CN109779894A - A kind of reciprocating compressor fault diagnosis system and method based on neural network algorithm - Google Patents
A kind of reciprocating compressor fault diagnosis system and method based on neural network algorithm Download PDFInfo
- Publication number
- CN109779894A CN109779894A CN201811643547.8A CN201811643547A CN109779894A CN 109779894 A CN109779894 A CN 109779894A CN 201811643547 A CN201811643547 A CN 201811643547A CN 109779894 A CN109779894 A CN 109779894A
- Authority
- CN
- China
- Prior art keywords
- compressor
- neural network
- module
- emulation
- parameter
- 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
Abstract
The invention discloses a kind of reciprocating compressor fault diagnosis system and method based on neural network algorithm, including local communication module, data acquisition module, emulation module neural network based, the expert system module based on standard failure feature, fault diagnosis host and host computer.The present invention obtains the simulation model under each operating condition of compressor in conjunction with the compressor assembly state parameter of actual measurement using neural network algorithm；Obtain standard failure feature by changing system parameter analog compression machine calculation of fault, after there is operating exception in compressor, by with standard failure Characteristic Contrast, trouble location can be accurately positioned and provide method for removing；It is higher that the present invention diagnoses success rate, and is not required to high precision apparatus Real Time Monitoring, reduces diagnosis cost.
Description
Technical field
The present invention relates to the fault diagnosis technology fields of equipment, particularly relate to a kind of based on neural network algorithm
Reciprocating compressor fault diagnosis system and method.
Background technique
The core equipment that reciprocating compressor is transported as oil gas field, stability, reliability and safety have very
Important meaning.Because compressor arrangement is more complicated, components are more, and assembly is complicated, sometimes will appear using same set of
The reciprocating compressor of drawing production is run at the scene there is the case where different operation characteristics, therefore is difficult to construct compressor quasi-
True whole mathematic simulated mode, so how to be reduced by fault diagnosis because compressor fault is made when failure symptom occurs
At direct or indirect loss, be always an important topic of compressor industry.
The detection method of mainstream has vibration detection, thermal parameter detection, ess-strain detection etc. at present, however examines above
It surveys and carries out detection judgement just for the specific failure of compressor, and compressor can not be compared whole, system state
And judgement.There are also dynamics and vibration that the fault diagnosis system of engine apparatus uses on-line real time monitoring engine at present
Dynamic performance parameters carry out fault diagnosis, but due to not having to carry out the amendment of thermodynamic model, and believe using in time-frequency domain
The method that fault signature spectrum is monitored in number, is easier to the case where reporting by mistake, ultimately causing technical staff mostly can only be by warp
It tests and handles the reason of causing failure, take time and effort and be also easier to malfunction.
Summary of the invention
It is a kind of based on mind technical problem to be solved by the present invention lies in view of the deficiency of the prior art, providing
Reciprocating compressor fault diagnosis system and method through network algorithm, can be accurately positioned trouble location and provide method for removing.
The present invention is to be achieved through the following technical solutions:
A kind of reciprocating compressor fault diagnosis system based on neural network algorithm, including local communication module, data
Acquisition module, emulation module neural network based, the expert system module based on standard failure feature, fault diagnosis host and
Host computer；
The data acquisition module is used to obtain the characterisitic parameter of each components of compressor and is sent to based on nerve net
The emulation module of network；
The local communication module is used to obtain the real time execution parameter of compressor from PLC monitoring system and be sent to and be based on
The emulation module of neural network；
Data during the emulation module neural network based is worked normally according to compressor design data and compressor
The real time execution parameter that the characterisitic parameter and local communication module of each components of acquisition module acquisition obtain, establishes compressor
Mathematic simulated mode and save to expert system module；By changing related trouble location parameter, mould in mathematic simulated mode
Intend the various failures of compressor different parts to obtain fault data feature and save to expert system module；
The expert system module includes integrated database and expert knowledge library, and integrated database includes the mathematics of compressor
Simulation model and fault data feature；Expert knowledge library includes the judgment criterion of failure and the processing method for various failures；
The fault diagnosis host is for user's read failure diagnosis process, result and eliminating measure；
The host computer reads for developer, adds and modify emulation module neural network based and expert system mould
Block.
Preferably, the data acquisition module includes temperature sensor, pressure sensor, strain-ga(u)ge transducer, vibration biography
Sensor and current vortex sensor.
