CN101666711B - Method for diagnosing engine integrated faults based on fuzzy semanteme network - Google Patents

Abstract

The invention discloses an engine comprehensive fault diagnosis method based on fuzzy semantic network, which is characterized in that an embedded platform based on an ARM embedded microprocessor and a data acquisition module based on a field programmable gate array FPGA are established on the embedded platform. A fault knowledge acquisition method based on a fuzzy semantic network model, a case-based reasoning subsystem and a guided interactive reasoning subsystem. Based on the fault knowledge acquisition method of the fuzzy semantic network model, the engine fault knowledge base is constructed. The case reasoning subsystem is used for the rapid diagnosis of common faults. In-depth diagnosis; after the data acquisition module automatically detects the fault symptoms or the user inputs the fault symptoms, it can start a single subsystem, or start two subsystems to run in parallel; each subsystem can realize fault diagnosis independently, or combine the two systems The diagnosis results are comprehensively diagnosed, which improves the accuracy of diagnosis.

Description

A Comprehensive Engine Fault Diagnosis Method Based on Fuzzy Semantic Network

Technical field:

The invention relates to the technical field of engine fault diagnosis, in particular to an engine comprehensive fault diagnosis method based on a fuzzy semantic network. the

Background technique:

At present, intelligent fault diagnosis methods for aero-engines mainly include rule-based reasoning, model-based reasoning, case-based reasoning, and artificial neural network-based diagnostic methods. This makes the reasoning mode single, unable to make full use of fault information, and affects the diagnostic accuracy. In addition, the problem of difficulty in obtaining fault knowledge has not been effectively resolved, and has become a bottleneck in fault diagnosis. the

Invention content:

The purpose of the present invention is to overcome the deficiencies of the above-mentioned prior art, and provide a kind of engine comprehensive fault diagnosis method based on fuzzy semantic network, mainly solve the reasoning mode of the existing intelligent fault diagnosis method of aeroengine, single inference mode, unable to make full use of fault information and fault Difficulty in acquiring knowledge. the

In order to achieve the above object, the present invention is achieved in that a kind of engine comprehensive fault diagnosis method based on fuzzy semantic network is characterized in that it comprises following processing steps:

a. Establish an embedded platform based on ARM embedded microprocessor and a data acquisition module based on field programmable gate array FPGA; the embedded platform consists of ARM embedded processor S3C2410, power circuit, FLASH memory, SDRAM memory, LCD interface, USB Interface, Ethernet interface, serial interface, data acquisition module interface, bus and reserved expansion interface, etc.; the data acquisition module is composed of main controller FPGA, AD converter and multi-channel analog switch;

b. The acquisition subroutine is set in the data acquisition module, and the database and application program are set in the memory of the embedded platform. The fault diagnosis program in the application program includes the knowledge acquisition method based on the fuzzy semantic network model, the case-based reasoning subsystem and the guided Interactive reasoning subsystem; the knowledge acquisition method based on the fuzzy semantic network model is based on the FMEA analysis data of the engine reliability design, and uses the concepts of concept nodes and semantic connections to clearly and comprehensively express the relationship between nodes such as faults and causes, and construct The engine fault knowledge base has been established, which has better solved the bottleneck problem of engine fault diagnosis knowledge acquisition; the case-based reasoning subsystem is mainly composed of symptom acquisition sub-module, case reasoning sub-module, case maintenance module and fault case library, and is used for common faults. The rapid diagnosis; the guided interactive reasoning subsystem is based on the engine fault knowledge base based on the fuzzy semantic network, and uses the human-computer interaction method to carry out fuzzy search reasoning, which can carry out in-depth fault diagnosis;

c. After the data acquisition module automatically detects the fault symptoms or the user inputs the fault symptoms, it transmits to the embedded platform, and the embedded platform starts the two subsystems to run in parallel;

d. When the two subsystems can get the preliminary diagnosis results, if the two results are the same, it is the final diagnosis result of the system. If the two results are not the same, it is necessary to use expert experience to make comprehensive decisions to give the final result ;

e. When one of the subsystems is diagnosed successfully and the other subsystem fails, the latter is guided to learn according to the diagnosis results of the former;

f. When neither of the two subsystems gives a diagnosis result, the system diagnosis fails. the

