CN110175369A - A kind of gear method for predicting residual useful life based on two-dimensional convolution neural network - Google Patents
A kind of gear method for predicting residual useful life based on two-dimensional convolution neural network Download PDFInfo
- Publication number
- CN110175369A CN110175369A CN201910360392.5A CN201910360392A CN110175369A CN 110175369 A CN110175369 A CN 110175369A CN 201910360392 A CN201910360392 A CN 201910360392A CN 110175369 A CN110175369 A CN 110175369A
- Authority
- CN
- China
- Prior art keywords
- gear
- neural network
- sample
- dimensional convolution
- useful life
- 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.)
- Withdrawn
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F30/00—Computer-aided design [CAD]
- G06F30/10—Geometric CAD
- G06F30/17—Mechanical parametric or variational design
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2119/00—Details relating to the type or aim of the analysis or the optimisation
- G06F2119/04—Ageing analysis or optimisation against ageing
Abstract
The present invention discloses a kind of gear method for predicting residual useful life based on two-dimensional convolution neural network, first building convolutional neural networks structure, and the health status information of gear is extracted to its degenerative character of the real-time vibration signal of the gear received；Classification recurrence is carried out to vibration signal again, the accurate remaining life estimation of gear is obtained with this；And deep learning prediction model is verified with the Real-time Monitoring Data of gear fatigue life testing stand, it is many and diverse to solve the content that vibration signal includes in real time in gear fatigue degenerative process, data scale is big, it is difficult to extract the characteristic information that effective, reflection gear is degenerated and carry out real-time remaining life Accurate Prediction problem.The prediction model that two-dimensional convolution neural metwork training of the invention obtains can quickly and accurately predict the remaining life of gear.
Description
Technical field
The invention belongs to Application of Neural Network technical field, specifically a kind of gear based on two-dimensional convolution neural network is surplus
Remaining life-span prediction method.
Background technique
During gear drive, long-term load be infected with dust or lack of lubrication etc. due to, can cause abrasion or
It scratches and wears.And the gear of different abrasions can all have oneself unique vibration signal.So can pass through in the ideal case
Gear vibration obtains life information.Ideally, when gear is not worn, Gearbox vibration signal is the shape of involute
Formula, vibration frequency are gear engagements.The single-frequency logs of frequency.If there is certain uniform abrasion occurs, then each gear
Smooth change and linear variation, vibration signal is based on muttering sum of fundamental frequencies rate as fundamental frequency, while the periodic signal that can be generated with harmonic wave.When
When serious wear, the curve of vibration signal is close to square wave, at this point, the amplitude of meshing frequency will increase, and order harmonic frequencies,
The amplitude of the time of fundamental frequency, number and time become readily apparent from.With the increase of order, the amplitude of higher hamonic wave becomes more next
Bigger, vibrational energy is increased to be more obvious；Gear it is easy to appear vibration frequency height, abrasion or breaks when long-term load is operated
The failures such as tooth, crackle, the study found that most gearbox faults is all the quality of running state of gear box as caused by gear
The normal operation of machinery equipment is directly affected, once equipment part is unable to normal operation, it is possible to damage whole equipment even shadow
Entire production process is rung, the economic losses such as shutdown is caused, even results in catastrophic casualties, therefore, gear is remained
Remaining life prediction is the important measures for ensureing mechanical equipment safe and efficient running and raising product quality.
The processing style of convolutional network is consistent with modern biology and computer vision understanding.Nowadays, convolution
Neural network is preferable method in pattern recognition system.This is clearly showed in System for Handwritten Character Recognition, convolution mind
The machine learning benchmark of many years is had become through network.
The Chinese patent of Publication No. CN108645615A discloses a kind of Adaptive Fuzzy Neural-network gear remaining longevity
Prediction technique is ordered, implementation step is as follows: 1, using vibrating sensor to gear degeneration real-time monitoring；2, to gear fatigue state
Feature extraction is carried out, slump evaluations are carried out to gear wear degraded performance；3, fuzzy system and neural network are combined, with nerve
The deficiency of network self-study mechanism Compensation Fuzzy control system, establishes a kind of fuzzy message fuzzy neural network；4, at fuzzy place
It manages all nodes of layer and memory units is added, by last moment imformation memory and be applied in output this moment, information is made to continue to protect
It deposits, reinforces information forward-backward correlation, reduce predicted value and actual value deviation, establish modified Adaptive Fuzzy Neural-network prediction system
System；5, gear remaining life is predicted according to training modified Adaptive Fuzzy Neural-network；Advantage is can effectively to predict gear
Degenerate state and real-time remaining life provide foundation for gear preventative maintenance.
