CN115423097A

CN115423097A - Detection method, terminal device and computer-readable storage medium

Info

Publication number: CN115423097A
Application number: CN202210938054.7A
Authority: CN
Inventors: 郑湃; 夏历翘; 周家樑
Original assignee: Beijing Aerospace Data Co ltd; Hong Kong Polytechnic University HKPU
Current assignee: Beijing Aerospace Data Co ltd; Hong Kong Polytechnic University HKPU
Priority date: 2022-08-05
Filing date: 2022-08-05
Publication date: 2022-12-02

Abstract

The application is applicable to the technical field of fault detection, and provides a detection method, terminal equipment and a computer readable storage medium, wherein the detection method comprises the following steps: acquiring real-time detection data of each part in equipment to be detected through a sensor; and obtaining the detection result of the equipment to be detected according to the real-time detection data and the trained detection model, wherein the trained detection model is obtained according to the hypergraph structure training of the equipment to be detected, the nodes of the hypergraph structure represent the parts of the equipment to be detected, and the edges of the hypergraph structure represent the physical relationship among the parts of the equipment to be detected. By the method, the accuracy and the reliability of equipment fault detection can be effectively improved.

Description

Detection method, terminal device and computer-readable storage medium

Technical Field

The present application belongs to the field of fault detection technologies, and in particular, to a detection method, a terminal device, and a computer-readable storage medium.

Background

The fault detection is an important link in equipment predictive maintenance, and aims to monitor whether a fault occurs or not, the fault position and identify the fault type. Real-time and efficient fault detection is of great significance to extend the service life of mechanical equipment, reduce maintenance costs and increase the safety of equipment operation.

The existing fault detection method generally utilizes a sensor to collect detection data of each component in the equipment, and then judges whether each detection data is abnormal or not, so as to judge whether the equipment has a fault or not and judge the fault position. In practical application, each component in the device has relevance, and sometimes, the detection data of a certain component in the device is abnormal, which may be caused by the faults of other components. In the existing method, all parts in the equipment are used as independent units for detection, and the physical association between the parts is ignored, so that the accuracy and reliability of the detection result are low.

Disclosure of Invention

The embodiment of the application provides a detection method, terminal equipment and a computer readable storage medium, which can effectively improve the accuracy and reliability of equipment fault detection.

In a first aspect, an embodiment of the present application provides a detection method, including:

acquiring real-time detection data of each part in equipment to be detected through a sensor;

and obtaining the detection result of the equipment to be detected according to the real-time detection data and the trained detection model, wherein the trained detection model is obtained according to the hypergraph structure training of the equipment to be detected, the nodes of the hypergraph structure represent the parts of the equipment to be detected, and the edges of the hypergraph structure represent the physical relationship among the parts of the equipment to be detected.

In the embodiment of the application, the incidence relation among all the parts in the equipment to be detected is represented through the hypergraph structure, the detection model is trained according to the hypergraph structure, namely the detection model is trained according to the incidence relation among all the parts in the equipment to be detected, so that the trained detection model can automatically identify the incidence relation among the real-time detection data acquired by the sensor, and the equipment fault detection is carried out according to the incidence relation. By the method, when the equipment fault is detected, the association among all parts in the equipment is fully considered, and the accuracy and reliability of the detection result are effectively improved.

In a possible implementation manner of the first aspect, the method further includes:

generating a hypergraph structure of the equipment to be detected according to the physical relationship among all parts in the equipment to be detected;

and training the detection model according to the hypergraph structure and multiple groups of training data to obtain the trained detection model, wherein each group of training data comprises a group of historical detection data and label types of all parts in the equipment to be detected.

In a possible implementation manner of the first aspect, the generating a hypergraph structure of the device to be detected according to a physical relationship between components in the device to be detected includes:

grouping the components in the equipment to be detected according to the physical relationship among the components in the equipment to be detected;

generating a sub-graph structure from the components contained in each group;

and generating the sub-graph structure into a super-graph structure of the equipment to be detected.

In a possible implementation manner of the first aspect, the physical relationship includes at least one mathematical model, each mathematical model includes at least 2 physical variables, the mathematical model is used for representing an operation mechanism of the device to be detected, and values of the physical variables are obtained through the sensor;

according to wait to examine the physical relation between each part among the equipment, right wait to examine each part in the equipment and divide into groups, include:

grouping the physical variables according to the relationship between the physical variables and the mathematical model;

and dividing the parts corresponding to the physical variables in each group into one group.

In a possible implementation manner of the first aspect, the generating a sub-graph structure according to the components included in each group includes:

for each group, if the number of the physical variables corresponding to the group is less than or equal to the number of the mathematical models, generating a first sub-graph by using the components contained in the group, wherein the first sub-graph is a directed sub-graph structure, and one end of each edge in the first sub-graph is connected with at most one node;

if the number of the physical variables corresponding to the group is larger than that of the mathematical models, and the number of coincident variables is larger than or equal to 2, generating a second sub-graph from the components contained in the group, wherein the second sub-graph is a non-directional sub-graph structure, and the number of nodes connected to one end of each edge in the second sub-graph is a positive integer larger than or equal to 0, and the coincident variables are the physical variables contained in each mathematical model corresponding to the group;

if the number of the physical variables corresponding to the group is larger than that of the mathematical model and the number of coincident variables is smaller than 2, generating a third sub-graph by using the components contained in the group, wherein the third sub-graph is a directed sub-graph structure and the number of nodes connected with one end of each edge in the third sub-graph is a positive integer larger than or equal to 0;

the edges in the sub-graph structure are determined according to the mathematical model to which the components contained in each group belong.