Preferably, the expert system module further includes inference machine, and inference machine is used to find with control strategy applicable
The judgment criterion of failure.
Preferably, integrated database and expert knowledge library can be read, add and modify.
Preferably, field communications module supports Modbus agreement, ICP/IP protocol and CAN bus communication protocol.
A kind of reciprocating compressor method for diagnosing faults based on neural network algorithm, includes the following steps:
S1, the characterisitic parameter that each components of compressor are obtained by data acquisition module, and be sent to based on nerve net
The emulation module of network；
S2, the real time execution parameter that compressor PLC monitoring system is obtained by local communication module, and be sent to based on mind
Emulation module through network；
S3, emulation module neural network based carry out analog simulation: compressor assembly and running-in period, neural network
Characterisitic parameter and compression of the emulation module using each components of compressor of compressor design data, data collecting module collected
Machine real time execution parameter, after filtering removes noise, establish the ideal thermodynamic equation of compressor, ideal kinetics equation and
Ideal vibration mechanics equation, obtains the ideal mathematics model of compressor；The compressor loads operation phase, into ideal mathematics model
The real time execution parameter for inputting the acquisition of local communication module establishes pressure in conjunction with BP neural network method and structure Dynamics Modification
Emulation thermodynamical model, emulation kinetic model and the emulation vibration mechanical model of contracting machine, obtain the emulation mathematical modulo of compressor
Type, and the simulation mathematical model of foundation is saved to expert system module；Change emulation mathematics by neural network algorithm iteration
Different related trouble location parameter in model, the various failures of analog compression machine different parts obtain fault data feature and protect
It deposits to expert system module；
S4, fault diagnosis: when failure sign, the real time execution parameter for the compressor that local communication module obtains is made
For fault data feature to be matched, the fault data feature that is saved in fault data feature to be matched and expert system module
It carries out judging whether to match according to the breakdown judge criterion in expert knowledge library；If successful match, fault diagnosis is realized, and give
Expert opinion is corresponded to out；If matching is unsuccessful, pass through having for neural network algorithm iterative modifications compressor simulation mathematical model
Trouble location parameter is closed, until the fault data feature and fault data characteristic matching success to be matched simulated, then is recalled
The method for removing of this failure in expert knowledge library completes fault diagnosis, and is by the fault data feature simulated deposit expert
System module.
Preferably, step S1 specifically: unit carry out process and assemble when, by data acquisition module to cylinder, middle body,
Fuselage, cylinder head support, the support of middle body, surge tank and washing tank are tested, and intrinsic frequency, Mode Shape and damping ratio ginseng are read
Number, unit read compressor acceleration, displacement, power and torque in running-in period, and unit after installation and debugging success, leads at the scene
It crosses the operating that data acquisition module carries out under unloaded, load, various inlet pressure and under different displacements compressor to test, and remembers
Record the operating parameter of each component.
Preferably, in step S3, the method for building up of simulation mathematical model specifically:
Establish ideal thermodynamic equation are as follows: P=f (n, q, T, T1, k, B, r1, r2, p1, p2, z, Q, t), after the completion of foundation
Thermodynamical equilibrium equation simultaneous composition equation groups at different levels are calculated, after the completion of calculating output dynamic pressure value P for dynamics calculation and
Vibration mechanics, which calculates, to be used, and enabling f (P) is unit ideal thermodynamic mathematical model, then exports f (P) and use for neural network；
Establish ideal kinetics equation are as follows: W=f (n, m, t, l, p1, p2, r2, c, P, z1), output torque after the completion of foundation
Value W is calculated for vibration mechanics and is used, and enabling f (W) is ideal kinetics mathematical model, then exports f (W) and use for neural network；
F (P) and f (W) combines BP neural network method, the real time execution parameter that input local communication module obtains, iteration
Optimal Fitting parameter z and z1 establish compressor emulation thermodynamical model and emulation kinetic model；
Ideal vibration mechanics equation: V=f (M, C, K, P, W, t) is established, enables f (V) for ideal vibration mechanics mathematical model,
Structural dynamic optimum is carried out in conjunction with the real time execution parameter that finite element analysis and local communication module obtain, enables f (V) ' it is SDM
Modified emulation vibration mechanical model then exports f (V) ' it is used for neural network and expert system；
Wherein, n is compressor rotary speed, and Q is capacity, and q is clearance volume, and t is the time, and T is intake air temperature, and T1 is exhaust
Temperature, k are gas adiabatic exponent, and B is cylinder unit time and external heat exchange value, and r1 is certain grade of cylinder bore value, and r2 is song
The axis radius of gyration, p1 are admission pressure, and p2 is pressure at expulsion, and Q is gas displacement, and z is thermodynamics fitting parameter, and P is in cylinder
Transient pressure, m is toward complex inertia mass power, and l is compressor columns, and c is that sliding friction damps, and z1 is that dynamics fitting is joined
Number, W are that motor exports transient torque value, and M is the mass matrix of each components of compressor, and C is each components of compressor
Damping matrix, K are the stiffness matrix of each components of compressor.
