CN108126987A - A kind of high-speed rod-rolling mill remote online monitoring and intelligent diagnosis system and method - Google Patents
A kind of high-speed rod-rolling mill remote online monitoring and intelligent diagnosis system and method Download PDFInfo
- Publication number
- CN108126987A CN108126987A CN201710307331.3A CN201710307331A CN108126987A CN 108126987 A CN108126987 A CN 108126987A CN 201710307331 A CN201710307331 A CN 201710307331A CN 108126987 A CN108126987 A CN 108126987A
- Authority
- CN
- China
- Prior art keywords
- data
- failure
- sample
- diagnosis
- rolling mill
- 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.)
- Granted
Links
Classifications
-
- B—PERFORMING OPERATIONS; TRANSPORTING
- B21—MECHANICAL METAL-WORKING WITHOUT ESSENTIALLY REMOVING MATERIAL; PUNCHING METAL
- B21B—ROLLING OF METAL
- B21B38/00—Methods or devices for measuring, detecting or monitoring specially adapted for metal-rolling mills, e.g. position detection, inspection of the product
-
- B—PERFORMING OPERATIONS; TRANSPORTING
- B21—MECHANICAL METAL-WORKING WITHOUT ESSENTIALLY REMOVING MATERIAL; PUNCHING METAL
- B21B—ROLLING OF METAL
- B21B33/00—Safety devices not otherwise provided for; Breaker blocks; Devices for freeing jammed rolls for handling cobbles; Overload safety devices
-
- B—PERFORMING OPERATIONS; TRANSPORTING
- B21—MECHANICAL METAL-WORKING WITHOUT ESSENTIALLY REMOVING MATERIAL; PUNCHING METAL
- B21B—ROLLING OF METAL
- B21B38/00—Methods or devices for measuring, detecting or monitoring specially adapted for metal-rolling mills, e.g. position detection, inspection of the product
- B21B38/006—Methods or devices for measuring, detecting or monitoring specially adapted for metal-rolling mills, e.g. position detection, inspection of the product for measuring temperature
-
- B—PERFORMING OPERATIONS; TRANSPORTING
- B21—MECHANICAL METAL-WORKING WITHOUT ESSENTIALLY REMOVING MATERIAL; PUNCHING METAL
- B21B—ROLLING OF METAL
- B21B38/00—Methods or devices for measuring, detecting or monitoring specially adapted for metal-rolling mills, e.g. position detection, inspection of the product
- B21B38/008—Monitoring or detecting vibration, chatter or chatter marks
Landscapes
- Engineering & Computer Science (AREA)
- Mechanical Engineering (AREA)
- Testing Of Devices, Machine Parts, Or Other Structures Thereof (AREA)
Abstract
A kind of high-speed rod-rolling mill remote online monitoring and intelligent diagnosis system, including sensor network, Centroid, industrial personal computer(In-situ processing machine), monitoring and diagnostic center, the subnet that the sensor network is made of data acquisition equipment in a milling train unit, the collected data transmission of data acquisition equipment is to industrial personal computer, dry, conversion process is carried out while the industrial personal computer stores data, industrial personal computer carries out treated data the judgement that high-speed rod-rolling mill failure whether there is to be diagnosed, and export result in advance;If the structural determination high-speed rod-rolling mill of industrial personal computer output is failure, then by this data transmission to monitoring and diagnostic center, it is on the contrary, industrial personal computer does not transmit data to monitoring and diagnostic center, monitoring further analyzes, and export result the pre- diagnosis of industrial personal computer after receiving faulty data with diagnostic center.The present invention can real-time online judge the type and abort situation of current rolling mill fault, extend service life of equipment.
Description
Technical field
The present invention relates to rod-rolling mill diagnostic techniques field, more particularly to a kind of high-speed rod-rolling mill remote online monitoring
With intelligent diagnosis system and method.
Background technology
Milling train is a kind of equipment for being widely used and realizing metal rolled process, and high-speed rod-rolling mill is generally used in milling train
One kind.It can be with the generation of various failures, including carrier wear, box deformation, axis cracking etc. in milling train operational process.It is right
It determining and positioning in this failure, it is common practice to, by professional technician by all kinds of test equipments, collect various
Parameter in particular for parameter, carries out analyzing determining fault type according to these parameters, and then determines it is which parts is sent
Failure.This process is all to carry out analysis when milling train is out of service by professional technician to determine, it is difficult to be divided online
Analysis diagnosis.
Invention content
Present invention aim to address the problems of the prior art, provide a kind of high-speed rod-rolling mill remote online monitoring and
Intelligent diagnosis system, realizing has high-speed rod-rolling mill the diagnosis of determining and fault type and positioning of fault-free.
The technical scheme is that:
A kind of high-speed rod-rolling mill remote online monitoring and intelligent diagnosis system, including data acquisition equipment, Centroid, industry control
Machine, monitoring and diagnostic center, it is characterised in that:After the collected data of data acquisition equipment are by being pooled to Centroid
By local area network or wireless network transmissions to industrial personal computer, this data is passed through the industrial personal computer local area network or wireless network passes
It is defeated to arrive monitoring and diagnostic center.
A kind of high-speed rod-rolling mill remote online monitoring and intelligent diagnosing method, which is characterized in that include the following steps:
S1 data collection steps:Each collected data of data acquisition equipment are by being pooled to after Centroid using local area network
Or wireless network transmissions are to industrial personal computer, the data acquisition equipment includes vibrating sensor in a milling train unit, speed
Sensor, three dimension acceleration sensor, displacement sensor, rotary encoder, temperature sensor, power meter and power sensor;
The pre- diagnosis algorithms of S2:The industrial personal computer is removed data while data are stored noise, data transformation and spy
Extraction process is levied, while treated data are carried out with the judgement that high-speed rod-rolling mill failure whether there is, and if exported as a result, work
The structural determination high-speed rod-rolling mill of control machine output is failure, then by this data transmission to monitoring and diagnostic center, industry control on the contrary
Machine does not transmit data to monitoring and diagnostic center then, is transmitted to diagnostic center to avoid by a large amount of trouble-free data, causes net
Network congestion, the burden for aggravating monitoring, diagnosing central server;
S3 Precise Diagnosis steps:The monitoring is received after faulty data with diagnostic center to the pre- diagnostic result of industrial personal computer
It is for further analysis again, to determine the type of failure and position.