A kind of engine comprehensive fault diagnosis method based on fuzzy semantic network of the present invention, the diagnosis step of its described case-based reasoning subsystem is:

a. The data acquisition module analyzes and processes the signal detected by the sensor, transmits it to the symptom acquisition sub-module of the embedded platform, compares it with the parameter threshold in the database, extracts the deviation or out-of-tolerance parameters, and constitutes the engine failure symptom information, and at the same time Combining the symptom information manually input, generate a fault symptom vector according to the method of case representation, and submit it to the case reasoning sub-module;

b. The case reasoning sub-module first indexes the fault cases belonging to the same category as the current fault from the case base according to the deterministic symptoms in the fault symptom vector, and then performs the similarity matching calculation based on the gray relational analysis method for the indexed cases , if the cases whose similarity meets a certain threshold are obtained, the fuzzy comprehensive evaluation will be made based on factors such as the number of failures and maintenance difficulty of these cases, and then the most suitable case will be selected for case reuse and correction, and the solution provided by the case will be submitted to the user. Finally, according to the user's feedback information, the case is submitted to the case maintenance module for learning and saved to the case library. If the case reasoning diagnosis fails, the fault symptom vector is submitted to the guided interactive reasoning subsystem. the

In the fuzzy semantic network-based engine comprehensive fault diagnosis method of the present invention, the case library includes a fault case basic information table, a fault case type table, and a fault case uncertainty symptom value table. the

A kind of engine comprehensive fault diagnosis method based on fuzzy semantic network of the present invention, the diagnosis step of its described guided interactive reasoning subsystem is:

a. When the system detects the fault symptoms, combined with the fault information manually input through the touch screen connected to the embedded platform, the fuzzy search reasoning is carried out based on the fault knowledge base built on the basis of the fuzzy semantic network, and the system prompts the user to perform the detection Operation and operation method, after the user completes the detection, input the obtained detection results to the system for further disposal suggestions after the system analyzes; the system continuously analyzes according to some feedback information obtained from the interaction with the user, and determines the final faulty part; locates the fault After the incident, the system will automatically propose reasonable troubleshooting methods;

b. Use the hyperlink to cross-link with the "Maintenance Manual" module in the database to directly view the corresponding maintenance and repair steps and their specific operation methods. Through the interactive mode of "Suggestion-Question and Answer", it can provide maintenance personnel with intelligent information Troubleshooting. the

Compared with the prior art, a kind of engine comprehensive fault diagnosis method based on fuzzy semantic network of the present invention has outstanding substantive characteristics and remarkable progress: 1, based on the knowledge acquisition method of fuzzy semantic network model, solves the problem of Bottleneck problem of engine fault diagnosis knowledge acquisition; 2. Comprehensive utilization of the diagnostic methods of two subsystems. If one subsystem fails to diagnose, it can be transferred to another subsystem model for diagnosis, which increases the success rate of fault diagnosis; 3. Two subsystems The reasoning process is parallel to each other, and each subsystem can implement fault diagnosis independently, and can also combine the diagnosis results of the two systems for fusion diagnosis, which improves the accuracy of diagnosis. the

Description of drawings:

Fig. 1 is the fuzzy semantic network model figure of the present invention;

Fig. 2 is the overall diagnosis flowchart of the present invention;

Fig. 3 is the diagnosis flowchart of case reasoning subsystem of the present invention;

Fig. 4 is an association diagram of the fault case data table of the present invention. the

Detailed ways:

In order to better understand and implement, the specific embodiments will be given below in conjunction with the accompanying drawings to describe in detail an engine comprehensive fault diagnosis method based on the fuzzy semantic network of the present invention. the

