CN112781903B - Blast furnace blower and TRT set fault diagnosis method based on digital twin system - Google Patents

Blast furnace blower and TRT set fault diagnosis method based on digital twin system Download PDF

Info

Publication number
CN112781903B
CN112781903B CN202011583058.5A CN202011583058A CN112781903B CN 112781903 B CN112781903 B CN 112781903B CN 202011583058 A CN202011583058 A CN 202011583058A CN 112781903 B CN112781903 B CN 112781903B
Authority
CN
China
Prior art keywords
frequency
blast furnace
data
digital twin
twin system
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Active
Application number
CN202011583058.5A
Other languages
Chinese (zh)
Other versions
CN112781903A (en
Inventor
柴秋子
龚亦昕
吴平
刘唐丁
李创
楼嗣威
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Hang Zhou Zeta Technology Co Lts
Original Assignee
Hang Zhou Zeta Technology Co Lts
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by Hang Zhou Zeta Technology Co Lts filed Critical Hang Zhou Zeta Technology Co Lts
Priority to CN202011583058.5A priority Critical patent/CN112781903B/en
Publication of CN112781903A publication Critical patent/CN112781903A/en
Application granted granted Critical
Publication of CN112781903B publication Critical patent/CN112781903B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Classifications

    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01M—TESTING STATIC OR DYNAMIC BALANCE OF MACHINES OR STRUCTURES; TESTING OF STRUCTURES OR APPARATUS, NOT OTHERWISE PROVIDED FOR
    • G01M99/00—Subject matter not provided for in other groups of this subclass
    • F—MECHANICAL ENGINEERING; LIGHTING; HEATING; WEAPONS; BLASTING
    • F04—POSITIVE - DISPLACEMENT MACHINES FOR LIQUIDS; PUMPS FOR LIQUIDS OR ELASTIC FLUIDS
    • F04D—NON-POSITIVE-DISPLACEMENT PUMPS
    • F04D27/00—Control, e.g. regulation, of pumps, pumping installations or pumping systems specially adapted for elastic fluids
    • F04D27/001—Testing thereof; Determination or simulation of flow characteristics; Stall or surge detection, e.g. condition monitoring
    • 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
    • G06N3/08—Learning methods
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T17/00—Three dimensional [3D] modelling, e.g. data description of 3D objects
    • Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
    • Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
    • Y02P—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN THE PRODUCTION OR PROCESSING OF GOODS
    • Y02P90/00—Enabling technologies with a potential contribution to greenhouse gas [GHG] emissions mitigation
    • Y02P90/30—Computing systems specially adapted for manufacturing

Landscapes

  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • Software Systems (AREA)
  • General Engineering & Computer Science (AREA)
  • Evolutionary Computation (AREA)
  • Computing Systems (AREA)
  • Computational Linguistics (AREA)
  • Data Mining & Analysis (AREA)
  • Biomedical Technology (AREA)
  • General Health & Medical Sciences (AREA)
  • Molecular Biology (AREA)
  • Biophysics (AREA)
  • Artificial Intelligence (AREA)
  • Mathematical Physics (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Health & Medical Sciences (AREA)
  • Computer Graphics (AREA)
  • Geometry (AREA)
  • Mechanical Engineering (AREA)
  • Testing And Monitoring For Control Systems (AREA)

Abstract

The invention discloses a blast furnace blower and TRT set fault diagnosis method based on a digital twin system, which comprises the following steps: constructing three-dimensional models of a blast furnace blower and a TRT unit system; constructing a blast furnace blower and a TRT set digital twin system by acquiring field real-time data and calculating virtual data; performing abnormal data elimination on data based on a digital twin system; extracting frequency domain features by using an improved vibration frequency spectrum ratio extraction method; constructing a neural network fault diagnosis algorithm based on an Adam algorithm according to the time-frequency domain characteristics and the operation parameters of the equipment; and finally, the output result of the neural network is stored in a digital twin system to form three-dimensional dynamic display of the fault information of the equipment fault component, generate a diagnosis report and push the diagnosis report to field management personnel. The method can effectively realize the fault diagnosis and the health analysis of the blast furnace blower and the TRT unit, and can guide the field operation by combining with the digital twin system, thereby ensuring the safe and efficient operation of field equipment.