In the data collection steps S1, vibrating sensor be mounted on babinet, shafting, gear, on parts of bearings position simultaneously
Acquisition includes the three-dimensional vibrating waveform of rolling bearing, transmission shaft, pinch roller, babinet, and rotary encoder is mounted in shaft and adopts
Collection includes rotational angle, the rotating speed of shaft, and temperature sensor is on bearing, motor and acquisition includes bearing portion and motor
Operating temperature, power sensor, which is mounted on motor and acquires, includes the power consumption of motor.
In the data collection steps S1, rotary encoder is sampled according to the following rules:The sampling period is divided into
Two periods of Twork and Twait, shaft per revolution sample 16 times, and the Twork periods are spent by 64 circle of shaft rotation
Time, sampled in the Twork periods 1024 times, the Twait periods do not sample, the time by shaft rotation 8 ~ 512 circle spend
Time.
The process of the pre- diagnosis algorithm S2 is:The data of data acquisition equipment are stored, the original number stored
According to can directly transfer to high in the clouds, while data are converted and feature extraction processing, pass through two graders(HSSVM)Selection contains
The feature vector of failure is simultaneously uploaded to cloud interface, this contains the feature vector of failure after Precise Diagnosis step confirms reliability again
It adds in HSSVM parameter libraries and carries out HSSVM learning trainings to optimize two graders, two grader(HSSVM)While also according to
Breakdown judge situation carries out controlling of sampling to each sensor of data acquisition equipment.
In the pre- diagnosis algorithm S2, the method for data transformation and feature extraction is:
(1)Data convert:The data series that one sensor is exported in the Twork periods are denoted as data [k], k=0,1 ..., ifCome from vibrating sensor, then one-dimensional multi-scale wavelet transformation is carried out to it, to extract the singular point under each scale,
Transformation results are denoted as data [k](s), whereinsIt is scale number,s=1,2 ..., N;The present invention takes 8 kinds of scales, i.e. N=8;
If the time that 1 circle of shaft rotation is spent(That is rotation period)ForTc, eachTcPositioning from the output number of rotary encoder
It can be obtained in, at oneTcIn period, it is intended to searchPresent onA singular pointIf certain is strange
Dissimilarity is 0 there is no its value is then enabled, and otherwise its value is in the periodModulus maximum;
If have in a sensor groupQA temperature sensor, 1 power sensor;Definitiond Dimensional feature vector P:
Wherein:M=2~5;By in the opinion period to the average value of each temperature sensor sampling, altogetherIt is a, Q=1 ~ 3;
By in the opinion period to the average value of power sensor sampling, 1.
In the pre- diagnosis algorithm S2, using two graders based on HSSVM, each component in production line of rolling mill is right
Two graders are answered, the state of each component is divided into two classes:Faulty or fault-free, while fault credibility R is provided, institute
The fault credibility R provided characterizes the severity of failure, for controlling the size of sampling intermittent phase Twait;The Twork periods
Containing 64 Tc periods, each Tc extracts a feature vector Pj, j=1,2 ..., and 64, Pj is sent to two graders, to judge whether
It is faulty;If in 64 judgements of Twork periods, faulty number occupies the majority, then it is faulty to judge the component;When sentencing
Certain fixed component in the event of failure, is selected with regard to log-on data, and 64 current Pj are transmitted to high in the clouds carries out Precise Diagnosis;When any
It waits, unit operation maintenance personnel also can be by 64 Pj or the initial data series of current Twork periodsIt is transferred to high in the clouds
It is diagnosed.
In the pre- diagnosis algorithm S2, the process of HSSVM learning trainings is:Kernel function is selected as Gauss functions, establishes
The feature space of HSSVM;First, acquisition a batch has the Positive training sample P+ of good cohesion, deposits in HSSVM training samples
In this library;Again by iteration, a minimum sphere face that can wrap up all P+ is found, so that it is determined that hyperspherical radius and the centre of sphere
Etc. parameters, it is established that have the HSSVM of tentative diagnosis ability;In practical applications, with the rewards and punishments mechanism of intensified learning, according to
Current training result optimizes training sample database, so as to which the diagnosis performance of system is continuously improved;Whenever collected sample
This obtains unreliable label through Precise Diagnosis(Including positive and negative samples)Afterwards, then training sample database is added into, and to HSSVM again
Training, to determine a new hypersphere;If new HSSVM has better classification performance, the sample that this is added in is effective,
Otherwise this addition is cancelled, cancels this training.
The process of the Precise Diagnosis step S3 is:The fault type sample database for being stored with training sample is first established, then
It establishes failure to make a definite diagnosis and disaggregated model, receive after the feature vector containing failure diagnosed in advance, the training sample according to storage
Training sample of data and lane database is compared for this, and failure modes are made a definite diagnosis according to failure and determined with disaggregated model
Analysis result is first updated in the training sample of fault type sample database and is optimized again by intensified learning by fault type.
Alternatively, the process of the Precise Diagnosis step S3 is:First establish be stored with training sample fault type sample database,
Then failure sorted knowledge base is established failure and is made a definite diagnosis and disaggregated model and failure fuzzy evaluation model, receives from pre- diagnosis
The feature vector containing failure after, the training sample of data and lane database is compared point the training sample according to storage
Analysis, wherein failure modes make a definite diagnosis according to failure and determine fault type with disaggregated model, and failure sorted is according to fuzzy evaluation model
It is classified come the severity to failure, while the development trend of failure is predicted according to fuzzy state model, it will analysis knot
It is optimized in fruit update to the training sample of fault type sample database and failure sorted knowledge base by intensified learning.
In the Precise Diagnosis step S3, to diagnosing the faulty feature vector of possibility primarily determined in advance, further into
Row failure modes, the fault type include bearing defect, carrier wear, bearing saddle bore abrasion, axis cracking, shaft distortion, tooth
Take turns broken teeth, box deformation.
In the Precise Diagnosis step S3, failure modes are realized using SOM, each component corresponds to a SOM, SOM knots
The number of nodes of the input layer of structure is d(The dimension of feature vector P), the node number of output layer is the number of training sample, is 16
× 16;Training process to SOM is:
(1)The selection of initial training sample:Initial training samples sources in:A, domain expert uses kinetic theory and experience pair
The sample that all kinds of typical faults are analyzed and are fitted;B, simulated failure, the sample of acquisition are set on milling train component;C, reality
The sample that failure acquires when occurring;By above-mentioned by way of 256 samples are obtained, it is denoted as Pl, l=1,2 ..., 256;These samples should
Fault type listed by table 1 can be represented, the number of samples of each fault type is roughly the same;
Again to PlThe N × M component derived from vibrating sensor make normalized, for the training of SOM;Training is completed, knot
Fruit is on output layer plane, and the corresponding output node position of similar input sample is adjacent, and a type of failure is distributed in one
Sub-regions;Distributing position fault type corresponding with its of this subregion is kept, in case determining to use during fault type;
(2)The optimization of training sample:In practical applications, whenever the true negative sample received, it is added into training sample
Library and re -training SOM;And with the rewards and punishments mechanism of intensified learning, training sample database is carried out according to current training result excellent
Change, eliminate disadvantage sample;So as to which the diagnosis performance of system is continuously improved.