Further, first discrete to compressor model progress to resolve into each sub- knot when establishing emulation vibration mechanical model
Then structure show that the master mode and Constrained mode of each minor structure obtain Modal Space multiplied by transformation matrix by finite element simulation
Polycondensation stiffness matrix and polycondensation mass matrix, by with actual measurement each minor structure of compressor modal data compare, be based on
The structural dynamic optimum of finite element analysis and experimental modal, then combination obtain complete machine Modal Space mass matrix, just
Spend matrix and damping matrix, introduce generalized force and carry out unit coupling analysis, extract the acceleration of compressor corresponding site, displacement,
Power and torque.
Preferably, in step S3, filtering uses Kalman filtering.
Compared with prior art, the invention has the following beneficial technical effects:
1, measured data is to the hot, dynamic of compressor model when the present invention uses neural network algorithm combination site installation test
Mechanic-mathematical model is fitted optimization, increases substantially mathematical model to the simulation precision of compressor.In failure diagnostic process
In, reference standard is provided for accurate judgement failure cause and position.
2, the present invention improves pressure by the way of the Dynamics Modification (SDM) based on finite element structure model and measured data
The accuracy of contracting machine vibration simulation can be simulated directly in conjunction with accurate heat, kinetic model emulation module neural network based
The vibration characteristics of compressor when all kinds of failures occur, comparison need from unit vibration characteristics when, extract in frequency domain it is corresponding therefore
The mode that the modes such as barrier characteristic spectrum carry out location determination failure substantially increases success rate and accuracy.
3, the present invention can be tracked during the production of every suit unit, assembly and debugging (or normal operation)
Detection extracts its actual measurement parameter and improves mathematical model (f (P) ', f (W) ', f (V) '), therefore the mathematical model is only in detection
The unit crossed, the compressor mathematical model after improving also improve while no longer there is universality its simulation accuracy (even
The compressor mathematical model of same model and batch also has fine distinction, so as to the operating of better analog compression unit).
4, the present invention can carry out fault diagnosis by way of real time on-line monitoring, can also only hand down to posterity in advance on unit
Sensor mounting hole (is generally found) to pacify again at this time by the PLC monitoring system or manual inspection of compressor when failure symptom occurs
It fills corresponding detection device and carries out field failure diagnosis.Come relative to other fault diagnosis systems for having to real time on-line monitoring
It says, greatly reduces the cost of fault diagnosis.
Detailed description of the invention
Fig. 1 is Troubleshooting Flowchart of the invention.
Fig. 2 thermodynamical model and kinetic model fit procedure schematic diagram.
Fig. 3 parameter fitting status diagram neural network based.
Specific embodiment
Below with reference to specific embodiment, the present invention is described in further detail, it is described be explanation of the invention and
It is not to limit.
The reciprocating compressor fault diagnosis system based on neural network algorithm that the present invention provides a kind of, including scene are logical
Interrogate module, data acquisition module, emulation module neural network based, the expert system module based on standard failure feature, event
Barrier diagnosis host computer and host；
Local communication module is used to obtain the real time execution parameter of compressor from PLC monitoring system；Field communications module branch
Hold Modbus agreement, ICP/IP protocol, CAN bus various communications protocols.