Summary of the invention
In response to the problems existing in the prior art, the purpose of the present invention is to provide a kind of teeth based on two-dimensional convolution neural network
Method for predicting residual useful life is taken turns, the health status of gear is extracted to its degenerative character of the real-time vibration signal of the gear received
Information；Classification recurrence is carried out to vibration signal again, the accurate remaining life estimation of gear is obtained with this；And with the gear tired longevity
The Real-time Monitoring Data of life testing stand verifies deep learning prediction model, predicts gear remaining life.
To achieve the above object, the technical solution adopted by the present invention is that:
A kind of gear method for predicting residual useful life based on two-dimensional convolution neural network, comprising the following steps:
S1 acquires training data, the vibration signal of gear is acquired by the acceleration transducer being mounted on gear-box, often
Primary every sampling in T seconds, the data of every N*N sampled point constitute one group of sample, acquire X group sample；The X group sample includes at least
70000 groups of unduplicated sample datas；
The X group sample is divided into C classification by the working time by S2, data prediction, is beaten according to classification to each sample
Upper label is as training sample；
S3 constructs neural network, one convolutional neural networks comprising 4 layers of convolutional layer and two layers of full articulamentum of building；
The neural network that training sample input is built is trained, it is surplus to obtain gear by S4, training neural network
The prediction model in remaining service life；
S5 predicts gear remaining life, acquires the multiple groups vibration signal of the same gear as prediction data sample, will be more
The prediction model that the group prediction data sample input training obtains, passes through the prediction model output gear multiple groups remaining working time
Average value, which is the prediction result of the gear remaining life.
Specifically, in step S1, the vibration signal of the gear is acquired by acquisition platform, and the acquisition platform includes:
Main examination case accompanies examination case, load motor, driving motor and control system；The main examination case accompanies and tries to be equipped with gear, sound biography in case
Sensor, acceleration transducer；The main examination case accompanies examination case to connect respectively with load motor, driving motor.The sonic transducer is used
In the noise for surveying gear vibration generation；The acceleration transducer is used to acquire the acceleration value of main examination case/accompany examination case vibration；Institute
State driving motor, load motor is connect with control system respectively, the control system is for controlling driving motor, load motor work
Make, while acquiring the detection data of the sonic transducer, acceleration transducer.
Specifically, in step S2, when classifying to the X group sample, remove the data close to classification line ± 5 minute
Sample, to eliminate the influence caused by classification results of continuity of the adjacent category at categorised demarcation line.
Specifically, in step S3, the convolutional neural networks include 4 convolutional layers, 4 pond layers, 4 active coatings, 2
Full articulamentum and 1 Softmax layers；The structure of the convolutional neural networks are as follows: convolutional layer → pond layer → active coating → convolution
Layer → pond layer → active coating → convolutional layer → pond layer → active coating → convolutional layer → pond layer → active coating → full articulamentum
→ articulamentum → Softmax layers complete (returning layer).
Specifically, in step S4, it is 0.5 that the learning rate when neural metwork training, which is 0.002, dropout probability, institute
The remaining life of prediction model prediction and the loss value of test sample are stated less than 0.1.
Specifically, in step S4, when training sample data input convolutional neural networks, convolutional layer can execute propagated forward
The step of；And the derivative of the error between the value and actual value of output is propagated backward to and updates neural network in neural network
Parameter and weight；The characteristic pattern of input can obtain new one layer of characteristic pattern after convolution kernel is inswept, and new characteristic pattern is using sharp
The output figure for combining other convolution kernels inswept after function living constitutes next multi input characteristic pattern；Calculation formula is as follows:
Wherein,Indicate a selection input feature vector mapping, L refers to that L layers of neural network, M are output after convolution
The size of feature, i, j are convolution input matrix size；K is the matrix of a S × S, and S is the size of convolution kernel；F be hyperbolic just
It cuts or sigmoid nonlinear activation function；Each output has added a specific biasOne is specifically inputted defeated
Data are carried out convolution from different convolution kernel K by data out.That is if the i of output is being added by m with n, to use
It is also not identical in the convolution kernel K of convolution.