In the Precise Diagnosis step S3, failure is exactly divided into 5 grades by failure sorted by severity;1 grade of event
Hinder minimum, 5 grades of failure highests.
In the Precise Diagnosis step S3, also have to being carried out from the initial data diagnosed in advance by Computer Aided Analysis System
The function of auxiliary diagnosis.
Beneficial effects of the present invention are:By rationally setting various kinds of sensors at each position of milling train, milling train operation is collected
Data carry out data diagnosis and Precise Diagnosis in advance, can real-time online judge the type and fault bit of current rolling mill fault
It puts, can find rolling mill fault in time, prevent rolling mill fault from extending service life of equipment toward the performance severe stage.
Description of the drawings
Fig. 1 is the networking structure figure of the present invention;
Fig. 2 is the disposed of in its entirety flow chart of the present invention;
Fig. 3 is the fault pre-diagnosing process chart of the present invention;
Fig. 4 is the failure Precise Diagnosis process chart of the present invention;
Fig. 5 is the structure chart of the SOM of the present invention.
Specific embodiment
The specific embodiment of the present invention is described further below in conjunction with the accompanying drawings:
As shown in Figure 1, the present invention is mainly by being located at each milling train production scene(Client)Fault pre-diagnosing subsystem, positioned at examining
Disconnected service centre(High in the clouds)Failure Precise Diagnosis subsystem composition.
Fault pre-diagnosing subsystem is contained in the M in Fig. 11, M2... MnDotted line frame.Each subsystem is by sensor network and now
Processing mechanism into.One production line of rolling mill is made of multiple critical components, and sensor network is responsible for acquiring in production line of rolling mill
Physical parameter during each critical component operation.The multiple scales installed on one component are a sensor group, are denoted as Gi,
I=0,1 ... m, is shown in Fig. 1.Fault pre-diagnosing is mainly carried out by standalone object of critical component.Sensor network type is RS485
With Zigbee hybrid networks, using one-to-many star structure networking.Wherein RS485 subnets acquisition rest point, Zigbee subnets are adopted
Collect motor point.The terminal node power supply of Zigbee is wirelessly fed using electromagnetic coupling.Sensor network is by a Centroid, more
A terminal node composition.Each terminal in sensor network is made of a sensor and its control circuit.
(1)Sensor type and purposes:The sensor type and purposes that the present invention uses such as table 1:
1 sensor type of table and purposes
Sensor type | Purposes | Installation position |
Vibrating sensor | Acquire the three-dimensional vibrating waveform of transmission shaft, pinch roller, babinet etc. | Babinet, shafting, gear, bearing |
Temperature sensor | Acquire the operating temperature of each bearing part, motor | Bearing, motor |
Power sensor | Acquire the power consumption of each motor | Motor |
Rotary encoder | Rotational angle, the rotating speed of shaft are acquired, to acquire and analyzing other sensors data when provides synchronizing information | Shaft |
(2)The control circuit of sensor:It is made of AVR Series Industrials microcontroller, Zigbee communication module, RS485 interfaces etc.,
Control sampling rate, data bit width, sampling configuration, network communication etc..The information control that each sensor is provided in rotary encoder
Lower synchronized sampling, the data acquired converge at Centroid.Sampling rate is:Shaft often rotates 1 circle, samples 16 times.Data
Bit wide is 16bit.Sampling configuration for etc. intermittent bursts formula, i.e.,:One collection period includes sampling period Twork and intermittent phase
Twait.The present invention takes Twork as 64 circle the time it takes of shaft rotation, about 1800ms.Twait takes shaft to rotate 8 ~ 512
Enclose the time it takes, about 200ms ~ 16000ms.The size of Twait is related with fault severity level, and fault severity level is got over
Height, Twait are smaller.
(3)The Centroid of sensor network:By RAM processors, Zigbee communication module, RS485 interfaces, Ethernet
The compositions such as interface.Its function is:Zigbee networkings;It manages and coordinates each terminal node data acquisition, transmission in an orderly manner;Profit
The initial data that each terminal acquires is collected with Ethernet interfaces and is transferred to in-situ processing machine etc..The center of sensor network
Ethernet interface transmission datas are used between node and in-situ processing machine.Initial data is also stored in local storage, preserves
The data of most a recent period of time, as historical archives, if high in the clouds can remotely be transferred when needing.Local storage is solid state disk,
Capacity is 1T bytes, can buffer the initial data of about 12 months.
As shown in Fig. 2, a kind of high-speed rod-rolling mill remote online monitoring and intelligent diagnosing method, which is characterized in that including
Following steps:
S1 data collection steps:Each collected data of data acquisition equipment are by being pooled to after Centroid using local area network
Or wireless network transmissions are to industrial personal computer;
The pre- diagnosis algorithms of S2:The industrial personal computer is removed data while data are stored noise, data transformation and spy
Extraction process is levied, while treated data are carried out with the judgement that high-speed rod-rolling mill failure whether there is, and if exported as a result, work
The structural determination high-speed rod-rolling mill of control machine output is failure, then by this data transmission to monitoring and diagnostic center, industry control on the contrary
Machine does not transmit data to monitoring and diagnostic center then, is transmitted to diagnostic center to avoid by a large amount of trouble-free data, causes net
Network congestion, the burden for aggravating monitoring, diagnosing central server;
S3 Precise Diagnosis steps:The monitoring is received after faulty data with diagnostic center to the pre- diagnostic result of industrial personal computer
It is for further analysis again, to determine the type of failure and position.
Further, the vibrating sensor is on babinet, shafting, gear, parts of bearings position and acquisition includes
Rolling bearing, transmission shaft, pinch roller, babinet three-dimensional vibrating waveform;The rotary encoder is mounted in shaft and acquires packet
Include rotational angle, the rotating speed of shaft;The temperature sensor is on bearing, motor and acquisition includes bearing portion and motor
Operating temperature;The power sensor is mounted on motor and acquires the power consumption for including motor;Further, the vibration wave
Shape includes bearing vibration waveform and rolling mill vibration waveform;The operating temperature include rolling mill temperature, bearing portion temperature and
Rolling bearing temperature.
