeseen loads in the future, or the like;
(2) environmental uncertainty affected by the operating environment and external conditions;
(3) the high-speed train running part is subjected to model building in the evaluation process and cognitive uncertainty caused by the imperfection of parameter knowledge;
in summary, there is a need to develop a health status assessment method that can effectively combine quantitative information and qualitative knowledge, and consider the influence of three layers of uncertainty in the assessment process, so as to perform online monitoring on large complex systems such as a high-speed train running gear, determine the health status of the running gear system in time, and provide support for later maintenance.
Disclosure of Invention
A design method for evaluating the health status of a CRH 380-type running gear system, comprising:
acquiring measurement data of a CRH380 type running gear system in different health states as training data, giving prior information of initial parameters in a confidence rule base by a domain expert, and establishing a training model of the initial parameters;
step two, according to the monitoring uncertainty and the environment uncertainty (both are related to the sensor and are divided into static factors and dynamic factors) existing in the quantitative evaluation process of the measured data;
thirdly, fusing the quantified static and dynamic factors based on an evidence discount method to obtain the corrected attribute reliability;
quantifying the cognitive uncertainty factors of the system according to the initial expert experience, and constructing an expert reliability model;
step five, collecting measurement data of the CRH380 type running gear system under real-time working conditions as test data;
and step six, the quantized three-layer uncertainty factors are fused into a confidence rule base evaluation model, and are fitted with a system real health state threshold value, and evaluation results of the test data are analyzed.
Preferably, in the step one, the measurement data of the CRH380 running gear system in different health states comprise two physical characteristic data of temperature and vibration; the initial value of the confidence rule base given by the domain expert is selected as shown in equation (1). Suppose there is a k rule in the model to represent the method as follows:
Figure RE-GDA0002850574280000021
wherein, x represents the input vector,
Figure RE-GDA0002850574280000022
denotes the input reference value, H, of the m-th attribute in the k-th rulen(N-1, 2, …, N) represents the nth evaluation scale, βn,kIndicates the evaluation level H relative to the nthnConfidence of (a), thetakThe rule weight representing the kth rule,
Figure RE-GDA0002850574280000023
a weight value representing the mth premise attribute, κ represents a discount operator,
Figure RE-GDA0002850574280000024
the range of tolerance is indicated and,
Figure RE-GDA0002850574280000025
rmrespectively expressed as the static reliability, dynamic reliability and attribute reliability of the mth sensor,
Figure RE-GDA0002850574280000026
indicating the expert reliability of the kth rule. If it is not
Figure RE-GDA0002850574280000027
The kth rule is complete, otherwise it is incomplete. In particular, it is possible to use, for example,
Figure RE-GDA0002850574280000028
indicating that the output of the kth rule is completely unknown.
Preferably, in the second step, the measured data in different health status levels of the system under the actual working condition is not only affected by the monitoring uncertainty, but also affected by the environmental uncertainty; in addition, the method is also influenced by the cognitive uncertainty of the system, and the traditional evaluation method is difficult to be applied to large-scale complex electromechanical equipment such as a high-speed train; equations (2) - (4) are selected as reliability factors after quantifying monitoring uncertainty:
Figure RE-GDA0002850574280000029
Figure RE-GDA00028505742800000210
Figure RE-GDA00028505742800000211
wherein x isij(i-1, …, M; j-1, …, N) is the j-th measurement data acquired by the i-th sensor,
Figure RE-GDA00028505742800000212
is that
Figure RE-GDA00028505742800000213
Represents the average of the N monitoring points.
Figure RE-GDA00028505742800000214
Representing all index data xijAnd the average value
Figure RE-GDA00028505742800000215
Measure of similarity therebetween, and ri sIs the reliability factor after the quantization of the ith monitoring uncertainty.
Similarly, equation (5) is selected as the reliability factor after quantifying the environmental uncertainty:
Figure RE-GDA00028505742800000217
wherein,
Figure RE-GDA00028505742800000319
for observation data xijHas a standard deviation of
Figure RE-GDA0002850574280000031
Figure RE-GDA0002850574280000032
Coefficients representing the adjustment tolerance range. If it is not
Figure RE-GDA0002850574280000033
Or
Figure RE-GDA0002850574280000034
The observed data is not reliable and,
Figure RE-GDA0002850574280000035
if not, then,
Figure RE-GDA0002850574280000036