In a possible implementation manner of the first aspect, the training the detection model according to the hypergraph structure and multiple sets of training data to obtain the trained detection model includes:

generating a hypergraph matrix according to the hypergraph structure;

and training the detection model according to the hypergraph matrix and the multiple groups of training data to obtain the trained detection model.

In a possible implementation manner of the first aspect, an abscissa of the hypergraph matrix represents a number of an edge in the hypergraph structure, and an ordinate of the hypergraph matrix represents a number of a node in the hypergraph structure;

the generating of the hypergraph matrix according to the hypergraph structure includes:

if the ith edge in the hypergraph structure is connected with the jth node, setting the value of the jth row and the jth column in the hypergraph matrix to be a non-zero preset value;

and if the ith edge in the hypergraph structure is not connected with the jth node, setting the value of the jth row and the jth column in the hypergraph matrix as 0.

In a possible implementation manner of the first aspect, the training the detection model according to the hypergraph matrix and a plurality of sets of training data to obtain the trained detection model includes:

carrying out normalization processing on the training detection data to obtain normalized data;

denoising the normalized data to obtain denoised data;

sorting the de-noising data according to the arrangement sequence of the components in the hypergraph matrix to obtain sorted data;

and inputting the sequencing data and the hypergraph matrix into the detection model for training to obtain the trained detection model.

In a second aspect, an embodiment of the present application provides a detection apparatus, including:

the data acquisition unit is used for acquiring real-time detection data of each part in the equipment to be detected through the sensor;

and the fault detection unit is used for obtaining a detection result of the equipment to be detected according to the real-time detection data and the trained detection model, wherein the trained detection model is obtained according to the hypergraph structure of the equipment to be detected, the nodes of the hypergraph structure represent the parts of the equipment to be detected, and the edges of the hypergraph structure represent the physical relationship among the parts of the equipment to be detected.

In a third aspect, an embodiment of the present application provides a terminal device, which includes a memory, a processor, and a computer program stored in the memory and executable on the processor, where the processor implements the detection method according to any one of the first aspect when executing the computer program.

In a fourth aspect, an embodiment of the present application provides a computer-readable storage medium, and the embodiment of the present application provides a computer-readable storage medium, where a computer program is stored, and the computer program, when executed by a processor, implements the detection method according to any one of the first aspect.

In a fifth aspect, embodiments of the present application provide a computer program product, which, when run on a terminal device, causes the terminal device to perform the method of any one of the first aspect.

It is understood that the beneficial effects of the second aspect to the fifth aspect can be referred to the related description of the first aspect, and are not described herein again.

Drawings

In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings needed to be used in the embodiments or the prior art descriptions will be briefly described below, and it is obvious that the drawings in the following description are only some embodiments of the present application, and it is obvious for those skilled in the art to obtain other drawings based on these drawings without inventive exercise.

FIG. 1 is a schematic flow chart diagram illustrating a model training method according to an embodiment of the present disclosure;

FIG. 2 is a schematic diagram of a hypergraph structure provided by an embodiment of the present application;

FIG. 3 is a schematic diagram of a variable matrix provided by an embodiment of the present application;

FIG. 4 is a schematic diagram of a sub-graph structure provided in an embodiment of the present application;

FIG. 5 is a schematic diagram of a hypergraph structure provided by another embodiment of the present application;

FIG. 6 is a schematic flow chart of a detection method provided in an embodiment of the present application;

FIG. 7 is a block diagram of a detecting apparatus provided in an embodiment of the present application;

fig. 8 is a schematic structural diagram of a terminal device according to an embodiment of the present application.

Detailed Description

In the following description, for purposes of explanation and not limitation, specific details are set forth, such as particular system structures, techniques, etc. in order to provide a thorough understanding of the embodiments of the present application. It will be apparent, however, to one skilled in the art that the present application may be practiced in other embodiments that depart from these specific details. In other instances, detailed descriptions of well-known systems, devices, circuits, and methods are omitted so as not to obscure the description of the present application with unnecessary detail.

It will be understood that the terms "comprises" and/or "comprising," when used in this specification and the appended claims, specify the presence of stated features, integers, steps, operations, elements, and/or components, but do not preclude the presence or addition of one or more other features, integers, steps, operations, elements, components, and/or groups thereof.

It should also be understood that the term "and/or" as used in this specification and the appended claims refers to and includes any and all possible combinations of one or more of the associated listed items.

As used in this specification and the appended claims, the term "if" may be interpreted contextually as "when", "upon" or "in response to" determining "or" in response to detecting ". Similarly, the phrase "if it is determined" or "if a [ described condition or event ] is detected" may be interpreted contextually to mean "upon determining" or "in response to determining" or "upon detecting [ described condition or event ]" or "in response to detecting [ described condition or event ]".

Furthermore, in the description of the present application and the appended claims, the terms "first," "second," "third," and the like are used for distinguishing between descriptions and not necessarily for describing or implying relative importance.

Reference throughout this specification to "one embodiment" or "some embodiments," or the like, means that a particular feature, structure, or characteristic described in connection with the embodiment is included in one or more embodiments of the present application. Thus, appearances of the phrases "in one embodiment," "in some embodiments," "in other embodiments," or the like, in various places throughout this specification are not necessarily all referring to the same embodiment, but rather "one or more but not all embodiments" unless specifically stated otherwise.

