CN108615047B

CN108615047B - Fault diagnosis knowledge model construction method for wind turbine generator equipment

Info

Publication number: CN108615047B
Application number: CN201810242945.2A
Authority: CN
Inventors: 王宏伟
Original assignee: Shaoxing Nuoleizhi Information Technology Co ltd
Current assignee: Shaoxing Nuoleizhi Information Technology Co ltd
Priority date: 2018-03-23
Filing date: 2018-03-23
Publication date: 2022-07-01
Anticipated expiration: 2038-03-23
Also published as: CN108615047A

Abstract

The invention discloses a method for constructing a fault diagnosis knowledge model for wind turbine equipment, which comprises the following steps of: (1) establishing top-level classification of wind turbine generator fault diagnosis, and establishing a fault system comprising different fault analysis problems; (2) aiming at a specific problem in the fault system, a cause chain is established firstly; (3) and judging whether the process chain or the resource chain needs to be established or not according to the specific knowledge nodes in the routing chain, and if so, establishing the process chain or the resource chain. According to the invention, explicit knowledge and implicit knowledge in the fault diagnosis of the unit equipment can be more conveniently and quickly represented in an integrated manner.

Description

Fault diagnosis knowledge model construction method for wind turbine generator equipment

Technical Field

The invention relates to the field of fault diagnosis, in particular to a method for constructing a fault diagnosis knowledge model for wind turbine equipment.

Background

Wind power is a clean renewable energy source, and wind energy utilization is vigorously advocated by governments of various countries in the world since the beginning of the century. China has abundant wind energy resources, has unique advantages in developing wind power technology, and the installed capacity of a wind turbine generator is continuously improved in the past 10 years. With more and more wind turbine generators being continuously installed and put into operation, operation and maintenance become a major problem faced by wind power plants. First, wind turbines are usually installed in suburbs or on hills, with the turbines very dispersed and far apart. Secondly, the main equipment of the wind turbine is arranged on a tower with the height of dozens of meters, and the close-range monitoring and debugging are very difficult. And thirdly, the operation conditions of the wind turbine generator are usually complex, and the prediction of the power generation and the operation conditions of the wind turbine generator is very difficult.

In this context, many wind power plants have begun to invest in hardware devices and software systems for condition monitoring and fault diagnosis, thereby providing efficient support for wind turbine operation and maintenance. The existing system and method focus on monitoring the running state by monitoring important index parameters in real time, fault diagnosis mainly depends on experts to complete, and fault diagnosis results and schemes are provided by writing reports. Fault diagnosis is a very complex process, highly dependent on expert knowledge and experience, and efficient capture and reuse of this knowledge is of great importance. The prior art such as fault tree analysis or root cause analysis can support classification analysis and fault root cause reasoning of faults to some extent, and is difficult to support capture and reuse of diagnosis knowledge. First, while it is possible to record the type and cause of a fault, it is not efficient to record diagnostic process knowledge. Secondly, the systematic integrated description of the explicit knowledge and the implicit knowledge cannot be carried out. Thirdly, the structured knowledge description is difficult to improve by reporting the document record diagnosis knowledge, and the information granularity is coarse, which is not beneficial to efficient reuse.

The invention provides a method for constructing a fault diagnosis knowledge model for wind turbine equipment, so that explicit knowledge and implicit knowledge in fault diagnosis of the wind turbine equipment can be expressed more conveniently and quickly in an integrated manner. Compared with the prior art, the method and the system solve the problem of model description of complex diagnosis knowledge, so that capture, storage and reuse of the knowledge are supported. Compared with a method based on a simple model and written report record knowledge, the method has the following excellent effects: firstly, different knowledge segments in fault diagnosis are efficiently integrated through a resource chain, a process chain and a cause chain, more complete knowledge information is provided, and knowledge understanding and reusing are facilitated; secondly, by means of structural graphic knowledge representation, fine-grained knowledge description is achieved, and capture, recording and reuse of knowledge are facilitated; third, diagnostic process knowledge is effectively captured and recorded through the process chain; fourthly, factual knowledge (such as measured data and understanding of parameters) and implicit knowledge (such as experience of analyzing reasons) which will be dominant through resource chains and cause chains are efficiently organized; fifthly, better support is provided for methods such as knowledge retrieval and intelligent knowledge recommendation based on calculation through connection among knowledge elements.