preferably, in step three, the selective equation (6) constructs a discount method based on the factor coordination fusion after the monitoring uncertainty and the environmental uncertainty are quantified, and the corrected attribute reliability is obtained.
ri＝f(ri s,ri d,κ) (6)
Wherein r isi sAnd ri dRespectively representing the static reliability and the dynamic reliability of the observation information of the ith sensor, riRepresenting the degree of reliability of the observation information of the ith sensor, and f (-) representing riAnd ri s、ri dK denotes the discount operator. Because the initial tolerance coefficient is given by an expert and cannot be well adapted to the change of the external environment, the achievement uses the tolerance coefficient
Figure RE-GDA0002850574280000037
And (4) the parameters are included in the parameter vector of the model and are trained.
Preferably, in step four, the formula (7) to the formula (15) are selected to quantify the cognitive uncertainty factor of the system according to the initial expert experience, and an expert reliability model is constructed:
Figure RE-GDA0002850574280000038
wherein
Figure RE-GDA0002850574280000039
And
Figure RE-GDA00028505742800000310
respectively representing the vector form of the expert description rule confidence allocation,Dis shown in (2)Θ×2ΘIn the form of a matrix of (a),
Figure RE-GDA00028505742800000318
the elements contained in (1) are defined as:
Figure RE-GDA00028505742800000312
since the confidence rules are connected by a logical or,Dis of a scale of 2Θ×2ΘThe identity matrix of (2). When there are multiple confidence rules, the distance of every two confidence rules can be represented in the form of a distance matrix DM, as follows:
Figure RE-GDA00028505742800000313
Sim(Ri,Rj)＝1-d(Ri,Rj) (10)
Figure RE-GDA00028505742800000314
the support for each confidence rule is as follows:
Figure RE-GDA00028505742800000315
Figure RE-GDA00028505742800000316
wherein, CrdkThe confidence level of the kth confidence rule is shown, and the larger Crd indicates that the confidence rule is supported by other experts.
Then introducing a Dun entropy calculation method of each confidence rule, and expressing the method as follows:
Figure RE-GDA00028505742800000317
and finally obtaining a calculation method of expert reliability:
Figure RE-GDA0002850574280000041
preferably, the step five is specifically: an online test sample is obtained in real time.
Preferably, the sixth step is specifically: and (3) the quantized three-layer uncertainty factors are fused into a confidence rule base evaluation model, a formula (16) is selected to evaluate the real-time state of the model, the model is fitted with a real health state threshold value of the system, and the evaluation result of the test data is analyzed.
Figure RE-GDA0002850574280000042
Wherein beta isnShows the nth evaluation result HnConfidence of (c), H { (H)n,βn) N-1, 2, …, N represents the overall confidence distribution after the overall confidence rules are integrated.
The invention has the following beneficial effects:
the method for evaluating the health state of the walking part system based on the multi-fold letter rule base utilizes semi-quantitative information to model the measurement data of the system in different states, simultaneously considers the influence of three layers of uncertainty existing in engineering practice, does not need to acquire an accurate analysis model of complex equipment, does not need to preprocess the measurement data, and is convenient for real-time monitoring; meanwhile, the health state evaluation of the CRH380 type running gear system under the actual working condition is realized.
Drawings
The technical solution of the present application is further illustrated in the form of the accompanying drawings and forms a part of the specification. The drawings for expressing the embodiments of the application will explain the technical solutions of the application together with the embodiments.
FIG. 1 is a flow chart illustrating a method for evaluating health status of a CRH380 running gear system according to an embodiment of the present invention;
FIG. 2 is a flow chart illustrating the modeling of the multi-fold trust rule base according to the present invention;
FIG. 3 is a flow chart illustrating attribute reliability discounting in a route according to the present technique;
FIG. 4 is a flow chart illustrating a process for optimizing parameters in the context of the present technology;
FIG. 5 is a diagram illustrating a health evaluation result of a running gear system according to an embodiment of the present invention;
FIG. 6 is a diagram illustrating the evaluation result of the health status of the running gear system according to an embodiment of the invention in FIG. 2.
Detailed Description
To realize the health state evaluation of a complex system such as a high-speed train running part, the influence of three layers of uncertainty factors in the evaluation process needs to be comprehensively considered. The invention proposes a multi-discount-based confidence rule base system health state evaluation model to realize the health state evaluation of the system.
Step one (S110), collecting the measurement data of the CRH380 type running gear system in different health states as training data, giving prior information of initial parameters in a confidence rule base by a domain expert, and establishing a training model of the initial parameters.