Specifically, the pond layer functions of the neural network are determined by following formula:
Wherein,For pond layer functions, down is down-sampling function；In input feature vector figure, the sampling function is to each
A different n × n input block is added, to reduce the Spatial Dimension of output Feature Mapping；β is biasing coefficient, and b is addition biasing.
Compared with prior art, the beneficial effects of the present invention are: the present invention passes through the vibration signal of real-time detection gear, and
Classification recurrence is carried out to vibration signal, the accurate remaining life estimation of gear is obtained with this, and acquire with Gearbox vibration signal
The Real-time Monitoring Data of platform verifies the prediction model of neural network, can effectively solve real in gear fatigue degenerative process
When the vibration signal content that includes it is many and diverse, data scale is big, it is difficult to extract characteristic information that effective, reaction gear is degenerated into
The real-time remaining life Accurate Prediction problem of row.
Detailed description of the invention
Fig. 1 is a kind of gear method for predicting residual useful life based on two-dimensional convolution neural network in the embodiment of the present invention 1
Schematic process flow diagram；
Fig. 2 is the flow chart of 1 middle gear method for predicting residual useful life of the embodiment of the present invention；
Fig. 3 is a kind of for acquiring the structural schematic block diagram of the acquisition platform of Gearbox vibration signal in the embodiment of the present invention 2.
Specific embodiment
Below in conjunction with the attached drawing in the present invention, technical solution of the present invention is clearly and completely described, it is clear that
Described embodiments are only a part of the embodiments of the present invention, instead of all the embodiments.Based on the implementation in the present invention
Example, those of ordinary skill in the art's all other embodiment obtained under the conditions of not making creative work belong to
The scope of protection of the invention.
The problem of gear distress, feature was described as follows:
(1) gear wear: ideally, when gear is not worn, Gearbox vibration signal is the form of involute, vibration
Dynamic frequency is gear engagement；
(2) gear fatigue: due to fatigue of materials, flank of tooth fatigue includes the spot corrosion and peeling being induced by it.Tooth surface fatigue
Mainly spot corrosion.Small shock response is produced when muttering and closing since spot corrosion makes to obtain gear.Its spectrum signature mainly include with
Centered on the frequency of conjunction of muttering, axis frequency is that side frequency is spaced, and sideband is few and sparse.Its vibrate with envelope energy have certain amplitude compared with
To increase；
(3) broken teeth: the characteristics of broken teeth vibration signal is apparent impact vibration, and frequency of impact is the axis frequency in axis.In frequency
Performance in spectrum concentrates on meshing frequency and its frequency multiplication appears at frequency band two sides, is divided into axis frequency.In general, snaggletooth
Sideband quantity is big, and amplitude is big, widely distributed, and impact force is excessive when serious fracture, centered on the intrinsic frequency for generating gear
Sideband modulation.
Embodiment 1
As shown in Figure 1, present embodiments providing a kind of gear predicting residual useful life side based on two-dimensional convolution neural network
Method, specifically includes the following steps:
S1 acquires training data, the vibration signal of gear is acquired by the acceleration transducer being mounted on gear-box, often
It is primary every 7ms sampling, collected vibration signal is normalized, is made with a sampled point of every 784 (N × N number of sampled value, N 28)
For a sample sequence, and not repeatedly acquire 70000 groups of sequence samples；
Sample is divided into 10 health levels by S2, data prediction at equal intervals；
S3 constructs neural network, one convolutional neural networks comprising 4 layers of convolutional layer and two layers of full articulamentum of building；
The neural network that training sample input is built is trained, it is surplus to obtain gear by S4, training neural network
The prediction model in remaining service life；Before training sample is inputted neural network, successively line by line by 784 sampled points of every group of sample first
It is arranged in 28 × 28 matrix, convolutional neural networks is facilitated directly to carry out convolution operation to matrix；
S5 predicts gear remaining life, acquires the multiple groups vibration signal of the same gear as prediction data sample, will be more
The prediction model that the group prediction data sample input training obtains, passes through the prediction model output gear multiple groups remaining working time
Average value, which is the prediction result of the gear remaining life.
