CN112069685A - Interpretability-considered complex electromechanical system health assessment method and system - Google Patents

Interpretability-considered complex electromechanical system health assessment method and system Download PDF

Info

Publication number
CN112069685A
CN112069685A CN202010938568.3A CN202010938568A CN112069685A CN 112069685 A CN112069685 A CN 112069685A CN 202010938568 A CN202010938568 A CN 202010938568A CN 112069685 A CN112069685 A CN 112069685A
Authority
CN
China
Prior art keywords
electromechanical system
complex electromechanical
interpretable
rule
system health
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
CN202010938568.3A
Other languages
Chinese (zh)
Inventor
曹友
周志杰
胡昌华
唐帅文
陈媛
张春潮
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.)
Rocket Force University of Engineering of PLA
Original Assignee
Rocket Force University of Engineering of PLA
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 Rocket Force University of Engineering of PLA filed Critical Rocket Force University of Engineering of PLA
Priority to CN202010938568.3A priority Critical patent/CN112069685A/en
Publication of CN112069685A publication Critical patent/CN112069685A/en
Pending legal-status Critical Current

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
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q10/00—Administration; Management
    • G06Q10/06—Resources, workflows, human or project management; Enterprise or organisation planning; Enterprise or organisation modelling
    • G06Q10/063—Operations research, analysis or management
    • G06Q10/0639—Performance analysis of employees; Performance analysis of enterprise or organisation operations
    • G06Q10/06393—Score-carding, benchmarking or key performance indicator [KPI] analysis
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F2113/00—Details relating to the application field
    • G06F2113/28—Fuselage, exterior or interior
    • 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]
    • 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

  • Business, Economics & Management (AREA)
  • Engineering & Computer Science (AREA)
  • Human Resources & Organizations (AREA)
  • Physics & Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Economics (AREA)
  • Development Economics (AREA)
  • Strategic Management (AREA)
  • Educational Administration (AREA)
  • Entrepreneurship & Innovation (AREA)
  • General Business, Economics & Management (AREA)
  • Tourism & Hospitality (AREA)
  • Quality & Reliability (AREA)
  • Operations Research (AREA)
  • Marketing (AREA)
  • Game Theory and Decision Science (AREA)
  • Computer Hardware Design (AREA)
  • Evolutionary Computation (AREA)
  • Geometry (AREA)
  • General Engineering & Computer Science (AREA)
  • Medical Treatment And Welfare Office Work (AREA)

Abstract

The invention relates to a complex electromechanical system health assessment method and system considering interpretability. The method comprises the following steps: constructing a complex electromechanical system health evaluation model based on a confidence rule base; respectively defining search intensity, rule activation factors and interpretable distribution constraints according to expert knowledge; selecting the activated parameters in the complex electromechanical system health evaluation model according to the rule activation factors; establishing an objective function according to the interpretable distribution constraint and the activated parameters; optimizing the activated parameters according to the search intensity, the interpretable distribution constraint and the objective function to obtain an optimized complex electromechanical system health evaluation model; and inputting the data of the complex electromechanical system into the optimized complex electromechanical system health assessment model for assessment to obtain the health assessment state of the complex electromechanical system. The method improves the evaluation precision and simultaneously ensures the interpretability of the model, so that the evaluation result is more reliable.