Embodiment 1, referring to Fig. 1, 2, 3, 4, at first set up the embedded platform based on ARM embedded microprocessor and the data acquisition module based on Field Programmable Gate Array FPGA, embedded platform consists of ARM embedded processor S3C2410, Power circuit, FLASH memory, SDRAM memory, LCD interface, USB interface, Ethernet interface, serial interface, data acquisition module interface, bus and reserved expansion interface, etc.; the data acquisition module consists of the main controller FPGA, AD converter Composed of multi-channel analog switches, etc., the ARM embedded processor S3C2410 is connected to the data acquisition module through the bus-connected data acquisition module interface, and the relevant components are connected by conventional technology;

Set the acquisition subroutine in the data acquisition module, set up the database and application program in the memory of the embedded platform, and the fault diagnosis program in the application program includes the knowledge acquisition method based on the fuzzy semantic network model, the case-based reasoning subsystem and guided interaction reasoning subsystem;

The case-based reasoning subsystem is mainly composed of symptom acquisition sub-module, case reasoning sub-module, case maintenance module and case library. For rapid diagnosis of common faults; the guided interactive reasoning subsystem is based on the engine fault knowledge base built on the basis of fuzzy semantic networks, and uses human-computer interaction to perform fuzzy search and reasoning, enabling in-depth fault diagnosis;

The specific method of knowledge acquisition of fuzzy semantic network model:

Aiming at the current situation of few engine fault samples, based on the FMEA analysis data of the engine reliability design, a fuzzy semantic network model is established by using the concept of concept nodes and semantic connections, which clearly and comprehensively expresses the relationship between nodes such as faults and causes. The knowledge base of engine faults has been established, and the bottleneck problem of engine fault diagnosis knowledge acquisition has been better solved. The establishment of the fuzzy semantic network model is carried out from three aspects: firstly, conceptual nodes are selected, such as engine failure modes and failure consequences. Secondly, determine the connection between concept nodes, and analyze whether the change of a certain node (value) has a significant impact on another node (value); then assign appropriate symbols and language strengths to each connection (A: always; O: often; So: sometimes; Se: rarely; +: increased impact; -: decreased impact; 0: negligible impact). the

The conceptual nodes of the FMEA of the engine air circuit unit based on the data of the engine factory test and troubleshooting and the design reliability FMEA can be expressed as shown in Table 1. the

Table 1

concept node describe concept node describe C1 Compressor aerodynamic fatigue damage C17 Decreased compressor efficiency C2 Compressor gas elastic instability fracture C18 Deterioration of fuel supply and atomization combustion process C3 The second stage blade tip of the compressor falls off C19 Broken blade C4 Combustion chamber nozzle carbon deposition and corrosion C20 blade twist C5 Compressor rectifier blade failure C21 Decreased circulation capacity C6 Combustion chamber afterburner cylinder crack C22 Decreased turbine efficiency C7 The first-stage vane burn failure of the guide C23 Break through the gate C8 The first-stage vane of the guide is ablated C24 Rotor scrap C9 Crack and ablation fault on the outer edge plate of the guide C25 The shaft is scrapped C10 Guider stress crack failure C26 Roulette Scrap C11 Damaged fan blades (injured by foreign objects) C27 Decreased fan efficiency C12 High pressure rotor blade elongation C28 The blade collided with the brake C13 Deformation of high pressure rotor blades C29 Turbo secondary damage C14 Deformation of turbine rotor blades C15 Turbine shaft crack

C16 Deformation of the turbine wheel

Taking Table 1 as an example, the necessary direction symbols and language intensity (language value) are applied to obtain the fuzzy semantic network model of a certain type of engine air circuit unit, as shown in Figure 1. the

After the data acquisition module automatically detects the fault symptoms or the user inputs the fault symptoms, it will be transmitted to the embedded platform. The embedded platform can start a single subsystem, or start two subsystems to run in parallel; the system has many possible output results, and finally The final result can be obtained through comprehensive decision-making, and the overall diagnosis process is shown in Figure 2; when both subsystems can obtain preliminary diagnosis results, if the two results are the same, it is the final diagnosis result of the system; if the two results are not the same, It is necessary to use expert experience to make comprehensive decisions to give the final result; when one of the subsystems is diagnosed successfully and the other subsystem fails, the latter is guided to learn according to the diagnosis results of the former; when the two subsystems are both If no diagnosis result is given, the system diagnosis fails. the