Disclosure of Invention

In order to realize the purpose of the invention, the following technical scheme is adopted for realizing the purpose:

a method for constructing a fault diagnosis knowledge model for wind turbine equipment comprises the following steps:

(1) establishing top-level classification of wind turbine generator fault diagnosis, and establishing a fault system comprising different fault analysis problems;

(2) aiming at a specific problem in the fault system, a cause chain is established firstly;

(3) and judging whether the process chain or the resource chain needs to be established or not according to the factor chain, and if so, establishing the process chain or the resource chain.

The construction method, wherein:

in step (2), a causal chain is established as follows: creating a top-level question node of the routing chain, and creating an answer node aiming at the question node; judging whether the answer node can be subjected to simple description for affirmation or overturn or needs to establish a detailed extended knowledge chain; and if the extension is needed, establishing a corresponding process chain or a corresponding resource chain according to the situation.

The construction method, wherein: if no expansion is required, positive and/or negative cause nodes diagnosing the problem node are established and corresponding information is filled in.

The construction method, wherein: the process of establishing the problem node continues to iterate until it can be determined that the cause of the fault has been found and it is confirmed that the diagnostic process can end.

The construction method, wherein: the process chain is used to describe a measurement, analysis or diagnostic process.

The construction method, wherein: the process chain includes step nodes, each node being used to describe the main information of the step.

The construction method, wherein: if more knowledge needs to be captured for a step, the node is further extended to a new cause chain and resource chain for that step.

The construction method, wherein: the resource chain includes interpretations of proper nouns, index parameters, standard methods, and theorem's laws, as well as invocations and interpretations of relevant data.

The construction method, wherein: a resource chain is made up of a connection of resource nodes, each describing an aspect of the resource, the resource nodes being interconnected to provide progressively refined information.

The construction method, wherein: besides establishing positive and/or negative cause nodes for diagnosing the problem node, the undetermined cause node is also established and filled with corresponding information.

The construction method, wherein: the resource chain describes explicit knowledge; the cause chain describes implicit knowledge and the process chain describes explicit knowledge and/or implicit knowledge.

The construction method, wherein: explicit knowledge includes formulas, data sheets, conceptual sketches, graphs, three-dimensional models and standard interpretation of terminology recorded by the computer during diagnosis; implicit knowledge includes knowledge and experience provided by experts to support decision making.

The construction method, wherein: specific nodes in each chain may extend other chains to complete the knowledge record.

The construction method is characterized in that: process or resource chains may be added by nodes in the chain to provide expanded information or evidence; a certain step node in the process chain may add a cause chain to explain how to do the cause.

The construction method, wherein: the method also comprises a step of carrying out term classification identification on the factor chain, the process chain or the resource chain.

The construction method, wherein: each node contained in the reason chain, the process chain and the resource chain can be called a knowledge node, the text information contained in each knowledge node describes knowledge content, and the term classification and identification step refers to element identification and classification of the knowledge content.

The construction method, wherein the knowledge content is classified and identified as one or more of 14 element forms, the 14 elements comprising: wind power plant, wind turbine name, professional organization, staff, wind turbine manufacturer, analysis/execution process, fault, analysis/execution step, subsystem, component, parameter name, cause, physical quantity, model/method.

The construction method, wherein: the semantic relationship of the 14 elements is described as follows: the wind power plant is an operation unit of the fan; the fan manufacturer is a manufacturing unit of the fan; professional institutions refer to research and engineering units related to wind turbine research, maintenance and fault handling; the above mechanisms all comprise staff; the name of the fan indicates a specific fan; one fan is composed of a plurality of subsystems; one subsystem is composed of a plurality of components; the subsystems or components all have a plurality of parameters; each parameter is associated with a certain physical quantity; the staff carries out the analysis/execution process; the analysis/execution process includes several analysis/execution steps; these steps require the use of a model/method; these models/methods are associated with specific physical quantities; the fault refers to the fault of a fan, a subsystem, a component and the like; cause refers to the possible cause of the failure.

Drawings

FIG. 1 is a schematic diagram of a wind turbine generator equipment fault diagnosis knowledge model;

FIG. 2 is a flow chart of a method of constructing a fault diagnosis knowledge model of wind turbine equipment;

FIG. 3 is a schematic diagram of a term recognition classification ontology model

Detailed Description

The invention is further illustrated by the following examples in conjunction with the drawings.

Fig. 1 is a structural diagram of a knowledge model for diagnosing a fault of a wind turbine, which is configured to construct a knowledge unit by using three knowledge chains and perform multidimensional and multi-engineering semantic connection, that is, a cause chain, a process chain and a resource chain. The model continuously performs reasoning, demonstration and support evidence and reason from top to bottom through the chain structures, supports deep knowledge decomposition and classification, can describe a high-level fault problem and support the gradual establishment of nodes to analyze specific subproblems and effectively capture and record the process, reason and fact knowledge in the fault problem.