Description

Interpretability-considered complex electromechanical system health assessment method and system
Technical Field
The invention relates to the field of complex electromechanical system health assessment, in particular to a complex electromechanical system health assessment method and system considering interpretability.
Background
In the modern industrialization process of China, the supporting effect of complex electromechanical systems represented by high-end numerical control machines, aircraft engines and the like on economic development is increasingly prominent, and the reliability and safety of the complex electromechanical systems are widely concerned. The health evaluation method can identify weak links and hidden dangers of a complex system, and provides a basis for improving the reliability of the system. However, unreliable and black box evaluation processes may result in potential risks, such as unreasonable maintenance decisions, delayed alarm messages, etc. Therefore, how to effectively fuse multi-source information such as expert knowledge and monitoring data and accurately and reliably evaluate the complex electromechanical health state in an interpretable manner becomes a problem to be solved urgently in various fields.
The existing research mainly adopts three types of health assessment models: 1) and a data-driven model built based on a large number of samples, such as a Support Vector Machine (SVM), an Artificial Neural Network (ANN) and the like. The rationality of the results of such models is difficult to convince due to the opacity of the modeling process; 2) and constructing a white box model based on mechanism information. Such models may provide a transparent modeling process and interpretable results. But the creation and application of white-box models becomes difficult to achieve due to the challenges of extracting accurate mathematical expressions from complex systems. 3) A gray box model established based on limited expert knowledge and data samples. The model can simultaneously obtain better modeling precision and interpretability to a certain extent. The Belief Rule Base (BRB) is a typical gray box model, and can effectively process uncertainty and realize the health assessment of a complex electromechanical system.
Expert knowledge is an important source of interpretability for BRB-based health assessment models. However, there are three problems in the model optimization process that may destroy the model's interpretability, resulting in unreliable output results. 1) Expert knowledge is not effectively utilized in the optimization process; 2) the optimized rule may not be consistent with common sense; 3) some rules are over-optimized, which may change the preliminary judgment of the expert. Therefore, to accurately and reliably realize the health assessment of the complex electromechanical system, the above three problems must be fully considered.
Disclosure of Invention
The invention aims to provide a complex electromechanical system health assessment method and system considering interpretability, which can ensure the interpretability of a model while improving assessment precision and ensure that an assessment result is more reliable.
In order to achieve the purpose, the invention provides the following scheme:
a method for health assessment of a complex electromechanical system taking interpretability into account, comprising:
constructing a complex electromechanical system health evaluation model based on a confidence rule base;
respectively defining search intensity, rule activation factors and interpretable distribution constraints according to expert knowledge;
selecting the activated parameters in the complex electromechanical system health evaluation model according to the rule activation factors;
establishing an objective function according to the interpretable distribution constraint and the activated parameters;
optimizing the activated parameters according to the search intensity, the interpretable distribution constraint and the objective function to obtain an optimized complex electromechanical system health evaluation model;
and inputting the data of the complex electromechanical system into the optimized complex electromechanical system health assessment model for assessment to obtain the health assessment state of the complex electromechanical system.
Optionally, the building of the complex electromechanical system health assessment model based on the confidence rule base specifically includes:
constructing a complex electromechanical system health evaluation model based on a confidence rule base:
Figure BDA0002672803740000024
with a rule weight θk and attribute weightsi(k＝1,2,...,L；i＝1,2,...,Tk)
wherein, X1,...,
Figure BDA0002672803740000025
Represents the monitoring index of the complex electromechanical system,
Figure BDA0002672803740000021
reference value, theta, representing the i-th indexkThe weight of the kth rule is used for representing the importance of the rule relative to other rules;i(i＝1,2,...,Tk) The weight of the ith index is expressed to represent the importance degree of the ith index relative to other indexes; l represents the number of rules, βnk(N ═ 1, 2.., N) denotes a reference level DnThe confidence of (c).
Optionally, the search strength is defined as:
Figure BDA0002672803740000022
wherein, P (omega) represents the search intensity and is used for describing the sampling probability in the optimization process; p (omega) satisfies both the properties of regularity and unimodal distribution,
Figure BDA0002672803740000023
and represents the credibility of the expert knowledge, and sigma is a covariance matrix used for reflecting and adjusting the credibility of the expert knowledge.
Optionally, the interpretable distribution constraint is defined as:
βk～Ck,(k＝1,...,L)
Figure BDA00026728037400000311
wherein, CkDenotes the interpretability distribution, betakIndicates the confidence in the kth rule with respect to a certain health level, N indicates the number of health levels, and L indicates the total number of rules.