The data acquisition module is used to obtain the characterisitic parameter of each components of compressor；The data acquisition module packet
Including temperature, pressure, stress, vibration and current vortex sensor, (each sensor is mounted on the reserved sensing of compressor all parts
In device mounting hole) and data collecting card and data acquisition software, for obtaining thermodynamics, the dynamics, vibration force of compressor
Learn each operating parameter and corresponding component required for equation property parameters (such as n, q, T, T1, k, B, r1, r2, p1, p2, m,
v,t,l,p2,r2,c).When unit carries out process and assemble, by data acquisition module to cylinder, middle body, fuselage, cylinder head branch
Support, the support of middle body, surge tank, washing tank these components are tested, and intrinsic frequency, Mode Shape and damping ratio parameter are read,
Unit reads compressor acceleration, displacement, power and torque in running-in period, and unit after installation and debugging success, passes through number at the scene
It tests, and records each according to the operating that acquisition module carries out under unloaded, load, various inlet pressure and under different displacements compressor
The operating parameter of component.
The data that the communication module and data acquisition module obtain are sent to emulation module neural network based.
Emulation module neural network based is that data are adopted during being worked normally using compressor design data and compressor
The real time execution parameter that the characterisitic parameter and local communication module for collecting each components of module acquisition obtain, after filtering,
Emulation thermodynamics, dynamics, vibration mechanics equation are established to it.Filtering mode can use Kalman filtering.
Thermodynamical equilibrium equation are as follows: P=f (n, q, T, T1, k, B, r1, r2, p1, p2, z, Q, t)
Wherein, n is compressor rotary speed, and Q is capacity, and q is clearance volume, and t is the time, and T is intake air temperature, and T1 is exhaust
Temperature, k are gas adiabatic exponent, and B is cylinder unit time and external heat exchange value, and r1 is cylinder radius, and r2 is crank up
Radius, p1 are admission pressure, and p2 is pressure at expulsion, and Q is gas displacement, and P is the transient pressure in cylinder, and z is thermodynamics fitting
Parameter (according to different operating conditions, chooses the cylinder that different z values collects equation result P close to site pressure sensor
Interior transient pressure value P), thermodynamical equilibrium equation simultaneous composition equation group at different levels is calculated after the completion of foundation, is exported after the completion of calculating
Gas-dynamic pressure P excitation parameters are calculated for vibration mechanics and are used, and output ideal thermodynamic model f (P) is used for neural network.
Thermodynamics fit procedure refers to that input is surveyed based on temperature, pressure, stress, vibration and the current vortex sensor installed in compressor set
When reality duty parameter and design parameter, fit compressor set difference by adjusting weight and the implicit number of plies of neural network
Row's temperature, discharge capacity and row pressure under operating condition.
Kinetics equation are as follows: W=f (n, m, t, l, p1, p2, r2, c, z1, P)
Wherein, W is the transient force of moving component, and n is compressor rotary speed, and m is toward complex inertia mass power, and t is the time, and l is
Compressor columns, c are sliding friction damping, z1 is that dynamics fitting parameter (according to different operating conditions, chooses different z1 values
The W for collecting equation result close to in-situ stresses sensor), output W is calculated for vibration mechanics after the completion of calculating uses, defeated
Ideal kinetics model f (W) is used for neural network out.Dynamics fit procedure, which refers to, inputs tested power, revolving speed, defeated
Enter torque parameter and design parameter, is fitted under compressor set difference operating condition by adjusting weight and the implicit number of plies of neural network
The variation of shafting stress.
To ideal thermodynamic model f (P) and ideal kinetics mathematical model f (W), in conjunction with BP neural network method, input
Local communication module obtain real time execution parameter, iteration optimization fitting parameter z, z1 enable f (P) ' be f (P) amendment form, f
(W) ' it is the amendment form of f (W), establishes energy accurate response compressor heat, kinetic characteristics and meet engineering precision demand
Simulation mathematical model (f (P) ', f (W) ').
Vibration mechanics equation: V=f (M, C, K, P, W, t)
Wherein, V is unit vibration value, and M is mass matrix, and C is damping matrix, and K is stiffness matrix.
Vibration mechanics calculate during, first to compressor model carry out it is discrete resolve into each minor structure, as cylinder, in
Then the minor structures such as body, body obtain the master mode and Constrained mode of each minor structure by finite element simulation, multiplied by transformation square
Battle array, obtains polycondensation stiffness matrix, the polycondensation mass matrix of Modal Space, passes through the modal data with actual measurement each minor structure of compressor
Comparison, carries out the structural dynamic optimum based on finite element analysis and experimental modal, and then combination obtains the Modal Space of complete machine
Quality, rigidity and damping matrix, introduce generalized force and carry out unit coupling analysis, i.e., the vibration of extractable compressor corresponding site
Acceleration, displacement, power, torque.