Description

Blast furnace blower and TRT set fault diagnosis method based on digital twin system
Technical Field
The invention relates to the fields of digital twin systems, fault diagnosis and the like, in particular to a fault diagnosis method for a blast furnace blower and a TRT (blast furnace gas turbine unit) based on a digital twin system.
Background
The blast furnace blast and TRT system is an important technological process for blast furnace iron making, can not only utilize the residual pressure of blast furnace gas to carry out high-efficiency power generation, but also effectively solve the noise pollution and pipeline vibration generated by the pressure reducing valve bank, and also plays an important role in stably controlling the top pressure of the blast furnace. Practice proves that the TRT power generation amount is about 40% of the power consumption of a blast furnace blower, so the health condition and the running state of the blast furnace blower and a TRT unit directly influence the yield and the safety of iron making, and the fault diagnosis of the blast furnace blower and the TRT unit is extremely important. However, in the blast furnace ironmaking process, the blast furnace blower and the TRT unit are often in the states of bad working conditions, instability, high power, heavy load and continuous operation, and malignant accidents caused by operation faults are rare.
With the research and the deepening of various intelligent algorithms, more and more algorithms are applied to fault diagnosis, but as the operation conditions of a blast furnace blower and a TRT unit are complex and changeable, the operation parameters are various, and the fault feature extraction also becomes a big difficulty in the field of fault diagnosis.
Disclosure of Invention
The invention overcomes the problems of complex and changeable operating conditions, various operating parameters and difficult extraction of fault characteristics of a blast furnace blower and a TRT unit, and provides a fault diagnosis method of the blast furnace blower and the TRT unit based on a digital twin system.
In order to achieve the above object, the present invention provides the following solutions:
a blast furnace blower and TRT set fault diagnosis method based on a digital twin system comprises the steps of firstly, constructing three-dimensional models of a blast furnace blower and a TRT set system; constructing a blast furnace blower and a TRT set digital twin system by acquiring field real-time data and calculating virtual data; performing abnormal data elimination on data based on a digital twin system; extracting time-frequency domain features by using an improved vibration frequency spectrum ratio extraction method; constructing a neural network fault diagnosis algorithm based on an Adam algorithm according to the time-frequency domain characteristics and other operating parameters; and finally, the output result of the neural network is stored in a digital twin system to form three-dimensional dynamic display of the fault information of the equipment fault component, generate a diagnosis report and push the diagnosis report to field management personnel.
In the above technical solution, preferably, in the step S1, three-dimensional modeling is performed based on Unity3D, the three-dimensional modeling includes three-dimensional modeling of the whole production line of the blast furnace blower and the TRT unit, and the on-site device scenario is reproduced and is consistent with the on-site process flow.
Preferably, the blast furnace blower and the TRT set digital twin system can collect real-time data on site and predict virtual data of data modeling, two scenes of a real-time system which operates synchronously with the site and a virtual twin system which can operate when the site is shut down are realized, and fault data of each scene are marked.
Preferably, the abnormal data elimination is to eliminate data with abnormal transmission or abnormal data caused by accidental variation factors in the environment by using a moving average algorithm, so as to avoid affecting the accuracy of fault diagnosis. The specific method comprises the following steps:
the operation parameter variable X at the time t after abnormal data are eliminated is marked as Xt，θtBefore abnormal data is eliminated, the value of a running parameter variable X at the time t is determined, beta is a moving average coefficient, beta belongs to [0,1 ], and when beta is 0, the moving average is not used, and X ist＝θt(ii) a After using the running average:
Xt＝β*Xt-1+(1-β)*θt
preferably, the improved vibration spectrum occupancy ratio extraction method in step S4 includes the following steps:
(1) acquiring a vibration time domain waveform signal x (n) ═ x of the device1,x2,x3…xN]Wherein x isNThe vibration acceleration value is N, and the number of sampling points is N;
(2) performing Fast Fourier Transform (FFT) on the vibration time domain waveform signal, and according to a Fourier transform formula:
Figure BDA0002866333300000021
wherein k is more than or equal to 0 and less than or equal to N-1, N is the nth data, k is the kth value in the frequency domain, WN＝e-j*2*π/NIs a twiddle factor, let N be 2rDividing x (N) into two halves to obtain two sequences with the length of N/2,the following are obtained through integrated calculation:
Figure BDA0002866333300000022
(3) obtaining a frequency domain value [ f1, | X (1) | of the vibration signal after FFT; f2, | X (2); … fN, | X (n) | ], wherein f1, f2, …, fN is the frequency value after FFT, | X (1) |, | X (2) |, …, | X (n) | is the frequency amplitude at the corresponding frequency;
(4) the characteristic frequencies of the computing device are respectively 0.5 frequency doubling, 1 frequency doubling, 1.5 frequency doubling, 2 frequency doubling, 2.5 frequency doubling, 3 frequency doubling, 3.5 frequency doubling, 4 frequency doubling and high frequency doubling, and are recorded as: [ f ] of0.5X,f1X,f1.5X,f2X,f2.5X,f3X,f3.5X,f4X,fnX]The frequency multiplication is the multiple of the fundamental frequency, and the fundamental frequency is equal to 1 frequency multiplication;
wherein f isnXFrequencies that are high multiples, i.e., frequencies greater than 4 multiples,
Figure BDA0002866333300000031
wherein f is1XThe frequency is 1 frequency doubling, Vr is the rotating speed of the equipment, and the obtained characteristic frequency of the equipment is as follows:
Figure BDA0002866333300000032
where n is the high frequency multiplication, meaning all frequencies greater than 4 frequency multiplication;
(5) calculating the spectrum ratio of the characteristic frequency:
if
Figure BDA0002866333300000033
If fi is the ith frequency value, selecting all the corresponding amplitudes of fi to be integrated as the amplitude of 0.5 frequency multiplication, and obtaining the amplitude ratio of 0.5 frequency multiplication as follows: (i is the ith frequency, or the amplitude corresponding to the ith frequency)
Figure BDA0002866333300000034
② in the same way, if
Figure BDA0002866333300000035
fj is the jth frequency value, and the amplitude ratio of the obtained 1 multiplied frequency is:
Figure BDA0002866333300000036
③ in the same way, if
Figure BDA0002866333300000037
fa is the a-th frequency value, and the amplitude ratio of the obtained 1.5 frequency multiplication is:
Figure BDA0002866333300000038
fourthly, if
Figure BDA0002866333300000039
fb is the b-th frequency value, and the obtained 2-frequency multiplication amplitude ratio is:
Figure BDA00028663333000000310
the same thing applies if
Figure BDA00028663333000000311
fc is the c frequency value, and the obtained 2.5 frequency multiplication amplitude ratio is:
Figure BDA0002866333300000041
sixthly, in the same way, if
Figure BDA0002866333300000042
fd is the d-th frequency value, and the obtained 3-frequency multiplication amplitude ratio is:
Figure BDA0002866333300000043
seventhly, in the same way, if
Figure BDA0002866333300000044
fg is the g-th frequency value, and the obtained 3.5 frequency multiplication amplitude ratio is:
Figure BDA0002866333300000045
eighthly, if
Figure BDA0002866333300000046
fe is the e-th frequency value, and the obtained 4-frequency multiplication amplitude ratio is:
Figure BDA0002866333300000047
ninthly, if
Figure BDA0002866333300000048
fn is the nth frequency value, and the amplitude ratio of the obtained high frequency multiplication is as follows:
Figure BDA0002866333300000049
wherein sum (| X (n) |) is: (
Figure BDA00028663333000000410
I x (n) i) the sum of the amplitudes of the first t maxima, t ═ e/8.
Preferably, the method for constructing the neural network fault diagnosis algorithm based on the Adam algorithm comprises the following steps:
(1) determining input and output of a neural network, wherein the input is operation parameter variables and vibration time-frequency domain data of the equipment, and the input comprises current, voltage, power, temperature, flow, triaxial vibration effective value, triaxial vibration time-domain index and triaxial frequency-domain index; the output is the fault type of the equipment, including the abrasion, unbalance, misalignment, looseness of a base, faults of a bearing inner ring, faults of a bearing outer ring and friction faults of a movable part and a static part of the equipment.
(2) The following parameters of the neural network are determined: the number of nodes of an input layer is 8, the number of nodes of an output layer is 1, the number of hidden layers is 1, the number of nodes of each hidden layer is 26, the network learning rate is 0.21, and the momentum coefficient is 0.01;
(3) optimizing the weight of the neural network by using an accelerated gradient algorithm (Adam algorithm), wherein the specific method comprises the following steps: introducing a square gradient on the basis of a gradient descent method of the amount of fluctuation, and correcting the deviation of the speed;
(4) constructing complete forward and reverse calculation neural network models, reading data with fault identification in the digital twin system, and training and testing;
(5) and reading real-time operation data to perform fault diagnosis and output.
The invention has the beneficial effects that:
according to the invention, three-dimensional models of the blower and the TRT are constructed, the blower and TRT neural network fault diagnosis algorithm is constructed by utilizing an improved frequency spectrum ratio method, real-time monitoring and fault diagnosis of the blower and the TRT can be realized, and the blower and the TRT are displayed in a three-dimensional visual form, so that the problems that the operation conditions of the blast furnace blower and the TRT unit are complicated and changeable, the operation parameters are various, and the fault characteristics are difficult to extract can be effectively solved. The method can effectively realize the fault diagnosis and the health analysis of the blast furnace blower and the TRT unit, and intuitively display the fault three-dimensional information of the equipment by combining with the digital twin system to guide the field operation, thereby ensuring the safe and efficient operation of the field equipment and having very high actual production value.
Drawings
FIG. 1 is a flow chart of a fault diagnosis method for a blast furnace blower and a TRT set based on a digital twin system.
FIG. 2 is a structural frame diagram of a blast furnace blower and a TRT digital twinning system of the present invention.
FIG. 3 is a schematic diagram of an input/output structure of a neural network according to the present invention.
Detailed Description
As shown in FIG. 1, the invention provides a blast furnace blast and TRT set fault diagnosis method based on a digital twin system, firstly, a three-dimensional model of a blast furnace blast and TRT set system is constructed; constructing a blast furnace blower and a TRT set digital twin system by acquiring field real-time data and calculating virtual data; performing abnormal data elimination on data based on a digital twin system; extracting time-frequency domain features by using an improved vibration frequency spectrum ratio extraction method; constructing a neural network fault diagnosis algorithm based on an Adam algorithm according to the time-frequency domain characteristics and other operating parameters; and finally, the output result of the neural network is stored in a digital twin system, so that the three-dimensional dynamic display of the fault information of the equipment fault part is formed, a diagnosis report is generated, and the diagnosis report is pushed to field management personnel.
The construction of the three-dimensional models of the blast furnace blower and the TRT set is based on Unity3D to carry out three-dimensional modeling, and the three-dimensional modeling comprises three-dimensional modeling of the whole production line of the blast furnace blower and the TRT set, and the on-site equipment scene is reproduced and is consistent with the on-site process flow.
The blast furnace blower and TRT unit digital twin system comprises functions of three-dimensional model real-time scene roaming, three-dimensional model virtual scene roaming, real-time operation data acquisition and storage, virtual data real-time prediction and storage, data marking and classification before failure, failure result storage, alarm pushing and the like, and two scenes of a real-time system which operates synchronously with a site and a virtual twin system which can operate when the site is shut down are realized.
The blast furnace blower and TRT unit digital twin system comprises an operation module and a simulation module, wherein the operation module is mainly used for acquiring field real-time data through communication of an intelligent sensor and a field database, and specifically comprises a blast furnace top pressure and blower control performance evaluation module, a blower and blast furnace top pressure control parameter optimization module, a blower surge identification module, a unit energy efficiency analysis module, a unit energy flow module, a turbine blade dust deposition module and a unit health diagnosis module, so that the real-time monitoring and state analysis of the blast furnace blower and the TRT unit are realized; the simulation module realizes the establishment of the blower control simulation module, the blast furnace top pressure control simulation module, the blower anti-surge control simulation module, the TRT startup process control simulation module, the TRT shutdown process control simulation module and the TRT emergency shutdown process control simulation module by setting relevant on-site working condition.
The invention eliminates the abnormal data of transmission or abnormal data caused by accidental variation factors in the environment by a moving average algorithm so as to avoid influencing the accuracy of fault diagnosis. The specific method comprises the following steps:
and recording an operation parameter variable X at the t moment after abnormal data are removed as Xt，θtBefore abnormal data is eliminated, the value of a running parameter variable X at the time t is determined, beta is a moving average coefficient, beta belongs to [0,1 ], and when beta is 0, the moving average is not used, and X ist＝θt(ii) a After using the sliding average:
Xt＝β*Xt-1+(1-β)*θt
the improved vibration frequency spectrum ratio extraction method comprises the following steps:
(1) acquiring a vibration time domain waveform signal x (n) ═ x of the device1,x2,x3…xN]Wherein x isNThe vibration acceleration value is N, and the number of sampling points is N;
(2) performing Fast Fourier Transform (FFT) on the vibration time domain waveform signal, and according to a Fourier transform formula:
Figure BDA0002866333300000061
wherein k is more than or equal to 0 and less than or equal to N-1, N is the nth data, k is the kth value in the frequency domain, WN＝e-j*2*π/NIs a twiddle factor, let N be 2rDividing x (N) into front and back halves to obtain two sequences with the length of N/2, and obtaining the sequence by integrated calculation:
Figure BDA0002866333300000071
(3) finally, obtaining a frequency domain value [ f1, | X (1) | of the vibration signal after FFT; f2, | X (2); … fN, | X (n) | ], wherein f1, f2, …, fN is the frequency value after FFT, | X (1) |, | X (2) |, …, | X (n) | is the frequency amplitude at the corresponding frequency;
(4) the characteristic frequencies of the computing device are respectively 0.5 frequency doubling, 1 frequency doubling, 1.5 frequency doubling, 2 frequency doubling, 2.5 frequency doubling, 3 frequency doubling, 3.5 frequency doubling, 4 frequency doubling and high frequency doubling (the frequencies greater than 4 frequency doubling are collectively referred to as high frequency doubling), and are recorded as: [ f ] of0.5X,f1X,f1.5X,f2X,f2.5X,f3X,f3.5X,f4X,fnX]，
Wherein f isnXFrequencies that are high multiples, i.e., frequencies greater than 4 multiples,
Figure BDA0002866333300000072
wherein f is1XThe frequency is 1 multiplied frequency, Vr is the rotating speed of the equipment, and the obtained characteristic frequency of the equipment is as follows:
Figure BDA0002866333300000073
where n is the high frequency multiplication, meaning all frequencies greater than 4 frequency multiplication;
(5) calculating the spectrum ratio of the characteristic frequency:
if
Figure BDA0002866333300000074
If fi is the ith frequency value, selecting all the corresponding amplitudes of fi to be integrated as the amplitude of 0.5 frequency multiplication, and obtaining the amplitude ratio of 0.5 frequency multiplication as follows: (i is the ith frequency, or the amplitude corresponding to the ith frequency)
Figure BDA0002866333300000075
② in the same way, if
Figure BDA0002866333300000076
fj is the jth frequency value, and the amplitude ratio of the obtained 1 multiplied frequency is:
Figure BDA0002866333300000077
thirdly, if all things are equal, if
Figure BDA0002866333300000078
fa is the a-th frequency value, and the amplitude ratio of the obtained 1.5 frequency multiplication is:
Figure BDA0002866333300000079
fourthly, if
Figure BDA00028663333000000710
fb is the b-th frequency value, and the obtained 2-frequency multiplication amplitude ratio is:
Figure BDA0002866333300000081
the same thing applies if
Figure BDA0002866333300000082
fc is the frequency value of the c, and the obtained amplitude ratio of 2.5 frequency multiplication is as follows:
Figure BDA0002866333300000083
sixth, if the same principle is adopted
Figure BDA0002866333300000084
fs is the d-th frequency value, and the obtained 3-frequency multiplication amplitude ratio is:
Figure BDA0002866333300000085
seventhly, in the same way, if
Figure BDA0002866333300000086
fg is the g-th frequency value, and the obtained 3.5 frequency multiplication amplitude ratio is:
Figure BDA0002866333300000087
eighthly, if
Figure BDA0002866333300000088
fe is the e-th frequency value, and the obtained 4-frequency multiplication amplitude ratio is:
Figure BDA0002866333300000089
ninthly, if
Figure BDA00028663333000000810
fn is the nth frequency value, and the amplitude ratio of the obtained high frequency multiplication is as follows:
Figure BDA00028663333000000811
wherein sum (| X (n) |) is: (
Figure BDA00028663333000000812
I x (n) i) the sum of the amplitudes of the first t maxima, t ═ e/8.
The method for constructing the neural network fault diagnosis algorithm based on the Adam algorithm comprises the following steps:
(1) determining input and output of a neural network, as shown in fig. 3, wherein the input is an operation parameter variable and vibration time-frequency domain data of the equipment, including current, voltage, power, temperature, flow, triaxial vibration effective value, triaxial vibration time-domain index and triaxial frequency-domain index; the output is the fault type of the equipment, including the abrasion of the rotor of the equipment, unbalance, misalignment, looseness of a base, faults of the inner ring of a bearing, faults of the outer ring of the bearing and friction faults of a movable part and a static part.
Wherein, the time domain indexes comprise a peak index, a kurtosis index, a skewness index, a margin index and a pulse index; the frequency domain indexes comprise 0.5 frequency doubling, 1 frequency doubling, 1.5 frequency doubling, 2 frequency doubling, 2.5 frequency doubling, 3 frequency doubling, 3.5 frequency doubling, 4 frequency doubling and high frequency doubling (the frequencies greater than 4 frequency doubling are collectively called high frequency doubling).
(2) The following parameters of the neural network are determined: the number of nodes of an input layer is 8, the number of nodes of an output layer is 1, the number of hidden layers is 1, the number of nodes of each hidden layer is 26, the network learning rate is 0.21, and the momentum coefficient is 0.01;
(3) optimizing the weight of the neural network by using an accelerated gradient algorithm (Adam algorithm), wherein the specific method is to introduce a square gradient on the basis of a gradient descent method with momentum and correct the deviation of the rate;
(4) constructing complete forward and reverse calculation neural network models, reading data with fault identification in the digital twin system, and training and testing;
(5) and reading real-time operation data to perform fault diagnosis, and outputting and storing the data in a corresponding module of the digital twin system.

Claims (5)

1. A blast furnace blower and TRT unit fault diagnosis method based on a digital twin system is characterized by comprising the following steps:
s1: constructing three-dimensional models of a blast furnace blower and a TRT unit system;
s2: constructing a blast furnace blower and a TRT set digital twin system by acquiring field real-time data and calculating virtual data;
s3: performing abnormal data elimination on data based on a digital twin system;
s4: extracting time-frequency domain features by using an improved vibration spectrum ratio extraction method;
s5: constructing a neural network fault diagnosis algorithm based on an Adam algorithm according to the time-frequency domain characteristics and the operation parameters of the equipment;
s6: the output result of the neural network is stored in the digital twin system, the three-dimensional dynamic display of the fault information of the equipment fault component is formed, a diagnosis report is generated, and the diagnosis report is pushed to field management personnel;
the improved vibration frequency spectrum ratio extraction method described in step S4 specifically includes the following steps:
s41: acquiring a vibration time domain waveform signal x (n) ═ x of the device1,x2,x3…xN]Wherein x isNThe vibration acceleration value is N, and the number of sampling points is N;
s42: performing Fast Fourier Transform (FFT) on the vibration time domain waveform signal, and according to a Fourier transform formula:
Figure FDA0003553428130000011
wherein k is more than or equal to 0 and less than or equal to N-1, N is nth sampling data, k is kth data on a frequency domain, and WN＝e-j*2*π/NIs a twiddle factor, let N be 2rDividing x (N) into front and back halves to obtain two sequences with the length of N/2, and obtaining the sequence by integrated calculation:
Figure FDA0003553428130000012
s43: obtaining a frequency domain value [ f1, | X (1) | of the vibration signal after FFT; f2, | X (2); … fN, | X (n) | ], wherein f1, f2, …, fN is the frequency value after FFT, | X (1) |, | X (2) |, …, | X (n) | is the frequency amplitude at the corresponding frequency;
s44: the characteristic frequencies of the computing device are respectively 0.5 frequency doubling, 1 frequency doubling, 1.5 frequency doubling, 2 frequency doubling, 2.5 frequency doubling, 3 frequency doubling, 3.5 frequency doubling, 4 frequency doubling and high frequency doubling, and are recorded as: [ f ] of0.5X,f1x,f1.5X,f2X,f2.5X,f3X,f3.5X,f4X,fnX]The frequency multiplication is the multiple of the fundamental frequency, and the fundamental frequency is equal to 1 frequency multiplication;
wherein f isnXFrequencies that are high multiples, i.e. frequencies greater than 4 multiples,
Figure FDA0003553428130000013
wherein f is1XThe frequency is 1 multiplied frequency, Vr is the rotating speed of the equipment, and the obtained characteristic frequency of the equipment is as follows:
Figure FDA0003553428130000021
where n is the high frequency multiplication, meaning all frequencies greater than 4 frequency multiplication;
s45: calculating the spectrum ratio of the characteristic frequency:
fi is the ith frequency value if
Figure FDA0003553428130000022
Then, selecting all the amplitudes corresponding to the fi to be used as the amplitude of the 0.5 frequency multiplication comprehensively, and obtaining the amplitude ratio of the 0.5 frequency multiplication as follows:
Figure FDA0003553428130000023
② in the same way, if
Figure FDA0003553428130000024
fj is the jth frequency value, and the amplitude ratio of the obtained 1 multiplied frequency is:
Figure FDA0003553428130000025
③ in the same way, if
Figure FDA0003553428130000026
fa is the a-th frequency value, and the amplitude ratio of the obtained 1.5 frequency multiplication is:
Figure FDA0003553428130000027
fourthly, if
Figure FDA0003553428130000028
fb is the b-th frequency value, and the obtained 2-frequency multiplication amplitude ratio is:
Figure FDA0003553428130000029
the same thing applies if
Figure FDA00035534281300000210
fc is the c frequency value, and the obtained 2.5 frequency multiplication amplitude ratio is:
Figure FDA00035534281300000211
sixthly, in the same way, if
Figure FDA00035534281300000212
fd is the d-th frequency value, and the obtained 3-frequency multiplication amplitude ratio is:
Figure FDA00035534281300000213
seventhly, in the same way, if
Figure FDA00035534281300000214
fg is the g-th frequency value, and the obtained 3.5 frequency multiplication amplitude ratio is:
Figure FDA0003553428130000031
the same thing is said, if
Figure FDA0003553428130000032
fe is the e-th frequency value, and the obtained 4-frequency multiplication amplitude ratio is:
Figure FDA0003553428130000033
ninthly, if
Figure FDA0003553428130000034
fn is the nth frequency value, and the amplitude ratio of the obtained high frequency multiplication is as follows:
Figure FDA0003553428130000035
wherein sum (| X (n) |) is
Figure FDA0003553428130000036
And the sum of the amplitudes of the first t maximum values, wherein t is e/8.
2. The method as claimed in claim 1, wherein in step S1, a Unity 3D-based three-dimensional modeling is performed, including a three-dimensional modeling of the whole production line of the blast furnace blower and the TRT unit, and a real scene of equipment on site is reproduced, which is consistent with a process flow on site.
3. The method for diagnosing the faults of the blast furnace blower and the TRT unit based on the digital twin system as claimed in claim 1, wherein the digital twin system of the blast furnace blower and the TRT unit in step S2 is used for acquiring real-time data on site and predicting virtual data of data modeling, two scenes of a real-time system which operates synchronously with the site and a virtual twin system which continues to operate when the site is shut down are realized, and the fault data of each scene is marked.
4. The method for diagnosing the faults of the blast furnace blower and the TRT unit based on the digital twin system as claimed in claim 1, wherein in the step S3, abnormal data are removed through a moving average algorithm, and the specific method is as follows:
and recording an operation parameter variable X at the t moment after abnormal data are removed as Xt，θtTo removeBefore abnormal data, the value of a running parameter variable X at the time t, beta is a moving average coefficient, beta belongs to [0,1 ], when beta is 0, the moving average is not used, and Xt＝θt(ii) a After using the running average:
Xt＝β*Xt-1+(1-β)*θt。
5. the method for diagnosing the faults of the blast furnace blower and the TRT unit based on the digital twin system as claimed in claim 1, wherein the step S5 of constructing the neural network fault diagnosis algorithm based on the Adam algorithm comprises the following steps:
s51: determining input and output of a neural network, wherein the input is operation parameter variables and vibration time-frequency domain data of the equipment, and the input comprises current, voltage, power, temperature, flow, triaxial vibration effective value, triaxial vibration time-domain index and triaxial frequency-domain index; the output is the fault type of the equipment, including the abrasion, unbalance, misalignment, looseness of a base, faults of an inner ring of a bearing, faults of an outer ring of the bearing and friction faults of a movable part and a static part of the equipment;
s52: determining parameters of the neural network: the number of nodes of an input layer is 8, the number of nodes of an output layer is 1, the number of hidden layers is 1, the number of nodes of each hidden layer is 26, the network learning rate is 0.21, and the momentum coefficient is 0.01;
s53: the method for optimizing the weight of the neural network by using the Adam algorithm comprises the following specific steps: introducing a square gradient on the basis of a gradient descent method of the amount of fluctuation, and correcting the deviation of the speed;
s54: constructing complete forward and reverse calculation neural network models, reading data of a fault mark in the digital twin system, and training and testing;
s55: and reading real-time operation data to perform fault diagnosis and output.
CN202011583058.5A 2020-12-28 2020-12-28 Blast furnace blower and TRT set fault diagnosis method based on digital twin system Active CN112781903B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202011583058.5A CN112781903B (en) 2020-12-28 2020-12-28 Blast furnace blower and TRT set fault diagnosis method based on digital twin system

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202011583058.5A CN112781903B (en) 2020-12-28 2020-12-28 Blast furnace blower and TRT set fault diagnosis method based on digital twin system

Publications (2)

Publication Number Publication Date
CN112781903A CN112781903A (en) 2021-05-11
CN112781903B true CN112781903B (en) 2022-07-08

Family

ID=75752999

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202011583058.5A Active CN112781903B (en) 2020-12-28 2020-12-28 Blast furnace blower and TRT set fault diagnosis method based on digital twin system

Country Status (1)

Country Link
CN (1) CN112781903B (en)

Families Citing this family (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN114757048B (en) * 2022-04-28 2023-03-24 北京千尧新能源科技开发有限公司 Health state assessment method, device, equipment and medium for fan foundation
CN115310490B (en) * 2022-08-17 2024-03-29 中国核动力研究设计院 Rotary equipment fault analysis method based on multi-domain feature and sensitive feature selection
CN116304695B (en) * 2023-03-07 2024-04-09 珠海紫燕无人飞行器有限公司 Method and system for predicting vibration faults of unmanned aerial vehicle based on convolutional neural network
CN117591831B (en) * 2024-01-16 2024-03-22 南京凯奥思数据技术有限公司 Fault diagnosis method and system based on fault frequency multiplication characteristics and neural network

Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN110530638A (en) * 2019-07-31 2019-12-03 西安交通大学 Based on number twin aeroplane engine main bearing damage check and diagnostic method
CN110851966A (en) * 2019-10-30 2020-02-28 同济大学 Digital twin model correction method based on deep neural network
CN111523260A (en) * 2020-03-18 2020-08-11 上海发电设备成套设计研究院有限责任公司 Steam turbine rotor digital twin body construction method and monitoring system

Patent Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN110530638A (en) * 2019-07-31 2019-12-03 西安交通大学 Based on number twin aeroplane engine main bearing damage check and diagnostic method
CN110851966A (en) * 2019-10-30 2020-02-28 同济大学 Digital twin model correction method based on deep neural network
CN111523260A (en) * 2020-03-18 2020-08-11 上海发电设备成套设计研究院有限责任公司 Steam turbine rotor digital twin body construction method and monitoring system

Non-Patent Citations (1)

* Cited by examiner, † Cited by third party
Title
数字孪生与深度学习融合驱动的采煤机健康状态预测;丁华等;《中国机械工程》;20200430;第31卷(第7期);815-821 *

Also Published As

Publication number Publication date
CN112781903A (en) 2021-05-11

Similar Documents

Publication Publication Date Title
CN112781903B (en) Blast furnace blower and TRT set fault diagnosis method based on digital twin system
Soualhi et al. Prognosis of bearing failures using hidden Markov models and the adaptive neuro-fuzzy inference system
CN106933778A (en) A kind of wind power combination forecasting method based on climbing affair character identification
CN103323772A (en) Wind driven generator operation state analyzing method based on neural network model
CN111898644B (en) Intelligent identification method for health state of aerospace liquid engine under fault-free sample
CN114841580A (en) Generator fault detection method based on hybrid attention mechanism
CN114741922B (en) Turbine blade creep-fatigue life prediction method based on Attention mechanism
CN113837432A (en) Power system frequency prediction method driven by physics-data combination
CN113255458A (en) Bearing fault diagnosis method based on multi-view associated feature learning
CN117371207A (en) Extra-high voltage converter valve state evaluation method, medium and system
Li et al. Aero-engine exhaust gas temperature prediction based on LightGBM optimized by improved bat algorithm
Qi et al. Feature classification method of frequency cepstrum coefficient based on weighted extreme gradient boosting
Liu et al. Application of the state deterioration evolution based on bi-spectrum entropy and HMM in wind turbine
CN113820132B (en) Fault alarm method based on multiscale scatter entropy construction threshold
Zhu et al. Prediction of air quality index based on wavelet transform combination model
CN115114964A (en) Sensor intermittent fault diagnosis method based on data driving
Du et al. Anomaly Root Cause Analysis for Wind Turbines Based on Denoising Autoencoder and Sparse Estimation
CN111814325A (en) Single variable alarm threshold value optimization method for non-steady operation of coal-fired power generation equipment
CN112987133B (en) Ground wind vector nowcasting method based on SVR and VAR
Zhang et al. Life Prediction of Rolling Bearing using Temporal Convolution Network and Attention Mechanism
CN114970044B (en) Rolling bearing fault diagnosis method and system based on threshold convolutional neural network
Wang et al. Decomposition Model of Wind Power Gear Fault Signal Based on IGWO-VMD Optimization
Gao et al. A fault diagnosis method of rolling bearing of CNC machine tool based on improved convolutional neural network
CN112653158B (en) Frequency response characteristic identification method for doubly-fed wind power plant with additional frequency control
CN118626948B (en) Wind turbine generator system fault diagnosis method based on data analysis

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