Optionally, the rule activation factor is defined as:
Figure BDA0002672803740000031
wherein, WkRepresenting a vector constructed by the activation weights of the kth rule, assuming a data set of size P, WkIs represented by Wk＝(w1,...,wp,...,wP),k＝1,...,L；p＝1,...,P。
Optionally, the selecting, according to the rule activation factor, an activated parameter in the complex electromechanical system health assessment model specifically includes:
activating a rule according to the rule activation factor;
when the activation rule is determined, selecting corresponding parameters as optimization vectors;
the optimization vector is represented as
Figure BDA0002672803740000032
Wherein the content of the first and second substances,
Figure BDA0002672803740000033
and
Figure BDA0002672803740000034
indicating the parameter that is activated.
Optionally, the objective function is:
Figure BDA0002672803740000035
Figure BDA0002672803740000036
Figure BDA0002672803740000037
wherein the content of the first and second substances,
Figure BDA0002672803740000038
the output error of the model is represented by,
Figure BDA0002672803740000039
is the first
Figure BDA00026728037400000310
Interpretable constraints of confidence distributions in the rules.
Optionally, the optimizing the activated parameter according to the search strength, the interpretable distribution constraint, and the objective function to obtain an optimized complex electromechanical system health assessment model specifically includes:
and optimizing the activated parameters by adopting a differential evolution algorithm according to the search intensity, the interpretable distribution constraint and the objective function to obtain an optimized complex electromechanical system health evaluation model.
A complex electromechanical systems health assessment system taking interpretability into account, comprising:
the complex electromechanical system health evaluation model establishing module is used for establishing a complex electromechanical system health evaluation model based on a confidence rule base;
the definition module is used for respectively defining the search intensity, the rule activation factor and the interpretable distribution constraint according to expert knowledge;
the activated parameter determination module is used for selecting the activated parameters in the complex electromechanical system health evaluation model according to the rule activation factors;
an objective function establishing module for establishing an objective function according to the interpretable distribution constraint and the activated parameter;
the optimization module is used for optimizing the activated parameters according to the search intensity, the interpretable distribution constraint and the objective function to obtain an optimized complex electromechanical system health assessment model;
and the health state evaluation module is used for inputting the complex electromechanical system data into the optimized complex electromechanical system health evaluation model for evaluation to obtain the health state of the complex electromechanical system.
According to the specific embodiment provided by the invention, the invention discloses the following technical effects:
firstly, constructing a complex electromechanical system health assessment model based on a confidence rule base, wherein the model can effectively combine expert knowledge and monitoring data; then, in order to ensure the interpretability of the model in the optimization process, respectively defining search intensity, rule activation factors and interpretable distribution constraints according to expert knowledge; and finally, an optimization model considering interpretability is constructed, the model is optimized by using monitoring data, so that the model interpretability is ensured while the evaluation precision is improved, and the evaluation result is more reliable.
Drawings
In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings needed in the embodiments will be briefly described below, and it is obvious that the drawings in the following description are only some embodiments of the present invention, and it is obvious for those skilled in the art to obtain other drawings without creative efforts.
FIG. 1 is a flow chart of a method for health assessment of a complex electromechanical system in accordance with the present invention, taking interpretability into account;
FIG. 2 is a block diagram of a complex electromechanical system health assessment system in accordance with the present invention in view of interpretability;
FIG. 3 is a diagram illustrating raw vibration data and characteristics;
FIG. 4 is a diagram illustrating the evaluation results and confidence distributions;
FIG. 5 is a comparison of rule weights;
FIG. 6 is a graph comparing confidence distributions.
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.
The invention aims to provide a complex electromechanical system health assessment method and system considering interpretability, which can ensure the interpretability of a model while improving assessment precision and ensure that an assessment result is more reliable.
In order to make the aforementioned objects, features and advantages of the present invention comprehensible, embodiments accompanied with figures are described in further detail below.
FIG. 1 is a flow chart of a method for evaluating health of a complex electromechanical system in consideration of interpretability according to the present invention. As shown in fig. 1, a method for health assessment of a complex electromechanical system considering interpretability includes:
step 101: the method comprises the following steps of constructing a complex electromechanical system health evaluation model based on a confidence rule base, and specifically comprises the following steps:
constructing a complex electromechanical system health evaluation model based on a confidence rule base:
Figure BDA0002672803740000051
with a rule weight θk and attribute weightsi(k＝1,2,...,L；i＝1,2,...,Tk)
(1)
wherein, X1, is.,
Figure BDA0002672803740000053