First, a training method of the detection model is described. Referring to fig. 1, which is a schematic flow chart of a model training method provided in an embodiment of the present application, by way of example and not limitation, the model training method includes the following steps:

and S101, generating a hypergraph structure of the equipment to be detected according to the physical relationship among all parts in the equipment to be detected.

In general, there are dependencies between components in a device, including mechanical structure dependencies, and operational mechanism dependencies. Therefore, the physical relationship in the embodiment of the present application includes a mechanical structure connection relationship between components in the equipment, and also includes mathematical models corresponding to the mechanism of the equipment, and each mathematical model includes at least 2 physical variables. The mathematical model is used to represent the operating mechanism of the plant, and the values of the physical variables can be obtained by means of sensors.

Taking a turbofan engine as an example, the mechanical structure of the turbofan engine comprises a turbofan, a supercharger, a turbine and a combustor, wherein the supercharger comprises a high-pressure compressor and a low-pressure compressor, and the turbine comprises a high-pressure turbine and a low-pressure turbine. In the mechanical structure, the turbofan is connected with the air compressor, the air compressor is connected with the combustor, and the combustor is connected with the supercharger. In the operation mechanism, the turbofan conveys sucked air to the air compressor; the gas compressor is used for pressurizing, and the pressurized high-temperature and high-pressure gas flow is conveyed to the combustor; the gas flow is heated by a combustor, and the heated and expanded gas flow impacts a rear turbine; the turbine is driven to rotate by high-temperature and high-pressure airflow, and the airflow is finally ejected out from the tail at a high speed to form the most main power of the engine.

For example, as can be seen from the physical relationship among the components of the turbofan engine, the rotating speed of the turbofan is related to the flow velocity of the air flow entering the compressor, the rotating speed of the turbofan is an independent variable, and the flow velocity of the air flow entering the compressor is a dependent variable, so as to form a mathematical model; for another example, the pressure of the air flow in the compressor is related to the power of the burner, the pressure of the air flow in the compressor is an independent variable, and the power of the burner is a dependent variable, so as to form a mathematical model. The physical variables of the rotating speed of the turbofan, the airflow flow rate of the compressor, the airflow pressure in the compressor and the power of the combustor can be detected and obtained through sensors arranged on all parts of the turbofan engine.

The hypergraph structure comprises nodes and edges, the nodes are connected through the edges, and the edges are used for representing associations between the nodes, such as attributes, categories, dependency relationships and the like. The essence of the hypergraph structure is a topological data structure that characterizes the extrinsic, intrinsic associations between objects. Fig. 2 is a schematic diagram of a hypergraph structure provided in the embodiment of the present application. As shown in fig. 2 (a), a conventional graph structure is provided, in which one end of each edge is connected to at most one node. As shown in fig. 2 (b), a hypergraph structure may be formed in which one end of each edge is connected to a plurality of nodes.

As can be seen from the above description, in step S101, the hypergraph structure of the device to be detected is generated according to the physical relationship between the components in the device to be detected, and the physical relationship between the components in the device to be detected, including the relationship between the connection relationship of the mechanical structure and the operation mechanism, can be represented. By the method, the abstract physical relationship among all parts in the equipment to be detected is embodied into the visual hypergraph structure through the hypergraph structure, so that subsequent data processing is facilitated.

S102, training the detection model according to the hypergraph structure and the multiple groups of training data to obtain the trained detection model.

Each set of training data comprises a set of historical detection data and label types of all parts in the device to be detected. It should be noted that which detection result is desired, which type of label is used in the training process. For example, if the expected detection result is whether the engine is failed or not, the label type corresponding to each group of historical detection data can be engine failure/normal during the training process. For another example, if the expected detection result is a fault location in the engine, the label type corresponding to each set of historical detection data during the training process may be a turbofan fault/a supercharger fault/a turbine fault/a combustor fault.

In the embodiment of the application, the incidence relation among all the parts in the equipment to be detected is represented through the hypergraph structure, the detection model is trained according to the hypergraph structure, which is equivalent to training the detection model according to the incidence relation among all the parts in the equipment to be detected, so that the trained detection model can automatically identify the incidence relation among the real-time detection data acquired by the sensor.

In one embodiment, the generation method of the hypergraph structure comprises the following steps:

grouping the components in the equipment to be detected according to the physical relationship among the components in the equipment to be detected; generating a sub-graph structure from the components contained in each group; and generating the hypergraph structure of the device to be detected by the sub-graph structure.

Optionally, the grouping manner may be: and dividing the components corresponding to the physical variables with the causal relationship into a group. As described in the example of the turbofan engine in S101, the rotation speed of the turbofan is related to the flow velocity of the air flow entering the compressor, the rotation speed of the turbofan is an independent variable, the flow velocity of the air flow of the compressor is a dependent variable, and a causal relationship exists between the rotation speed of the turbofan and the flow velocity of the air flow of the compressor, so that a component corresponding to the rotation speed of the turbofan and a component corresponding to the flow velocity of the air flow of the compressor are divided into a group.

Accordingly, the way of generating the sub-graph structure may be: the nodes represent physical variables and the direction of the edges points from the independent variables to the dependent variables.