Specifically, the measurement data of the CRH380 running gear system in different health states comprise two physical characteristic data of temperature and vibration; the initial value of the confidence rule base given by the domain expert is selected as shown in equation (1). Suppose there is a k rule in the model to represent the method as follows:
Figure RE-GDA0002850574280000051
wherein, x represents the input vector,
Figure RE-GDA0002850574280000052
denotes the input reference value, H, of the m-th attribute in the k-th rulen(N-1, 2, …, N) represents the nth evaluation scale, βn,kIndicates the evaluation level H relative to the nthnConfidence of (a), thetakThe rule weight representing the kth rule,
Figure RE-GDA0002850574280000053
a weight value representing the mth premise attribute, κ represents a discount operator,
Figure RE-GDA0002850574280000054
the range of tolerance is indicated and,
Figure RE-GDA0002850574280000055
rmrespectively expressed as the static reliability, dynamic reliability and attribute reliability of the mth sensor,
Figure RE-GDA0002850574280000056
indicating the expert reliability of the kth rule. If it is not
Figure RE-GDA0002850574280000057
The kth rule is complete, otherwise it is incomplete. In particular, it is possible to use, for example,
Figure RE-GDA0002850574280000058
indicating that the output of the kth rule is completely unknown.
And step two (S120), the monitoring uncertainty and the environment uncertainty existing in the evaluation process are quantified according to the measured data.
In particular, the monitoring uncertainty in the pre-knowledge evaluation process is mainly generated by the static factors of the sensors. Thus, static factors may be intended to quantify static reliability through a minimum distance-based approach. Static reliability for describing sensingThe quality of the device itself (also called evidence), and the difference between the evidences can be used to determine the difference degree of the observed information between different sensors, assuming xij(i-1, …, M; j-1, …, N) is the j-th observed information acquired by the i-th sensor,
Figure RE-GDA0002850574280000059
represents the average of the N monitoring points. Then, xijAnd
Figure RE-GDA00028505742800000510
is a distance of
Figure RE-GDA00028505742800000511
And all index data x in N monitoring pointsijAnd the average value
Figure RE-GDA00028505742800000512
Measure the similarity between them
Figure RE-GDA00028505742800000513
Thus, the static reliability as defined by this work is as follows:
Figure RE-GDA00028505742800000514
the dynamic reliability is used for describing that the sensor is interfered by external uncertain factors (such as poor contact, environmental noise and the like) in the operation process of the sensor, so that the information obtained by the sensor is inaccurate and even leads to errors.
Figure RE-GDA00028505742800000515
For observation data xijHas a standard deviation of
Figure RE-GDA00028505742800000516
Figure RE-GDA00028505742800000517
Coefficients representing the adjustment tolerance range. The different tolerance ranges depend on the system. After determining the tolerance range, unreliable data can be calculated, if
Figure RE-GDA00028505742800000518
Or
Figure RE-GDA00028505742800000519
The observed data is not reliable and,
Figure RE-GDA00028505742800000520
if not, then,
Figure RE-GDA00028505742800000521
Figure RE-GDA00028505742800000522
because the initial tolerance coefficient is given by an expert and cannot be well adapted to the change of the external environment, the achievement uses the tolerance coefficient
Figure RE-GDA00028505742800000523
And (4) the parameters are included in the parameter vector of the model and are trained.
And step three (S130), fusing the quantified static and dynamic factors based on an evidence discount method to obtain the corrected attribute reliability.
Specifically, the attribute reliability is determined by both static factors and dynamic factors, and based on the method, an attribute reliability discount method based on the coordination and fusion of the static reliability and the dynamic reliability is established, and the calculation of the attribute reliability includes more interference factors, specifically as follows:
ri＝f(ri s,ri d,κ) (5)
wherein r isi sAnd ri dRespectively representing the static reliability and the dynamic reliability of the observation information of the ith sensor, riRepresenting the degree of reliability of the observation information of the ith sensor, and f (-) representing riAnd ri s、ri dK denotes the discount operator.
The attribute reliability is shown in fig. 3. The attribute reliability factors after coordination and fusion are used for counteracting the influence caused by various interference factors, and the expression of experts in the confidence rule base on knowledge is favorably enhanced, so that the overall robustness of the model is improved, the uncertainty of the system is reduced, and the evaluation capability of the model is enhanced.