represents the monitoring index of the complex electromechanical system,
Figure BDA0002672803740000052
reference value, theta, representing the i-th indexkThe weight of the kth rule is used for representing the importance of the rule relative to other rules;i(i＝1,2,...,Tk) The weight of the ith index is expressed to represent the importance degree of the ith index relative to other indexes; l represents the number of rules, βnk(N ═ 1, 2.., N) denotes a reference level DnThe confidence of (c).
The reasoning steps of the complex electromechanical system health assessment model based on the confidence rule library structure are as follows:
step 1: converting quantitative and qualitative information into a confidence distribution:
S(xi)＝{(Ai,j,ai,j),i＝1,...,Tk；j＝1,...,Ji} (2)
wherein A isi,jRepresenting an input xiThe jth reference value of (a). ai, j represents Ai,jThe degree of matching of (2).
Step 2: the rule activation weights are calculated as follows:
Figure BDA0002672803740000061
Figure BDA0002672803740000062
wherein, thetak∈[0,1]Is the weight of the kth rule.
Figure BDA0002672803740000063
Is xiNormalized weights of (1). L is the number of rules.
Figure BDA0002672803740000064
Representing input x in the k-th ruleiRelative to its jth reference value Ai,jThe degree of matching of (2).
And step 3: the final confidence is generated using a evidence reasoning algorithm, as follows:
Figure BDA0002672803740000065
and 4, step 4: the final distributed result is expressed as:
S(X)＝{(Dn,βn)；n＝1,...,N} (5)
where X represents the actual input vector. S (X) represents a healthy state, the utility of which can be calculated by the following formula:
Figure BDA0002672803740000066
step 102: the search strength, rule activation factor and interpretable distribution constraint are defined separately according to expert knowledge.
In order to effectively utilize expert knowledge in the model optimization process, the method for evaluating the health of the complex electromechanical system considering interpretability, provided by the invention, has the definition of the search strength as follows:
Figure BDA0002672803740000067
in the above equation, P (Ω) represents the search strength, which is used to describe the sampling probability in the optimization process. P (omega) satisfies both the properties of regularity and unimodal distribution,
Figure BDA0002672803740000068
representing the trustworthiness of the expert knowledge. σ is a covariance matrix that can be used to reflect and adjust the confidence of expert knowledge. Expert knowledge accumulated in long-term practice is the basis for model interpretability. The search intensity ensures the characteristic of local optimization, and can effectively improve the modeling precision of the BRB-based health assessment model under the interpretability requirement.
In the method for evaluating the health of the complex electromechanical system considering interpretability, in order to ensure the consistency of rules and common knowledge, the interpretable distribution constraint is defined as:
βk～Ck,(k＝1,...,L) (8)
in the field of health assessment, the conflicting health states cannot simultaneously obtain a high degree of confidence. The interpretable distribution of the rule conclusion part may thus be specifically:
Figure BDA0002672803740000072
the invention relates to a complex electromechanical system health assessment method considering interpretability, which comprises the following steps of calculating and applying rule activation factors:
generally, the expert knowledge accumulated in long-term practice contains information of all working states of the real system, and the training data set is limited and can only contain information of partial working states. That is, the initial BRB-based model is built taking into account all the operating conditions of the system, but the optimization process can only adjust parameters associated with some of the operating conditions reflected in the data set. Thus, it may be determined that only the parameters of the activation rule may be fine-tuned by the training data set. Over-optimization refers to the process of optimizing the non-activation parameters, which may change the initial judgment information of the expert, resulting in loss of interpretability.
To efficiently optimize the model parameters, the active rules need to be distinguished from all rules. The rule activation factor ω is used to mark rules for activation. If ω isk1 means that the kth rule is activated. The activation factor can thus be defined as:
Figure BDA0002672803740000071
wherein, WkRepresenting the vector constructed by the activation weights of the kth rule. Assume that the size of the data set is P. WkCan be expressed as
Wk＝(w1,...,wp,...,wP),k＝1,...,L；p＝1,...,P (11)
Step 103: selecting the activated parameters in the complex electromechanical system health evaluation model according to the rule activation factors, and specifically comprises the following steps:
activating a rule according to the rule activation factor;
when the activation rule is determined, selecting corresponding parameters as optimization vectors;
the optimization vector is represented as:
Figure BDA0002672803740000081
wherein the content of the first and second substances,
Figure BDA0002672803740000082
and
Figure BDA0002672803740000083
indicating the parameter that is activated.
Step 104: and establishing an objective function according to the interpretable distribution constraint and the activated parameters.
An interpretable objective function is considered. In order to achieve a reasonable optimization process, the objective function needs to be modified. The modified objective function is:
Figure BDA0002672803740000084
wherein the content of the first and second substances,
Figure BDA0002672803740000085
the output error of the model is represented by,
Figure BDA0002672803740000086
is the first
Figure BDA0002672803740000087
Interpretable constraints of confidence distributions in the rules.