Specifically, in step S2, when classifying to 70000 groups of samples, remove close to classification line ± 5 minute
Data sample, to eliminate the influence caused by classification results of continuity of the adjacent category at categorised demarcation line.
Specifically, in step S3, the convolutional neural networks include 4 convolutional layers, 4 pond layers, 4 active coatings, 2
Full articulamentum and 1 Softmax layers；The structure of the convolutional neural networks are as follows: convolutional layer → pond layer → active coating → convolution
Layer → pond layer → active coating → convolutional layer → pond layer → active coating → convolutional layer → pond layer → active coating → full articulamentum
→ articulamentum → Softmax layers complete (returning layer)；
Specifically, in step S4, the learning rate when neural metwork training is 0.002,30000 epoch of training, often
64 batch are chosen in secondary training, and sample class is 5 classes, and dropout probability is 0.5；Finally obtained prediction model prediction remains
The loss value of remaining service life and test sample is less than 0.1.Dropout, which refers to, random when network model study allows some
Implicit neural unit abandons current output, and weight is temporary not to be will be updated.
The over-fitting degree of model can be effectively reduced using mode as dropout.Because inputting net in data
Implicit node occurs at random according to given probability during network updates weight, the update of weight not depend on some or
Multiple key nodes effectively avoids the case where some feature occurs dependent on other feature and occurred.
Specifically, in the training process of neural network, each sample is inputted into convolution according to 28 × 28 size first
Layer extracts the characteristic information that gear is degenerated；The data input pond layer of convolutional layer output is subjected to pond again, pondization can be mind
Stronger robust is provided through network and prevents over-fitting；Again by pond layer output data input active coating, active coating be in order to
Neuron is allowed to have nonlinear characteristic；Because the Linear Network of multilayer and the Linear Network expression power of single layer are approximately equal
, the feature that show non-linear factor must just add nonlinear nervous layer, i.e. active coating, the activation in the present embodiment
Relu function can be selected as activation primitive in layer；After 4 take turns convolution, Chi Hua, activation, the data of output are inputted two layers and are connected entirely
It connects layer and makees last life prediction.
As shown in Fig. 2, the step of life prediction, is as follows:
The first step acquires the vibration signal of gear-box, and the acceleration value of acquisition is normalized to N × N number of sampled value, and (N is
28), then by 28 × 28 sampled values a matrix is constituted；
Second step judges whether 28 × 28 sampled values are test set data；If so, by Input matrix two-dimensional convolution
Neural network exports recognition result by full articulamentum；If it is not, then according to the gear the working time by the sampled value into
Row classification, gives label；Sorted sampled value input two-dimensional convolution neural network is trained again；
Whether third step, training of judgement number meet the requirements, if so, terminating program；Otherwise, the first step is returned to continue to adopt
Collect the vibration signal of gear-box.
The gear method for predicting residual useful life of the present embodiment, it is only necessary to by the vibration signal classification input neural network of gear
It can quickly and accurately judge the remaining life of gear, and then the health level of gear is assessed.
Embodiment 2
As shown in figure 3, present embodiments providing a kind of for acquiring the acquisition platform of Gearbox vibration signal, the acquisition platform
The data of acquisition can be used for the training dataset of neural network in gear method for predicting residual useful life；The acquisition platform of the present embodiment
Including main examination case, accompany examination case, load motor, driving motor and control system；The main examination case, accompany examination case in be equipped with gear,
Sonic transducer, acceleration transducer；The main examination case accompanies examination case to connect respectively with load motor, driving motor.The sound sensing
Device is used to survey the noise of gear vibration generation；The acceleration transducer is used to acquire the acceleration of main examination case/accompany examination case vibration
Value；The driving motor, load motor are connect with control system respectively, and the control system is for controlling driving motor, load
Motor work, while acquiring the detection data of the sonic transducer, acceleration transducer.
Specifically, the number of teeth of the main examination case internal gear is 50 teeth, and the number of teeth for accompanying examination case internal gear is 24 teeth, described
Driving motor, load motor revolving speed be 1200r/min, it is described it is main examination case internal gear meshing frequency be 1000Hz, it is described
The meshing frequency for accompanying examination case internal gear is 480Hz.