Diagnosis process of the case-based reasoning subsystem is shown in Figure 3. The fault diagnosis steps are: a. The data acquisition module analyzes and processes the signal detected by the sensor, and transmits it to the symptom acquisition sub-module of the embedded platform, which is performed with the parameter threshold in the database. Comparing, extracting the deviation or out-of-tolerance parameters to form the engine fault symptom information, and combining the manually input symptom information, generating a fault symptom vector according to the method of case representation, and submitting it to the case reasoning sub-module; b, the case reasoning sub-module firstly according to For deterministic symptoms in the fault symptom vector, the fault cases belonging to the same category as the current fault are indexed from the case base, and then the similarity matching calculation based on the gray relational analysis method is performed on the indexed cases. If the obtained similarity meets a certain threshold For the cases, the most appropriate case is selected for case reuse and correction after fuzzy comprehensive evaluation based on factors such as the number of failures and maintenance difficulty of these cases, and the solution provided by the case is submitted to the user, and finally the case is reported according to the user’s feedback Submit it to the case maintenance module for learning and save it to the case library; if the case reasoning diagnosis fails, submit the fault symptom vector to the guided interactive reasoning diagnosis subsystem. the

It can be seen from the diagnosis process that the case reasoning diagnosis subsystem is mainly composed of symptom acquisition sub-module, case reasoning sub-module, case maintenance module and case library. the

The case library includes the fault case basic information table, the fault case type table, and the fault case uncertainty symptom value table. Design of case base: basic information table of fault cases, table of types of fault cases, and table of uncertainty symptom value of fault cases. The relationship between these three data tables is shown in Figure 4. The structure of the three data tables and the relationship between them:

(1) Fault case type table: In this table, each record corresponds to a typical case type in an engine fault case. It consists of the category number of the fault case, the index number and index value of the deterministic symptom of the case category, the name of the uncertain symptom contained in the case category and the corresponding symptom weight. Among them, the fault case type number field is the primary key. (2) Fault case basic information table: In this table, each record corresponds to a fault case basic information, which includes the case category number, fault phenomenon, fault cause, troubleshooting measures, case occurrence times, detection The degree of difficulty and the time spent on troubleshooting. Wherein, the fault case number field is the primary key; the case category number field to which the fault case belongs is the foreign key of the fault case category number field in the fault case category table. (3) Failure case uncertainty symptom value table: In this table, the attribute value of each uncertainty symptom of the failure case is stored, and the failure case number field is the primary key of this table, and it is also the failure Foreign key to the case number field. the

The diagnostic steps of the guided interactive reasoning subsystem are as follows: a. When the system detects the fault symptoms, combined with the fault information manually input through the touch screen connected to the embedded platform, fuzzy search reasoning is carried out based on the fault knowledge base, and the system Prompt the user for the detection operation and operation method that need to be performed. After the user completes the detection, input the obtained detection result to the system, and provide further disposal suggestions after analysis by the "troubleshooting expert system"; Continuously analyze the information to determine the final faulty part; after locating the faulty part, the system will automatically propose reasonable troubleshooting methods; b, use the hyperlink to cross-link with the "Maintenance Manual" module in the database to directly view the corresponding maintenance and repair steps And its specific operation method can provide maintenance personnel with intelligent fault diagnosis through the interactive mode of "suggestion-question and answer". the

The fuzzy semantic network-based engine comprehensive fault diagnosis method of the present invention is not limited to aero-engines, and can also be used for fault diagnosis in other fields. the

Claims (4)