Step 105: optimizing the activated parameters according to the search strength, the interpretable distribution constraint and the objective function to obtain an optimized complex electromechanical system health assessment model specifically comprises:
and optimizing the activated parameters by adopting a differential evolution algorithm according to the search intensity, the interpretable distribution constraint and the objective function to obtain an optimized complex electromechanical system health evaluation model.
Taking a differential evolution algorithm (DE) as an example, the search strength is introduced in the initial operation, and the interpretable constraint is introduced in the newly added control operation. The modified DE algorithm is denoted by P-DE-I and comprises the following steps:
step 1 (initialization operation): and generating an initial population according to the search intensity.
Figure BDA0002672803740000088
Wherein
Figure BDA0002672803740000089
Representing the initial parameter vector. The covariance matrix is σ. λ represents the overall scale. P (-) represents the search strength under normal distribution.
Step 2 (mutation operation): the new solution is generated using the mutation operation, and is represented as:
Figure BDA00026728037400000810
wherein p is1≠p2≠p3≠k。F∈[0,1]Is the scaling factor.
Figure BDA00026728037400000811
Is the optimal individual of the g generation. μ denotes the center of the population, calculated from:
Figure BDA0002672803740000091
step 3 (crossover operation): cross-operations were introduced to increase population diversity, denoted as
Figure BDA0002672803740000092
Wherein Cr is within 0,1]Indicating the crossover rate.
Figure BDA0002672803740000093
Is the updated population.
Step 4 (control operation): resampling the erroneous confidence distributions in the kth solution of the g-th generation until all confidence distributions satisfy the distribution constraint, which is described in detail as follows:
Figure BDA0002672803740000094
wherein the content of the first and second substances,
Figure BDA0002672803740000095
representing the kth solution of the g-th generation, which may contain a false confidence distribution.
Figure BDA0002672803740000096
Is a newly generated confidence distribution that satisfies the constraint.
Step 5 (projection operation): to satisfy the equality constraint, the equality constraint is converted to an equality constraint in the hyperplane using a projection operation:
Figure BDA0002672803740000097
the projection operation is as follows:
Figure BDA0002672803740000098
step 6 (selection operation): performing a selection operation to update the best individual and population as:
Figure BDA0002672803740000099
Figure BDA00026728037400000910
step 7 (termination criteria): go to step 2 until the maximum number of iterations is reached.
Step 106: and inputting the data of the complex electromechanical system into the optimized complex electromechanical system health assessment model for assessment to obtain the health assessment state of the complex electromechanical system.
The invention relates to a complex electromechanical system health assessment method considering interpretability. The present invention also provides a complex electromechanical system health assessment system considering interpretability, as shown in fig. 2, the system including:
the complex electromechanical system health assessment model establishing module 201 is used for establishing a complex electromechanical system health assessment model based on the confidence rule base.
And the defining module 202 is used for respectively defining the search strength, the rule activation factor and the interpretable distribution constraint according to expert knowledge.
An activated parameter determination module 203, configured to select the activated parameter in the complex electromechanical system health assessment model according to the rule activation factor.
An objective function establishing module 204, configured to establish an objective function according to the interpretable distribution constraint and the activated parameter.
And the optimization module 205 is configured to optimize the activated parameters according to the search strength, the interpretable distribution constraint, and the objective function, so as to obtain an optimized complex electromechanical system health assessment model.
And the health state evaluation module 206 is configured to input the complex electromechanical system data into the optimized complex electromechanical system health evaluation model for evaluation, so as to obtain the health state of the complex electromechanical system.
Example 1:
in this embodiment, the health of the aircraft engine is evaluated. The aircraft engine is a precise and complex dynamic system consisting of an electromagnetic unit and a mechanical unit, and the health state of the aircraft engine is influenced by high temperature, high pressure and strong stress change and has uncertainty. Because of the particularity of aircraft engines, they cannot be tested frequently, and therefore the test data samples obtained are limited. The health state of the aircraft engine is the key of the safety of the aircraft, and the health state of the engine can be estimated in an interpretable and reliable manner by using limited data samples and knowledge, so that potential safety hazards in engineering practice can be reduced. In this experiment, 5000 sets of vibration data were collected using an aircraft engine internal vibration sensor at 0%, 5%, 10% and 20% reduction in high altitude valve spring pressure (corresponding to a healthy state of normal (N), minor fault (S), medium fault (M) and major fault (SE), respectively.) the raw vibration data are shown in fig. 3- (a) using Kurtosis and Skewness (Skewness) of the vibration data as characteristics describing the aircraft engine' S healthy state:
Figure BDA0002672803740000101
for every 100 data, the characteristics were calculated using equation (15), as shown in fig. 3- (b) and 3- (c). In each state, 30 data were selected as a training set, and all data were selected as a test set.