Step four (S140), quantifying the cognitive uncertainty factors of the system according to the initial expert experience, and constructing an expert reliability model.
Specifically, the rules described by the experts are expressed in the form of vector space, and the distance between two rules in the confidence rule base is defined as:
Figure RE-GDA0002850574280000061
wherein
Figure RE-GDA0002850574280000062
And
Figure RE-GDA0002850574280000063
respectively representing the vector form of the expert description rule confidence allocation,Dis shown in (2)Θ×2ΘIn the form of a matrix of (a),
Figure RE-GDA0002850574280000067
the elements contained in (A) are defined as
Figure RE-GDA0002850574280000065
Since the confidence rules are connected by a logical or,Dis of a scale of 2Θ×2ΘThe identity matrix of (2). When there are multiple confidence rules, the distance of every two confidence rules can be represented in the form of a distance matrix DM, as follows:
Figure RE-GDA0002850574280000066
the final decision of the confidence rule base is realized by aggregating the activated rules, and it is noted that in the process, the aggregated rules are the input end of knowledge inference in the confidence rule base, and the meaning of the attribute is similar. In general, the higher the degree to which an activated rule is supported by other rules, the more reliable the rule is. To this end, a similarity measure Sim is constructed by constructing each rulei,jAnd confidence level CrdkTo measure the reliability of the rule:
Sim(Ri,Rj)＝1-d(Ri,Rj) (9)
accordingly, a similarity metric matrix is derived, represented as follows:
Figure RE-GDA0002850574280000071
the support for each confidence rule is as follows:
Figure RE-GDA0002850574280000072
Figure RE-GDA0002850574280000073
wherein, CrdkThe confidence level of the kth confidence rule is shown, and the larger Crd indicates that the confidence rule is supported by other experts.
Then introducing a Dun entropy calculation method of each confidence rule, and expressing the method as follows:
Figure RE-GDA0002850574280000074
the information quantity of the confidence rules is measured by utilizing the Dun entropy, and if a certain confidence rule contains a large amount of correct information, the confidence rule should be fully supported by other experts, and the confidence rule is distributed with a great component, so that a method for calculating the reliability of the experts is finally obtained:
Figure RE-GDA0002850574280000075
by reasonably quantifying the unreliable part of the initial expert experience (confidence, rule weight, premise attribute weight) and integrating the expert reliability factor into the reasoning part of the confidence rule base knowledge:
Figure RE-GDA0002850574280000076
wherein beta isnShows the nth evaluation result HnConfidence of (c), H { (H)n,βn) N-1, 2, …, N represents the overall confidence distribution after the overall confidence rules are integrated. The parameter feedback optimization process is shown in fig. 4, and the parameter feedback optimization method integrating the expert reliability realizes the readjustment of the parameters, corrects the error brought by the unreliable data to the forward parameter training, and further improves the modeling precision of the model.
And step five (S150), collecting the measurement data of the CRH380 type running gear system under the real-time working condition as test data. The process is similar to the step one (S110) of acquiring training data.
Step six (S160), the quantified three layers of uncertainty factors are merged into a confidence rule base evaluation model, and are fitted with a system real health state threshold value, and evaluation results of the test data are analyzed.
Specifically, a confidence rule base system state evaluation method model based on multi-order discount is shown in fig. 3. Since the different information presentation forms may be different, the information provided by the sensors is first converted into the same form by observing the data
Figure RE-GDA0002850574280000077
Establishing attribute reference values
Figure RE-GDA0002850574280000078
The matching degree of (b) is specifically expressed as follows:
Figure RE-GDA0002850574280000079
wherein
Figure RE-GDA0002850574280000081
Observation data, x, representing the ith prerequisite AttributeikAnd xi(k+1)Respectively represent the reference level of the ith precondition attribute in the kth rule and the (k + 1) th rule, | xiAnd | represents the number of rules contained in the ith premise attribute.
And correspondingly giving a calculation method of reliability discount in the model provided by the result according to an evidence reliability discount process in evidence reasoning. Combining the premise attribute weight and the attribute reliability as follows:
Figure RE-GDA0002850574280000082
wherein
Figure RE-GDA0002850574280000083
And
Figure RE-GDA0002850574280000084
Ykindicating the number of attributes in the kth rule.
And using the factor combining the weight of the precondition attribute and the reliability of the attribute in the calculation of the matching degree of the kth rule, and finishing the first discount of the model.
Figure RE-GDA0002850574280000085
