阅读说明：本技术 一种基于混沌算法的泵站故障树建立方法 (Pump station fault tree establishment method based on chaotic algorithm ) 是由 施伟 成立 蔡瑞民 夏臣智 王希晨 徐文涛 罗灿 杨登俊 范雪梅 刘佳佳 沈冲 于 2020-12-29 设计创作，主要内容包括：本发明涉及一种基于混沌算法的泵站故障树建立方法,选取一个顶事件；将顶事件逐渐分解为底事件,建立基本故障树；基本故障树建立后选取故障原因不同但表现类似的难以判别的故障,并且该故障的故障信息具有混沌特性；相空间重构要确定时间延迟τ和嵌入维数m,计算时间延迟；嵌入维数的意义是选择合适的维数,使看似无序的时间序列规律容易被看见,嵌入维数的选择非常重要,嵌入维数过低,复杂的信息不能在相应的最佳维度展开,可能导致相空间重构的失败；选择好时间延迟τ和嵌入维数m后进行相空间重构,之后需要进而求取关联维数；通过本发明可建立更加全面准确的泵站故障树。(The invention relates to a pump station fault tree building method based on a chaotic algorithm, which comprises the steps of selecting a top event; gradually decomposing the top event into bottom events, and establishing a basic fault tree; after the basic fault tree is established, faults which have different fault reasons but are similar in appearance and difficult to distinguish are selected, and fault information of the faults has chaotic characteristics; determining a time delay tau and an embedding dimension m for phase space reconstruction, and calculating the time delay; the significance of the embedding dimension is to select a proper dimension, so that a disordered time sequence rule is easy to see, the selection of the embedding dimension is very important, the embedding dimension is too low, complex information cannot be expanded in the corresponding optimal dimension, and the failure of phase space reconstruction can be caused; selecting a time delay tau and an embedding dimension m, then carrying out phase space reconstruction, and then solving a correlation dimension; by the method and the system, a more comprehensive and accurate pump station fault tree can be established.)

1. A pump station fault tree building method based on a chaos algorithm is characterized by comprising the following steps:

(1) selecting a top event;

(2) gradually decomposing the top event into bottom events, and establishing a basic fault tree;

(3) after the basic fault tree is established, faults which have different fault reasons but are similar in appearance and difficult to distinguish are selected, and fault information of the faults has chaotic characteristics;

(4) determining a time delay tau and an embedding dimension m for phase space reconstruction, and calculating the time delay;

(5) the significance of the embedding dimension is to select a proper dimension, so that a disordered time sequence rule is easy to see, the selection of the embedding dimension is very important, the embedding dimension is too low, complex information cannot be expanded in the corresponding optimal dimension, and the failure of phase space reconstruction can be caused;

(6) selecting a time delay tau and an embedding dimension m, then carrying out phase space reconstruction, and then solving a correlation dimension; for the embedding dimension m, time delay τ, reconstructed phase space vector X_iThe associated integral of (a) is defined as:

wherein h is a Heaviside function; x_i-X_jIs a two-phase point X_iAnd X_jThe distance of (d); c (r) represents the probability that the distance between two points in the phase space is less than r; n is a natural number set; the slopes of InC (r) and Inr correspond to the correlation coefficients; d is a correlation dimension; the correlation dimension can be obtained by a linear fitting method;

(7) and obtaining the relation of the correlation dimensions of different fault categories through analysis and classification, and embedding the relation into a fault tree to refine and perfect the fault tree.

2. The method for building the pump station fault tree based on the chaotic algorithm according to claim 1, wherein in the step (1), a certain fault event which is most influenced and is to be analyzed is selected as a top event, and the fault event which is most influenced comprises that the pump station stops running.

3. The method for building the pump station fault tree based on the chaotic algorithm according to claim 1, wherein in the step (2), the possible causes of the fault are decomposed into intermediate events in a step-by-step manner until a bottom event.

4. The method for establishing the pump station fault tree based on the chaos algorithm according to claim 1, wherein in the step (3), some same faults possibly caused by different reasons are analyzed and refined through a chaos theory, such as a motor fault in an electrical appliance fault, the motor fault reason may be divided into a motor broken bar, a motor eccentric starting or a motor overload reason, but the fault reason is difficult to accurately locate from a fault expression, so that fault stator current information needs to be collected, a proper time delay τ and an embedding dimension are selected to perform phase space reconstruction, a correlation dimension is further obtained, the motor fault reason is judged by comparing the correlation dimensions of different conditions and is written into the fault tree, and then the fault reason can be quickly located by retrieving the fault tree for comparison.

5. The method for building the pump station fault tree based on the chaotic algorithm according to claim 1, wherein in the step (4), the calculation time delay can be calculated by an autocorrelation function method, a complex autocorrelation method, a Shaw mutual information method or a C-C algorithm.

6. The method for building the pump station fault tree based on the chaotic algorithm according to claim 1, wherein in the step (5), the dimension embedding method comprises the following steps: the method comprises a trial algorithm, a false neighbor point method, a Cao method, an association integral method, a singular decomposition method and the like, wherein the Cao method is a mature method, and different methods have different advantages and can be selected according to conditions.