Specifically:
The mass matrix M under the modal coordinate of each minor structure is obtained by finite element simulationss, stiffness matrix Kmm, neutron
The freedom degree positioning that structure docks boundary point is main freedom degree xm, the freedom degree of non-boundary point is defined as from freedom degree xs, X is son
The Degree of Structure Freedom, then the movement of minor structure can be used following equation to indicate:
Wherein Mms, MsmAnd Kms, KsmThe respectively coupling matrix of principal and subordinate freedom degree, fmFor to relay, f0mFor to relay width
Value, enables D=K- ω2M is dynamic matrix, fm=f0meiωtWhen, there is response displacement x=Xeiωt, then solving the first formula in above formula hasThen minor structure freedom degree is signable isIt enablesThen
TdFor transformation matrix, then the polycondensation mass matrix under i-th of minor structure principal coordinate is obtainedWith polycondensation stiffness matrixFor
Then it is obtained by the modal response of each minor structure of test intrinsic under each minor structure principal coordinate
Frequency, principal mode, damping ratio parameter, comparing calculation obtain the polycondensation mass matrix based on Modal TestPolycondensation rigidity square
Battle arrayAnd damping ratio, i.e., structural dynamic optimum (SDM) is carried out to minor structure.
It combines to obtain the quality, rigidity and damping matrix of the Modal Space of entire compressor by substructure mode, introduce
Generalized force carries out unit coupling analysis, i.e., the acceleration of extractable compressor corresponding site, displacement, power, torque, output emulation vibration
Kinetic model f (V) ' is used for neural network and expert system.See Fig. 3, XX minor structure is represented as the support of middle body, intake and exhaust are slow
Rush the sub-structure model of the other components such as tank, pedestal.
Expert system module, including integrated database, expert knowledge library and inference machine three parts.Wherein, integrated database
Comprising emulation module fitting reciprocating compressor normal work during emulation thermodynamical model, emulation kinetic model,
Emulate vibration mechanical model and above-mentioned fault signature data.Expert knowledge library contains the judgment criterion of failure and for each
The processing method of kind failure.Inference machine is then responsible for finding the judgment criterion of applicable failure with control strategy.This expert system
System module can be read, adds and modify.Integrated database and expert knowledge library can be read, and added and modified.
Fitting parameter z, the z1 for adjusting two mathematical models are established accurate in conjunction with the method (see Fig. 2) of BP neural network
Compressor simulation mathematical model saves the model to expert system, compares for fault diagnosis.Then, change different emulation
Mathematical model parameter, the corresponding fault data feature of the various failures of analog compression machine different parts is (as reduced level-one exhaust pressure
Power p1 reduces level-one cylinder volumetric efficiency, simulates level-one exhaust valve principal fault；When piston motion is to top dead centre, reciprocal
Increase transient state impact force simulation piston knock failure on inertia force m；The K value simulation coupling bolt for reducing shaft coupling loosens event
Barrier；Increase C value simulation crankshaft secondary series main bearing lubrication failure etc. at body secondary series), and be saved to and be based on expert
In the failure library module of system, this failure library module can be read, adds and modify.
Fault diagnosis host is that failure diagnostic process, result can be read and shown by this interface for user oriented
And eliminating measure etc.；The host computer be for towards developer, by this interface can be read, add and modify emulation module and
Expert system module.
As shown in Figure 1, the present invention also provides a kind of reciprocating compressor fault diagnosis side based on neural network algorithm
Method includes the following steps:
S1, the characterisitic parameter that each components of compressor are obtained by data acquisition module, and be sent to based on nerve net
The emulation module of network；
Specifically: when unit carries out process and assemble, by data acquisition module to cylinder, middle body, fuselage, cylinder head branch
The components such as support, the support of middle body, surge tank, washing tank are tested, and the immutable property parameters such as intrinsic frequency, Mode Shape are read
(for calculating the parameters such as M, C, K and z2 together in conjunction with such as length, width and height, density, elasticity modulus design parameter)；Unit is in test run
Stage reads compressor acceleration, displacement, power and torque；Unit after installation and debugging success, carries out compressor empty at the scene
It carries, load, the operating test under various inlet pressure and under different displacements, passing through PLC controller and data acquisition module detects
The parameters such as compressor heat, mechanics and vibration characteristics (such as n, v (t), q, T, k, B, r1, r2, p1, p2, m, v, t, l, p2, r2, c)；
All parameters are sent to emulation module neural network based.