In practical applications, the number of physical variables related to each component in the device may be large, and the relationship among the physical variables is chaotic. If according to the method, the edges in the generated hypergraph structure may intersect with each other, and the structure is disordered.

In order to solve the above problem, optionally, another grouping manner provided in the embodiment of the present application is as follows:

grouping the physical variables according to the relationship between the physical variables and the mathematical model; and dividing the parts corresponding to the physical variables in each group into one group.

Illustratively, the specific steps are as follows:

A. firstly, a variable matrix is established according to the relationship between the physical variable and the mathematical model.

The abscissa of the variable matrix represents the number of the mathematical model, and the ordinate represents the physical variable. Referring to fig. 3, a schematic diagram of a variable matrix provided in the embodiment of the present application is shown. The abscissa E1 to E7 shown in fig. 3 (a) represents 7 mathematical models, and the ordinate v1 to v8 represents 8 physical variables. The physical variable v1 belongs to a mathematical model E1 and a mathematical model E2, and elements exist in the position of the 1 st row and the 1 st column of the variable matrix and the position of the 1 st row and the 2 nd column of the variable matrix; the physical variable v2 belongs to the mathematical models E2 and E3, and there are elements in the position of the 2 nd column in the 2 nd row and the 3 rd column in the variable matrix. And so on. The relationship between the physical variables and the mathematical model can be visually seen from the figure.

The term "belonging relation" refers to that the mathematical model X represents a relation between the physical variables a and b, and if X = a + b, the physical variables a and b belong to the mathematical model X.

B. And carrying out region division on the variable matrix according to a preset rule.

The preset rules include:

in the overdetermined area, the number of the physical variables is less than that of the mathematical model;

in the positive definite area, the number of the physical variables is equal to the number of the mathematical models;

in the underdetermined region, the number of physical variables is greater than the number of mathematical models.

As shown in fig. 3 (a), the overdetermined area includes physical variables v1 and v2, and the mathematical models E1 to E3, i.e., the number of physical variables is smaller than the number of mathematical models. The positive definite region includes physical variables v3 and v4 and mathematical models E4 and E5, i.e. the number of physical variables is equal to the number of mathematical models. The physical variables in the underdetermined area are v5-v8, the mathematical models are E6 and E7, namely the number of the physical variables is larger than that of the mathematical models.

C. And dividing the parts corresponding to the physical variables in each area into a group.

Taking a turbofan engine as an example, the physical relationship includes the following mathematical model:

wherein F is a forward pushing force,

is the turbofan flow rate, V _exitfan Is the speed of the airflow at the outlet of the turbofan,

is the flow rate of the low-pressure gas compressor, V ₀ Is the air flow speed, P, of the low-pressure air compressor _fan For turbofan pressure, P ₀ Is the pressure at the inlet of the low-pressure gas compressor, P _core Is the pressure at the outlet of the low-pressure compressor, A _fan To extend the air flow for the turbofan A _core Expanding the flow for the low-pressure compressor, c _pc Is a constant of the high-pressure compressor, c _pt Is a high pressure turbine constant, T _HPT Is the high pressure turbine temperature, T _HPC For the temperature of the high-pressure compressor, h _pr Is energy per unit produced, eta _b For the combustor energy consumption, f for the high pressure turbine energy consumption, T _LPT To low pressure turbine temperature, T _LPC Is the low pressure turbine temperature, BPR is the bypass ratio, η _t Energy consumption of the high-pressure turbine.

Fig. 3 (b) shows a variable matrix corresponding to the turbofan engine. According to the above method, the variable matrix is divided into 3 regions, and the 3 regions are all underdetermined regions. And for the region (1), the parts corresponding to the physical variables comprise a turbofan and a low-pressure compressor, and the turbofan and the low-pressure compressor are divided into a group. And for the region (2), the parts corresponding to the physical variables comprise a high-pressure compressor, a high-pressure turbine and a combustor, and the high-pressure compressor, the high-pressure turbine and the combustor are divided into a group. And for the region (3), the parts corresponding to the physical variables comprise a low-pressure turbine, a low-pressure compressor, a high-pressure turbine and a high-pressure compressor, and the low-pressure turbine, the low-pressure compressor, the high-pressure turbine and the high-pressure compressor are divided into a group.

The values of the physical variables can all be obtained by sensors. For example, an airflow speed sensor is arranged at the outlet of the turbofan, and the airflow speed V can be detected as the outlet airflow speed of the turbofan _exitfan . For another example, a pressure sensor is installed at the inlet of the low pressure gas compressor, and the pressure P at the inlet of the low pressure gas compressor can be detected ₀ 。

Optionally, the generation manner of the sub-graph structure includes:

if the number of the physical variables corresponding to the group is larger than that of the mathematical models, and the number of coincident variables is larger than or equal to 2, generating a second sub-graph from the components contained in the group, wherein the second sub-graph is a non-directional sub-graph structure, and the number of connection nodes at one end of each edge in the second sub-graph is an integer larger than or equal to 0, and the coincident variables are the physical variables contained in each mathematical model corresponding to the group;

if the number of the physical variables corresponding to the group is larger than that of the mathematical model and the number of coincident variables is smaller than 2, generating a third sub-graph by using the components contained in the group, wherein the third sub-graph is a directed sub-graph structure and the number of connecting nodes at one end of each edge in the third sub-graph is an integer larger than or equal to 0.