Specifically, there are two the sonic transducer is set, respectively 9# sonic transducer and 10# sonic transducer, the 9# sound are passed
Sensor is arranged in main examination case, and the 10# sonic transducer setting is being accompanied in examination case.
Specifically, the acceleration transducer is equipped with 8, respectively 1#, 2#, 3#, 4#, 5#, 6#, 7# and 8#, the 5#
It is arranged in and is accompanied on examination case with 6# acceleration transducer, remaining acceleration transducer is arranged on main examination case.
The acquisition platform of the present embodiment accompanies examination case as reference by setting, can eliminate the systematic error of acquisition data,
To improve the accuracy of data acquisition.
It although an embodiment of the present invention has been shown and described, for the ordinary skill in the art, can be with
A variety of variations, modification, replacement can be carried out to these embodiments without departing from the principles and spirit of the present invention by understanding
And modification, the scope of the present invention is defined by the appended.
Claims (7)
1. a kind of gear method for predicting residual useful life based on two-dimensional convolution neural network, which comprises the following steps:
S1 acquires training data, the vibration signal of gear is acquired by the acceleration transducer being mounted on gear-box, every T
Second sampling is primary, and the data of every N*N sampled point constitute one group of sample, acquires X group sample；
The X group sample is divided into C classification by the working time by S2, data prediction, stamps mark to each sample according to classification
Label are used as training sample；
S3 constructs neural network, one convolutional neural networks comprising 4 layers of convolutional layer and two layers of full articulamentum of building；
The neural network that training sample input is built is trained by S4, training neural network, obtains the gear remaining longevity
The prediction model of life；
S5 predicts gear remaining life, acquires the multiple groups vibration signal of the same gear as prediction data sample, by multiple groups institute
The prediction model that the input training of prediction data sample obtains is stated, putting down for prediction model output gear multiple groups remaining working time is passed through
Mean value, the average value are the prediction result of the gear remaining life.
2. a kind of gear method for predicting residual useful life based on two-dimensional convolution neural network according to claim 1, special
Sign is, in step S1, the vibration signal of the gear is acquired by acquisition platform, and the acquisition platform includes: main examination case, accompanies
Try case, load motor, driving motor and control system；The main examination case accompanies and tries to be equipped with gear, sonic transducer, acceleration in case
Spend sensor；The main examination case accompanies examination case to connect respectively with load motor, driving motor.The sonic transducer is for surveying gear
Vibrate the noise generated；The acceleration transducer is used to acquire the acceleration value of main examination case/accompany examination case vibration；The driving electricity
Machine, load motor are connect with control system respectively, and the control system is for controlling driving motor, load motor work, simultaneously
Acquire the detection data of the sonic transducer, acceleration transducer.
3. a kind of gear method for predicting residual useful life based on two-dimensional convolution neural network according to claim 1, special
Sign is, in step S2, when classifying to the X group sample, removes the data sample close to classification line ± 5 minute, with
Eliminate continuity of the adjacent category at categorised demarcation line influences caused by classification results.
4. a kind of gear method for predicting residual useful life based on two-dimensional convolution neural network according to claim 1, special
Sign is, in step S3, the convolutional neural networks include 4 convolutional layers, 4 pond layers, 4 active coatings, 2 full articulamentums
With 1 Softmax layers.
5. a kind of gear method for predicting residual useful life based on two-dimensional convolution neural network according to claim 1, special
Sign is, in step S4, it is 0.5 that the learning rate when neural metwork training, which is 0.002, dropout probability,.
6. a kind of gear method for predicting residual useful life based on two-dimensional convolution neural network according to claim 1, special
Sign is, in step S4, when training sample data input convolutional neural networks, and the step of convolutional layer can execute propagated forward；
And by the derivative of the error between the value and actual value of output propagate backward in neural network update neural network parameter and
Weight；The characteristic pattern of input can obtain new one layer of characteristic pattern after convolution kernel is inswept, and new characteristic pattern is using activation primitive
The output figure for combining other convolution kernels inswept afterwards constitutes next multi input characteristic pattern；Calculation formula is as follows:
Wherein,Indicate a selection input feature vector mapping, L refers to that L layers of neural network, M are that convolution exports feature later
Size, i, j be convolution input matrix size；K is the matrix of a S × S, and S is the size of convolution kernel；F be tanh or
Sigmoid nonlinear activation function.