7. The method for building the pump station fault tree based on the chaotic algorithm according to the claim 1, wherein in the step (6), the correlation dimension is an important index for describing the chaotic system, and the GP method proposed by Grass-berger and Procaccia simplifies the calculation of the correlation dimension.

Technical Field

The invention relates to a pump station fault tree establishment method based on a chaotic algorithm, and belongs to the field of hydraulic engineering.

Background

The pump station fault detection has a focus and difficulty in the operation process of a pump station all the time, the operation environment of a plurality of large, medium and small pump stations is complex, the fault occurrence reason has diversity and complexity, the fault occurrence reason is probably caused by external factors in the management aspect and the like besides the hydraulic, mechanical and electrical reasons and the like, the fault occurrence is uncertain, so that the fault is difficult to predict, a fault tree model with clear logic and clear arrangement is greatly helpful for the operation management of the pump station, and the fault tree model can assist pump station operation and maintenance personnel to quickly and accurately judge the fault category and take effective measures to avoid the occurrence of accidents.

At present, a traditional fault tree construction method is generally implemented through a fault tree analysis method (FTA), the method is traditional and effective, a pump station fault tree is generally implemented based on the method, but due to the particularity of a pump station fault, the fault tree constructed by the method generally has the defects of ambiguous direction and incomplete completeness, because the faults of the pump station are generally various and complex in cause, two types of faults which are often irrelevant but have the same expression form, and at the moment, the fault tree established only through the general method is difficult to distinguish the faults, so that the accuracy of the fault tree is influenced, and further the judgment of a pump station operation and maintenance worker is influenced. Therefore, a fault tree building method capable of more accurately distinguishing and classifying various faults is needed.

Disclosure of Invention

The invention aims to solve the defects of the prior art and provides an accurate and comprehensive pump station fault tree establishing method to help pump station operation and maintenance personnel to better distinguish and process pump station faults with similar expression forms and avoid accidents caused by fault judgment errors, and particularly relates to a chaotic algorithm-based pump station fault tree establishing method.

The invention aims to realize the following technical scheme, and discloses a pump station fault tree establishing method based on a chaotic algorithm, which is characterized by comprising the following steps of:

(1) selecting a top event;

(2) gradually decomposing the top event into bottom events, and establishing a basic fault tree;

(3) after the basic fault tree is established, faults which have different fault reasons but are similar in appearance and difficult to distinguish are selected, and fault information of the faults has chaotic characteristics;

(4) determining a time delay tau and an embedding dimension m for phase space reconstruction, and calculating the time delay; .

(5) The significance of the embedding dimension is to select a proper dimension, so that a disordered time sequence rule is easy to see, the selection of the embedding dimension is very important, the embedding dimension is too low, complex information cannot be expanded in the corresponding optimal dimension, and the failure of phase space reconstruction can be caused;

(6) selecting a time delay tau and an embedding dimension m, then carrying out phase space reconstruction, and then solving a correlation dimension; for the embedding dimension m, time delay τ, reconstructed phase space vector X_iThe associated integral of (a) is defined as:

wherein h is a Heaviside function; x_i-X_jIs a two-phase point X_iAnd X_jThe distance of (d); c (r) represents the probability that the distance between two points in the phase space is less than r; n is a natural number set; the slopes of InC (r) and Inr correspond to the correlation coefficients; d is a correlation dimension; the correlation dimension can be obtained by a linear fitting method;

(7) and obtaining the relation of the correlation dimensions of different fault categories through analysis and classification, and embedding the relation into a fault tree to refine and perfect the fault tree.

In the step (1), a certain fault event which is most influenced and is to be analyzed is selected as a top event, and the fault event which is most influenced comprises the stop of the pump station.

In the step (2), the possible reasons of the fault are gradually decomposed into intermediate events until a bottom event.

In the step (3), some same faults possibly caused by different reasons are analyzed and refined through a chaos theory, for example, motor faults in electrical appliance faults can be divided into motor broken bars, motor eccentric starting or motor overload reasons, but the fault reasons are difficult to accurately locate from fault expression, so that fault stator current information needs to be collected, appropriate time delay tau and embedding dimension are selected to carry out phase space reconstruction, then correlation dimension is obtained, the motor fault reasons are judged by comparing the correlation dimensions of different conditions and are written into a fault tree, and then the fault reasons can be quickly located through searching the fault tree and comparing.

In step (4), the calculation time delay can be calculated by an autocorrelation function method, a complex autocorrelation method, a Shaw mutual information quantity method or a C-C algorithm.

In the step (5), the method for embedding the dimension number comprises the following steps: the method comprises a trial algorithm, a false neighbor point method, a Cao method, an association integral method, a singular decomposition method and the like, wherein the Cao method is a mature method, and different methods have different advantages and can be selected according to conditions.