Further, the rotary encoder is sampled according to the following rules:Sampling period is divided into Twork and Twait two
A period, shaft per revolution sample 16 times, and sampled data bit wide is 16bit;The Twork periods are 64 circle of shaft rotation
The time it takes, as most preferred embodiment, value 1800ms is sampled 1024 times, the Twait periods in the Twork periods
Do not sample, the time is 8 ~ 512 circle the time it takes of shaft rotation, as most preferred embodiment, value range for 200ms ~
16000ms, specific time can set, and sampling period optimal duty ratio is 50%.
As shown in figure 3, the detailed process of the pre- diagnosis algorithm S2 is:The data of data acquisition equipment are stored,
The initial data stored can be transferred directly to high in the clouds, at the same data are converted and feature extraction processing, pass through two classification
Device(HSSVM)Feature vector of the selection containing failure is simultaneously uploaded to cloud interface, this contains the feature vector of failure through Precise Diagnosis step
HSSVM parameter libraries progress HSSVM learning trainings are rejoined to optimize two graders, two grader after confirming reliability
(HSSVM)Controlling of sampling is carried out to each sensor of data acquisition equipment also according to breakdown judge situation simultaneously.
Further, in the pre- diagnosis algorithm S2, data transformation is used simultaneously using the wavelet transformation of one-dimensional eight scale
Two graders of support vector machines in the present embodiment using selection Tree Classifier, carry out failure whether there is to collected data
Judgement, if it is determined that be faulty, then fault data is transferred to monitoring and diagnostic center, and divided into next step
Analysis.
Further, in the pre- diagnosis algorithm S2, the method for data transformation and feature extraction is:
(1)Data convert:The data series that one sensor is exported in the Twork periods are denoted asIfCome from vibrating sensor, then one-dimensional multi-scale wavelet transformation is carried out to it, to extract the singular point under each scale,
Transformation results are denoted as, whereinsIt is scale number,s=1,2,…N;The present invention takes 8 kinds of scales, i.e. N=8;
(2)Feature extraction:It is located in a sensor group, hasMA vibrating sensor correspondingly, has,
If the time that 1 circle of shaft rotation is spent(That is rotation period)ForTc, eachTcPositioning from the output data of rotary encoder
It can obtain, at oneTcIn period, it is intended to searchPresent onA singular pointIf certain singular point
There is no its value is then enabled to be 0, otherwise its value is in the periodModulus maximum;
If have in a sensor groupQA temperature sensor, 1 power sensor;Definitiond Dimensional feature vector P:
Wherein:By in the opinion period to the average value of each temperature sensor sampling, altogetherIt is a;To be discussed in the period
To power sensor sampling average value, 1;
In the present invention, M=2 ~ 5, Q=1 ~ 3;That is, for different components, the dimension of feature vectordIt is general different.
Further, it is every in production line of rolling mill using two graders based on HSSVM in the pre- diagnosis algorithm S2
A component corresponds to two graders, and the state of each component is divided into two classes:Faulty or fault-free, while be out of order
Confidence level R, given fault credibility R characterize the severity of failure, and the big of intermittent phase Twait is sampled for controlling
It is small;The Twork periods contain 64 Tc periods, and each Tc extracts a feature vector Pj, j=1,2 ..., and 64, send Pj to two classification
Device, it is faulty to judge whether;If in 64 judgements of Twork periods, faulty number occupies the majority, then judges the component
It is faulty;When judging certain component in the event of failure, log-on data is selected, 64 current Pj are transmitted to high in the clouds carries out accurately
Diagnosis;Whenever, unit operation maintenance personnel can also be by 64 Pj or the initial data series of current Twork periodsHigh in the clouds is transferred to be diagnosed.
In the present invention, two graders use a kind of hypersphere support vector machines(Hyper-Sphere Support Vector
Machine, HSSVM)It realizes, features described above vector P is its sample.
Further, in the pre- diagnosis algorithm S2, using HSSVM learning training systems, but in general, one SVM of training
Need positive sample and negative sample;Feature vector when positive sample is milling train component normal operation, can largely collect;Negative sample
It is feature vector when milling train component operates with failure, acquisition is difficult;Solution to this problem of the present invention is as follows:
(1)In view of a kind of pre- diagnosis only diagnosis early period, only failure judgement whether there is, and since positive sample has good cohesion
Property, therefore the present invention realizes this two classification using HSSVM;SVM has very strong generalization ability, can be divided into faulty mistake without reason
The risk control of barrier is to very little;
(2)The training of HSSVM:Kernel function is selected as Gauss functions, establishes the feature space of HSSVM;First, acquisition a batch has
The Positive training sample P+ of good cohesion, is deposited in HSSVM training sample databases;Again by iteration, finding one can wrap up
The minimum sphere face of all P+, so that it is determined that the parameters such as hyperspherical radius and the centre of sphere, it is established that have tentative diagnosis ability
HSSVM;
In practical applications, with the rewards and punishments mechanism of intensified learning, training sample database is optimized according to current training result,
So as to which the diagnosis performance of system is continuously improved;Whenever collected sample obtains unreliable label through Precise Diagnosis(Including positive and negative
Sample)Afterwards, then training sample database is added into, and to HSSVM re -trainings, to determine a new hypersphere;If new
HSSVM has better classification performance, then the sample of this addition is effective, otherwise cancels this addition, cancels this training.
As shown in figure 4, failure Precise Diagnosis is completed by the server catalyst Catalyst for being located at high in the clouds, function is:To diagnosing knot in advance
Fruit makees further Precise Diagnosis, obtains fault type and failure rank, and the detailed process of the Precise Diagnosis step S3 is:First build
Then the vertical fault type sample database for being stored with training sample is established failure and is made a definite diagnosis and disaggregated model, receives from pre- diagnosis
The feature vector containing failure after, the training sample of data and lane database is compared point the training sample according to storage
Analysis, failure modes make a definite diagnosis according to failure and determine fault type with disaggregated model, and analysis result is first updated to failure classes pattern
It is optimized again by intensified learning in the training sample in this library.
In the Precise Diagnosis step S3, first establish and be stored with the fault type sample database of training sample, failure sorted is known
Know library, then establish failure and make a definite diagnosis and disaggregated model and failure fuzzy evaluation model, receive from diagnosing in advance containing failure
After feature vector, the training sample of data and lane database is compared the training sample according to storage, wherein failure
Classification foundation failure is made a definite diagnosis determines fault type with disaggregated model, and failure sorted comes according to fuzzy evaluation model to the tight of failure
Weight degree is classified, while predict the development trend of failure according to fuzzy state model, by analysis result update to failure
It is optimized in the training sample of type sample database and failure sorted knowledge base by intensified learning.