S2, the real time execution parameter that compressor PLC monitoring system is obtained by the local communication module, and it is sent to base
In the emulation module of neural network；
S3, analog simulation: emulation module neural network based carries out analog simulation: compressor assembly and running-in period,
The emulation module of neural network uses the characteristic of each components of compressor of compressor design data, data collecting module collected
Parameter and compressor real time execution parameter, after filtering removes noise, ideal thermodynamic equation, the ideal for establishing compressor are dynamic
Mechanical equation and ideal vibration mechanics equation, obtain the ideal mathematics model of compressor；The compressor loads operation phase, to ideal
The real time execution parameter that the acquisition of local communication module is inputted in mathematical model, is repaired in conjunction with BP neural network method and structure dynamics
Change, establishes the emulation thermodynamical model, emulation kinetic model and emulation vibration mechanical model of compressor, obtain the imitative of compressor
True mathematical model, and the simulation mathematical model of foundation is saved to expert system module；Changed by neural network algorithm iteration
Different related trouble location parameter in simulation mathematical model, the various failures of analog compression machine different parts obtain fault data
Feature is simultaneously saved to expert system module
Specifically: combine neural network to calculate when emulation module neural network based calculates compressor heat, mechanics
Method and field test data repair existing thermodynamics and kinetics Empirical Calculating Method f (P), f (W) and parameter z, z1
Just, the mathematical model (f (P) ', f (W) ') of energy accurate response compressor heat, kinetic characteristics is established；To vibration of compressor characteristic
When being calculated, compressor is first split into several minor structures, compressor minor structure finite element model is established, is obtained in conjunction with actual measurement
It is dynamic to carry out the structure based on experiment test to sub-structure model for the parameters such as intrinsic frequency, Mode Shape and the damping ratio of minor structure
Mechanics corrects (SDM), combines to obtain the quality of the Modal Space of entire compressor, rigidity and damping square by substructure mode
Battle array introduces generalized force and carries out unit coupling, establishes the mathematical model f (V) ' of energy accurate response vibration of compressor characteristic.
S4, failure diagnostic process, when failure sign, the real time execution ginseng for the compressor that local communication module obtains
Number is used as fault data feature to be matched, the fault data saved in fault data feature to be matched and expert system module
Feature carries out judging whether to match according to the breakdown judge criterion in expert knowledge library；If successful match, fault diagnosis is realized,
And provide corresponding expert opinion；If matching is unsuccessful, pass through neural network algorithm iterative modifications compressor simulation mathematical model
The related trouble location parameter of (f (P) ', f (W) ', f (V) '), until the fault data feature and failure to be matched simulated
Data characteristics successful match, then recall the method for removing of this failure in expert knowledge library, complete fault diagnosis, and will simulate
Fault data feature is stored in expert system module, in case occurring next time to select when similar failure.
Specifically: when compressor is when a certain operating condition is run, failure symptom has occurred suddenly, such as compressor PLC control cabinet
Certain cylinder delivery temperature of upper display increases extremely, and the power of the assembling unit reduces, and the vibration of on-site test cylinder head increases.But cause such abnormal event
A possibility that barrier, has very much, is difficult directly to judge to cause failure cause and trouble location.It at this time can be by the sensor of this system
Mounted in the position that compressor is reserved, the cylinder pressure of system detection compressor, temperature, pulsation, the horizontal bounce of piston rod, crankshaft
Twisting vibration, body, cylinder, support the data such as vibration；In conjunction with the normal mathematical model (f of compressor under this operating condition
(P) ', f (W) ', f (V) '), comparison can determine whether trouble location or failure cause (breaking down than example exhaust valve position), then adjust
Expert system compares lookup failure cause out, completes fault diagnosis.If (in such as fault database not without this fault signature
Deposit valve fault) it can then pass through the related trouble location of neural network algorithm iterative modifications healthy compressor simulation mathematical model
Parameter (such as modifies clearance volume, the modification exhaust valve in vibration mechanical model (f (V) ') in thermodynamical model (f (P) ')
Rigidity and mass parameter (K, M)), until accurate simulation goes out the fault data feature that live compressor occurs；Expert is recalled again to know
The method for removing (such as replacement exhaust valve spring and valve block) for knowing this failure in library, completes fault diagnosis, and its fault data is special
Sign deposit failure expert knowledge library, when in case similar failure occurring for next time, system can quickly judge, complete fault database accumulation.
