CN109375608B - Engine fault diagnosis method based on graph model - Google Patents
Engine fault diagnosis method based on graph model Download PDFInfo
- Publication number
- CN109375608B CN109375608B CN201811187223.8A CN201811187223A CN109375608B CN 109375608 B CN109375608 B CN 109375608B CN 201811187223 A CN201811187223 A CN 201811187223A CN 109375608 B CN109375608 B CN 109375608B
- Authority
- CN
- China
- Prior art keywords
- fault
- node
- engine
- parts
- path
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Images
Classifications
-
- G—PHYSICS
- G05—CONTROLLING; REGULATING
- G05B—CONTROL OR REGULATING SYSTEMS IN GENERAL; FUNCTIONAL ELEMENTS OF SUCH SYSTEMS; MONITORING OR TESTING ARRANGEMENTS FOR SUCH SYSTEMS OR ELEMENTS
- G05B23/00—Testing or monitoring of control systems or parts thereof
- G05B23/02—Electric testing or monitoring
- G05B23/0205—Electric testing or monitoring by means of a monitoring system capable of detecting and responding to faults
- G05B23/0218—Electric testing or monitoring by means of a monitoring system capable of detecting and responding to faults characterised by the fault detection method dealing with either existing or incipient faults
- G05B23/0221—Preprocessing measurements, e.g. data collection rate adjustment; Standardization of measurements; Time series or signal analysis, e.g. frequency analysis or wavelets; Trustworthiness of measurements; Indexes therefor; Measurements using easily measured parameters to estimate parameters difficult to measure; Virtual sensor creation; De-noising; Sensor fusion; Unconventional preprocessing inherently present in specific fault detection methods like PCA-based methods
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01M—TESTING STATIC OR DYNAMIC BALANCE OF MACHINES OR STRUCTURES; TESTING OF STRUCTURES OR APPARATUS, NOT OTHERWISE PROVIDED FOR
- G01M15/00—Testing of engines
Landscapes
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Engineering & Computer Science (AREA)
- Automation & Control Theory (AREA)
- Testing And Monitoring For Control Systems (AREA)
- Combined Controls Of Internal Combustion Engines (AREA)
Abstract
The invention relates to an engine fault diagnosis method based on a graph model, which is characterized in that a fault node graph model among all parts of an engine is established according to an engine structure and a historical fault database, all the parts are nodes, the interaction relation among the parts is a connecting line among the nodes, a fault is decomposed into a plurality of fault characteristics, each fault characteristic is a fault leaf node, when the fault occurs, a fault leaf node corresponding to the fault characteristic is positioned, a node path leading to the leaf node is found by searching the graph, and the reason of the fault is judged through each node path sink. The invention utilizes the data in the historical fault database, fully utilizes the prior experience knowledge and ensures the accuracy of the diagnosis model; the invention decomposes the fault characteristics, searches the corresponding fault parts for each fault characteristic, clearly explains the relation between each part and the corresponding fault characteristics, and quickly and accurately diagnoses the fault of the engine.
Description
Technical Field
The invention relates to the field of engine diagnosis method control, in particular to an engine fault diagnosis method based on a graph model.
Background
The engine is the heart of car, agricultural machinery, has contained a large amount of spare parts, and the structure is complicated, can help people to practice thrift cost of maintenance to its quick, accurate diagnosis of trouble, improves work efficiency.
In order to diagnose the engine fault rapidly, the following two schemes are proposed:
the method comprises the following steps: simulation fault data are generated through an electric vehicle simulation model, and then the neural network is trained by using the data obtained through simulation.
An electric vehicle simulation model is built through Matlab/Simulink, parameters of a sensor during fault are simulated through the simulation model, and a neural network is trained to carry out fault diagnosis according to the sensor parameters and the fault type. However, in the method, the fault diagnosis system and the physical model are separated, the diagnosis is carried out through the neural network, the existing diagnosis experience is not utilized, the diagnosis process takes all faults as a whole for diagnosis, the transmission principle of the power system cannot be clearly explained, when a new fault type occurs, the neural network needs to be retrained when the model is updated, the method is complicated and time-consuming, and the diagnosis model is difficult to update.