The cause chain is used for describing a problem analysis and fault reasoning process, and is specifically described by four types of nodes: question nodes, answer nodes, positive reason nodes and negative reason nodes. Where the problem node is a description of a particular fault problem (e.g., a drop in fan output ratio, gearbox noise anomaly, etc.). The answer node is a possible answer to the fault problem described by the problem node, the answer node is connected with the problem node through a connecting line, and one problem node can have a plurality of answer nodes. The answer node may simply describe an answer, and if the answer is more complex, a sub-answer node may be established for the answer node. The positive reason node and the negative reason node are the basis for providing support for a certain answer node or overriding the answer node, the positive node or the negative node can simply describe the positive reason or the negative reason, and can further establish detailed description of child nodes or establish other extended knowledge chains, namely, a factor chain, a process chain and a resource chain can be further constructed, so that all considered factors in the diagnosis process including the reason that the factors are accepted or rejected finally can be recorded. As shown in fig. 1, when a node of a cause chain needs to be further explained, such as a detailed description of a measurement process or a key data structure for supporting understanding, the further explanation can be extended to an external process chain or resource chain.

As mentioned above, the present invention creatively provides the concept of negative reason node, and the positive reason node and the negative node are used together, and the positive node and the negative node are compared to the advantage and the disadvantage of a scheme, wherein the diagnosis of the answer is that for an answer, both positive and negative cause are existed, so that the negative answer is recorded, and the negative cause is also existed, so that the detailed knowledge can be captured and recorded.

In addition, a reason to be determined node can be set, because when the positive or negative result of the answer node is diagnosed or evaluated, the positive or negative result is sometimes difficult to obtain, and the answer is certainly valuable, so that the reason to be determined node is constructed in addition to the positive node and the negative node, the reason to be determined or the basis of suspicion is simply described in the node, and half of the reason to be determined node needs to further establish detailed description of the child node or establish an extended other knowledge chain to finally obtain the result that the answer is positive or negative.

A process chain can be used to describe a measurement, analysis or diagnostic process, which is linked through a series of steps to accurately record detailed information about the process. The process chain is formed by step nodes, and main information of the step is briefly described in text in each node. If more knowledge needs to be captured for a step, the node can be further extended into new cause chains and resource chains for that step, the same reasoning as described above for extending other knowledge chains from cause chains.

The resource chain comprises interpretations of proper nouns, index parameters, standard methods, theorem laws and the like, and can also describe and interpret data of the index parameters. Resource chains are made up of connections of resource nodes, each of which can simply describe some aspect of a resource, providing richer information through the interconnections. The resources are: for interpretation of terms, for interpretation of parameters, for interpretation of data, etc., resources are used to supplement factors by nodes inside, provide evidence, or provide additional information. After the resource nodes are connected with each other, the connection relations also become useful information, and deep knowledge and rich context can be recorded.

In summary, the knowledge model constructed based on the three knowledge chains only includes six types of nodes (i.e., question nodes, answer nodes, positive nodes, negative nodes, step nodes, and resource nodes), and can flexibly describe multi-level and multi-dimensional fault diagnosis knowledge.

Another innovation of the knowledge model construction method provided by the invention is that the integration of explicit knowledge and implicit knowledge is effectively supported through structural model description. Explicit knowledge refers to formulas, data sheets, conceptual sketches, graphs, three-dimensional models, and standard interpretation of terms that are easy to record by a computer during a diagnostic procedure, and the like, which are used to provide data and evidence support for the diagnostic procedure. Implicit knowledge refers to the knowledge and experience that experts have to support decision making decisions, which are generally not easily computer-recordable for supporting diagnostic process implementation and interpretation of critical decision making decisions, such as knowing which factors to analyze and how to develop, affirm or negate certain possible causes of failure, why to do something, and so on. The resource chain in the knowledge construction method can describe explicit knowledge, the implicit knowledge can be described by the resource chain, and the two types of knowledge can be described by the process chain. In the actual knowledge construction process, specific nodes in each chain can be extended to other chains to complete knowledge recording, for example, a node in a process chain can be added with the process chain or a resource chain to provide expanded information or evidence, and a node at a certain step in the process chain can be added with a reason for explaining the process. Through the free connection of the nodes, the diagnosis process knowledge record driven by the integration of explicit knowledge and implicit knowledge can be effectively realized.