In the above generation manner, it is equivalent to generating a conventional graph structure for the overdetermined region and the positive definite region; and generating a hypergraph structure aiming at the underdetermined area. In a conventional graph structure, one end of each edge is connected with at most one node, while in a hypergraph structure, one end of each edge can be connected with a plurality of nodes. Compared with the traditional graph structure, the hypergraph structure has stronger characterization and mining capabilities of nonlinear high-order association between data, and can more accurately model the multivariate relation. The method adopts the mode of fusing the hypergraph structure and the traditional graph structure, and can accurately express the physical relationship among all the parts in the equipment.

The edges in the sub-graph structure are determined according to the mathematical model to which the components contained in each group belong. Specifically, the mathematical model to which each node belongs in common is used as an edge connecting each node.

Exemplarily, refer to fig. 4, which is a schematic diagram of a sub-graph structure provided in the embodiment of the present application. As shown in fig. 4 (a), the sub-graph structure is generated according to the overdetermined region in fig. 3 (a), nodes in the sub-graph structure are divided into n1 (a component corresponding to the physical variable v 1) and n2 (a component corresponding to the physical variable v 2), a mathematical model to which the two nodes belong together is E2, and the two nodes are connected by a common directed edge, and the edge is E2. As shown in fig. 4 (b), the sub-graph structure is generated from the positive definite region in fig. 3 (a), nodes in the sub-graph structure are n2 (component corresponding to physical variable v 3) and n3 (component corresponding to physical variable v 4), respectively, a common mathematical model to which the nodes belong is E4, and the two nodes are connected by a common directed edge, and the edge is E4. As shown in fig. 4 (c), a sub-graph structure is generated according to the underdetermined region in fig. 3 (a), nodes in the sub-graph structure are n3 (component corresponding to physical variable v 5), n4 (component corresponding to physical variable v 6), n5 (component corresponding to physical variable v 7) and n6 (component corresponding to physical variable v 8), the number of overlapped variables is 2, v5-v7 belong to the mathematical model E6 in common, v6-v8 belong to the mathematical model E7 in common, and four nodes are connected by using undirected super edges, where an edge between n3-n5 is E6 and an edge between n4-n6 is E7. Finally, the sub-graph structures are combined into a super graph structure, as shown in (d) of fig. 4.

Fig. 5 is a schematic diagram of a hypergraph structure according to another embodiment of the present application. The hypergraph structure shown in (a) in fig. 5 is a hypergraph structure generated from the variable matrix shown in (b) in fig. 3. The three regions shown in (b) in fig. 3 are all underdetermined regions, the part corresponding to the physical variable included in the region (1) is a turbofan and a low-pressure compressor, the mathematical model to which the turbofan and the low-pressure compressor belong together is E1, and because the region only includes one mathematical model, the turbofan and the low-pressure compressor can be connected by using a common directed edge, and the edge is E1, so that a subgraph structure is obtained. The parts corresponding to the physical variables in the region (2) are a high-pressure compressor, a high-pressure turbine and a combustor, the high-pressure compressor, the high-pressure turbine and the combustor belong to a mathematical model E2, the number of coincident variables is 3, so that the high-pressure compressor, the high-pressure turbine and the combustor are connected by using a non-directional overcide, and the overcide is E2 to obtain a sub-graph structure. The parts corresponding to the physical variables in the region (3) are a low-pressure turbine, a low-pressure compressor, a high-pressure turbine and a high-pressure compressor, the mathematical models of the four are E4, the number of coincident variables is 4, the low-pressure turbine, the low-pressure compressor, the high-pressure turbine and the high-pressure compressor are connected by using a non-directional super edge, and the super edge is E4, so that a subgraph structure is obtained. The three sub-graph structures are then combined into a hypergraph structure.

In one embodiment, S102 may further include:

I. and generating a hypergraph matrix according to the hypergraph structure.

II. And training the detection model according to the hypergraph matrix and multiple groups of training data to obtain the trained detection model.

The hypergraph structure is an intuitive map structure and cannot be identified by the processor. To address this issue, the hypergraph structure may be generated into a hypergraph matrix of data structures for recognition by the processor.

Optionally, the abscissa of the hypergraph matrix represents the number of an edge in the hypergraph structure, and the ordinate of the hypergraph matrix represents the number of a node in the hypergraph structure.

In the prior art, the generation method of the hypergraph matrix is generally as follows: setting elements in the hypergraph matrix to be 0 or 1; the elements in the hypergraph matrix may also be set to random values.

In the embodiment of the present application, a generation method of a hypergraph matrix is provided:

Wherein i and j are positive integers.

Compared with the conventional generation mode of the hypergraph matrix, the hypergraph matrix generated by the method provided by the embodiment of the application can more accurately express the hypergraph structure, and further the accuracy of subsequent detection is ensured.

Exemplarily, since the hypergraph structure shown in (a) of fig. 5 involves 3 sides E1, E2, and E3, accordingly, the abscissa of the hypergraph matrix shown in (b) of fig. 5 is E1, E2, and E3, respectively. Since the hypergraph structure shown in fig. 5 (a) involves 6 components (i.e., 6 nodes) of the turbofan, the low-pressure air compressor, the high-pressure turbine, the low-pressure turbine, and the combustor, in the hypergraph matrix shown in fig. 5 (b), the vertical coordinates of the hypergraph structure are the turbofan, the low-pressure air compressor, the high-pressure compressor, the combustor, the high-pressure turbine, and the low-pressure turbine in sequence according to the connection sequence of the hypergraph structure (the mechanical connection sequence of the components in the equipment can also be referred to). As the turbofan and the low-pressure compressor are connected in a directed mode, the numerical value of the element in the 1 st row and the 1 st column in the hypergraph matrix is set to be-1 according to the predefined direction; correspondingly, the value of the 1 st column in the 2 nd row of the hypergraph matrix is 1. Similarly, if the value of the element in the 1 st row and the 1 st column in the hypergraph matrix is set to be 1; correspondingly, the value of the 2 nd row and 1 st column in the hypergraph matrix is-1.