In the Precise Diagnosis step S3, transmitted for the data between high in the clouds and user terminal and order, specially:(1)
The feature vector that user terminal uploads is received, for Precise Diagnosis;Feed back Precise Diagnosis result;(2)It can direct command use when needing
Family end uploads the primary fault data segment specified, as the auxiliary and supplement analysis data diagnosed to certain;(3)It is handed over user terminal
Mutually, various controls and the transmission of status information are realized.
In the Precise Diagnosis step S3, to diagnosing the faulty feature vector of possibility primarily determined in advance, further into
Row failure modes, fault type break including bearing defect, carrier wear, bearing saddle bore abrasion, axis cracking, shaft distortion, gear
Tooth, box deformation;The major critical component of high-speed rod-rolling mill production line has:Horizontal conveyor case, vertical type transmission case, gearbox,
Cone case subtracts sizing gearbox, cone case, pinch roller, Laying head ontology;The event of listed type in table 2 may occur for all parts
Barrier.
2 fault type table of table
Trouble point | Fault type |
Babinet | Bearing saddle bore is worn, box deformation |
Shafting | Axis cracks, shaft distortion, and lazy axis engagement is abnormal |
Gear | Beat tooth, broken teeth, fissure line |
Bearing | Bearing internal external circle or retainer damage, roll volume defect |
In the Precise Diagnosis step S3, using the Fault Classification based on SOM:
Due to self-organizing map neural network(Self-Organizing Maps, SOM)The tolerance of distortion and noise to sample
Greatly, the present invention realizes failure modes using SOM, and each component corresponds to a SOM;
The structure of SOM is as shown in figure 5, the number of nodes of its input layer is d(The dimension of feature vector P), the node number of output layer
It it is 16 × 16 for the number of training sample.
In the Precise Diagnosis step S3, the training process to SOM is:
(1)The selection of initial training sample:Initial training samples sources in:A, domain expert uses kinetic theory and experience pair
The sample that all kinds of typical faults are analyzed and are fitted;B, simulated failure, the sample of acquisition are set on milling train component;C, reality
The sample that failure acquires when occurring;
By above-mentioned by way of 256 samples are obtained, it is denoted as Pl, l=1,2 ..., 256;These samples should be able to represent failure listed by table 1
Type, the number of samples of each fault type are roughly the same;
Again to PlThe N × M component derived from vibrating sensor make normalized, for the training of SOM;Training is completed, knot
Fruit is on output layer plane, and the corresponding output node position of similar input sample is adjacent, and a type of failure is distributed in one
Sub-regions;Distributing position fault type corresponding with its of this subregion is kept, in case determining to use during fault type;
(2)The optimization of training sample:In practical applications, whenever the true negative sample received, it is added into training sample
Library and re -training SOM;And with the rewards and punishments mechanism of intensified learning, training sample database is carried out according to current training result excellent
Change, eliminate disadvantage sample;So as to which the diagnosis performance of system is continuously improved.
In the Precise Diagnosis step S3, failure is exactly divided into 5 grades by failure sorted by severity;1 grade of event
Hinder minimum, 5 grades of failure highests.
In the Precise Diagnosis step S3, also to being assisted from the initial data diagnosed in advance by Computer Aided Analysis System
Diagnosis, assistant diagnosis system provide the interface of an intervention diagnosis for domain expert;When necessary, domain expert can carry out such as
Lower work:
(1)Evaluation is provided to automatic diagnostic result, machine learning system to be helped to complete the optimization of sample database or knowledge base;
(2)Domain expert can analyze exceptional fault, as a result can carry out optimization system as expertise;
(3)The functions such as software analysis platform, common profession measurement and analytical instrument interface are provided for domain expert.
Described above is only presently preferred embodiments of the present invention, is not intended to limit the invention, all essences in the present invention
God and any modification, equivalent substitution, improvement and etc. within principle, done, should according to the spirit identical with us and principle
This is considered as the present invention should all be within protection scope of the present invention.
Claims (10)
1. a kind of high-speed rod-rolling mill remote online monitoring and intelligent diagnosis system, including data acquisition equipment, Centroid, work
Control machine, monitoring and diagnostic center, it is characterised in that:The collected data of data acquisition equipment are by being pooled to Centroid
Afterwards by local area network or wireless network transmissions to industrial personal computer, this data is passed through local area network or wireless network by the industrial personal computer
It is transferred to monitoring and diagnostic center.
2. a kind of high-speed rod-rolling mill remote online monitoring and intelligent diagnosing method, which is characterized in that include the following steps:
S1 data collection steps:Each collected data of data acquisition equipment are by being pooled to after Centroid using local area network
Or wireless network transmissions are to industrial personal computer, the data acquisition equipment includes vibrating sensor in a milling train unit, speed
Sensor, three dimension acceleration sensor, displacement sensor, rotary encoder, temperature sensor, power meter and power sensor;
The pre- diagnosis algorithms of S2:The industrial personal computer is removed data while data are stored noise, data transformation and spy
Extraction process is levied, while treated data are carried out with the judgement that high-speed rod-rolling mill failure whether there is, and if exported as a result, work
The structural determination high-speed rod-rolling mill of control machine output is failure, then by this data transmission to monitoring and diagnostic center, industry control on the contrary
Machine does not transmit data to monitoring and diagnostic center then, is transmitted to diagnostic center to avoid by a large amount of trouble-free data, causes net
Network congestion, the burden for aggravating monitoring, diagnosing central server;
S3 Precise Diagnosis steps:The monitoring is received after faulty data with diagnostic center to the pre- diagnostic result of industrial personal computer
It is for further analysis again, to determine the type of failure and position.
3. high-speed rod-rolling mill remote online monitoring according to claim 2 and intelligent diagnosing method, it is characterised in that:Institute
It states in data collection steps S1, rotary encoder is sampled according to the following rules:The sampling period is divided into Twork and Twait
Two periods, shaft per revolution sample 16 times, and the Twork periods are 64 circle the time it takes of shaft rotation,
The Twork periods sample 1024 times, and the Twait periods do not sample, and the time is 8 ~ 512 circle the time it takes of shaft rotation.