1. a kind of engine comprehensive fault diagnosis method based on fuzzy semantic network, it is characterized in that it comprises following process step: a. Establish an embedded platform based on ARM embedded microprocessor and a data acquisition module based on Field Programmable Gate Array FPGA, which are connected by bus; b. The data acquisition module is equipped with an acquisition subroutine, and the memory of the embedded platform is equipped with a database and an application program. The fault diagnosis program in the application program includes a fault knowledge acquisition method based on a fuzzy semantic network model, a case-based reasoning subsystem, and a guided Interactive reasoning subsystem; the fault knowledge acquisition method based on the fuzzy semantic network model is based on the engine reliability design FMEA analysis data, and uses the concepts of concept nodes and semantic connections to clearly and comprehensively express the relationship between faults and cause nodes, and construct The engine fault knowledge base is established; the case-based reasoning subsystem is mainly composed of fault symptom acquisition sub-module, case reasoning sub-module, case maintenance module and fault case library, which are used for rapid diagnosis of common faults; the guided interactive reasoning subsystem is based on Based on the engine fault knowledge base built by the fuzzy semantic network, human-computer interaction is used to carry out fuzzy search and reasoning, and in-depth fault diagnosis can be carried out; c. After the data acquisition module automatically detects the fault symptoms or the user inputs the fault symptoms, it transmits to the embedded platform, and the embedded platform starts the two subsystems to run in parallel; d. When the two subsystems can get the preliminary diagnosis results, if the two results are the same, it is the final diagnosis result of the system. If the two results are not the same, it is necessary to use expert experience to make comprehensive decisions to give the final result ; e. When one of the subsystems is diagnosed successfully and the other subsystem fails, guide the latter to learn according to the diagnosis results of the former; f. When neither of the two subsystems gives a diagnosis result, the system diagnosis fails. 2. a kind of engine comprehensive fault diagnosis method based on fuzzy semantic network according to claim 1, is characterized in that the diagnosis step based on case reasoning subsystem is: a. The data acquisition module analyzes and processes the signal detected by the sensor, transmits it to the symptom acquisition sub-module of the embedded platform, compares it with the parameter threshold in the database, extracts the deviation or out-of-tolerance parameters, and constitutes the engine failure symptom information, and at the same time Combining the symptom information entered manually, generate a fault symptom vector according to the method of case representation, and submit it to the case reasoning sub-module; b. The case reasoning sub-module first indexes the fault cases belonging to the same category as the current fault from the case base according to the deterministic symptoms in the fault symptom vector, and then performs the similarity matching calculation based on the gray relational analysis method for the indexed cases , if the cases whose similarity meets a certain threshold are obtained, the fuzzy comprehensive evaluation will be made based on the number of failures and maintenance difficulty factors of these cases, and then the most suitable case will be selected for case reuse and correction, and the solution provided by the case will be submitted to the user, and finally According to the user's feedback information, the case is submitted to the case maintenance module for learning, and saved to the fault case library; if the case reasoning diagnosis fails, the fault symptom vector is submitted to the guided interactive reasoning subsystem. 3. A kind of engine comprehensive fault diagnosis method based on fuzzy semantic network according to claim 1 or 2, it is characterized in that described case library comprises fault case basic information table, fault case category table, fault case uncertainty symptom table of values. 4. a kind of engine comprehensive fault diagnosis method based on fuzzy semantic network according to claim 1, is characterized in that the diagnosis step of described guided interactive reasoning subsystem is: a. When the system detects the fault symptoms, combined with the fault information input manually, it performs fuzzy search and reasoning based on the engine fault knowledge base. The system prompts the user to perform the detection operation and its operation method. Input the test results obtained for the system to analyze and put forward further disposal suggestions; the system will continuously analyze the feedback information obtained from the interaction with the user to determine the final faulty part; after locating the faulty part, the system will automatically propose reasonable troubleshooting methods ; b. Use the hyperlink to cross-link with the "Maintenance Manual" module in the database to directly view the corresponding maintenance and repair steps and their specific operation methods. Through the interactive mode of "Suggestion-Question and Answer", it can provide maintenance personnel with intelligent information Troubleshooting.

Images