It should be noted that, in the hypergraph matrix shown in fig. 5, the non-zero element is set to 1/-1, and in practical application, the non-zero element may be set to different values according to practical needs, which is not specifically limited herein.

Optionally, step II may further include:

1. and carrying out normalization processing on the training detection data to obtain normalized data.

In order to reject extreme data, the training detection data needs to be normalized. Illustratively, it can be according to a formula

Carrying out normalization treatment, wherein Xnol is data after normalization treatment, X is data to be treated, and X is _max Is the maximum value, X, in the same attribute data _min Is the minimum value in the same attribute data. It should be noted that, in general, the normalization process is performed on the same attribute data obtained by the same sensor.

2. And denoising the normalized data to obtain denoised data.

In order to reduce noise mixed in the sensor detection data and improve the detection accuracy, the data is subjected to denoising processing in the embodiment of the application. The existing data denoising processing method may be adopted, and is not specifically limited herein.

Illustratively, a fast Fourier transform denoising method can be used, according to a formula

Performing data de-noising processing, wherein X _k For denoised data, k is a fourier coefficient, and N represents the number of times a certain sensor is sampled.

3. And sorting the de-noising data according to the arrangement sequence of the components in the hypergraph matrix to obtain sorted data.

Illustratively, the hypergraph matrix shown in fig. 5 (b) is arranged in the order of a turbofan, a low pressure gas compressor, a high pressure compressor, a combustor, a high pressure turbine, and a low pressure turbine. Correspondingly, the physical variable corresponding to the turbofan is F, A _fan 、

V _exitfan And P _fan (ii) a Physical variables corresponding to low-pressure compressorHas A _core 、

V ₀ 、P ₀ 、P _core And T _LPC (ii) a And arranging the sensor detection value of the physical variable corresponding to the turbofan before the sensor detection value of the physical variable corresponding to the low-pressure compressor.

4. And inputting the sequencing data and the hypergraph matrix into the detection model for training to obtain the trained detection model.

The inspection model in the embodiment of the present application may include a hypergraph convolution layer, a graph aggregation layer, and a full connection layer.

Optionally, the hypergraph convolution layer in the detection model may adopt a residual convolution neural network, a graph convolution neural network, a deep neural network, a random forest model, a logistic regression model, a support vector machine model, and the like. But in contrast, the detection accuracy of the residual convolutional neural network is higher.

In the embodiment of the application, the hypergraph convolution layer adopts a residual convolution neural network, and the mathematical model of the hypergraph convolution layer is as follows:

X ^l+1 ＝X ^l +F(X _l ,W _l )；

wherein F () represents the corresponding forward network in the residual convolutional neural network, l is the number of layers of the residual convolutional neural network, W _l Is the weight of the l-th layer, X ^l Is data of layer I, X ^l+1 Is the data of layer l + 1. In a residual convolutional neural network, the output of each layer serves as the input to the next layer.

The iterative formula of the residual convolutional neural network is as follows:

wherein alpha is _l And beta _l Is a model parameter, I is an identity matrix,

to represent the hypergraph matrix, Θ ^l Represents the convolution kernel of the l-th layer.

Since the dimensions of the sensor detection data corresponding to each of the plurality of components may be different, it is necessary to aggregate the sensor detection data by the graph aggregation layer.

The iterative formula of the full connection layer is:

wherein, the first and the second end of the pipe are connected with each other,

is the vector after graph aggregation, W _o And b _o Respectively, the weight and bias terms of the fully connected layer, and Y is the fully connected output.

The termination condition for training the detection model can be iteration times or a loss value smaller than a preset value.

In the training process, multiple groups of historical detection data can be divided into training detection data and testing detection data, the training detection data is used for training the detection model, and when the training termination condition is reached, the testing detection data is used for testing the trained detection model so as to ensure the detection precision of the detection model.

The neural network has good learning ability, and the trained detection model can accurately identify the physical relationship among all the parts in the equipment through training the detection model, so that the accuracy of the detection result is ensured.

Based on the model training method described in the above embodiment, a trained detection model is obtained, and the model is used for equipment fault detection. The fault detection procedure is described below. Referring to fig. 6, which is a schematic flow chart of a detection method provided in the embodiment of the present application, by way of example and not limitation, the method may include the following steps:

s601, acquiring real-time detection data of each part in the equipment to be detected through a sensor.

S602, obtaining the detection result of the equipment to be detected according to the real-time detection data and the trained detection model, wherein the trained detection model is obtained according to the hypergraph structure training of the equipment to be detected, the nodes of the hypergraph structure represent the parts of the equipment to be detected, and the edges of the hypergraph structure represent the physical relationship among the parts of the equipment to be detected.

In step S602, the real-time detection data also needs to be preprocessed, and the preprocessing process is the same as the preprocessing process of the training detection data, which may be specifically referred to the description in step II above, and is not described herein again.