The specific implementation steps for the health assessment of the aircraft engine are as follows:
step 1: and constructing an initial evaluation model.
Degradation of critical components of an aircraft engine, such as valve springs, can result in slight variations in the engine vibration signal and manifest as variations in kurtosis and skewness. In engineering practice, kurtosis and skewness are described by four semantic values, "low (L)", "medium (M)", "high (H)" and "Very High (VH)". In addition, based on the knowledge accumulated from long-term testing of aircraft engines, reference values and initial attribute weights were determined, as shown in table 1, consistent with the design principles.
TABLE 1 Attribute reference values and weights
Figure BDA0002672803740000111
In order to describe the relationship between the characteristics and the state of health, initial rules need to be given through analysis of the aircraft engine. Under normal conditions, the kurtosis and skewness of the aircraft engine vibrations are both at the VH level. In this case, the state of health becomes worse because the degree of skew becomes "M". When the skewness level is "M" and the kurtosis level is "L" or "M", the healthy state is at the "SE" level. In general, as kurtosis and skewness levels decrease, health conditions become worse. Based on the empirical knowledge described above, the initial rules are shown in Table 2.
TABLE 2 initial rule base
Figure BDA0002672803740000112
Figure BDA0002672803740000121
Step 2: an optimization model is constructed that takes interpretability into account.
Due to the limitation of expert knowledge, the initial model needs to be optimized by using monitoring data. First, a final parameter vector is determined
Figure BDA0002672803740000122
In the present case, the rule activation factor is calculated using a training set and equations (10) - (12), which are expressed as
ω＝{ω1,...,ωk}＝{1,1,1,0,1,1,1,1,1,1,1,1,0,0,1,1}(16)
This means that rules No. 4, 13 and 14 are not activated by the training set, i.e. the corresponding parameters do not need to be optimized. The final parameter vector is represented as
Figure BDA0002672803740000123
The objective function is determined by equation (13), and the output error is calculated as the mean square error. The parameters of P-DE-I are shown in Table 3.
TABLE 3P-DE-I parameters
Figure BDA0002672803740000124
And step 3: and optimizing the model and testing results.
The optimized evaluation model is shown in table 4, and the optimized attribute weights are 1 and 0.7864, respectively.
TABLE 4 optimized evaluation model
Figure BDA0002672803740000125
Figure BDA0002672803740000131
In the test section, a test data set is taken as input. The state of health assessed is shown in fig. 4- (a), with a mean square error of 0.0054, indicating that the optimized BRB is able to accurately estimate the state of health of the aircraft engine. The confidence corresponding to the state of health is shown in fig. 4- (b), and it can be seen that the aircraft engine can still operate normally in the normal state and the minor fault state. To verify the robustness of P-DE-I, experiments were performed 25 times. The variance of MSEs (1.826E-07) is much smaller than the mean of MSEs (0.00564).
And 4, step 4: and (5) carrying out comparative study.
To verify the validity of the above models, comparative studies were performed on an initial model (denoted BRB 0), a model optimized with the P-DE-I algorithm (denoted BRB-1), a model optimized with the conventional DE algorithm (denoted BRB-2), a fuzzy rule base model (FRB), and an Extreme Learning Machine (ELM).
Table 5 shows the corresponding mean square error in terms of modeling accuracy. It can be seen that the optimized BRB models (BRB1, BRB2) have higher modeling accuracy than other models. Compared with FRB and ELM, the modeling precision of BRB1 is improved by 41.30% and 15.63%. BRB2 has similar modeling accuracy as BRB 1.
TABLE 5 evaluation error of five models
Figure BDA0002672803740000141
In terms of model interpretability, rule weights and confidence distributions of BRB0, BRB1, and BRB2 were compared, as shown in fig. 5 and 6. As can be seen from fig. 4, the rule weights in BRB1 are similar to those in BRB0, while most of the rule weights in BRB2 are far from those in BRB 0. Taking the third rule as an example, the rule weights of BRB0 and BRB1 are 0.7 and 0.62, respectively, which indicates that the third rule is important. However, the rule weight of the third rule in BRB2 is 0.0278, indicating that the rule is not important. This is in full conflict with the expert's preliminary judgment. As shown in fig. 6, the confidence distributions in BRB0 and BRB1 are close. Most of the confidence distributions in BRB2 are different from those in BRB 0. Some distributions, such as the confidence distribution in the eighth rule, do not meet the interpretability requirements. The results show that the BRB1 improves the modeling precision and fine-tunes the parameters determined by expert knowledge, namely the optimization process of the BRB1 is based on the initial judgment of experts, so that the interpretability of the evaluation process is effectively ensured.
The embodiments in the present description are described in a progressive manner, each embodiment focuses on differences from other embodiments, and the same and similar parts among the embodiments are referred to each other. For the system disclosed by the embodiment, the description is relatively simple because the system corresponds to the method disclosed by the embodiment, and the relevant points can be referred to the method part for description.
The principles and embodiments of the present invention have been described herein using specific examples, which are provided only to help understand the method and the core concept of the present invention; meanwhile, for a person skilled in the art, according to the idea of the present invention, the specific embodiments and the application range may be changed. In view of the above, the present disclosure should not be construed as limiting the invention.