Claims (10)
1. a kind of reciprocating compressor fault diagnosis system based on neural network algorithm, it is characterised in that: including local communication
Module, data acquisition module, emulation module neural network based, the expert system module based on standard failure feature, failure
Diagnose host and host computer；
The data acquisition module is used to obtain the characterisitic parameter of each components of compressor and is sent to neural network based
Emulation module；
The local communication module is used to obtain the real time execution parameter of compressor from PLC monitoring system and be sent to based on nerve
The emulation module of network；
Data acquire during the emulation module neural network based is worked normally according to compressor design data and compressor
The real time execution parameter that the characterisitic parameter and local communication module of each components of module acquisition obtain, establishes the number of compressor
It learns simulation model and saves to expert system module；By changing related trouble location parameter in mathematic simulated mode, simulation pressure
The various failures of contracting machine different parts obtain fault data feature and save to expert system module；
The expert system module includes integrated database and expert knowledge library, and integrated database includes the mathematical simulation of compressor
Model and fault data feature；Expert knowledge library includes the judgment criterion of failure and the processing method for various failures；
The fault diagnosis host is for user's read failure diagnosis process, result and eliminating measure；
The host computer reads for developer, adds and modify emulation module neural network based and expert system module.
2. the reciprocating compressor fault diagnosis system according to claim 1 based on neural network algorithm, feature exist
In: the data acquisition module includes temperature sensor, pressure sensor, strain-ga(u)ge transducer, vibrating sensor and current vortex
Sensor.
3. the reciprocating compressor fault diagnosis system according to claim 1 or 2 based on neural network algorithm, feature
Be: the expert system module further includes inference machine, and inference machine is used to find the judgement of applicable failure with control strategy
Criterion.
4. the reciprocating compressor fault diagnosis system according to claim 1 or 2 based on neural network algorithm, feature
Be: integrated database and expert knowledge library can be read, add and modify.
5. the reciprocating compressor fault diagnosis system according to claim 1 or 2 based on neural network algorithm, feature
Be: field communications module supports Modbus agreement, ICP/IP protocol and CAN bus communication protocol.
6. a kind of reciprocating compressor method for diagnosing faults based on neural network algorithm, which comprises the steps of:
S1, the characterisitic parameter that each components of compressor are obtained by data acquisition module, and be sent to neural network based
Emulation module；
S2, the real time execution parameter that compressor PLC monitoring system is obtained by local communication module, and be sent to based on nerve net
The emulation module of network；
S3, emulation module neural network based carry out analog simulation: compressor assembly and running-in period, the emulation of neural network
Module is real using the characterisitic parameter and compressor of each components of compressor of compressor design data, data collecting module collected
When operating parameter, by filtering remove noise after, establish the ideal thermodynamic equation, ideal kinetics equation and ideal of compressor
Vibration mechanics equation obtains the ideal mathematics model of compressor；It the compressor loads operation phase, is inputted into ideal mathematics model
The real time execution parameter of local communication module acquisition establishes compressor in conjunction with BP neural network method and structure Dynamics Modification
Emulation thermodynamical model, emulation kinetic model and emulation vibration mechanical model, obtain the simulation mathematical model of compressor, and
The simulation mathematical model of foundation is saved to expert system module；Changed in simulation mathematical model by neural network algorithm iteration
Different related trouble location parameters, the various failures of analog compression machine different parts obtain fault data feature and save to special
Family's system module；
S4, fault diagnosis: when failure sign, local communication module obtain compressor real time execution parameter be used as to
Matched fault data feature, the fault data feature saved in fault data feature to be matched and expert system module according to
Breakdown judge criterion in expert knowledge library carries out judging whether to match；If successful match, fault diagnosis is realized, and provide pair
Answer expert opinion；If matching is unsuccessful, pass through the related event of neural network algorithm iterative modifications compressor simulation mathematical model
Hinder positional parameter, until the fault data feature and fault data characteristic matching success to be matched simulated, then recalls expert
The method for removing of this failure in knowledge base completes fault diagnosis, and the fault data feature simulated is stored in expert system mould
Block.