After some rules are activated in the confidence rule base, the expert reliability is integrated in the knowledge reasoning process of the model, and then the second discount of the model is completed. And finally obtaining the expected utility of the final health state according to the statistical utility:
Figure RE-GDA0002850574280000086
by utilizing the sequential discount process of attribute reliability-expert reliability, the problem that the evaluation process is influenced by objective unreliable information when the traditional confidence rule base model is applied to engineering practice is well solved. By introducing the thought of sequential discount for a plurality of times, the modeling precision of the confidence rule base can be improved, and the health state of the system can be effectively evaluated.
The invention is explained below by means of a Matlab tool with respect to actual operating condition data of a CRH380 type running gear system, and the effects of the invention are shown in combination with diagrams.
(1) Selecting the time from the time when the walking part system changes parts to the time when the fault occurs as a training data set and a calibration data set according to the fault log;
the temperature and vibration characteristics of the running gear system are respectively collected under different health states to serve as a training data set and a testing data set, and each data set comprises 150 groups of data (the system data are averaged by days). The training part is recorded as X1Similarly, the test portion is denoted X2。
(2) Based on training set X1And implementing the multi-discount trust rule base model. By giving a rule R as in formula (1)kConstructing an initial rule base;
in this example, there are 25 rules in the initial rule base and the dimension of the measurement sample is 2.
(3) Calculating and quantifying three layers of uncertainty factors existing in the evaluation process;
according to training set X1And initial expert experience, calculating r in the rule base by the expression (1-14)i s,ri dAnd
Figure RE-GDA0002850574280000087
corresponding to factors after quantitative monitoring, environmental and expert uncertainty, respectively.
(4) Training initial parameters of a multi-fold letter configuration rule base;
by giving an objective function based on minimum mean square error and utilizing a projection covariance matrix adaptive strategy method to train initial parameters in a model, the structure of the model becomes more compact.
(5) Collecting measurement data under real-time working conditions of different health states as test data;
it is noted that the measured variables of the test data are consistent with the measured variables in step (1) of this example, and are also two-dimensional measured data of temperature and vibration characteristics.
(6) And testing the test data through the trained multi-discount trust rule base model to evaluate the health state of the system.
Fig. 5 shows a health status evaluation result chart after considering three layers of uncertainty effects, the CRH380 type running part system is divided into 5 health statuses, which are respectively replaced by 5,4,3,2, and 1, and correspond to the maintenance stop, the warehouse maintenance, the temporary maintenance, the round maintenance, and the health. Specific definitions of five health states are given below:
stopping repair: and each part of the train is seriously damaged, and the operation is stopped at any time. In this state, emergency braking is required and the monitored train components should be repaired immediately to prevent a dangerous situation;
and (4) warehousing and repairing: the running function of the high-speed train is reduced, and partial components are in bad states, so that faults can be caused. As inspections seek to improve the safety of trains, it is necessary to replace or repair the monitored train components;
repairing: high speed trains reach critical points of health and failure. According to the relevant regulation of < train repair and repair system >, in the next maintenance, under the condition of not disassembling wheels, the train needs to repair and adjust the system and components regularly to ensure the normal operation of the train;
wheel repair: parts wear slightly during operation, requiring inspection of the appearance, working condition and general performance of the various parts working for preventive and corrective maintenance. Under the state, the normal operation of the train is not influenced;
health: all parts of the high-speed train work well, the fastening pieces are not loosened, all indexes meet the requirements of a factory, and the high-speed train can ensure safe operation.
FIG. 6 shows a result graph of a three-dimensional coordinate system, and analysis of the result graph shows that the real health state curve of the running gear system can be well fitted.
It is understood that the above description is not intended to limit the present invention, but rather, the present invention is not limited to the above examples, and any changes, modifications, additions and substitutions which may be made by one skilled in the art within the spirit and scope of the present invention are included therein.