4. high-speed rod-rolling mill remote online monitoring according to claim 2 and intelligent diagnosing method, it is characterised in that:Institute
The process for stating pre- diagnosis algorithm S2 is:The data of data acquisition equipment are stored, the initial data stored can be adjusted directly
Take to high in the clouds, at the same data are converted and feature extraction processing, pass through two graders(HSSVM)Select the feature containing failure
Vector is simultaneously uploaded to cloud interface, which rejoins HSSVM after Precise Diagnosis step confirms reliability
Parameter library carries out HSSVM learning trainings to optimize two graders, two grader(HSSVM)Simultaneously also according to breakdown judge feelings
Condition carries out controlling of sampling to each sensor of data acquisition equipment.
5. high-speed rod-rolling mill remote online monitoring according to claim 4 and intelligent diagnosing method, it is characterised in that:Institute
It states in pre- diagnosis algorithm S2, the method for data transformation and feature extraction is:
(1)Data convert:The data series that one sensor is exported in the Twork periods are denoted asIfCome from vibrating sensor, then one-dimensional multi-scale wavelet transformation is carried out to it, to extract the singular point under each scale,
Transformation results are denoted as, whereinsIt is scale number,s=1,2,…N;The present invention takes 8 kinds of scales, i.e. N=8;
(2)Feature extraction:It is located in a sensor group, hasMA vibrating sensor correspondingly, hasIf the time that 1 circle of shaft rotation is spent(That is rotation period)ForTc, eachTcPositioning from rotation
It can be obtained in the output data of encoder, at oneTcIn period, it is intended to searchPresent onIt is a strange
Dissimilarity, it is 0 that its value is enabled if certain singular point is not present, and otherwise its value is in the periodModulus maximum;
If have in a sensor groupQA temperature sensor, 1 power sensor;Definitiond Dimensional feature vector P:
Wherein:M=2~5;By in the opinion period to the average value of each temperature sensor sampling, altogetherIt is a, Q=1 ~ 3;
By in the opinion period to the average value of power sensor sampling, 1.
6. high-speed rod-rolling mill remote online monitoring according to claim 5 and intelligent diagnosing method, it is characterised in that:Institute
It states in pre- diagnosis algorithm S2, using two graders based on HSSVM, each component in production line of rolling mill is one two points corresponding
The state of each component is divided into two classes by class device:Faulty or fault-free, while provide fault credibility R, given failure
Confidence level R characterizes the severity of failure, for controlling the size of sampling intermittent phase Twait;The Twork periods contain 64 Tc
Period, each Tc extract a feature vector Pj, j=1,2 ..., and 64, send Pj to two graders, it is faulty to judge whether;If
In 64 judgements of Twork periods, faulty number occupies the majority, then it is faulty to judge the component;When certain component of judgement
In the event of failure, log-on data is selected, 64 current Pj are transmitted to high in the clouds carries out Precise Diagnosis;Whenever, unit fortune
Row maintenance personnel also can be by 64 Pj or the initial data series of current Twork periodsHigh in the clouds is transferred to be diagnosed.
7. the high-speed rod-rolling mill remote online monitoring and intelligent diagnosing method, feature according to claim 4 or 6 exist
In:In the pre- diagnosis algorithm S2, the process of HSSVM learning trainings is:Kernel function is selected as Gauss functions, establishes the spy of HSSVM
Levy space;First, acquisition a batch has the Positive training sample P+ of good cohesion, deposits in HSSVM training sample databases;Again
By iteration, a minimum sphere face that can wrap up all P+ is found, so that it is determined that the parameters such as hyperspherical radius and the centre of sphere, build
Erect the HSSVM for having tentative diagnosis ability;In practical applications, with the rewards and punishments mechanism of intensified learning, according to current training
As a result training sample database is optimized, so as to which the diagnosis performance of system is continuously improved;Whenever collected sample is through accurate
Diagnosis obtains unreliable label(Including positive and negative samples)Afterwards, then training sample database is added into, and to HSSVM re -trainings, with true
A fixed new hypersphere;If new HSSVM has better classification performance, the sample that this is added in is effective, otherwise cancels
This is added in, and cancels this training.
8. high-speed rod-rolling mill remote online monitoring according to claim 4 and intelligent diagnosing method, it is characterised in that:Institute
The process for stating Precise Diagnosis step S3 is:The fault type sample database for being stored with training sample is first established, it is true then to establish failure
Examine and disaggregated model, receive after the feature vector containing failure diagnosed in advance, the training sample according to storage by data and
The training sample of lane database is compared, and failure modes make a definite diagnosis according to failure and determine fault type with disaggregated model,
Analysis result is first updated and is optimized again by intensified learning in the training sample of fault type sample database.
9. high-speed rod-rolling mill remote online monitoring according to claim 4 and intelligent diagnosing method, it is characterised in that:Institute
The process for stating Precise Diagnosis step S3 is:Fault type sample database, the failure sorted knowledge base for being stored with training sample are first established,
Then failure is established to make a definite diagnosis and disaggregated model and failure fuzzy evaluation model, receive from the feature containing failure diagnosed in advance to
After amount, according to storage training sample the training sample of data and lane database is compared, wherein failure modes according to
It is made a definite diagnosis according to failure and determines fault type with disaggregated model, failure sorted carrys out the severity to failure according to fuzzy evaluation model
It is classified, while the development trend of failure is predicted according to fuzzy state model, by analysis result update to failure classes patterns
It is optimized in the training sample of this library and failure sorted knowledge base by intensified learning.