The second method comprises the following steps: and training a neural network by using the fault data, and predicting the fault by using the neural network.
And according to the engine parameters and the corresponding faults, building a neural network model, and sending the engine parameters and the corresponding fault codes into a neural network for training to realize the function of predicting the engine faults by the model. However, in the method, the neural network model lacks interpretability, the neural network maps the fault type and the fault generation reason to establish a corresponding relation, the mapping and the corresponding relation diagnose each fault as a whole, the mathematical relation and the thermodynamic principle between parts of the engine cannot be clearly explained, and a diagnosis system lacks a link of knowledge storage and continuous learning. When a new fault type occurs, the neural network needs to be retrained, which is more complicated and difficult to update the diagnostic model.
Therefore, according to the two schemes, each fault is diagnosed as a whole, the relationship between each part and the corresponding fault feature cannot be clearly explained, when a certain fault feature appears, the part corresponding to the fault feature cannot be quickly found, and after a new fault feature appears, the neural network needs to be retrained, which is more complicated, and the diagnostic model is more difficult to update, which is not beneficial to improving the working efficiency of enterprises.
Disclosure of Invention
Aiming at the problems in the prior art, the invention provides the engine fault diagnosis method based on the graph model, which can clearly explain the relation between each part and the corresponding fault characteristics, quickly and accurately diagnose the faulted part, improve the working efficiency and facilitate the updating and maintenance of the diagnosis model.
In order to achieve the above object, the present invention provides a map model-based engine fault diagnosis method, comprising the steps of:
step ①, decomposing the engine fault into fault characteristics based on the historical fault data of the engine, establishing an engine fault database, defining each fault characteristic as a leaf node, each engine part as a node, each node having a unique identification number, and establishing an engine fault node map model based on the engine fault database;
step ②, connecting the parts related to each fault feature by connecting lines according to the sequence of the interaction relation among the parts according to the parts of the engine corresponding to the fault feature, and establishing a fault path, wherein each node in a single fault path only appears once in the path, and no annular sub-path exists in the fault path;
step ④, calculating the probability of the fault cause of the parts according to the maintenance history and the working time of the parts and the times of the parts in all fault paths of the fault, sequencing the parts from high to low according to the probability of the fault cause, and troubleshooting from the parts with the front sequence;
and ⑤, if the engine fault inference is not accurate during maintenance, updating the fault map model according to the methods of steps ① - ② according to actual maintenance parts and processes.
Regarding the engine fault as a plurality of monitoring indexes including engine surface vibration acceleration, engine fuel rate, engine displacement, engine oil pressure and coolant temperature, wherein each abnormal monitoring index is a fault characteristic, such as overhigh coolant temperature, overhigh engine oil pressure, overlow engine fuel rate and the like, and judging whether the monitoring index is abnormal through a formula (1):
wherein z is0Is the observed value of the current index, hereinafter referred to as the observed value, znFor normal index value, g is eachSet value of index, when observed value z0And a normal index znThe difference is divided by znThe obtained result is more than or equal to g percent, namely the observed value is considered as an abnormal value, and if the observed value is less than g percent, the observed value is considered as a normal value.
The search method in step ③ includes the steps of:
1) find and fail feature leaf node yiConnected nodes, denoted as list yi[s]；
2) Find and list yi[s]Of each node siThe connected nodes record the search result as a list si[q]；
3) Search and list si[q]In each node qiThe connected nodes record the search result as a list qi[p]；
4) Repeatedly executing the step (2) and the step (3) until the root node is searched;
5) get all leaf nodes y connected to the failure featureiThe fault path of (2);
6) performing the steps (1) to (5) on all the fault feature leaf nodes to obtain a fault path set of all the fault feature leaf nodes;
7) calculating all nodes which repeatedly appear in a fault path to be path cross points, wherein the node with the most occurrence times is the node with the most possibility of fault occurrence, and the calculating method comprises the following steps:
a) traversing nodes of all fault paths;
b) if the node appears for the first time, recording the number of times of appearance of the node as 1;
c) if the node appears again, adding 1 to the appearance frequency;
d) until all node traversals of all failed paths are complete.