Claims (7)

1. A method for assessing the health of a CRH 380-type running gear system, comprising:
acquiring measurement data of a CRH380 type running gear system in different health states as training data, giving prior information of initial parameters in a confidence rule base by a domain expert, and establishing a training model of the initial parameters;
step two, according to the monitoring uncertainty and the environment uncertainty (both are related to the sensor and are divided into static factors and dynamic factors) existing in the quantitative evaluation process of the measured data;
thirdly, fusing the quantified static and dynamic factors based on an evidence discount method to obtain the corrected attribute reliability;
quantifying the cognitive uncertainty factors of the system according to the initial expert experience, and constructing an expert reliability model;
step five, collecting measurement data of the CRH380 type running gear system under real-time working conditions as test data;
and step six, the quantized three-layer uncertainty factors are fused into a confidence rule base evaluation model, and are fitted with a system real health state threshold value, and evaluation results of the test data are analyzed.
2. The method for evaluating the health status of the CRH380 running gear system according to claim 1, wherein in the first step, the measured data of the CRH380 running gear system in different health statuses comprise two physical characteristic data of temperature and vibration; the initial value of the confidence rule base given by the domain expert is selected as shown in equation (1). Suppose there is a k rule in the model to represent the method as follows:
Figure RE-FDA0002850574270000011
wherein, x represents the input vector,
Figure RE-FDA0002850574270000012
denotes the input reference value, H, of the m-th attribute in the k-th rulen(N-1, 2, …, N) represents the nth evaluation scale, βn,kIndicates the evaluation level H relative to the nthnConfidence of (a), thetakThe rule weight representing the kth rule,
Figure RE-FDA0002850574270000013
a weight value representing the mth premise attribute, κ represents a discount operator,
Figure RE-FDA0002850574270000014
the range of tolerance is indicated and,
Figure RE-FDA0002850574270000015
rmrespectively expressed as the static reliability, dynamic reliability and attribute reliability of the mth sensor,
Figure RE-FDA0002850574270000016
express the expert reliability of the kth rule; if it is not
Figure RE-FDA0002850574270000017
The kth rule is complete, otherwise it is incomplete; in particular, it is possible to use, for example,
Figure RE-FDA0002850574270000018
indicating that the output of the kth rule is completely unknown.
3. The method for evaluating the health status of the CRH380 running gear system according to claim 1, wherein in the second step, the measured data in different health status levels of the system under actual working conditions are influenced by not only monitoring uncertainty but also environmental uncertainty; in addition, the method is also influenced by the cognitive uncertainty of the system, and the traditional evaluation method is difficult to be applied to large-scale complex electromechanical equipment such as a high-speed train; equations (2) - (4) are selected as reliability factors after quantifying monitoring uncertainty:
Figure RE-FDA0002850574270000019
Figure RE-FDA00028505742700000110
Figure RE-FDA0002850574270000021
wherein x isij(i-1, …, M; j-1, …, N) is the j-th measurement data acquired by the i-th sensor,
Figure RE-FDA0002850574270000022
is that
Figure RE-FDA0002850574270000023
Represents the average of the N monitoring points.
Figure RE-FDA0002850574270000024
Representing all index data xijAnd the average value
Figure RE-FDA0002850574270000025
Measure of similarity therebetween, and ri sIs the reliability factor after the quantization of the ith monitoring uncertainty;
similarly, equation (5) is selected as the reliability factor after quantifying the environmental uncertainty:
Figure RE-FDA0002850574270000026
wherein,
Figure RE-FDA0002850574270000027
for observation data xijHas a standard deviation of
Figure RE-FDA0002850574270000028
Figure RE-FDA0002850574270000029
A coefficient indicating an adjustment tolerance range; if it is not
Figure RE-FDA00028505742700000210
Or
Figure RE-FDA00028505742700000211
The observed data is not reliable and,
Figure RE-FDA00028505742700000212
if not, then,
Figure RE-FDA00028505742700000220
4. the method for assessing the health status of a CRH 380-type running gear system according to claim 1, wherein in the third step, the selection formula (6) is used to construct a discounted method based on a factor coordination fusion after the monitoring uncertainty and the environmental uncertainty are quantified, and the reliability of the modified attribute is obtained;
Figure RE-FDA00028505742700000213
wherein r isi sAnd ri dRespectively representing the static reliability and the dynamic reliability of the observation information of the ith sensor, riRepresenting the degree of reliability of the observation information of the ith sensor, and f (-) representing riAnd ri s、ri dκ represents the discount operator;
because the initial tolerance coefficient is given by an expert and cannot be well adapted to the change of the external environment, the achievement uses the tolerance coefficient
Figure RE-FDA00028505742700000214
And (4) the parameters are included in the parameter vector of the model and are trained.
5. The method for assessing the health status of a CRH 380-type running gear system according to claim 1, wherein in the fourth step, the formula (7) to formula (15) is selected to quantify the cognitive uncertainty factor of the system based on the initial expert experience and to construct an expert reliability model:
Figure RE-FDA00028505742700000215
wherein
Figure RE-FDA00028505742700000216
And
Figure RE-FDA00028505742700000217
respectively representing the vector form of the expert description rule confidence allocation,Dis shown in (2)Θ×2ΘIn the form of a matrix of (a),Dthe elements contained in (A) are defined as
Figure RE-FDA00028505742700000218
Since the confidence rules are connected by logical or,Dis of a scale of 2Θ×2ΘThe identity matrix of (2). When there are multiple confidence rules, the distance of every two confidence rules can be represented in the form of a distance matrix DM, as follows:
Figure RE-FDA00028505742700000219
Sim(Ri,Rj)＝1-d(Ri,Rj) (10)
Figure RE-FDA0002850574270000031
the support for each confidence rule is as follows:
Figure RE-FDA0002850574270000032
Figure RE-FDA0002850574270000033
wherein, CrdkExpress confidence to the kThe greater the Crd, the more the confidence rule is supported by other experts;
then introducing a Dun entropy calculation method of each confidence rule, and expressing the method as follows:
Figure RE-FDA0002850574270000034
and finally obtaining a calculation method of expert reliability:
Figure RE-FDA0002850574270000035
6. the method for assessing the health of a CRH 380-type running gear system according to claim 1, wherein the fifth step is specifically: an online test sample is obtained in real time.
7. The method for assessing the health of a CRH 380-type running gear system according to claim 1, wherein the sixth step is specifically: the quantized three layers of uncertainty factors are fused into a confidence rule base evaluation model, a formula (16) is selected to evaluate the real-time state of the model, the model is fitted with a real health state threshold value of the system, and evaluation results of test data are analyzed;
Figure RE-FDA0002850574270000036
wherein beta isnShows the nth evaluation result HnConfidence of (c), H { (H)n,βn) N-1, 2, …, N represents the overall confidence distribution after the overall confidence rules are integrated.
CN202011187692.7A 2020-10-30 2020-10-30 Method for evaluating health state of CRH380 type running gear system Pending CN112729878A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202011187692.7A CN112729878A (en) 2020-10-30 2020-10-30 Method for evaluating health state of CRH380 type running gear system

