CN105043776A - Aircraft engine performance monitoring and fault diagnosis method - Google Patents
Aircraft engine performance monitoring and fault diagnosis method Download PDFInfo
- Publication number
- CN105043776A CN105043776A CN201510491460.3A CN201510491460A CN105043776A CN 105043776 A CN105043776 A CN 105043776A CN 201510491460 A CN201510491460 A CN 201510491460A CN 105043776 A CN105043776 A CN 105043776A
- Authority
- CN
- China
- Prior art keywords
- engine
- detecting device
- space
- engine performance
- fault diagnosis
- 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.)
- Pending
Links
Landscapes
- Combined Controls Of Internal Combustion Engines (AREA)
Abstract
The invention discloses an aircraft engine performance monitoring and fault diagnosis method. According to the aircraft engine performance monitoring and fault diagnosis method, the negative selection mechanism of an immune system and an artificial neural network are used in combination to determine the degree of deviation of engine performance from a normal value, namely, the anomaly degree of the engine performance, and therefore, monitoring on engine performance trends can be realized. With the method adopted, the change situation of the overall performance of an engine can be reflected flexibly and accurately, and the recognition rate of engine performance can be improved, and therefore, potential early-stage faults of the engine can be found, and the expansion of the faults can be prevented, and aircraft engine fault detection effects can be significantly improved.
Description
Technical field
The present invention relates to a kind of aircraft engine performance monitoring and fault diagnosis method, specifically a kind of aircraft engine performance monitoring and fault diagnosis method based on Theory of Artificial Immunization.
Background technology
The complexity of present generation aircraft increases, and which results in people also more strong to the demand that automatically can detect airplane fault system.These fault detection systems are devised, and in order to monitor aircraft state in such systems, to detect potential fault, thus potential fault are processed before may causing the more serious system failure.
The fault detection method of aircraft engine is an important component part in airplane fault detection system, develop the multiple method being applied to the performance monitoring and fault diagnosis of aircraft engine both at home and abroad at present, comprise statistical analysis method, neural network and comprehensive parameters method etc.For comprehensive parameters method, by overall engine multinomial performance index, obtain the comprehensive parameters of a quantitative response overall engine performance, therefore need to determine the influence degree of each parameter to overall engine performance, namely the weights of each parameter are determined, but no matter adopt which kind of Weighting, all need abundant engine abnormity and fault data as training sample, otherwise, the generalization of the weights obtained is very poor, can not reflect exception and the fault of other types, thus the comprehensive parameters that weighted method is determined loses meaning.On the other hand, the fault sample for calculation training mostly is laboratory simulation and gets, and real engine is abnormal and fault data is less, and not easily obtains, thus limits the application of these class methods.
In sum, owing to lacking fault sample data, there is aircraft engine fault sample and obtain difficulty in current fault detection method, is only confined to detect fixing fault mode, be difficult to problems such as failure mode all standings, unsatisfactory to the effect of aircraft engine fault detect.
Summary of the invention
For above-mentioned prior art Problems existing, the invention provides a kind of aircraft engine performance monitoring and fault diagnosis method, can sensitive, the situation of change that reflects overall engine performance exactly, improve the discrimination that whether normal engine performance is, and find the potential initial failure of engine with this, prevent the expansion of fault, significantly improve the effect of aircraft engine fault detect.
To achieve these goals, this aircraft engine performance monitoring and fault diagnosis method specifically comprises the following steps:
Step 1: the state space of definition engine;
Step 2: abnormality degree detects;
Abnormality degree test problems can be defined as: known normal sample set
ask for oneself space, i.e. the membership function μ of the proper space
self, utilize this function can carry out abnormality degree detection to unknown sample, and with the form quantitative analysis result of abnormality degree;
Step 3: generating detector;
Original negative selection algorithm adopts scale-of-two to encode to the data in oneself space, and therefore detecting device exists with the form of binary string; In order to improve the formation speed of detecting device, adopt the proper vector of reflection engine behavior here
with this, oneself space and non-own space are described;
Detecting device has the dimension identical with normal mode vector, but it is distributed in non-own space;
For detecting device d demand fulfillment with lower inequality:
E(d,s)＞r
In formula, E () represents Euclidean distance, and s represents any normal mode vector in oneself space, and r is threshold value;
All be distributed in non-own space according to the detecting device that above rule produces;
Step 4: the abnormality degree curve generating reflection engine performance; Neural network has very strong non-linear mapping capability, can as exception monitoring function.
Further, to define the concrete steps of the state space of engine as follows for described step 1:
Step 1.1: carry out record to engine work in every parameter, these parameters are respectively high pressure rotor corrected spool speed n
hcor, low pressure rotor corrected spool speed n
lcor, low pressure guide vane angle alpha
1, high pressure guide vane angle alpha
2, vibration values B, lube use rate ph, turbine rear exhaust temperature T
4, revolutional slip S and nozzle indicated value le;
Step 1.2: Power Function
value reflection function of the engine quality, and
when engine is in optimum Working,
when engine is in malfunction,
Step 1.3: Power Function forms the proper vector of reflection engine behavior
it is the function of time, forms engine condition space S thus.
Further, in described step 3 in the production process of detecting device, for the random detecting device produced, by following steps, the detecting device produced at random is adjusted to non-own space:
Step 3.1: regulation calculates step number p;
Step 3.2: to each detecting device d, finds out and the k of its arest neighbors normal mode vector set N
c;
Step 3.3: material calculation Δ:
Step 3.4: calculate as follows, wherein η is calculation rate:
d＝d+η·Δ
Step 3.5: after often completing a step 3.2 to step 3.4, checks whether d meets the demands, if meet matched rule, then d is valid detector, is joined in set of effective detectors D; If reach step number p, detecting device still can not meet the demands, then remove this detecting device;
By above process, produce the set of effective detectors D that can cover non-own space.
Further, the generating step of described detecting device is as follows:
Step 3.6.1: the quantity n that required detecting device is set
d;
Step 3.6.2: threshold values r is set;
Step 3.6.3: a random generation real-valued vectors d;
Step 3.6.4: calculate E;
Step 3.6.5: judge whether detector number is n
d.
Further, the concrete steps in described step 4 are as follows:
Step 4.1: under engine work state, fully collects its normal sample, and calculates corresponding Power Function, form oneself space;
Step 4.2: according to reverse side system of selection generating detector, i.e. exceptional sample.These detecting devices do not match with oneself space, and only match with non-own space, what therefore detecting device represented in fact is exceptional sample;
Step 4.3: utilize the normal sample of engine operation and exceptional sample obtained in the previous step, neural network is trained;
Step 4.4: utilize the neural network trained to carry out exception monitoring;
In the training process, make the output of normal sample be 1, the output of exceptional sample is 0; In exception monitoring process, the output of neural network reflects the intensity of anomaly of overall engine performance, and its value is less, represents that the overall performance of engine more departs from normal value;
In order to make the output of neural network clearly show by curve form, adopt the smoothing process of following formula;
In formula: s is level and smooth window width;
for the mean value exported, i.e. sharpening result;
The duty of engine can be differentiated according to the change of abnormality degree curve and whether break down.
Compared with prior art, this aircraft engine performance monitoring and fault diagnosis method utilizes immune negative selection mechanism, and in conjunction with artificial neural network, determine that engine performance departs from the degree (abnormality degree) of normal value, realize the monitoring of engine performance trend.The using and satisfied experience equipped by outfield, most of fault of aircraft engine is gradual, is formed often because certain defect constantly expands via exception and then further develop.From extremely to failure phase, the relevant parameters of engine more and more can depart from normal value, and in many situations, engine condition Parameters variation is the beginning producing catastrophic failure.Therefore the present invention is according to the situation of change of engine behavior parameter, determines that engine performance departs from the degree (abnormality degree) of normal value, in this, as an evaluation criteria of the overall engine general level of the health.Adopt this method can sensitive, the situation of change that reflects overall engine performance exactly, improve the discrimination that whether normal engine performance is, and find the potential initial failure of engine with this, prevent the expansion of fault, significantly improve the effect of aircraft engine fault detect.
Accompanying drawing explanation
Fig. 1 is that in the present invention, detecting device produces algorithm flow chart;
Fig. 2 is training and the observation process schematic diagram of neural network in the present invention;
Fig. 3 is the engine performance curve figure mono-of abnormality degree and comprehensive parameters reflection in the present invention;
Fig. 4 is the engine performance curve figure bis-of abnormality degree and comprehensive parameters reflection in the present invention.
Embodiment
Below in conjunction with accompanying drawing, the present invention will be further described.
This aircraft engine performance monitoring and fault diagnosis method specifically comprises the following steps:
Step 1: the state space of definition engine;
Step 2: abnormality degree detects;
Negative selection algorithm can only monitor oneself space sample to be had unchanged, can not monitoring the degree of oneself space sample change, by improving negative selection algorithm, detecting the intensity of anomaly of sample to be tested.The boundary of abnormality degree to oneself space and non-own space of fuzzy space is adopted to carry out obfuscation, by the membership function μ in oneself space
selfthe state space S of engine is mapped to interval [0,1], i.e. μ
self: [0,1]
n→ [0,1], in this case, the corresponding normal degree of numeric representation engine: 1 represents normal, 0 represents abnormal, and the value between 0 ~ 1 represents abnormality degree, is worth less, represents that intensity of anomaly is larger.
Therefore, abnormality degree test problems can be defined as: known normal sample set
ask for the membership function μ in oneself space (proper space)
self, utilize this function can carry out abnormality degree detection to unknown sample, and with the form quantitative analysis result of abnormality degree.
Step 3: generating detector;
Original negative selection algorithm adopts scale-of-two to encode to the data in oneself space, and therefore detecting device exists with the form of binary string; In order to improve the formation speed of detecting device, do not adopt binary coding form here, but adopt the proper vector of reflection engine behavior
with this, oneself space and non-own space are described;
Detecting device has the dimension identical with normal mode vector, but it is distributed in non-own space;
For detecting device d demand fulfillment with lower inequality:
E(d,s)＞r
In formula, E () represents Euclidean distance, and s represents any normal mode vector in oneself space, and r is threshold value;
All be distributed in non-own space according to the detecting device that above rule produces;
Step 4: the abnormality degree curve generating reflection engine performance; Neural network has very strong non-linear mapping capability, can as exception monitoring function.
Further, to define the concrete steps of the state space of engine as follows for described step 1:
Step 1.1: carry out record to engine work in every parameter, these parameters are respectively high pressure rotor corrected spool speed n
hcor, low pressure rotor corrected spool speed n
lcor, low pressure guide vane angle alpha
1, high pressure guide vane angle alpha
2, vibration values B, lube use rate ph, turbine rear exhaust temperature T
4, revolutional slip S and nozzle indicated value le, totally 9 parameters;
Step 1.2: the standardization whether completing engine individual event running parameter according to these independent running parameters close to ideal value, namely obtains Power Function
power Function
value reflection function of the engine quality, and
when engine is in optimum Working,
when engine is in malfunction,
Step 1.3: Power Function forms the proper vector of reflection engine behavior
it is the function of time, forms engine condition space S thus.
Further, in described step 3 in the production process of detecting device, for the random detecting device produced, by following steps, the detecting device produced at random is adjusted to non-own space:
Step 3.1: regulation calculates step number p;
Step 3.2: to each detecting device d, finds out and the k of its arest neighbors normal mode vector set N
c;
Step 3.3: material calculation Δ:
Step 3.4: calculate as follows, wherein η is calculation rate:
d＝d+η·Δ
Step 3.5: after often completing a step 3.2 to step 3.4, checks whether d meets the demands, if meet matched rule, then d is valid detector, is joined in set of effective detectors D; If reach step number p, detecting device still can not meet the demands, then remove this detecting device;
By above process, produce the set of effective detectors D that can cover non-own space.The quantity of detecting device can according to practical problems need determine, detector number is more, and monitoring effect is better, but can increase the training time of neural network below too much.
By above process, produce the set of effective detectors D that can cover non-own space.
Further, as shown in Figure 1, the generating step of described detecting device is as follows:
Step 3.6.1: the quantity n that required detecting device is set
d;
Step 3.6.2: threshold values r is set;
Step 3.6.3: a random generation real-valued vectors d;
Step 3.6.4: calculate E;
Step 3.6.5: judge whether detector number is n
d.
Further, as shown in Figure 2, the concrete steps in described step 4 are as follows:
Step 4.1: under engine work state, fully collects its normal sample, and calculates corresponding Power Function, form oneself space;
Step 4.2: according to reverse side system of selection generating detector, i.e. exceptional sample.These detecting devices do not match with oneself space, and only match with non-own space, what therefore detecting device represented in fact is exceptional sample;
Step 4.3: utilize the normal sample of engine operation and exceptional sample obtained in the previous step, neural network is trained;
Step 4.4: utilize the neural network trained to carry out exception monitoring;
In the training process, make the output of normal sample be 1, the output of exceptional sample is 0; In exception monitoring process, the output of neural network reflects the intensity of anomaly of overall engine performance, and its value is less, represents that the overall performance of engine more departs from normal value;
In order to make the output of neural network clearly show by curve form, adopt the smoothing process of following formula:
In formula: s is level and smooth window width;
for the mean value exported, i.e. sharpening result;
The duty of engine can be differentiated according to the change of abnormality degree curve and whether break down.
According to above thinking and countermeasure step, the performance trend of certain fanjet is analyzed.First by the calculating of engine performance trend analysis program, the Power Function value of 9 parameters of wherein 200 subnormal work can be obtained, form the space of oneself with this.Then according to reverse side system of selection generating detector (exceptional sample), the quantity of detecting device can be determined by experiment, when the quantity of detecting device is greater than 400, just can obtain good effect.Neural network adopts the BP network of three layers of multiple input single output, and structure is 10-18-1.400 exceptional samples of existing 200 normal samples and generation are used to train neural network.The neural network trained just can carry out exception monitoring as exception monitoring function to this engine, with the performance change trend of monitoring engine.
Experiment one: we monitor this engine, have recorded engine operation 109 times, the working time is the parameter value of 122 hours, for carrying out exception monitoring.
First calculate the Power Function value that each parameter is corresponding, as the input of neural network, the smoothing process of Window width s=6 is adopted to the Output rusults of neural network, obtain the abnormality degree curve reflecting engine performance.In order to contrast with comprehensive parameters method, adopt comprehensive parameters method to carry out Capability trend analyse to recording parameters simultaneously, have also been obtained the comprehensive parameters curve of a reflection engine performance.By these two kinds of methods this engine performance variation tendency determined as shown in Figure 3, declining to a great extent all appearred in the performance parameter value that two kinds of methods obtain, and showed that significantly worsening appearred in the performance of this engine.Actual conditions be this engine installation work 36 hours, electronics synthesis regulator fault, has changed electronics synthesis regulator.In Fig. 3, performance parameter value declines to a great extent is due to before regulator fault, the Related Work parameter drift of regulator, be in abnormal operation, therefore cause the relevant parameters of engine also to depart from normal value, along with the continuation of regulator parameter is drifted about, finally cause in work by 36 hours, regulator fault, performance parameter also drops to minimum point.
Although the performance change of comprehensive parameters reflection engine is more obvious, time abnormal and normal, more greatly, this can not illustrate that comprehensive parameters method is more effective than method in this paper in aggregate parameter value change.Because when calculating comprehensive parameters, the normal sample in this group parameter and the weights of fault sample data to each monitored parameter are used to be optimized, therefore the comprehensive parameters obtained is only responsive to the fault type used during training, and for the fault of other type or emerging fault, its effect may be very poor, even inspection does not measure abnormality, and this can be confirmed from experiment two.
Experiment two: continue to monitor the engine of another same model, obtains one group of new engine operation 102 times, and the working time is the parameter value of 114 hours.At 69 hours, there is a new fault: low pressure guide vane fault (alpha1 fault) in this engine.Abnormality degree and comprehensive parameters is adopted to draw engine performance change curve, as shown in Figure 4.Wherein, comprehensive parameters adopts the weights identical with experiment one to calculate.As seen from Figure 4, abnormality degree curve correctly can reflect the change of engine performance trend, has occurred declining to a great extent of performance parameter value at 65 hours, shows that exception has appearred in engine, finally causes within 69 hours, occurring low pressure guide vane fault.But comprehensive parameters curve does not but reflect this variation tendency, but at 33 hours with within 82 hours, give the false-alarm of property abnormality (fault).
In sum, this aircraft engine performance monitoring and fault diagnosis method utilizes immune negative selection mechanism, and in conjunction with artificial neural network, determine that engine performance departs from the degree (abnormality degree) of normal value, realize the monitoring of engine performance trend.Adopt this method can sensitive, the situation of change that reflects overall engine performance exactly, improve the discrimination that whether normal engine performance is, and find the potential initial failure of engine with this, prevent the expansion of fault, significantly improve the effect of aircraft engine fault detect.
Claims (5)
1. an aircraft engine performance monitoring and fault diagnosis method, is characterized in that,
Specifically comprise the following steps:
Step 1: the state space of definition engine;
Step 2: abnormality degree detects, known normal sample set
ask for oneself space, i.e. the membership function μ of the proper space
self, utilize this function can carry out abnormality degree detection to unknown sample, and with the form quantitative analysis result of abnormality degree;
Step 3: generating detector;
Original negative selection algorithm adopts scale-of-two to encode to the data in oneself space, and therefore detecting device exists with the form of binary string; In order to improve the formation speed of detecting device, adopt the proper vector of reflection engine behavior here
with this, oneself space and non-own space are described;
Detecting device has the dimension identical with normal mode vector, but it is distributed in non-own space;
For detecting device d demand fulfillment with lower inequality:
E(d,s)＞r
In formula, E () represents Euclidean distance, and s represents any normal mode vector in oneself space, and r is threshold value;
All be distributed in non-own space according to the detecting device that above rule produces;
Step 4: the abnormality degree curve generating reflection engine performance; Neural network has very strong non-linear mapping capability, as exception monitoring function.
2. a kind of aircraft engine performance monitoring and fault diagnosis method according to claim 1, is characterized in that,
The concrete steps defining engine condition space in described step 1 are as follows:
Step 1.1: carry out record to engine work in every parameter, these parameters are respectively high pressure rotor corrected spool speed n
hcor, low pressure rotor corrected spool speed n
lcor, low pressure guide vane angle alpha
1, high pressure guide vane angle alpha
2, vibration values B, lube use rate ph, turbine rear exhaust temperature T
4, revolutional slip S and nozzle indicated value le;
Step 1.2: Power Function
value reflection function of the engine quality, and
when engine is in optimum Working,
when engine is in malfunction,
Step 1.3: Power Function forms the proper vector of reflection engine behavior
it is the function of time, forms engine condition space S thus.
3. a kind of aircraft engine performance monitoring and fault diagnosis method according to claim 1, is characterized in that,
In described step 3 in the production process of detecting device, for the random detecting device produced, by following steps, the detecting device produced at random is adjusted to non-own space:
Step 3.1: regulation calculates step number p;
Step 3.2: to each detecting device d, finds out and the k of its arest neighbors normal mode vector set N
c;
Step 3.3: material calculation Δ:
Step 3.4: calculate as follows, wherein η is calculation rate:
d＝d+η·Δ
Step 3.5: after often completing a step 3.2 to step 3.4, checks whether d meets the demands, if meet matched rule, then d is valid detector, is joined in set of effective detectors D; If reach step number p, detecting device still can not meet the demands, then remove this detecting device;
By above process, produce the set of effective detectors D that can cover non-own space.
4. a kind of aircraft engine performance monitoring and fault diagnosis method according to claim 1 and 2, is characterized in that,
The generating step of described detecting device is as follows:
Step 3.6.1: the quantity n that required detecting device is set
d;
Step 3.6.2: threshold values r is set;
Step 3.6.3: a random generation real-valued vectors d;
Step 3.6.4: calculate E;
Step 3.6.5: judge whether detector number is n
d.
5. a kind of aircraft engine performance monitoring and fault diagnosis method according to claim 1, is characterized in that,
Concrete steps in described step 4 are as follows:
Step 4.1: under engine work state, fully collects its normal sample, and calculates corresponding Power Function, form oneself space;
Step 4.2: according to reverse side system of selection generating detector, i.e. exceptional sample;
Step 4.3: utilize the normal sample of engine operation and exceptional sample obtained in the previous step, neural network is trained;
Step 4.4: utilize the neural network trained to carry out exception monitoring;
In the training process, make the output of normal sample be 1, the output of exceptional sample is 0; In exception monitoring process, the output of neural network reflects the intensity of anomaly of overall engine performance, and its value is less, represents that the overall performance of engine more departs from normal value;
In order to make the output of neural network clearly show by curve form, adopt the smoothing process of following formula:
In formula: s is level and smooth window width;
for the mean value exported, i.e. sharpening result;
Whether the change according to abnormality degree curve differentiates the duty of engine and breaks down.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201510491460.3A CN105043776A (en) | 2015-08-12 | 2015-08-12 | Aircraft engine performance monitoring and fault diagnosis method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201510491460.3A CN105043776A (en) | 2015-08-12 | 2015-08-12 | Aircraft engine performance monitoring and fault diagnosis method |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN105043776A true CN105043776A (en) | 2015-11-11 |
Family
ID=54450494
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201510491460.3A Pending CN105043776A (en) | 2015-08-12 | 2015-08-12 | Aircraft engine performance monitoring and fault diagnosis method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN105043776A (en) |
Cited By (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105784380A (en) * | 2016-04-19 | 2016-07-20 | 上海交通大学 | Aero-engine gas compressor rotor blade systematic fault detection method |
| CN106368816A (en) * | 2016-10-27 | 2017-02-01 | 中国船舶工业系统工程研究院 | Method for online abnormity detection of low-speed diesel engine of ship based on baseline deviation |
| CN107917811A (en) * | 2016-10-10 | 2018-04-17 | 通用电气公司 | Aviation aircraft engine health forecast |
| CN109163911A (en) * | 2018-09-21 | 2019-01-08 | 昆明理工大学 | A kind of fault of engine fuel system diagnostic method based on improved bat algorithm optimization ELM |
| CN111259993A (en) * | 2020-03-05 | 2020-06-09 | 沈阳工程学院 | Fault diagnosis method and device based on neural network |
| CN112541597A (en) * | 2019-09-04 | 2021-03-23 | 上海杰之能软件科技有限公司 | Multi-equipment maintenance method and device, storage medium and terminal |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN1151475A (en) * | 1995-10-26 | 1997-06-11 | 三洋电机株式会社 | Failure diagnosing apparatus for combustion systems |
| JP2003065163A (en) * | 2001-08-23 | 2003-03-05 | Toyota Motor Corp | Diagnostic system for evaporative emission purge system |
| CN101415256A (en) * | 2008-11-28 | 2009-04-22 | 南京航空航天大学 | Method of diagnosing wireless sensor network fault based on artificial immunity system |
| CN103245373A (en) * | 2013-04-09 | 2013-08-14 | 哈尔滨工程大学 | Method for diagnosing faults of sensor of underwater robot |
| CN103678881A (en) * | 2013-11-22 | 2014-03-26 | 张清华 | Composite fault diagnosis method based on combination of artificial immunity and evidence theory |
| CN104280253A (en) * | 2013-07-08 | 2015-01-14 | 广州中国科学院先进技术研究所 | Immune detector based fault diagnosis method and system |
-
2015
- 2015-08-12 CN CN201510491460.3A patent/CN105043776A/en active Pending
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN1151475A (en) * | 1995-10-26 | 1997-06-11 | 三洋电机株式会社 | Failure diagnosing apparatus for combustion systems |
| JP2003065163A (en) * | 2001-08-23 | 2003-03-05 | Toyota Motor Corp | Diagnostic system for evaporative emission purge system |
| CN101415256A (en) * | 2008-11-28 | 2009-04-22 | 南京航空航天大学 | Method of diagnosing wireless sensor network fault based on artificial immunity system |
| CN103245373A (en) * | 2013-04-09 | 2013-08-14 | 哈尔滨工程大学 | Method for diagnosing faults of sensor of underwater robot |
| CN104280253A (en) * | 2013-07-08 | 2015-01-14 | 广州中国科学院先进技术研究所 | Immune detector based fault diagnosis method and system |
| CN103678881A (en) * | 2013-11-22 | 2014-03-26 | 张清华 | Composite fault diagnosis method based on combination of artificial immunity and evidence theory |
Non-Patent Citations (2)
| Title |
|---|
| 侯胜利 等: "一种航空发动机性能监控的免疫神经网络模型", 《航空动力学报》 * |
| 王威 等: "一种基于人工免疫理论的性能监控与故障诊断方法", 《科技信息》 * |
Cited By (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105784380A (en) * | 2016-04-19 | 2016-07-20 | 上海交通大学 | Aero-engine gas compressor rotor blade systematic fault detection method |
| CN107917811A (en) * | 2016-10-10 | 2018-04-17 | 通用电气公司 | Aviation aircraft engine health forecast |
| CN106368816A (en) * | 2016-10-27 | 2017-02-01 | 中国船舶工业系统工程研究院 | Method for online abnormity detection of low-speed diesel engine of ship based on baseline deviation |
| CN109163911A (en) * | 2018-09-21 | 2019-01-08 | 昆明理工大学 | A kind of fault of engine fuel system diagnostic method based on improved bat algorithm optimization ELM |
| CN112541597A (en) * | 2019-09-04 | 2021-03-23 | 上海杰之能软件科技有限公司 | Multi-equipment maintenance method and device, storage medium and terminal |
| CN111259993A (en) * | 2020-03-05 | 2020-06-09 | 沈阳工程学院 | Fault diagnosis method and device based on neural network |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN105043776A (en) | Aircraft engine performance monitoring and fault diagnosis method | |
| Yang et al. | Real-time condition monitoring and fault detection of components based on machine-learning reconstruction model | |
| KR101955305B1 (en) | Gas turbine sensor failure detection utilizing a sparse coding methodology | |
| Zhao et al. | Advanced correlation-based anomaly detection method for predictive maintenance | |
| CN101701940B (en) | On-line transformer fault diagnosis method based on SVM and DGA | |
| Rashidi et al. | Data-driven root-cause fault diagnosis for multivariate non-linear processes | |
| EP2171598B1 (en) | Fuzzy classification approach to fault pattern matching | |
| Soylemezoglu et al. | Mahalanobis-Taguchi system as a multi-sensor based decision making prognostics tool for centrifugal pump failures | |
| US10613153B2 (en) | Method and apparatus for defect pre-warning of power device | |
| CN111474919B (en) | Aeroengine control system sensor fault diagnosis method based on AANN network group | |
| CN111506049B (en) | Multiple fault diagnosis method for aero-engine control system based on AANN network system | |
| US8751423B2 (en) | Turbine performance diagnostic system and methods | |
| CN104390657A (en) | Generator set operating parameter measuring sensor fault diagnosis method and system | |
| KR100997009B1 (en) | The method for dynamic detection and on-time warning of industrial process | |
| GB2476246A (en) | Diagnosing an operation mode of a machine | |
| KR102362159B1 (en) | Method for Fault Detection and Fault Diagnosis in Semiconductor Manufacturing Process | |
| Zhang et al. | Remaining useful life prediction of rolling bearings using electrostatic monitoring based on two-stage information fusion stochastic filtering | |
| Li et al. | An LSTM based method for stage performance degradation early warning with consideration of time-series information | |
| Amin et al. | Fuzzy inference and fusion for health state diagnosis of hydraulic pumps and motors | |
| Sanayha et al. | Fault detection for circulating water pump using time series forecasting and outlier detection | |
| Du et al. | A SOM based Anomaly detection method for wind turbines health management through SCADA data | |
| Li et al. | Design of comprehensive diagnosis system in nuclear power plant | |
| Wu et al. | An improved LSTM neural network with uncertainty to predict remaining useful life | |
| CN111913463B (en) | State monitoring method for chemical volume control system of nuclear power plant | |
| Muhammad et al. | Fault diagnostic model for rotating machinery based on principal component analysis and neural network |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| WD01 | Invention patent application deemed withdrawn after publication | ||
| WD01 | Invention patent application deemed withdrawn after publication |
Application publication date: 20151111 |