In the embodiment of the application, the incidence relation among all the parts in the equipment to be detected is represented through the hypergraph structure, the detection model is trained according to the hypergraph structure, namely the detection model is trained according to the incidence relation among all the parts in the equipment to be detected, so that the trained detection model can automatically identify the incidence relation among the real-time detection data acquired by the sensor, and the equipment fault detection is carried out according to the incidence relation. By the method, when the equipment fault is detected, the association among all parts in the equipment is fully considered, and the accuracy and reliability of the detection result are effectively improved. In addition, the detection model can be trained offline, and online real-time detection is carried out by using the trained detection model, so that the detection time can be greatly saved, and the detection efficiency is effectively improved.

It should be understood that, the sequence numbers of the steps in the foregoing embodiments do not imply an execution sequence, and the execution sequence of each process should be determined by its function and inherent logic, and should not constitute any limitation to the implementation process of the embodiments of the present application.

Fig. 7 is a block diagram of a detecting apparatus provided in the embodiment of the present application, corresponding to the detecting method described in the foregoing embodiment, and only the relevant portions of the embodiment of the present application are shown for convenience of description.

Referring to fig. 7, the apparatus includes:

and the data acquisition unit 71 is used for acquiring real-time detection data of each part in the device to be detected through the sensor.

And the fault detection unit 72 is used for obtaining the detection result of the equipment to be detected according to the real-time detection data and the trained detection model, wherein the trained detection model is obtained according to the hypergraph structure training of the equipment to be detected, the nodes of the hypergraph structure represent the parts of the equipment to be detected, and the edges of the hypergraph structure represent the physical relationship among the parts in the equipment to be detected.

Optionally, the apparatus 7 further comprises:

the model training unit 73 is used for generating a hypergraph structure of the equipment to be detected according to the physical relationship among all parts in the equipment to be detected; and training the detection model according to the hypergraph structure and multiple groups of training data to obtain the trained detection model, wherein each group of training data comprises a group of historical detection data and label types of all parts in the equipment to be detected.

Optionally, the model training unit 73 is further configured to:

generating a sub-graph structure from the components contained in each group;

and generating the hypergraph structure of the device to be detected by the sub-graph structure.

Optionally, the physical relationship includes at least one mathematical model, each mathematical model includes at least 2 physical variables, the mathematical model is used for representing an operation mechanism of the device to be detected, and values of the physical variables are obtained through the sensor.

Correspondingly, the model training unit 73 is further configured to:

Optionally, the model training unit 73 is further configured to:

generating a hypergraph matrix according to the hypergraph structure;

Correspondingly, the model training unit 73 is further configured to:

Optionally, the model training unit 73 is further configured to:

denoising the normalized data to obtain denoised data;

It should be noted that, for the information interaction, execution process, and other contents between the above-mentioned devices/units, the specific functions and technical effects thereof are based on the same concept as those of the embodiment of the method of the present application, and specific reference may be made to the part of the embodiment of the method, which is not described herein again.

The detection device shown in fig. 7 may be a software unit, a hardware unit, or a combination of software and hardware unit built in the existing terminal device, may be integrated into the terminal device as a separate pendant, or may exist as a separate terminal device.

It will be apparent to those skilled in the art that, for convenience and brevity of description, only the above-mentioned division of the functional units and modules is illustrated, and in practical applications, the above-mentioned function distribution may be performed by different functional units and modules according to needs, that is, the internal structure of the apparatus is divided into different functional units or modules to perform all or part of the above-mentioned functions. Each functional unit and module in the embodiments may be integrated in one processing unit, or each unit may exist alone physically, or two or more units are integrated in one unit, and the integrated unit may be implemented in a form of hardware, or in a form of software functional unit. In addition, specific names of the functional units and modules are only for convenience of distinguishing from each other, and are not used for limiting the protection scope of the present application. The specific working processes of the units and modules in the system may refer to the corresponding processes in the foregoing method embodiments, and are not described herein again.

Fig. 8 is a schematic structural diagram of a terminal device according to an embodiment of the present application. As shown in fig. 8, the terminal device 8 of this embodiment includes: at least one processor 80 (only one shown in fig. 8), a memory 81, and a computer program 82 stored in the memory 81 and executable on the at least one processor 80, the processor 80 implementing the steps in any of the various detection method embodiments described above when executing the computer program 82.

The terminal device can be a desktop computer, a notebook, a palm computer, a cloud server and other computing devices. The terminal device may include, but is not limited to, a processor, a memory. Those skilled in the art will appreciate that fig. 8 is merely an example of the terminal device 8, and does not constitute a limitation of the terminal device 8, and may include more or less components than those shown, or combine some components, or different components, such as an input-output device, a network access device, and the like.

The Processor 80 may be a Central Processing Unit (CPU), and the Processor 80 may be other general purpose Processor, a Digital Signal Processor (DSP), an Application Specific Integrated Circuit (ASIC), an off-the-shelf Programmable Gate Array (FPGA) or other Programmable logic device, discrete Gate or transistor logic, discrete hardware components, etc. A general purpose processor may be a microprocessor or the processor may be any conventional processor or the like.