Fig. 2 is a flowchart of a method for constructing a wind turbine generator-oriented fault diagnosis knowledge model, which is provided by the present invention, and specifically includes the following steps: firstly, establishing top-level classification of wind turbine generator fault diagnosis, so as to unify various fault analysis problems into a fault system of a system; for a specific problem in the system, an integrated knowledge model can be used for describing and capturing diagnosis knowledge, and a cause chain is established: creating a top problem node of the routing chain; aiming at the question node, creating an answer node; judging whether the answer node can be subjected to simple description for affirmation, overturn or undetermined or whether a detailed extended knowledge chain needs to be established for further recording the related analysis and diagnosis knowledge; if extension is needed, a corresponding process chain or a resource chain is established according to the situation, the specific process is similar to the establishment process of the cause chain, if extension is not needed, positive and negative or undetermined cause nodes for diagnosing the problem node are established, and corresponding information is filled in. The process of establishing the problem node will continue to iterate until it can be determined that the cause of the fault has been found and it is confirmed that the diagnostic process can end. In this way, the entire diagnostic process can be recorded, i.e. the reasons for the negation and the work done for them can be recorded. The establishment process of the knowledge chain expanded from a certain node is similar to the process, and a new knowledge chain can be further expanded from the internal node.

On the basis, a fault diagnosis knowledge model (which can be called as a primary fault diagnosis model) is obtained, and further, the fault can be subjected to term classification identification on the knowledge model so as to construct an advanced fault diagnosis model.

The method comprises the steps of carrying out term classification recognition on a cause chain, a process chain and a resource chain in a primary barrier diagnosis knowledge model according to a term recognition classification ontology model shown in FIG. 3, wherein each node contained in the cause chain, the process chain and the resource chain can be called a "knowledge node", and text information contained in each knowledge node describes "knowledge content". Fig. 3 shows an ontology structure of term classification recognition, where the ontology structure includes a plurality of main elements involved in a fault diagnosis process, such as a fan name, a subsystem, a component, a parameter name, a fault, and the like, and terms in the term classification recognition refer to vocabularies appearing in knowledge content and the like, and by classifying the vocabularies according to the ontology structure, a computer can judge and understand information content in a cause chain, a process chain, and a resource chain, and support more efficient knowledge retrieval and reuse.

As shown in fig. 3, a total of 14 elements constitute a semantic description framework of the fault diagnosis overall process. The semantic relationship is described as follows: the wind power plant is an operation unit of the fan; the fan manufacturer is a manufacturing unit of the fan; professional institutions refer to research and engineering units related to wind turbine research, maintenance, fault handling and the like; the mechanisms comprise workers; the name of the fan indicates a specific fan, such as a No. 1 unit of an A power plant; a fan is composed of a plurality of subsystems, such as a pitch system, a yaw system, blades, a tower and the like; one subsystem is composed of a plurality of components, such as a low-speed shaft, a high-speed shaft and the like in a gear box; subsystems or components all have many parameters such as power generation, pitch angle, oil temperature, high speed shaft vibration, etc.; each parameter is associated with some physical quantity, such as oil temperature and temperature; the staff carries out analysis/execution processes, such as spindle frequency analysis; the analysis/execution process comprises a plurality of analysis/execution steps such as theoretical generated power calculation, data calibration and the like; models/methods such as time domain analysis, theoretical power calculation models and the like are needed in the steps, and the models/methods are related to specific physical quantities; in addition, the system also comprises two types of faults (such as overlarge pitch error and the like) and reasons (such as sensor failure and the like), wherein the faults can refer to faults occurring on a fan, a subsystem, a component and the like, and comprise mechanical, control or electrical faults and the like; cause refers to a possible cause, a reason for elimination, or a reason for undetermined cause of the occurrence of a fault. The computer can establish some important associations for the knowledge content according to the semantic relations, so as to form intelligent judgment on the knowledge content in a single node and the content of a certain knowledge file, and retrieve the knowledge needed by the user more accurately.

In conclusion, the construction method of the integrated knowledge model provided by the invention effectively structures and concretes the multi-level and multi-dimensional wind turbine generator fault diagnosis knowledge from top to bottom and in a transverse and longitudinal cross mode through three knowledge chains, so that the content description of the integrated diagnosis knowledge and the deep capture record are realized. The method not only can record the final conclusion, but also can effectively describe the diagnosis process and effectively capture the denied scheme or factor. By describing and connecting the fact knowledge, the process knowledge and the factor knowledge, the high-efficiency integration of the explicit knowledge and the implicit knowledge is realized, and the understanding and the reusing of the knowledge are greatly improved. The model only comprises six or seven types of nodes, and the content of a single node is short but the information is rich, so that the information granularity is effectively improved. The connection relations among the nodes can be effectively recorded, the relations contain rich engineering semantics, understanding and reusing of knowledge can be better supported, and the knowledge can be intelligently recommended through a computer.