10. high-speed rod-rolling mill remote online monitoring and intelligent diagnosing method, feature according to claim 8 or claim 9 exists
In:In the Precise Diagnosis step S3, failure modes are realized using SOM, each component corresponds to a SOM, SOM structures it is defeated
The number of nodes for entering layer is d, and the node number of output layer is the number of training sample, is 16 × 16;To the training process of SOM
For:
(1)The selection of initial training sample:Initial training samples sources in:A, domain expert uses kinetic theory and experience pair
The sample that all kinds of typical faults are analyzed and are fitted;B, simulated failure, the sample of acquisition are set on milling train component;C, reality
The sample that failure acquires when occurring;By above-mentioned by way of 256 samples are obtained, it is denoted as Pl, l=1,2 ..., 256;These samples should
Fault type listed by table 1 can be represented, the number of samples of each fault type is roughly the same;
Again to PlThe N × M component derived from vibrating sensor make normalized, for the training of SOM;Training is completed, as a result
On output layer plane, the corresponding output node position of similar input sample is adjacent, and a type of failure is distributed in one
Subregion;Distributing position fault type corresponding with its of this subregion is kept, in case determining to use during fault type;
(2)The optimization of training sample:In practical applications, whenever the true negative sample received, it is added into training sample
Library and re -training SOM;And with the rewards and punishments mechanism of intensified learning, training sample database is carried out according to current training result excellent
Change, eliminate disadvantage sample;So as to which the diagnosis performance of system is continuously improved.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201710307331.3A CN108126987B (en) | 2017-05-04 | 2017-05-04 | Remote online monitoring and intelligent diagnosis method for high-speed wire rolling mill |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201710307331.3A CN108126987B (en) | 2017-05-04 | 2017-05-04 | Remote online monitoring and intelligent diagnosis method for high-speed wire rolling mill |
Publications (2)
Publication Number | Publication Date |
---|---|
CN108126987A true CN108126987A (en) | 2018-06-08 |
CN108126987B CN108126987B (en) | 2020-07-28 |
Family
ID=62388440
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201710307331.3A Active CN108126987B (en) | 2017-05-04 | 2017-05-04 | Remote online monitoring and intelligent diagnosis method for high-speed wire rolling mill |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN108126987B (en) |
Cited By (15)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN108667927A (en) * | 2018-05-07 | 2018-10-16 | 山推工程机械股份有限公司 | Method for remote management, device, system and the storage medium of engineering machinery failure |
CN108994090A (en) * | 2018-07-06 | 2018-12-14 | 南京钢铁股份有限公司 | A kind of diagnostic method of down-pressing system of rolling mill failure |
CN109002672A (en) * | 2018-09-30 | 2018-12-14 | 珠海市运泰利自动化设备有限公司 | A kind of design method of the intelligent producing line based on optic test |
CN109143851A (en) * | 2018-07-11 | 2019-01-04 | 佛山科学技术学院 | The method of the identification of multiple labeling failure deep learning and its result intelligent expression |
CN109598309A (en) * | 2018-12-14 | 2019-04-09 | 广东寰球智能科技有限公司 | A kind of detection system and its monitoring method of mental package press machine |
CN109683565A (en) * | 2018-12-12 | 2019-04-26 | 电子科技大学 | A kind of instrument and meter fault detection method based on multi-method fusion |
CN109894476A (en) * | 2019-02-28 | 2019-06-18 | 北京首钢股份有限公司 | A kind of cold rolled silicon steel production line hydraulic equipment fault diagnosis method and device |
CN109909306A (en) * | 2019-03-11 | 2019-06-21 | 北京首钢自动化信息技术有限公司 | A kind of monitoring method of acid rolling plant on-line monitoring and degradation trend intelligent early-warning |
CN110825059A (en) * | 2019-11-29 | 2020-02-21 | 浙江工业大学 | Intelligent monitoring system and monitoring method for elevator wire rod production based on MQTT intelligent gateway |
CN111122155A (en) * | 2019-12-31 | 2020-05-08 | 湖南大学 | Gear fault diagnosis method based on telescopic shifting super-disc |
CN111438184A (en) * | 2020-03-20 | 2020-07-24 | 中冶赛迪工程技术股份有限公司 | System and method for handling accidents among high-speed wire rod module units |
CN111652227A (en) * | 2020-05-21 | 2020-09-11 | 哈尔滨市科佳通用机电股份有限公司 | Method for detecting damage fault of bottom floor of railway wagon |
CN111922095A (en) * | 2020-07-14 | 2020-11-13 | 上海数深智能科技有限公司 | Vibration diagnosis method for abnormal torsional vibration fault of roller of cold rolling mill |
CN112857797A (en) * | 2019-11-27 | 2021-05-28 | 睿捷国际股份有限公司 | Rolling mill equipment torsional vibration detection system and torsional vibration detection method |
CN113280910A (en) * | 2021-04-27 | 2021-08-20 | 圣名科技(广州)有限责任公司 | Real-time monitoring method and system for long product production line equipment |
Citations (7)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
KR100900676B1 (en) * | 2002-11-14 | 2009-06-01 | 주식회사 포스코 | Apparatus for diagnosing fault sucsessive vernier in mill |
CN102059256A (en) * | 2010-11-26 | 2011-05-18 | 首钢总公司 | Equipment operating and process quality state monitoring system for high-speed rolling mill |
CN202951709U (en) * | 2012-09-21 | 2013-05-29 | 浙江联鑫板材科技有限公司 | Continuous rolling system state remote monitoring device |
CN203343182U (en) * | 2013-07-13 | 2013-12-18 | 吉林大学 | Magnesium alloy sheet rolling parameter monitoring and fault diagnosis system |
CN203380190U (en) * | 2013-05-08 | 2014-01-08 | 浙江富春江环保热电股份有限公司 | Device for monitoring operating state of rolling mill |
CN103521531A (en) * | 2013-11-07 | 2014-01-22 | 天津理工大学 | Fault diagnosis and feedback system according to third octave flutter of high-speed cold rolling mill |
CN106292591A (en) * | 2016-08-20 | 2017-01-04 | 广州明珞汽车装备有限公司 | A kind of equipment big data platform monitoring diagnosis system |
-
2017
- 2017-05-04 CN CN201710307331.3A patent/CN108126987B/en active Active
Patent Citations (7)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
KR100900676B1 (en) * | 2002-11-14 | 2009-06-01 | 주식회사 포스코 | Apparatus for diagnosing fault sucsessive vernier in mill |
CN102059256A (en) * | 2010-11-26 | 2011-05-18 | 首钢总公司 | Equipment operating and process quality state monitoring system for high-speed rolling mill |
CN202951709U (en) * | 2012-09-21 | 2013-05-29 | 浙江联鑫板材科技有限公司 | Continuous rolling system state remote monitoring device |
CN203380190U (en) * | 2013-05-08 | 2014-01-08 | 浙江富春江环保热电股份有限公司 | Device for monitoring operating state of rolling mill |
CN203343182U (en) * | 2013-07-13 | 2013-12-18 | 吉林大学 | Magnesium alloy sheet rolling parameter monitoring and fault diagnosis system |
CN103521531A (en) * | 2013-11-07 | 2014-01-22 | 天津理工大学 | Fault diagnosis and feedback system according to third octave flutter of high-speed cold rolling mill |