Claims (9)

1. A method for health assessment of a complex electromechanical system taking interpretability into account, comprising:
constructing a complex electromechanical system health evaluation model based on a confidence rule base;
respectively defining search intensity, rule activation factors and interpretable distribution constraints according to expert knowledge;
selecting the activated parameters in the complex electromechanical system health evaluation model according to the rule activation factors;
establishing an objective function according to the interpretable distribution constraint and the activated parameters;
optimizing the activated parameters according to the search intensity, the interpretable distribution constraint and the objective function to obtain an optimized complex electromechanical system health evaluation model;
and inputting the data of the complex electromechanical system into the optimized complex electromechanical system health assessment model for assessment to obtain the health assessment state of the complex electromechanical system.
2. The interpretable complex electromechanical system health assessment method of claim 1, wherein the building of the complex electromechanical system health assessment model based on the confidence rule base specifically comprises:
constructing a complex electromechanical system health evaluation model based on a confidence rule base:
Figure FDA0002672803730000011
with a rule weightθk and attribute weightsi(k＝1,2,...,L；i＝1,2,...,Tk)
wherein the content of the first and second substances,
Figure FDA0002672803730000012
representing complex electromechanical systemsThe monitoring index of (a) is,
Figure FDA0002672803730000013
reference value, theta, representing the i-th indexkThe weight of the kth rule is used for representing the importance of the rule relative to other rules;i(i＝1,2,...,Tk) The weight of the ith index is expressed to represent the importance degree of the ith index relative to other indexes; l represents the number of rules, βnk(N ═ 1, 2.., N) denotes a reference level DnThe confidence of (c).
3. The interpretable complex electromechanical system health assessment method of claim 2, wherein the search strength is defined as:
Figure FDA0002672803730000014
wherein, P (omega) represents the search intensity and is used for describing the sampling probability in the optimization process; p (omega) satisfies both the properties of regularity and unimodal distribution,
Figure FDA0002672803730000015
and represents the credibility of the expert knowledge, and sigma is a covariance matrix used for reflecting and adjusting the credibility of the expert knowledge.
4. The interpretable complex electromechanical systems health assessment method of claim 3, wherein the interpretable distribution constraint is defined as:
Figure FDA0002672803730000021
wherein, CkDenotes the interpretability distribution, betakIndicates the confidence in the kth rule with respect to a certain health level, N indicates the number of health levels, and L indicates the total number of rules.
5. The interpretable complex electromechanical system health assessment method of claim 4, wherein the rule activation factor is defined as:
Figure FDA0002672803730000022
wherein, WkRepresenting a vector constructed by the activation weights of the kth rule, assuming a data set of size P, WkIs represented by Wk＝(w1,...,wp,...,wP),k＝1,...,L；p＝1,...,P。
6. The interpretable complex electromechanical system health assessment method of claim 5, wherein the selecting parameters to be activated in the complex electromechanical system health assessment model according to the rule activation factor specifically comprises:
activating a rule according to the rule activation factor;
when the activation rule is determined, selecting corresponding parameters as optimization vectors;
the optimization vector is represented as
Figure FDA0002672803730000023
Wherein the content of the first and second substances,
Figure FDA0002672803730000024
and
Figure FDA0002672803730000025
indicating the parameter that is activated.
7. The interpretable complex electromechanical system health assessment method of claim 6, wherein the objective function is:
Figure FDA0002672803730000026
Figure FDA0002672803730000027
wherein the content of the first and second substances,
Figure FDA0002672803730000028
the output error of the model is represented by,
Figure FDA0002672803730000029
is the first
Figure FDA00026728037300000210
Interpretable constraints of confidence distributions in the rules.
8. The method for evaluating the health of a complex electromechanical system considering interpretability according to claim 1, wherein the optimizing the activated parameters according to the search strength, the interpretable distribution constraint and the objective function to obtain an optimized complex electromechanical system health evaluation model specifically comprises:
and optimizing the activated parameters by adopting a differential evolution algorithm according to the search intensity, the interpretable distribution constraint and the objective function to obtain an optimized complex electromechanical system health evaluation model.
9. A complex electromechanical systems health assessment system taking interpretability into account, comprising:
the complex electromechanical system health evaluation model establishing module is used for establishing a complex electromechanical system health evaluation model based on a confidence rule base;
the definition module is used for respectively defining the search intensity, the rule activation factor and the interpretable distribution constraint according to expert knowledge;
the activated parameter determination module is used for selecting the activated parameters in the complex electromechanical system health evaluation model according to the rule activation factors;
an objective function establishing module for establishing an objective function according to the interpretable distribution constraint and the activated parameter;
the optimization module is used for optimizing the activated parameters according to the search intensity, the interpretable distribution constraint and the objective function to obtain an optimized complex electromechanical system health assessment model;
and the health state evaluation module is used for inputting the complex electromechanical system data into the optimized complex electromechanical system health evaluation model for evaluation to obtain the health state of the complex electromechanical system.
CN202010938568.3A 2020-09-09 2020-09-09 Interpretability-considered complex electromechanical system health assessment method and system Pending CN112069685A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202010938568.3A CN112069685A (en) 2020-09-09 2020-09-09 Interpretability-considered complex electromechanical system health assessment method and system