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202011187692.7A CN112729878A (en) 2020-10-30 2020-10-30 Method for evaluating health state of CRH380 type running gear system

Publications (1)

Publication Number Publication Date
CN112729878A true CN112729878A (en) 2021-04-30

Family

ID=75597351

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202011187692.7A Pending CN112729878A (en) 2020-10-30 2020-10-30 Method for evaluating health state of CRH380 type running gear system

Country Status (1)

Country Link
CN (1) CN112729878A (en)

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN115712841A (en) * 2022-11-18 2023-02-24 南京航空航天大学 Spacecraft component state evaluation method based on data distribution characteristics of periodic data
CN115996082A (en) * 2023-03-24 2023-04-21 中国人民解放军火箭军工程大学 Traceable airborne bus network health state evaluation method and system

Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN105868536A (en) * 2016-03-24 2016-08-17 中国人民解放军海军航空工程学院 Test data based health state assessment method for solid rocket engine
CN106679675A (en) * 2016-12-29 2017-05-17 北京理工大学 Mars final approaching section autonomous navigation method based on relative measurement information
CN109195104A (en) * 2018-08-27 2019-01-11 上海市计量测试技术研究院 A kind of indoor orientation method combined based on support vector regression and Kalman filtering
WO2019080367A1 (en) * 2017-10-25 2019-05-02 南京航空航天大学 Method for evaluating health status of mechanical device
CN109711453A (en) * 2018-12-21 2019-05-03 广东工业大学 A kind of equipment dynamical health state evaluating method based on multivariable