The calculation formula of the failure probability of the component in the step ④ is as follows:
p＝f(n，m，t) (2)
wherein p is the probability of the fault of the part, n is the historical maintenance frequency of the part, m is the frequency of the part in all fault paths at the time, and t is the historical accumulated working time of the part.
Compared with the prior art, the invention has the following advantages:
1) data in a historical fault database is fully utilized;
2) the diagnostic model is convenient to update and maintain, and the accuracy of new fault diagnosis is ensured;
3) the invention decomposes the fault characteristics, searches the corresponding fault parts for each fault characteristic, clearly explains the relation between each part and the corresponding fault characteristics, and quickly and accurately diagnoses the fault of the engine.
Drawings
FIG. 1 is a schematic view of an engine fault broken down into fault signatures in accordance with the present invention;
FIG. 2 is a schematic diagram of a fault node map model of an engine according to the present invention, where p1-pn, q1-qn, s1-sn are component nodes, and y1-yn are fault signature leaf nodes;
fig. 3 is a schematic diagram of a fault path and a fault path crossing node, the connection between nodes is the interaction relationship between the components in the fault, i.e. the fault path, q2 represents a fault path crossing node, and q1 and q3 belong to a single fault path and are not the fault path crossing node.
Detailed Description
The invention will be further explained with reference to the drawings.
As shown in the figure, the invention provides an engine fault diagnosis method based on a graph model, which comprises the following steps:
step ①, based on the collected engine fault characteristics, fault reasons and historical data of the conduction chains of the fault reasons from a plurality of vehicles, decomposing the engine fault into a plurality of fault characteristics, defining each fault characteristic as a fault characteristic leaf node, each engine part as a node, establishing an engine fault node graph model, wherein each node has a unique identification number, each identification number corresponds to one engine part, and the corresponding engine part is conveniently and quickly found according to the identification number, and the engine fault decomposition method comprises the following steps:
regarding engine faults as a plurality of monitoring indexes, including engine surface vibration acceleration, engine fuel oil rate, engine displacement, engine oil pressure and coolant temperature, wherein each abnormal monitoring index is fault characteristics, such as overhigh coolant temperature, overhigh engine oil pressure, overlow engine fuel oil rate and the like, and judging whether the monitoring index is abnormal through a formula (1):
wherein z is0Is the observed value of the current index, hereinafter referred to as the observed value, znG is the set value of each index when the observed value z is normal index value0And a normal index znThe difference is divided by znThe obtained result is more than or equal to g percent, namely the observed value is considered as an abnormal value, and if the observed value is less than g percent, the observed value is considered as a normal value.
For each fault signature, the values of g are as follows:
step ②, connecting the parts related to each fault feature by connecting lines according to the sequence of the interaction relationship among the parts according to the parts of the engine corresponding to the fault feature, and establishing a node path, namely a fault path, wherein each node in a single fault path only appears once in the path, and a ring-shaped sub-path does not exist in the fault path, wherein the connecting lines among the nodes represent the interaction relationship among the parts, for example:
when the component a fails, the component b cannot normally operate, the component c cannot normally operate, and finally the monitoring index y generates an abnormal value, the nodes a, b and c and the monitoring index y are connected by a connecting line, namely 'a-b-c-y', so that a fault path is formed.
1) find and fail feature leaf node yiConnected nodes, denoted as list yi[s]；
2) Find and list yi[s]Of each node siThe connected nodes record the search result as a list si[q]；
3) Search and list si[q]In each node qiThe connected nodes record the search result as a list qi[p]；
4) Repeatedly executing the step (2) and the step (3) until the root node is searched;
5) get all leaf nodes y connected to the failure featureiThe fault path of (2);
6) performing the steps (1) to (5) on all the fault feature leaf nodes to obtain a fault path set of all the fault feature leaf nodes; .
7) Calculating all nodes which repeatedly appear in a fault path to be path cross points, wherein the node with the most occurrence times is the node which is most likely to have faults, and the calculation method comprises the following steps:
1) traversing nodes of all fault paths;
2) if the node appears for the first time, recording the number of times of appearance of the node as 1;
3) if the node appears again, adding 1 to the appearance frequency;
4) until all node traversals of all failed paths are complete.