CN106292591A (en) * | 2016-08-20 | 2017-01-04 | 广州明珞汽车装备有限公司 | A kind of equipment big data platform monitoring diagnosis system |
Cited By (24)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN108667927A (en) * | 2018-05-07 | 2018-10-16 | 山推工程机械股份有限公司 | Method for remote management, device, system and the storage medium of engineering machinery failure |
CN108994090B (en) * | 2018-07-06 | 2020-03-27 | 南京钢铁股份有限公司 | Method for diagnosing faults of rolling mill pressing system |
CN108994090A (en) * | 2018-07-06 | 2018-12-14 | 南京钢铁股份有限公司 | A kind of diagnostic method of down-pressing system of rolling mill failure |
CN109143851A (en) * | 2018-07-11 | 2019-01-04 | 佛山科学技术学院 | The method of the identification of multiple labeling failure deep learning and its result intelligent expression |
CN109143851B (en) * | 2018-07-11 | 2021-06-01 | 佛山科学技术学院 | Method for recognizing multi-mark fault deep learning and intelligently expressing result thereof |
CN109002672A (en) * | 2018-09-30 | 2018-12-14 | 珠海市运泰利自动化设备有限公司 | A kind of design method of the intelligent producing line based on optic test |
CN109683565A (en) * | 2018-12-12 | 2019-04-26 | 电子科技大学 | A kind of instrument and meter fault detection method based on multi-method fusion |
CN109598309A (en) * | 2018-12-14 | 2019-04-09 | 广东寰球智能科技有限公司 | A kind of detection system and its monitoring method of mental package press machine |
CN109598309B (en) * | 2018-12-14 | 2022-02-15 | 广东寰球智能科技有限公司 | Detection system and monitoring method of metal packaging punching machine |
CN109894476A (en) * | 2019-02-28 | 2019-06-18 | 北京首钢股份有限公司 | A kind of cold rolled silicon steel production line hydraulic equipment fault diagnosis method and device |
CN109894476B (en) * | 2019-02-28 | 2020-06-26 | 北京首钢股份有限公司 | Fault diagnosis method and device for hydraulic equipment of cold-rolled silicon steel production line |
CN109909306B (en) * | 2019-03-11 | 2020-08-21 | 北京首钢自动化信息技术有限公司 | Monitoring method for on-line monitoring and intelligent degradation trend early warning of acid rolling equipment |
CN109909306A (en) * | 2019-03-11 | 2019-06-21 | 北京首钢自动化信息技术有限公司 | A kind of monitoring method of acid rolling plant on-line monitoring and degradation trend intelligent early-warning |
CN112857797B (en) * | 2019-11-27 | 2023-08-08 | 睿捷国际股份有限公司 | Torsional vibration detection system and method for rolling mill equipment |
CN112857797A (en) * | 2019-11-27 | 2021-05-28 | 睿捷国际股份有限公司 | Rolling mill equipment torsional vibration detection system and torsional vibration detection method |
CN110825059A (en) * | 2019-11-29 | 2020-02-21 | 浙江工业大学 | Intelligent monitoring system and monitoring method for elevator wire rod production based on MQTT intelligent gateway |
CN111122155B (en) * | 2019-12-31 | 2021-10-12 | 湖南大学 | Gear fault diagnosis method based on telescopic shifting super-disc |
CN111122155A (en) * | 2019-12-31 | 2020-05-08 | 湖南大学 | Gear fault diagnosis method based on telescopic shifting super-disc |
CN111438184A (en) * | 2020-03-20 | 2020-07-24 | 中冶赛迪工程技术股份有限公司 | System and method for handling accidents among high-speed wire rod module units |
CN111438184B (en) * | 2020-03-20 | 2024-05-17 | 中冶赛迪工程技术股份有限公司 | Accident handling system and method between high-speed wire rod module units |
CN111652227B (en) * | 2020-05-21 | 2021-01-22 | 哈尔滨市科佳通用机电股份有限公司 | Method for detecting damage fault of bottom floor of railway wagon |
CN111652227A (en) * | 2020-05-21 | 2020-09-11 | 哈尔滨市科佳通用机电股份有限公司 | Method for detecting damage fault of bottom floor of railway wagon |
CN111922095A (en) * | 2020-07-14 | 2020-11-13 | 上海数深智能科技有限公司 | Vibration diagnosis method for abnormal torsional vibration fault of roller of cold rolling mill |
CN113280910A (en) * | 2021-04-27 | 2021-08-20 | 圣名科技(广州)有限责任公司 | Real-time monitoring method and system for long product production line equipment |
Also Published As
Publication number | Publication date |
---|---|
CN108126987B (en) | 2020-07-28 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN108126987A (en) | A kind of high-speed rod-rolling mill remote online monitoring and intelligent diagnosis system and method | |
CN107003663B (en) | The monitoring of device with movable part | |
CN106682814B (en) | Wind turbine generator fault intelligent diagnosis method based on fault knowledge base | |
CN109144014A (en) | The detection system and method for industrial equipment operation conditions | |
CN104329222B (en) | A kind of on-line condition monitoring being integrated in fan master control system and method for diagnosing faults | |
CN111046945A (en) | Fault type and damage degree diagnosis method based on combined convolutional neural network | |
EP1808768A2 (en) | Automatic remote monitoring and diagnostics system and communication method for communicating between a programmable logic controller and a central unit | |
CN107797537A (en) | A kind of prognostic and health management method applied to automatic production line | |
Wang et al. | Collaborative deep learning framework for fault diagnosis in distributed complex systems | |
CN106951197A (en) | A kind of printing equipment remote fault diagnosis method and system based on signal testing | |
CN106769051B (en) | A kind of rolling bearing remaining life prediction technique based on MCEA-KPCA and combination S VR | |
Hoffman et al. | The application of neural networks to vibrational diagnostics for multiple fault conditions | |
CN107473036A (en) | Elevator traction machine remote online detects and diagnostic system and its checkout and diagnosis method | |
CN112329520B (en) | Truck bearing fault identification method based on generation countermeasure learning | |
CN112200263B (en) | Self-organizing federal clustering method applied to power distribution internet of things | |
CN115563853A (en) | Rolling bearing fault diagnosis method based on digital twinning | |
CN110056640A (en) | Speed reducer wireless malfunction diagnostic method based on acceleration signal and edge calculations | |
CN116880402A (en) | Intelligent factory cooperative control system and method thereof | |
CN113496061A (en) | SOM network clustering electromechanical device bearing fault analysis method based on transfer learning and manifold distance | |
CN113114748A (en) | Intelligent prediction system based on data analysis | |
CN110160781B (en) | Test set reconstruction and prediction method for rotary machine fault classification | |
Guo et al. | Personalized federated learning for multi-task fault diagnosis of rotating machinery | |
Kolar et al. | Identification of inability states of rotating machinery subsystems using industrial IoT and convolutional neural network–initial research | |
CN116244640A (en) | Unbalanced bearing fault diagnosis method and system based on federal learning | |
Wang et al. | Wavelet Packet Transform‐Assisted Least Squares Support Vector Machine for Gear Wear Degree Diagnosis |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
GR01 | Patent grant | ||
GR01 | Patent grant |