The memory 81 may in some embodiments be an internal storage unit of the terminal device 8, such as a hard disk or a memory of the terminal device 8. The memory 81 may also be an external storage device of the terminal device 8 in other embodiments, such as a plug-in hard disk, a Smart Media Card (SMC), a Secure Digital (SD) Card, a Flash memory Card (Flash Card), and the like, provided on the terminal device 8. Further, the memory 81 may also include both an internal storage unit and an external storage device of the terminal device 8. The memory 81 is used for storing an operating system, an application program, a Boot Loader (Boot Loader), data, and other programs, such as program codes of the computer programs. The memory 81 may also be used to temporarily store data that has been output or is to be output.

The embodiments of the present application further provide a computer-readable storage medium, where a computer program is stored, and when the computer program is executed by a processor, the computer program implements the steps in the above-mentioned method embodiments.

The embodiments of the present application provide a computer program product, which when running on a terminal device, enables the terminal device to implement the steps in the above method embodiments when executed.

The integrated unit, if implemented in the form of a software functional unit and sold or used as a stand-alone product, may be stored in a computer readable storage medium. Based on such understanding, all or part of the processes in the methods of the embodiments described above can be implemented by a computer program, which can be stored in a computer-readable storage medium and can implement the steps of the embodiments of the methods described above when the computer program is executed by a processor. Wherein the computer program comprises computer program code, which may be in the form of source code, object code, an executable file or some intermediate form, etc. The computer-readable medium may include at least: any entity or device capable of carrying computer program code to an apparatus/terminal device, recording medium, computer Memory, read-Only Memory (ROM), random-Access Memory (RAM), electrical carrier wave signals, telecommunications signals, and software distribution medium. Such as a usb-disk, a removable hard disk, a magnetic or optical disk, etc. In certain jurisdictions, computer-readable media may not be an electrical carrier signal or a telecommunications signal in accordance with legislative and patent practice.

In the above embodiments, the description of each embodiment has its own emphasis, and reference may be made to the related description of other embodiments for parts that are not described or recited in any embodiment.

Those of ordinary skill in the art will appreciate that the various illustrative elements and algorithm steps described in connection with the embodiments disclosed herein may be implemented as electronic hardware or combinations of computer software and electronic hardware. Whether such functionality is implemented as hardware or software depends upon the particular application and design constraints imposed on the implementation. Skilled artisans may implement the described functionality in varying ways for each particular application, but such implementation decisions should not be interpreted as causing a departure from the scope of the present application.

In the embodiments provided in the present application, it should be understood that the disclosed apparatus/terminal device and method may be implemented in other ways. For example, the above-described embodiments of the apparatus/terminal device are merely illustrative, and for example, the division of the modules or units is only one type of logical function division, and other division manners may be available in actual implementation, for example, multiple units or components may be combined or integrated into another system, or some features may be omitted, or not executed. In addition, the shown or discussed mutual coupling or direct coupling or communication connection may be an indirect coupling or communication connection through some interfaces, devices or units, and may be in an electrical, mechanical or other form.

The units described as separate parts may or may not be physically separate, and parts displayed as units may or may not be physical units, may be located in one place, or may be distributed on a plurality of network units. Some or all of the units can be selected according to actual needs to achieve the purpose of the solution of the embodiment.

The above-mentioned embodiments are only used to illustrate the technical solutions of the present application, and not to limit the same; although the present application has been described in detail with reference to the foregoing embodiments, it should be understood by those of ordinary skill in the art that: the technical solutions described in the foregoing embodiments may still be modified, or some technical features may be equivalently replaced; such modifications and substitutions do not substantially depart from the spirit and scope of the embodiments of the present application and are intended to be included within the scope of the present application.

Claims

1. A method of detection, comprising:

2. The detection method of claim 1, further comprising:

3. The inspection method of claim 2, wherein said generating a hypergraph structure of said device under inspection based on physical relationships between components in said device under inspection comprises:

generating a sub-graph structure from the components contained in each group;

4. The detection method according to claim 3, characterized in that said physical relationship comprises at least one mathematical model, each mathematical model comprising at least 2 physical variables, said mathematical model being intended to represent the operating mechanism of the device under test, the values of said physical variables being obtained by means of said sensors;

according to the physical relation between each part in the equipment to be detected, each part in the equipment to be detected is grouped, and the method comprises the following steps:

5. The detection method according to claim 4, wherein generating a sub-graph structure from the components contained in each group comprises:

for each group, if the number of the physical variables corresponding to the group is less than or equal to the number of the mathematical model, generating a first subgraph from the components contained in the group, wherein the first subgraph is a directed subgraph structure, and one end of each edge in the first subgraph is connected with at most one node;

6. The method as claimed in claim 4, wherein said training said detection model based on said hypergraph structure and a plurality of sets of training data to obtain said trained detection model comprises:

generating a hypergraph matrix according to the hypergraph structure;

7. The inspection method of claim 6, wherein the abscissa of the hypergraph matrix represents the number of edges in the hypergraph structure and the ordinate of the hypergraph matrix represents the number of nodes in the hypergraph structure;

8. The method as claimed in claim 6, wherein said training said detection model based on said hypergraph matrix and a plurality of sets of training data to obtain said trained detection model comprises:

denoising the normalized data to obtain denoised data;

9. A terminal device comprising a memory, a processor and a computer program stored in the memory and executable on the processor, characterized in that the processor implements the method according to any of claims 1 to 8 when executing the computer program.

10. A computer-readable storage medium, in which a computer program is stored which, when being executed by a processor, carries out the method according to any one of claims 1 to 8.