7. a kind of gear method for predicting residual useful life based on two-dimensional convolution neural network according to claim 1, special
Sign is that the pond layer functions of the neural network are determined by following formula:
Wherein,For pond layer functions, down is down-sampling function；In input feature vector figure, the sampling function to each not
Same n × n input block is added, to reduce the Spatial Dimension of output Feature Mapping；β is biasing coefficient, and b is addition biasing.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910360392.5A CN110175369A (en) | 2019-04-30 | 2019-04-30 | A kind of gear method for predicting residual useful life based on two-dimensional convolution neural network |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910360392.5A CN110175369A (en) | 2019-04-30 | 2019-04-30 | A kind of gear method for predicting residual useful life based on two-dimensional convolution neural network |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN110175369A true CN110175369A (en) | 2019-08-27 |
Family
ID=67690375
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910360392.5A Withdrawn CN110175369A (en) | 2019-04-30 | 2019-04-30 | A kind of gear method for predicting residual useful life based on two-dimensional convolution neural network |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110175369A (en) |
Cited By (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110705176A (en) * | 2019-09-02 | 2020-01-17 | 北京市燃气集团有限责任公司 | Method and device for predicting residual life of gas pipeline |
| CN110988537A (en) * | 2019-12-08 | 2020-04-10 | 中国航空综合技术研究所 | Electric steering engine residual life prediction method based on position feedback |
| CN111008485A (en) * | 2019-12-25 | 2020-04-14 | 中国石油大学(华东) | Neural network-based multi-parameter life prediction method for three-phase alternating current asynchronous motor |
| CN111024147A (en) * | 2019-12-26 | 2020-04-17 | 玳能科技(杭州)有限公司 | Component mounting detection method and device based on CNNs, electronic equipment and storage medium |
| CN111680800A (en) * | 2020-05-12 | 2020-09-18 | 埃睿迪信息技术（北京）有限公司 | Method for realizing equipment predictive maintenance model based on deep learning |
| CN112035779A (en) * | 2020-09-03 | 2020-12-04 | 郑州机械研究所有限公司 | Method for judging residual life of gear |
| CN112580169A (en) * | 2020-12-24 | 2021-03-30 | 广东省科学院智能制造研究所 | Harmonic reducer health assessment method based on machine learning |
| CN113533945A (en) * | 2021-06-30 | 2021-10-22 | 桂林电子科技大学 | Analog circuit fault diagnosis method based on two-dimensional convolutional neural network |
| CN114154269A (en) * | 2021-12-08 | 2022-03-08 | 西人马(深圳)科技有限责任公司 | Gear service life prediction method and device |
| CN114896873A (en) * | 2022-04-26 | 2022-08-12 | 盐城工学院 | Traveling wave ultrasonic motor service life prediction method and system based on convolutional neural network |
| CN117350174A (en) * | 2023-12-04 | 2024-01-05 | 国网天津市电力公司营销服务中心 | Method, system, electronic equipment and medium for predicting residual life of intelligent ammeter |
-
2019
- 2019-04-30 CN CN201910360392.5A patent/CN110175369A/en not_active Withdrawn
Cited By (14)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110705176A (en) * | 2019-09-02 | 2020-01-17 | 北京市燃气集团有限责任公司 | Method and device for predicting residual life of gas pipeline |
| CN110988537A (en) * | 2019-12-08 | 2020-04-10 | 中国航空综合技术研究所 | Electric steering engine residual life prediction method based on position feedback |
| CN111008485A (en) * | 2019-12-25 | 2020-04-14 | 中国石油大学(华东) | Neural network-based multi-parameter life prediction method for three-phase alternating current asynchronous motor |
| CN111008485B (en) * | 2019-12-25 | 2021-05-11 | 中国石油大学(华东) | Neural network-based multi-parameter life prediction method for three-phase alternating current asynchronous motor |
| CN111024147A (en) * | 2019-12-26 | 2020-04-17 | 玳能科技(杭州)有限公司 | Component mounting detection method and device based on CNNs, electronic equipment and storage medium |
| CN111680800A (en) * | 2020-05-12 | 2020-09-18 | 埃睿迪信息技术（北京）有限公司 | Method for realizing equipment predictive maintenance model based on deep learning |
| CN112035779B (en) * | 2020-09-03 | 2021-06-22 | 郑州机械研究所有限公司 | Method for judging residual life of gear |