In the step (6), the correlation dimension is an important index for describing the chaotic system, and the GP method proposed by Grass-berger and Procaccia simplifies the calculation of the correlation dimension.

The method is advanced and scientific, and according to the method, a certain fault event which is most influenced and needs to be analyzed is selected as a top event, such as the stop of a pump station, and then possible reasons of the fault are gradually decomposed into intermediate events until a bottom event. The method comprises the steps of analyzing and refining the same faults possibly caused by different reasons through a chaos theory, wherein the same faults possibly caused by different reasons, such as motor faults in electrical appliance faults, the motor fault reasons can be divided into motor broken bars, motor eccentric starting or motor overload reasons, and the like, but the fault reasons are difficult to accurately locate from fault expression, so that fault stator current information needs to be collected, proper time delay tau and embedding dimension are selected to carry out phase space reconstruction, then association dimension is obtained, the motor fault reasons are judged by comparing the association dimensions of different conditions and are written into a fault tree, and then the fault reasons can be quickly located through searching the fault tree and comparing.

The method is combined with the chaos theory to analyze and judge the pump station faults, and the relation between the pump station faults and the correlation dimension is embedded into the fault tree, so that the more comprehensive and accurate pump station fault tree is established.

The chaos theory is a method with both qualitative and qualitative analysis, and is used to discuss the behavior of a dynamic system that must be explained and predicted by using an integral, continuous rather than a single data relationship.

The phase space reconstruction process is realized based on the delay embedding theorem of f.takens and r.mane. The phase space of the reconstructed time series is divided into two steps, the first step is to calculate the time delay tau of the time series, and the second step is to estimate the embedding dimension. The meaning of calculating the time delay is to fully expose the information of the time series, and in the ideal case, τ can be arbitrarily selected, but the time data has errors of different degrees. The choice of delay is crucial to the reconstruction phase space. The significance of the embedding dimension is to select a suitable dimension so that the time series regularity that appears unordered is easily seen.

By the method and the device, a more comprehensive pump station fault tree can be established more conveniently and quickly. The pump station fault diagnosis and analysis is an important ring in the operation of a pump station, and the establishment of a pump station fault tree is a difficult point due to various pump station faults and complex and various reasons, so that a reliable and comprehensive pump station fault tree establishment method is necessary and is required by the market.

Has the advantages that: the invention is based on the chaos theory, can better analyze and classify the pump station fault reason, thereby improving the accuracy and comprehensiveness of the pump station fault tree and greatly helping pump station operation and maintenance personnel to judge and process faults and safely operate the pump station.

Drawings

FIG. 1 is a flow chart.

FIG. 2 is an example fault tree.

Detailed Description

The detailed flow chart is shown in fig. 1, which is an example shown in fig. 2; the top event selected in this example is pump station failure shutdown (in this example, the relationship between the correlation dimension and the failure is a relationship under a certain condition, and the relationship is not universal, and the specific correlation needs to be obtained through sampling analysis of sampling points of a specific case).

(1) Selecting a suitable top event.

(2) And gradually decomposing the top event into bottom events, and establishing a basic fault tree.

(3) After the basic fault tree is established, faults which have different fault reasons but are similar in performance and difficult to distinguish are selected, and fault information of the faults has chaotic characteristics, such as several fault types of the motor are selected for analysis in an example.

(4) The phase space reconstruction needs to determine the time delay tau and the embedding dimension m, the time delay can be calculated through an autocorrelation function method, a complex autocorrelation method, a Shaw mutual information quantity method, a C-C algorithm and the like, different methods have different characteristics, and the mutual information quantity method is a generally accepted method and can be selected according to conditions.

(5) The significance of the embedding dimension is to select a proper dimension so that a disordered time sequence rule is easy to see, the selection of the embedding dimension is very important, the embedding dimension is too low, complex information cannot be expanded in the corresponding optimal dimension, and the phase space reconstruction can fail, and the method for embedding the dimension comprises the following steps: the method comprises a trial algorithm, a false neighbor point method, a Cao method, an association integral method, a singular decomposition method and the like, wherein the Cao method is a mature method, and different methods have different advantages and can be selected according to conditions.

(6) And selecting the time delay tau and the embedding dimension m, then carrying out phase space reconstruction, and then obtaining the correlation dimension which is an important index for describing the chaotic system, wherein the GP method proposed by Grass-berger and Procaccia simplifies the calculation of the correlation dimension. For the embedding dimension m, time delay τ, reconstructed phase space vector X_iThe associated integral of (a) is defined as:

wherein h is a Heaviside function; x_i-X_jIs a two-phase point X_iAnd X_jThe distance of (d); c (r) represents the probability that the distance between two points in the phase space is less than r; n is a natural number set; the slopes of InC (r) and Inr correspond to the correlation coefficients. D is the correlation dimension. The correlation dimension can be obtained by a linear fitting method.

(7) And obtaining the relation of the correlation dimensions of different fault categories through analysis and classification, and embedding the relation into a fault tree to refine and perfect the fault tree.