Step ④, combining the maintenance history and the working time of the parts themselves and the times of occurrence in all fault paths of the current fault, calculating the probability of the parts becoming the fault cause based on the formula (2), sorting the parts from top to bottom according to the probability of becoming the fault cause, and troubleshooting from the parts with the front ranking, wherein the parts with the front ranking are more likely to be the cause of the fault, and the calculation formula is as follows:
p＝f(n，m，t) (2)
wherein p is the probability of the fault of the part, n is the historical maintenance frequency of the part, m is the frequency of the part in all fault paths at the time, and t is the historical accumulated working time of the part.
And ⑤, if the engine fault inference is not accurate during maintenance, updating the fault path and updating the fault map model according to the method of steps ① - ② according to the process of actually maintaining the parts, namely the sequence of the interdependencies among the parts corresponding to the fault characteristics in the actual maintenance process.
Claims (5)
1. An engine fault diagnosis method based on a graph model is characterized by comprising the following steps:
step ①, decomposing the engine fault into fault features based on the engine structure and historical fault data, defining each fault feature as a leaf node, each engine part as a node, each node having a unique identification number, and establishing an engine fault node graph model;
step ②, connecting the parts related to each fault feature by connecting lines according to the sequence of the interaction relation among the parts according to the parts of the engine corresponding to the fault feature, and establishing a fault path, wherein each node in a single fault path only appears once in the path, and no annular sub-path exists in the fault path;
step ③, when a fault occurs, firstly, the fault is decomposed into a plurality of fault characteristics, the fault characteristics are positioned to a fault characteristic leaf node, and for each fault characteristic leaf node, all fault paths which can reach the leaf node are searched;
step ④, calculating the probability of the fault cause of the parts according to the maintenance history and the working time of the parts and the times of the parts in all fault paths of the fault, sequencing the parts from high to low according to the probability of the fault cause, and troubleshooting from the parts in the front of the sequence;
and ⑤, if the engine fault inference is not accurate during maintenance, updating the fault path and updating the fault node map model according to the methods of steps ① - ② according to actual maintenance parts and processes.
2. The method of claim 1, wherein the engine fault decomposition method in step ① is to regard the engine fault as several monitoring indexes, each abnormal monitoring index is a fault feature, and determine whether the monitoring index is abnormal according to formula (1):
wherein z is0Is the observed value of the current index, hereinafter referred to as the observed value, znG is the set value of each index when the observed value z is normal index value0And a normal index znThe difference is divided by znThe obtained result is more than or equal to g percent, namely the observed value is considered as an abnormal value, and if the observed value is less than g percent, the observed value is considered as a normal value.
3. The map model-based engine fault diagnosis method according to claim 1, wherein the search method in step ③ comprises the steps of:
(1) find and fail feature leaf node yiConnected nodes, denoted as list yi[s]；
(2) Find and list yi[s]Of each node siThe connected nodes record the search result as a list si[q]；
(3) Search and list si[q]In each node qiThe connected nodes record the search result as a list qi[p]；
(4) Repeatedly executing the step (2) and the step (3) until the root node is searched;
(5) get all leaf nodes y connected to the failure featureiThe fault path of (2);
(6) performing the steps (1) to (5) on all the fault feature leaf nodes to obtain a fault path set of all the fault feature leaf nodes;
(7) and calculating all nodes which repeatedly appear in the fault path as path cross points, wherein the node with the most occurrence times is the node with the most possibility of fault occurrence.
4. The map model-based engine fault diagnosis method according to claim 3, wherein the calculation method in step (7) includes the steps of:
a) traversing nodes of all fault paths;
b) if the node appears for the first time, recording the number of times of appearance of the node as 1;
c) if the node appears again, adding 1 to the appearance frequency;
d) until all node traversals of all failed paths are complete.