| CN112035779A (en) * | 2020-09-03 | 2020-12-04 | 郑州机械研究所有限公司 | Method for judging residual life of gear |
| CN112580169A (en) * | 2020-12-24 | 2021-03-30 | 广东省科学院智能制造研究所 | Harmonic reducer health assessment method based on machine learning |
| CN113533945A (en) * | 2021-06-30 | 2021-10-22 | 桂林电子科技大学 | Analog circuit fault diagnosis method based on two-dimensional convolutional neural network |
| CN114154269A (en) * | 2021-12-08 | 2022-03-08 | 西人马(深圳)科技有限责任公司 | Gear service life prediction method and device |
| CN114896873A (en) * | 2022-04-26 | 2022-08-12 | 盐城工学院 | Traveling wave ultrasonic motor service life prediction method and system based on convolutional neural network |
| CN117350174A (en) * | 2023-12-04 | 2024-01-05 | 国网天津市电力公司营销服务中心 | Method, system, electronic equipment and medium for predicting residual life of intelligent ammeter |
| CN117350174B (en) * | 2023-12-04 | 2024-04-02 | 国网天津市电力公司营销服务中心 | Method, system, electronic equipment and medium for predicting residual life of intelligent ammeter |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110175369A (en) | A kind of gear method for predicting residual useful life based on two-dimensional convolution neural network | |
| CN105973594B (en) | A kind of rolling bearing fault Forecasting Methodology based on continuous depth confidence network | |
| CN109063308B (en) | Health assessment method based on deep quantum learning | |
| CN108645615B (en) | A kind of Adaptive Fuzzy Neural-network gear method for predicting residual useful life | |
| Bafroui et al. | Application of wavelet energy and Shannon entropy for feature extraction in gearbox fault detection under varying speed conditions | |
| Abu-Mahfouz | A comparative study of three artificial neural networks for the detection and classification of gear faults | |
| Gawde et al. | Multi-fault diagnosis of Industrial Rotating Machines using Data-driven approach: A review of two decades of research | |
| Yan et al. | A dynamic multi-scale Markov model based methodology for remaining life prediction | |
| CN109343505A (en) | Gear method for predicting residual useful life based on shot and long term memory network | |
| CN104820716A (en) | Equipment reliability evaluation method based on data mining | |
| Gecgel et al. | Gearbox fault diagnostics using deep learning with simulated data | |
| Wang et al. | An evolving neuro-fuzzy technique for system state forecasting | |
| CN109443766A (en) | A kind of heavy-duty vehicle gearbox gear Safety Analysis Method | |
| CN102207998B (en) | Method for evaluating service life and reliability of product based on zero-failure data | |
| Qin et al. | Remaining useful life prediction for rotating machinery based on optimal degradation indicator | |
| Pichika et al. | Optimal sensor placement for identifying multi-component failures in a wind turbine gearbox using integrated condition monitoring scheme | |
| Gecgel et al. | Machine learning in crack size estimation of a spur gear pair using simulated vibration data | |
| Zhu et al. | A simulation-data-driven subdomain adaptation adversarial transfer learning network for rolling element bearing fault diagnosis | |
| Cao et al. | Deterioration state diagnosis and wear evolution evaluation of planetary gearbox using vibration and wear debris analysis | |
| Yen et al. | Application of a neural network integrated with the internet of things sensing technology for 3D printer fault diagnosis | |
| CN115358533A (en) | Predictive maintenance management system based on big data | |
| Yaghmaei-Sabegh | Earthquake ground-motion duration estimation using general regression neural network | |
| Kumar et al. | Latest innovations in the field of condition-based maintenance of rotatory machinery: A review | |
| Su et al. | Diagnosis of gradual faults in high-speed gear pairs using machine learning | |
| Jiang et al. | A novel ensemble fuzzy model for degradation prognostics of rolling element bearings |
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 | ||
| WW01 | Invention patent application withdrawn after publication | ||
| WW01 | Invention patent application withdrawn after publication |
Application publication date: 20190827 |