Patent Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN105868536A (en) * 2016-03-24 2016-08-17 中国人民解放军海军航空工程学院 Test data based health state assessment method for solid rocket engine
CN106679675A (en) * 2016-12-29 2017-05-17 北京理工大学 Mars final approaching section autonomous navigation method based on relative measurement information
WO2019080367A1 (en) * 2017-10-25 2019-05-02 南京航空航天大学 Method for evaluating health status of mechanical device
CN109195104A (en) * 2018-08-27 2019-01-11 上海市计量测试技术研究院 A kind of indoor orientation method combined based on support vector regression and Kalman filtering
CN109711453A (en) * 2018-12-21 2019-05-03 广东工业大学 A kind of equipment dynamical health state evaluating method based on multivariable

Non-Patent Citations (4)

* Cited by examiner, † Cited by third party
Title
CHAO CHENG ET.AL: "Health Status Prediction Based on Belief Rule Base for High-Speed Train Running Gear System", 《IEEE ACCESS》 *
丰世林: "民航飞机健康维修管理评估系统设计仿真", 《计算机仿真》 *
程超 等: "一种基于凸优化的WSN障碍环境下定位算法", 《吉林大学学报（理学版）》 *
赵丽琴 等: "基于动态权重的设备健康状态评估方法", 《计算机系统应用》 *

Cited By (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN115712841A (en) * 2022-11-18 2023-02-24 南京航空航天大学 Spacecraft component state evaluation method based on data distribution characteristics of periodic data
CN115712841B (en) * 2022-11-18 2023-08-15 南京航空航天大学 Spacecraft part state evaluation method based on data distribution characteristics of periodic data
CN115996082A (en) * 2023-03-24 2023-04-21 中国人民解放军火箭军工程大学 Traceable airborne bus network health state evaluation method and system
CN115996082B (en) * 2023-03-24 2023-05-23 中国人民解放军火箭军工程大学 Traceable airborne bus network health state evaluation method and system

Similar Documents

Publication Publication Date Title
CN109472110B (en) Method for predicting residual service life of aeroengine based on LSTM network and ARIMA model
CN107991870B (en) Fault early warning and service life prediction method for escalator equipment
CN112785091B (en) Method for carrying out fault prediction and health management on oil field electric submersible pump
Meeker et al. Reliability: The other dimension of quality
CN112729878A (en) Method for evaluating health state of CRH380 type running gear system
CN111815136A (en) Modeling method for full life cycle maturity of aviation product
CN111241629B (en) Intelligent prediction method for performance change trend of hydraulic pump of airplane based on data driving
Liu Maintenance model of aircraft structure based on three-stage degradation process
Li et al. A decision-based framework for predictive maintenance technique selection in industry 4.0
CN114139589A (en) Fault diagnosis method, device, equipment and computer readable storage medium
Daher et al. New prognosis approach for preventive and predictive maintenance—Application to a distillation column
Islam et al. Human factors in marine and offshore systems
Verna et al. Towards Zero Defect Manufacturing: probabilistic model for quality control effectiveness
CN117575027A (en) Sensor fault diagnosis method for enhancing space-time causality by using knowledge
CN116258467B (en) Electric power construction management and control system
Pang Intelligent belt conveyor monitoring and control
Ling et al. Information gain-based inspection scheduling for fatigued aircraft components
CN112990275B (en) High-speed train running gear system fault diagnosis method based on semi-quantitative information fusion
CN114154266A (en) Gas turbine fault prediction method based on partial rank correlation flow causal structure learning
Scott et al. A structural PHM framework for aircraft control surface free-play
Henry et al. An equivalent age model for condition-based maintenance
CN112365022A (en) Engine bearing fault prediction method based on multiple stages
Khan et al. Prognostics of crack propagation in structures using time delay neural network
CN110705833A (en) Reliability evaluation method for train control and management system
CN114841000B (en) Soft measurement modeling method based on modal common feature separation

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
AD01 Patent right deemed abandoned

Effective date of abandoning: 20230811

AD01 Patent right deemed abandoned