Claims

1. A method for constructing a fault diagnosis knowledge model for wind turbine equipment is characterized by comprising the following steps:

(2) for a specific problem in the fault system, a cause chain is established first, and the cause chain is used for describing a problem analysis and fault reasoning process and is specifically described by four types of nodes: a question node, an answer node, a positive reason node and a negative reason node; the problem node is used for describing a specific fault problem, the answer node is a possible answer to the fault problem described by the problem node, the answer node is connected with the problem node through a connecting line, and one problem node is provided with a plurality of answer nodes; the answer node can simply describe an answer, and if the answer is more complex, a sub-answer node is established for the answer node; the positive reason node and the negative reason node are the basis for providing support for a certain answer node or overriding the answer node, the positive node or the negative node can simply describe the positive reason or the negative reason, and can further establish detailed description of child nodes or establish other extended knowledge chains, namely further establish a factor chain, a process chain and a resource chain, so that all considered factors in the diagnosis process including the reason that the factors are accepted or rejected finally are recorded;

(3) judging whether a process chain or a resource chain needs to be established or not according to the factor chain, if so, establishing the process chain or the resource chain, wherein the process chain is used for describing a measurement, analysis or diagnosis process, the detailed information of the process is accurately recorded through the connection of a series of steps, the process chain is established by step nodes, main information of the step is briefly described in each node by characters, and if more knowledge needs to be captured for a certain step, the step nodes are further expanded into a new factor chain and a new resource chain; the resource chain comprises interpretations of proper nouns, index parameters, standard methods and theorem laws and calls and interpretations of related data, the resource chain is formed by connection of resource nodes, each resource node can simply describe some aspect of a resource, and richer information is provided through interconnection, and the resource refers to: for the interpretation of terms, parameters and data, resources are used for supplementing, providing evidence or providing additional information for nodes inside, and after the nodes of the resources are connected with each other, the connection relations also become useful information, and deep knowledge and rich context can be recorded; therefore, the knowledge model constructed based on the three knowledge chains only comprises six types of nodes, namely a question node, an answer node, a positive node, a negative node, a step node and a resource node; the resource chain describes explicit knowledge; the cause chain describes implicit knowledge, the process chain describes explicit knowledge and/or implicit knowledge, and the explicit knowledge comprises formulas, data tables, conceptual sketches, graphs, three-dimensional models and standard term explanations recorded by a computer in the diagnosis process; the implicit knowledge comprises knowledge and experience provided by experts and supporting decision making judgment;

specific nodes in each chain can extend other chains to complete knowledge recording, wherein process chains or resource chains can be added by the nodes in the chains to provide expanded information or evidence; a certain step node in the process chain can be added with a cause chain to explain the reason for doing so;

in step (2), a causal chain is established as follows: creating a top-level question node of the routing chain, and creating an answer node aiming at the question node; judging whether the answer node can be subjected to simple description for affirmation or overturn or needs to establish a detailed extended knowledge chain; if the expansion is needed, establishing a corresponding process chain or a corresponding resource chain according to the situation; if the expansion is not needed, establishing positive and/or negative cause nodes for diagnosing the problem node, filling corresponding information, and continuously iterating the process of establishing the problem node until the failure cause can be found and the diagnosis process can be finished;

each node contained in the factor chain, the process chain and the resource chain is called a knowledge node, text information contained in each knowledge node describes knowledge content, and the step of term classification and identification refers to the step of element identification and classification of the knowledge content; the knowledge content is categorically identified as one or more of 14 element forms, the 14 elements including: wind power plant, fan name, professional organization, worker, fan manufacturer, analysis/execution process, fault, analysis/execution step, subsystem, component, parameter name, cause, physical quantity, model/method; the semantic relationship of the 14 elements is described as follows: the wind power plant is an operation unit of the fan; the fan manufacturer is a manufacturing unit of the fan; professional institutions refer to research and engineering units related to wind turbine research, maintenance and fault handling; the above mechanisms all comprise staff; the name of the fan indicates a specific fan; one fan is composed of a plurality of subsystems; one subsystem is composed of a plurality of components; the subsystems or components all have a plurality of parameters; each parameter is associated with a certain physical quantity; the staff carries out the analysis/execution process; the analysis/execution process includes several analysis/execution steps; these steps require the use of models/methods; these models/methods are associated with specific physical quantities; the fault refers to the fault of the fan, the subsystem and the component; cause refers to the possible cause of the failure.