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202010938568.3A CN112069685A (en) 2020-09-09 2020-09-09 Interpretability-considered complex electromechanical system health assessment method and system

Publications (1)

Publication Number Publication Date
CN112069685A true CN112069685A (en) 2020-12-11

Family

ID=73662940

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202010938568.3A Pending CN112069685A (en) 2020-09-09 2020-09-09 Interpretability-considered complex electromechanical system health assessment method and system

Country Status (1)

Country Link
CN (1) CN112069685A (en)

Cited By (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN112861403A (en) * 2021-02-09 2021-05-28 中国人民解放军火箭军工程大学 Safety evaluation method for large liquid storage tank structure
CN115619106A (en) * 2022-12-19 2023-01-17 中国人民解放军火箭军工程大学 Method and system for determining quantity of spare parts of laser gyroscope in consideration of performance degradation
CN115964907A (en) * 2023-03-17 2023-04-14 中国人民解放军火箭军工程大学 Complex system health trend prediction method and system, electronic device and storage medium
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
CN116451912A (en) * 2023-06-19 2023-07-18 西北工业大学 Complex electromechanical system performance evaluation method and system under condition of influence of replacement

Cited By (10)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN112861403A (en) * 2021-02-09 2021-05-28 中国人民解放军火箭军工程大学 Safety evaluation method for large liquid storage tank structure
CN112861403B (en) * 2021-02-09 2022-12-20 中国人民解放军火箭军工程大学 Safety evaluation method for large liquid storage tank structure
CN115619106A (en) * 2022-12-19 2023-01-17 中国人民解放军火箭军工程大学 Method and system for determining quantity of spare parts of laser gyroscope in consideration of performance degradation
CN115619106B (en) * 2022-12-19 2023-05-16 中国人民解放军火箭军工程大学 Method and system for determining number of spare parts of laser gyroscope in consideration of performance degradation
CN115964907A (en) * 2023-03-17 2023-04-14 中国人民解放军火箭军工程大学 Complex system health trend prediction method and system, electronic device and storage medium
CN115964907B (en) * 2023-03-17 2023-12-01 中国人民解放军火箭军工程大学 Complex system health trend prediction method, system, electronic equipment and storage medium
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
CN116451912A (en) * 2023-06-19 2023-07-18 西北工业大学 Complex electromechanical system performance evaluation method and system under condition of influence of replacement
CN116451912B (en) * 2023-06-19 2023-09-19 西北工业大学 Complex electromechanical system performance evaluation method and system under condition of influence of replacement

Similar Documents

Publication Publication Date Title
CN112069685A (en) Interpretability-considered complex electromechanical system health assessment method and system
Wang et al. Data-driven mode identification and unsupervised fault detection for nonlinear multimode processes
El-Midany et al. A proposed framework for control chart pattern recognition in multivariate process using artificial neural networks
CN110348752B (en) Large industrial system structure safety assessment method considering environmental interference
CN107273924B (en) Multi-data fusion power plant fault diagnosis method based on fuzzy clustering analysis
CN109740859A (en) Transformer condition evaluation and system based on Principal Component Analysis and support vector machines
Liu et al. An adversarial bidirectional serial–parallel LSTM-based QTD framework for product quality prediction
CN116448419A (en) Zero sample bearing fault diagnosis method based on depth model high-dimensional parameter multi-target efficient optimization
CN110766320A (en) Method and device for evaluating operation safety of airport intelligent power grid
CN115688581A (en) Oil gas gathering and transportation station equipment parameter early warning method, system, electronic equipment and medium
Zhang et al. Fault diagnosis of oil-immersed power transformer based on difference-mutation brain storm optimized catboost model
Cao et al. On the robustness of belief-rule-based expert systems
Sun et al. A wireless sensor network node fault diagnosis model based on belief rule base with power set
Wang et al. Recognition method of equipment state with the FLDA based Mahalanobis–Taguchi system
Yousefpour et al. Unsupervised anomaly detection via nonlinear manifold learning
Zhang et al. An enhanced temporal algorithm-coupled optimized adaptive sparse principal component analysis methodology for fault diagnosis of chemical processes
CN109829115B (en) Search engine keyword optimization method
Xia A systematic graph-based methodology for cognitive predictive maintenance of complex engineering equipment
CN114358244B (en) Big data intelligent detection system of pressure based on thing networking
CN115630582A (en) Multi-sliding-window model fused soft rock tunnel surrounding rock deformation prediction method and equipment
CN112949524B (en) Engine fault detection method based on empirical mode decomposition and multi-core learning
Baur et al. Explainability and interpretability in electric load forecasting using machine learning techniques–A review
Liu et al. Network traffic big data prediction model based on combinatorial learning
Zheng et al. Research on Predicting Remaining Useful Life of Equipment Based on Health Index
Wang et al. A multi‐block NMF model for non‐Gaussian process monitoring based on the adaptive partition non‐negative matrix factorization and B ayesian 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