5. The map model-based engine fault diagnosis method according to claim 1 or 4, wherein the calculation formula of the probability that the component becomes the cause of the fault in step ④ is:
p＝f(n，m，t) (2)
wherein p is the probability of the fault of the part, n is the historical maintenance frequency of the part, m is the frequency of the part in all fault paths at the time, and t is the historical accumulated working time of the part.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811187223.8A CN109375608B (en) | 2018-10-12 | 2018-10-12 | Engine fault diagnosis method based on graph model |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811187223.8A CN109375608B (en) | 2018-10-12 | 2018-10-12 | Engine fault diagnosis method based on graph model |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN109375608A CN109375608A (en) | 2019-02-22 |
| CN109375608B true CN109375608B (en) | 2020-02-07 |
Family
ID=65403509
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201811187223.8A Active CN109375608B (en) | 2018-10-12 | 2018-10-12 | Engine fault diagnosis method based on graph model |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109375608B (en) |
Family Cites Families (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102095588A (en) * | 2010-12-28 | 2011-06-15 | 中国人民解放军军事交通学院 | Intelligent fault diagnosis method for diesel engine |
| US10778748B2 (en) * | 2015-06-05 | 2020-09-15 | Apple Inc. | Rapid reconfiguration of device location system |
| CN106019195B (en) * | 2016-07-22 | 2019-02-05 | 国网浙江省电力公司电力科学研究院 | Electric power measurement automation verification assembly line fault diagnosis system |
| CN107656518A (en) * | 2016-11-30 | 2018-02-02 | 中国航空工业集团公司沈阳飞机设计研究所 | A kind of system failure inference method |
| CN107655693A (en) * | 2017-11-02 | 2018-02-02 | 哈尔滨理工大学 | A kind of engine of boat and ship fault diagnosis system and method |
| CN108398637B (en) * | 2018-01-29 | 2020-04-21 | 合肥工业大学 | Fault diagnosis method for nonlinear electromechanical system |
| CN108469802B (en) * | 2018-01-31 | 2021-04-13 | 中汽研汽车检验中心(武汉)有限公司 | Vehicle information remote monitoring and fault diagnosis system |
-
2018
- 2018-10-12 CN CN201811187223.8A patent/CN109375608B/en active Active
Also Published As
| Publication number | Publication date |
|---|---|
| CN109375608A (en) | 2019-02-22 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109086889B (en) | Terminal fault diagnosis method, device and system based on neural network | |
| CN105955243B (en) | A kind of aircraft diagnosis and repair method and its device based on fault case | |
| JP4928532B2 (en) | Vehicle fault diagnosis device | |
| JP4414470B1 (en) | Generating reference values for vehicle fault diagnosis | |
| JP2008536219A (en) | Diagnosis and prediction method and system | |
| CN110296124B (en) | Remote fault diagnosis system and method based on expert system | |
| CN109783500A (en) | A kind of method for diagnosing faults based on fault tree and case similitude | |
| CN110705812A (en) | Industrial fault analysis system based on fuzzy neural network | |
| JPH08202444A (en) | Method and device for diagnosing abnormality of machine facility | |
| CN104268375A (en) | Ship electric power station fault diagnosing method based on knowledge petri network | |
| CN112749744B (en) | Bayesian network rail passenger car traction system fault diagnosis method and system based on observation module | |
| CN117763449A (en) | Multi-source fault autonomous positioning and classifying method | |
| CN111598467A (en) | Reliability evaluation method and system for gathering and transportation combined station and key equipment | |
| CN115481726A (en) | Industrial robot complete machine health assessment method and system | |
| CN113395182B (en) | Intelligent network equipment management system and method with fault prediction | |
| CN108825482B (en) | Fault detection method and system for axial plunger pump of airplane | |
| CN109375608B (en) | Engine fault diagnosis method based on graph model | |
| CN110674893B (en) | Self-adaptive correction method for diagnosis experience in rotary machine fault diagnosis knowledge base | |
| CN115150255B (en) | Self-adaptive knowledge-graph-based automatic root cause positioning method for application faults | |
| CN113887452B (en) | Fault diagnosis optimization method based on correlation matrix | |
| CN115237091B (en) | Electromechanical equipment fault tracing method and system | |
| CN114386603B (en) | Thermal power plant generator fault early warning method based on time sequence causal network | |
| CN112182743B (en) | Fault transmission feature matching-based aircraft system fault diagnosis method | |
| CN115469643A (en) | Nuclear power station rotating machinery health management method, system and medium | |
| EP3303835B1 (en) | Method for windmill farm monitoring |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |