CN105466670A - Method for health state monitoring of multi-blade collimator based on current signals of blade motor - Google Patents
Method for health state monitoring of multi-blade collimator based on current signals of blade motor Download PDFInfo
- Publication number
- CN105466670A CN105466670A CN201510987945.1A CN201510987945A CN105466670A CN 105466670 A CN105466670 A CN 105466670A CN 201510987945 A CN201510987945 A CN 201510987945A CN 105466670 A CN105466670 A CN 105466670A
- Authority
- CN
- China
- Prior art keywords
- blade
- health status
- current
- signal
- signal characteristic
- 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
Classifications
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01M—TESTING STATIC OR DYNAMIC BALANCE OF MACHINES OR STRUCTURES; TESTING OF STRUCTURES OR APPARATUS, NOT OTHERWISE PROVIDED FOR
- G01M11/00—Testing of optical apparatus; Testing structures by optical methods not otherwise provided for
- G01M11/08—Testing mechanical properties
Landscapes
- Chemical & Material Sciences (AREA)
- Analytical Chemistry (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Measuring And Recording Apparatus For Diagnosis (AREA)
Abstract
The invention discloses a method for health state monitoring of a multi-blade collimator based on current signals of a blade motor. The method comprises the steps that output currents of the blade current are collected and converted into current digital signals; the current digital signals are pre-processed, so that noise and null drift factors during collection of the currents can be eliminated; the pre-processed current digital signals are processed in different segments by a wavelet packet decomposition method, so that a plurality of signal characteristics can be obtained; by a superior characteristic selection technology, the signal characteristics or combinations of the characteristics closely related to different health states of the multi-blade collimator are selected separately, so that a series of health state tendency signal characteristic combinations are formed; a series of neural network models are established and then trained; and the trained neutral network models corresponding to drive current signal input of the blade motor which is monitored in real time are recognized to obtain classification results, so that a health state of the multi-blade collimator is obtained. According to the method, motion states of the multi-blade collimator can be monitored conveniently and rapidly in real time.
Description
Technical field
The invention belongs to the monitoring technique field of multi-diaphragm collimator, particularly relate to a kind of monitoring method of health state and system of the multi-diaphragm collimator based on paddle motor current signal.
Background technology
Multi-diaphragm collimator is used to the mechanical moving element producing conformal radiation field size, is commonly called as multi-leaf optical grating, multi-leaf collimator etc., is widely used in medical domain.
Radiotherapy apparatus relies on the motion positions of multi-diaphragm collimator blade to form launched field, multiple blade is comprised in a multi-diaphragm collimator, directly contact between blade with blade, and have relative motion, in Long-Time Service process, blade owing to getting loose, stuck, wearing and tearing, abnormal vibrations, motor damage etc. reason is the common cause causing the multi-diaphragm collimator life-span to reduce, how to ensure the operation that blade is long-term, reliable and stable, and the health status understanding multi-diaphragm collimator is in real time the problem that radiotherapy apparatus manufacturer needs solution always.In addition, how blade, in installation process, ensures the consistance of the mounting process of each blade, thus ensures that the motor imagination consistance of each blade is also perplex a difficult problem for equipment manufacturers always.
There is no leafy collimation performance method for monitoring state not yet in effect at present.
Summary of the invention
The technical problem to be solved in the present invention is the defect for there is no leafy collimation performance method for monitoring state not yet in effect in prior art, provides one can effectively to leafy collimation performance method for monitoring state and system.
The technical solution adopted for the present invention to solve the technical problems is:
A kind of method for supervising of the multi-diaphragm collimator health status based on blade driving motor current signal is provided, it is characterized in that, comprise the following steps:
The output current of S1, collection paddle motor;
S2, the output current gathered change respectively by amplification, filtering and A/D, obtain the current digital signal of expression blade state signal;
S3, pre-service is carried out to current digital signal, the noise produced when removing gathers electric current and drift factor;
S4, utilize WAVELET PACKET DECOMPOSITION method to process current digital signal after pretreatment, obtain multiple signal characteristic;
S5, by advantageous characteristic selecting technology, select respectively the signal characteristic of health status strong correlation different from multi-diaphragm collimator or its combine, and the trend relation of this signal characteristic and time is set up by fitting of a polynomial, form a series of health status trend signal characteristic combination (as state of wear trend signal combination, get loose state trend signal characteristic combination etc.);
S6, set up the model (comprise state of wear identification neural network, get loose state recognition neural network etc.) of a series of neural network for identifying some health status of multi-diaphragm collimator, for each neural network model, utilize the health status trend signal characteristic mix vector and object vector that have established, and input this neural network model it is trained, make its classification accuracy reach the accuracy rate of expectation;
S7, Real-Time Monitoring paddle motor driving current signal, repeat the step of S2 ~ S5, set up for judging that multi-diaphragm collimator health status trend signal characteristic combines, the neural network model trained that input is corresponding respectively identifies, obtain classification results, thus obtain the health status of multi-diaphragm collimator.
In method of the present invention, the output current of described paddle motor is obtained by the current measurement module on Driver Card.
In method of the present invention, in step S2, preprocess method comprises averaging method and FFT.
In method of the present invention, in step S4, multiple signal characteristic comprises amplitude, frequency, phase place and wavelet coefficient.
In method of the present invention, described neural network comprises BP neural network and RBF neural.
In method of the present invention, the health status of multi-diaphragm collimator, comprise whether get loose, stuck, wearing and tearing, abnormal vibrations and motor damage.
The present invention also provides a kind of supervisory system of the multi-diaphragm collimator health status based on blade driving motor current signal, it is characterized in that, comprising:
Current acquisition module, for gathering the output current of paddle motor;
Analog-to-digital conversion module, for changing respectively by amplification, filtering and A/D the output current gathered, obtains the current digital signal representing blade state signal
Pretreatment module, carries out pre-service for current digital signal, the noise of removing sensor self and drift factor;
Signal characteristic abstraction module, for utilizing WAVELET PACKET DECOMPOSITION method to process current digital signal after pretreatment, obtains multiple signal characteristic; And by advantageous characteristic selecting technology, select respectively the signal characteristic of health status strong correlation different from multi-diaphragm collimator or its combine, and the trend relation of this signal characteristic and time is set up by fitting of a polynomial, form the combination of a series of health status trend signal characteristic;
Neural network module, for setting up the prototype of a series of neural network, be respectively used to the some health status identifying multi-diaphragm collimator, for each neural network model, utilize blade state measured value composition input vector and the object vector of the time period residing for trend signal characteristic of health status trend signal characteristic combination and the correspondence established, and input this neural network model it is trained, make its classification accuracy reach the accuracy rate of expectation;
Identification module, for identifying measuring and process a series of health status trend signal characteristic combination neural network models trained that input is corresponding respectively obtained in real time, obtaining classification results, thus obtaining the health status of multi-diaphragm collimator.
The beneficial effect that the present invention produces is: the health status of electric current to multi-diaphragm collimator that the present invention is based on blade driving motor is monitored in real time, by the blade state measured value of the time period residing for the trend signal characteristic of the combination of the health status trend signal characteristic that establishes and correspondence, neural network model is trained, identify eventually through in input real-time measurement values to neural network model, thus the performance state realized when working to multi-diaphragm collimator is assessed, detect, effectively can predict the fault of multi-diaphragm collimator, guarantee multi-diaphragm collimator is long-term, reliable operation.By the analyzing and processing to current signal, all multimodes in multi-diaphragm collimator blade movement process can be drawn, as vibrational state, damping state etc.Can realize monitoring in real time the motion state of multi-diaphragm collimator quickly and easily, and applicability is strong, is easy to promote.
Accompanying drawing explanation
Below in conjunction with drawings and Examples, the invention will be further described, in accompanying drawing:
Fig. 1 is the system structural representation of embodiment of the present invention multi-diaphragm collimator;
Fig. 2 is the method for supervising process flow diagram of the embodiment of the present invention based on the multi-diaphragm collimator of blade driving motor current signal;
Fig. 3 is the supervisory system structural representation of the embodiment of the present invention based on the multi-diaphragm collimator of blade driving motor current signal.
Summary of the invention
In order to make object of the present invention, technical scheme and advantage clearly understand, below in conjunction with drawings and Examples, the present invention is further elaborated.Should be appreciated that specific embodiment described herein only in order to explain the present invention, be not intended to limit the present invention.
As shown in Figure 1, multi-diaphragm collimator mainly comprises blade driving motor 1, screw mandrel 2, blade 3, multi-diaphragm collimator casing 4 and guide rail 5.The present invention monitors the health status of blade mainly through the current signal of blade driving motor.The health status of blade mainly comprises whether blade gets loose, stuck, wearing and tearing, abnormal vibrations and motor damage etc.
The embodiment of the present invention, based on the method for supervising of the multi-diaphragm collimator of blade driving motor current signal, as shown in Figure 2, comprises the following steps:
The output current of S1, the current acquisition module measurement paddle motor utilized on Driver Card;
S2, the electric current gathered to be changed respectively by amplification, filtering and A/D, obtain current digital signal, this current digital signal and blade state signal;
S3, carry out pre-service to blade movement status signal, the noise of self that removing current acquisition module produces when gathering electric current and drift factor, preprocess method can use the method such as average, FFT;
S4, utilize WAVELET PACKET DECOMPOSITION method to process current signal section during multi-diaphragm collimator blade movement, obtain multiple signal characteristic, as amplitude, frequency, phase place, wavelet coefficient etc.,
S5, by advantageous characteristic selecting technology, select respectively to combine with the signal characteristic of different health status strong correlation or its, and set up the trend relation of this signal characteristic and time by fitting of a polynomial, form a series of health status trend signal characteristic and combine;
S6, set up a series of neural network (as BP neural network, RBF neural etc.) model, be respectively used to the some health status identifying multi-diaphragm collimator, as got loose, stuck, wearing and tearing, abnormal vibrations, motor damage etc.For each neural network model, the blade state measured value composition input vector and object vector of the trend signal characteristic the established combination utilized and the time period residing for trend signal characteristic of correspondence, and input this neural network model it is trained.Its classification accuracy is made to reach the accuracy rate of expectation.Blade state measured value mainly refers to each health status that blade is corresponding, each health status can set one or more parameter and carry out quantitative description, such as stuck, get loose, motor damage uses and 0 or 1 to distinguish, wearing and tearing then use the wear extent measured value of a certain position of blade to represent, abnormal vibrations uses vibration frequency and amplitude to represent.
S7, Real-Time Monitoring paddle motor driving current signal, and analyze extraction current signal feature, the step repeating S2 ~ S5 is set up for judging that multi-diaphragm collimator health status trend signal characteristic combines, the neural network model trained that input is corresponding respectively identifies, obtain classification results by model, thus obtain the health status of multi-diaphragm collimator.
Illustrate to monitor blade wear for example.
Utilize WAVELET PACKET DECOMPOSITION method to process study blade state signal segment in step S4, obtain multiple signal characteristic; Detailed process is as follows:
The blade state signal segment that first S4.1 will obtain in step S1, is decomposed in multiple frequency band by WAVELET PACKET DECOMPOSITION technology.Original frequency band can be divided into two by one deck WAVELET PACKET DECOMPOSITION, and primary frequency band can be decomposed into 2 by k layer wavelet packet
kindividual frequency band, realizes subdivision thereof section, improves the resolution of frequency domain.K value is obtained by following formula usually:
In formula, f is the sample frequency of signal.
S4.2 calculates blade state signal segment respectively in the average of each frequency band, variance and gross energy etc., draws multiple signal characteristic.
In step S5,4 can be selected with the signal characteristic of blade wear strong correlation as learning signal feature; Analyze the correlativity of blade wear and each signal characteristic, select wherein 4 with the signal characteristic of blade wear strong correlation, as the prediction signal feature learning blade wear state.Dependency analysis process is: first make the running mean curve of each signal characteristic with run duration, obtains the running mean curve S of blade 2 life cycles
1(x), S
2(x); Then calculate
the correlativity of e more novel clear signal feature and blade wear is stronger, selects 4 signal characteristics that wherein correlativity is the strongest, is learning signal feature.X
irepresent the blade movement time, N represents the number in the blade fortune motor point of monitoring in the whole life cycle of blade.
In step S6, each blade movement time point obtains 4 learning signal features, and when blade experience is from new blade to this process of wearing and tearing, total N number of blade movement time point, obtain 4 groups of learning signal features, often group has N number of learning signal feature.Utilize and often organize the learning signal feature one group blade movement time point corresponding with this group learning signal feature, pass through fitting of a polynomial, set up the relation curve often organizing learning signal feature and blade movement time point, obtain study blade from new blade to the study blade signals changing features trend curve of wearing and tearing, fitting of a polynomial adopts 3 order polynomials usually, blade movement time point is x, and learning signal is characterized as y.Obtain study blade signals changing features trend curve, on this curve all process times point y value (i.e. ordinate) namely form one group and learn trend signal characteristic.4 groups of learning signal feature correspondences obtain 4 groups of study blade signals changing features trend curves, finally obtain 4 groups of study trend signal characteristics; Utilize 4 groups to learn trend signal characteristic and learn with these 4 groups a series of blade wear amounts that the corresponding one group of blade movement time point of trend signal characteristic measures acquisition respectively, using the input of 4 groups of study trend signal characteristics as neural network, corresponding a series of blade wear amount VB are as the output of neural network, pass through neural metwork training, set up the relation between 4 study trend signal characteristics and blade wear amount, study blade, repeatedly from using this process of wearing and tearing for the first time, obtains blade study wear law.The specific implementation process of neural metwork training is: first determine 3 layers of neural network node in hidden layer, generally select 3 ~ 5 layers; Then according to neural metwork training principle, the initial value of relevant weights or threshold value is set; Learn trend signal characteristics as input using 4 groups again, corresponding a series of blade wear amounts, as output, carry out the training of neural network.
In step S7, the step of repetition S1 ~ S4, obtains the trend signal characteristic needing to differentiate, is input to the neural network model trained, obtains Output rusults, thus can differentiate the wear extent of blade.
The embodiment of the present invention, based on the supervisory system of the multi-diaphragm collimator health status of blade driving motor current signal, as shown in Figure 3, comprising:
Current acquisition module, for gathering the output current of paddle motor;
Analog-to-digital conversion module, for changing respectively by amplification, filtering and A/D the output current gathered, obtains the current digital signal representing blade state signal
Pretreatment module, carries out pre-service for current digital signal, the noise of removing current acquisition module self and drift factor;
Signal characteristic abstraction module, for utilizing WAVELET PACKET DECOMPOSITION method to process current digital signal after pretreatment, obtains multiple signal characteristic; And by advantageous characteristic selecting technology, select respectively the signal characteristic of health status strong correlation different from multi-diaphragm collimator or its combine, and the trend relation of this signal characteristic and time is set up by fitting of a polynomial, form the combination of a series of health status trend signal characteristic;
Neural network module, for setting up the prototype of a series of neural network, be respectively used to the some health status identifying multi-diaphragm collimator, for each neural network model, utilize blade state measured value composition input vector and the object vector of the time period residing for trend signal characteristic of health status trend signal characteristic combination and the correspondence established, and input this neural network model it is trained, make its classification accuracy reach the accuracy rate of expectation;
Identification module, for identifying measuring and process a series of health status trend signal characteristic combination neural network models trained that input is corresponding respectively obtained in real time, obtaining classification results, thus obtaining the health status of multi-diaphragm collimator.
Should be understood that, for those of ordinary skills, can be improved according to the above description or convert, and all these improve and convert the protection domain that all should belong to claims of the present invention.
Claims (7)
1., based on a method for supervising for the multi-diaphragm collimator health status of blade driving motor current signal, it is characterized in that, comprise the following steps:
The output current of S1, collection paddle motor;
S2, the output current gathered change respectively by amplification, filtering and A/D, obtain the current digital signal of expression blade state signal;
S3, pre-service is carried out to current digital signal, produce when removing gathers electric current noise and drift factor;
S4, utilize WAVELET PACKET DECOMPOSITION method to process current digital signal after pretreatment, obtain multiple signal characteristic;
S5, by advantageous characteristic selecting technology, select respectively the signal characteristic of health status strong correlation different from multi-diaphragm collimator or its combine, and the trend relation of signal characteristic and time is set up by fitting of a polynomial, form the combination of a series of health status trend signal characteristic;
S6, set up the model of a series of neural network, be respectively used to the some health status identifying multi-diaphragm collimator, for each neural network model, utilize blade state measured value composition input vector and the object vector of the time period residing for trend signal characteristic of health status trend signal characteristic combination and the correspondence established, and input this neural network model it is trained, make its classification accuracy reach the accuracy rate of expectation;
S7, Real-Time Monitoring paddle motor driving current signal, repeat the step of S2 ~ S5, set up for judging that multi-diaphragm collimator health status trend signal characteristic combines, the neural network model trained that input is corresponding respectively identifies, obtain classification results, thus obtain the health status of multi-diaphragm collimator.
2. method according to claim 1, is characterized in that, the output current of described paddle motor is obtained by the current measurement module on Driver Card.
3. method according to claim 1, is characterized in that, in step S2, preprocess method comprises averaging method and FFT.
4. method according to claim 1, is characterized in that, in step S4, multiple signal characteristic comprises amplitude, frequency, phase place and wavelet coefficient.
5. method according to claim 1, is characterized in that, described neural network comprises BP neural network and RBF neural.
6. method according to claim 1, is characterized in that, the health status of multi-diaphragm collimator comprise whether get loose, stuck, wearing and tearing, abnormal vibrations and motor damage.
7., based on a supervisory system for the multi-diaphragm collimator health status of blade driving motor current signal, it is characterized in that, comprising:
Current acquisition module, for gathering the output current of paddle motor;
Analog-to-digital conversion module, for changing respectively by amplification, filtering and A/D the output current gathered, obtains the current digital signal representing blade state signal
Pretreatment module, carries out pre-service for current digital signal, the noise of removing current acquisition module sensors self and drift factor;
Signal characteristic abstraction module, for utilizing WAVELET PACKET DECOMPOSITION method to process current digital signal after pretreatment, obtains multiple signal characteristic; And by advantageous characteristic selecting technology, select respectively the signal characteristic of health status strong correlation different from multi-diaphragm collimator or its combine, and the trend relation of signal characteristic and time is set up by fitting of a polynomial, form the combination of a series of health status trend signal characteristic;
Neural network module, for setting up the prototype of a series of neural network, be respectively used to the some health status identifying multi-diaphragm collimator, for each neural network model, utilize blade state measured value composition input vector and the object vector of the time period residing for trend signal characteristic of health status trend signal characteristic combination and the correspondence established, and input this neural network model it is trained, make its classification accuracy reach the accuracy rate of expectation;
Identification module, for identifying measuring and process a series of health status trend signal characteristic combination neural network models trained that input is corresponding respectively obtained in real time, obtaining classification results, thus obtaining the health status of multi-diaphragm collimator.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201510987945.1A CN105466670A (en) | 2015-12-24 | 2015-12-24 | Method for health state monitoring of multi-blade collimator based on current signals of blade motor |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201510987945.1A CN105466670A (en) | 2015-12-24 | 2015-12-24 | Method for health state monitoring of multi-blade collimator based on current signals of blade motor |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN105466670A true CN105466670A (en) | 2016-04-06 |
Family
ID=55604612
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201510987945.1A Pending CN105466670A (en) | 2015-12-24 | 2015-12-24 | Method for health state monitoring of multi-blade collimator based on current signals of blade motor |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN105466670A (en) |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109974849A (en) * | 2019-04-03 | 2019-07-05 | 上海交通大学 | Without under reference signal based on the blade vibration on-line monitoring method of Tip-Timing technology |
| WO2021092845A1 (en) | 2019-11-14 | 2021-05-20 | Elekta (Shanghai) Technology Co., Ltd. | Predictive maintenance of dynamic leaf guide based on deep learning |
| US11549354B2 (en) * | 2018-03-06 | 2023-01-10 | The Texas A&M University System | Methods for real-time optimization of drilling operations |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN2703310Y (en) * | 2004-05-24 | 2005-06-01 | 肖若冰 | Full-automatic multi-blade collimator blade unit with real-time monitoring and self-protection |
| CN102091972A (en) * | 2010-12-28 | 2011-06-15 | 华中科技大学 | Numerical control machine tool wear monitoring method |
| CN202119467U (en) * | 2011-05-12 | 2012-01-18 | 北京工业大学 | Self-adaptive wavelet neural network categorizing system of anomaly detection and fault diagnosis |
| CN104147712A (en) * | 2014-08-28 | 2014-11-19 | 山东交通学院 | Vane displacement error detection device of multi-vane collimator |
| CN104595112A (en) * | 2013-10-30 | 2015-05-06 | 通用电气公司 | Wind turbine and method for evaluating health status of blades thereon |
| WO2015107651A1 (en) * | 2014-01-16 | 2015-07-23 | 三菱重工業株式会社 | Leaf position detection device, multi-leaf collimator, and radiation therapy device |
-
2015
- 2015-12-24 CN CN201510987945.1A patent/CN105466670A/en active Pending
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN2703310Y (en) * | 2004-05-24 | 2005-06-01 | 肖若冰 | Full-automatic multi-blade collimator blade unit with real-time monitoring and self-protection |
| CN102091972A (en) * | 2010-12-28 | 2011-06-15 | 华中科技大学 | Numerical control machine tool wear monitoring method |
| CN202119467U (en) * | 2011-05-12 | 2012-01-18 | 北京工业大学 | Self-adaptive wavelet neural network categorizing system of anomaly detection and fault diagnosis |
| CN104595112A (en) * | 2013-10-30 | 2015-05-06 | 通用电气公司 | Wind turbine and method for evaluating health status of blades thereon |
| WO2015107651A1 (en) * | 2014-01-16 | 2015-07-23 | 三菱重工業株式会社 | Leaf position detection device, multi-leaf collimator, and radiation therapy device |
| CN104147712A (en) * | 2014-08-28 | 2014-11-19 | 山东交通学院 | Vane displacement error detection device of multi-vane collimator |
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US11549354B2 (en) * | 2018-03-06 | 2023-01-10 | The Texas A&M University System | Methods for real-time optimization of drilling operations |
| CN109974849A (en) * | 2019-04-03 | 2019-07-05 | 上海交通大学 | Without under reference signal based on the blade vibration on-line monitoring method of Tip-Timing technology |
| WO2021092845A1 (en) | 2019-11-14 | 2021-05-20 | Elekta (Shanghai) Technology Co., Ltd. | Predictive maintenance of dynamic leaf guide based on deep learning |
| CN115243759A (en) * | 2019-11-14 | 2022-10-25 | 医科达(上海)科技有限公司 | Predictive maintenance of dynamic blade guides based on deep learning |
| EP4058139A4 (en) * | 2019-11-14 | 2023-08-02 | Elekta (Shanghai) Technology Co., Ltd. | Predictive maintenance of dynamic leaf guide based on deep learning |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108731923B (en) | Fault detection method and device for rotary mechanical equipment | |
| CN102809493B (en) | Abnormal sound diagnosis device | |
| CN104819846B (en) | Rolling bearing sound signal fault diagnosis method based on short-time Fourier transform and sparse laminated automatic encoder | |
| CN109214355B (en) | Mechanical monitoring data abnormal section detection method based on kernel estimation LOF | |
| CN103559888B (en) | Based on non-negative low-rank and the sound enhancement method of sparse matrix decomposition principle | |
| CN105758644A (en) | Rolling bearing fault diagnosis method based on variation mode decomposition and permutation entropy | |
| CN109977920A (en) | Fault Diagnosis of Hydro-generator Set method based on time-frequency spectrum and convolutional neural networks | |
| CN103962888A (en) | Tool abrasion monitoring method based on wavelet denoising and Hilbert-Huang transformation | |
| KR102018330B1 (en) | Apparatus and method for detecting anomaly behavior in plant pipe using multiple meta-learning | |
| CN105424359A (en) | Sparse-decomposition-based hybrid fault feature extraction method of gear wheel and bearing | |
| CN104390781A (en) | Gear fault diagnosis method based on LMD and BP neural network | |
| CN103995237A (en) | Satellite power supply system online fault diagnosis method | |
| CN105466670A (en) | Method for health state monitoring of multi-blade collimator based on current signals of blade motor | |
| CN105099759A (en) | Detection method and device | |
| CN103217248A (en) | Method for detecting tensile force of bridge steel cable | |
| CN103034170A (en) | Numerical control machine tool machining performance prediction method based on intervals | |
| Hoseinzadeh et al. | Modifying the Hilbert-Huang transform using the nonlinear entropy-based features for early fault detection of ball bearings | |
| CN109002026B (en) | Hydroelectric generating set full-condition comprehensive parameter degradation trend analysis method | |
| CN105626502A (en) | Plunger pump health assessment method based on wavelet packet and Laplacian Eigenmap | |
| CN103454526A (en) | Method for judging fault type of winding based on power transformer | |
| CN106441843A (en) | Rotating machinery fault waveform recognition method | |
| KR102404498B1 (en) | Industrial gearbox failure diagnosis apparatus and method using convolutional neural network based on adaptive time-frequency representation | |
| CN108107437A (en) | A kind of marine environmental monitoring method using normal mode coupling interference | |
| CN102928211A (en) | Fault diagnosing device and diagnosing method of high-speed rotation device | |
| CN110956112B (en) | Novel high-reliability slewing bearing service life assessment method |
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: 20160406 |