7. the reciprocating compressor method for diagnosing faults according to claim 6 based on neural network algorithm, feature exist
In: step S1 specifically: when unit carries out process and assemble, by data acquisition module to cylinder, middle body, fuselage, cylinder head branch
Support, the support of middle body, surge tank and washing tank are tested, and are read intrinsic frequency, Mode Shape and damping ratio parameter, unit and are being tried
In the vehicle stage, compressor acceleration, displacement, power and torque are read, unit after installation and debugging success, passes through data acquisition module at the scene
The operating that block carries out under unloaded, load, various inlet pressure and under different displacements compressor is tested, and records the fortune of each component
Row parameter.
8. the reciprocating compressor method for diagnosing faults according to claim 6 based on neural network algorithm, feature exist
In: in step S3, the method for building up of simulation mathematical model specifically:
Establish ideal thermodynamic equation are as follows: P=f (n, q, T, T1, k, B, r1, r2, p1, p2, z, Q, t), it will be each after the completion of foundation
The thermodynamical equilibrium equation simultaneous composition equation group of grade calculates, and output dynamic pressure value P is for dynamics calculation and vibration after the completion of calculating
Mechanics Calculation uses, and enabling f (P) is unit ideal thermodynamic mathematical model, then exports f (P) and use for neural network；
Establish ideal kinetics equation are as follows: W=f (n, m, t, l, p1, p2, r2, c, P, z1), output torque value W after the completion of foundation
It calculates and uses for vibration mechanics, enabling f (W) is ideal kinetics mathematical model, then exports f (W) and use for neural network；
F (P) and f (W) combines BP neural network method, the real time execution parameter that input local communication module obtains, iteration optimization
Fitting parameter z and z1 establish compressor emulation thermodynamical model and emulation kinetic model；
Ideal vibration mechanics equation: V=f (M, C, K, P, W, t) is established, enables f (V) for ideal vibration mechanics mathematical model, in conjunction with
The real time execution parameter that finite element analysis and local communication module obtain carries out structural dynamic optimum, enables f (V) ' it is that SDM is modified
Emulation vibration mechanical model afterwards then exports f (V) ' it is used for neural network and expert system；
Wherein, n is compressor rotary speed, and Q is capacity, and q is clearance volume, and t is the time, and T is intake air temperature, and T1 is exhaust temperature
Degree, k are gas adiabatic exponent, and B is cylinder unit time and external heat exchange value, and r1 is certain grade of cylinder bore value, and r2 is crankshaft
The radius of gyration, p1 are admission pressure, and p2 is pressure at expulsion, and Q is gas displacement, and z is thermodynamics fitting parameter, and P is in cylinder
Transient pressure, m are toward complex inertia mass power, and l is compressor columns, and c is sliding friction damping, and z1 is dynamics fitting parameter, W
Transient torque value is exported for motor, M is the mass matrix of each components of compressor, and C is the damping of each components of compressor
Matrix, K are the stiffness matrix of each components of compressor.
9. the reciprocating compressor method for diagnosing faults according to claim 8 based on neural network algorithm, feature exist
In: it is first discrete to compressor model progress to resolve into each minor structure when establishing emulation vibration mechanical model, then by limited
Member emulation show that the master mode and Constrained mode of each minor structure obtain the polycondensation stiffness matrix of Modal Space multiplied by transformation matrix
It is carried out by being compared with the modal data of actual measurement each minor structure of compressor based on finite element analysis and reality with polycondensation mass matrix
The structural dynamic optimum of mode is tested, then combination obtains the mass matrix, stiffness matrix and damping square of the Modal Space of complete machine
Battle array introduces generalized force and carries out unit coupling analysis, extracts acceleration, displacement, power and the torque of compressor corresponding site.
10. the reciprocating compressor method for diagnosing faults according to claim 6 based on neural network algorithm, feature exist
In: in step S3, filtering uses Kalman filtering.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811643547.8A CN109779894B (en) | 2018-12-29 | 2018-12-29 | Reciprocating compressor fault diagnosis system and method based on neural network algorithm |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811643547.8A CN109779894B (en) | 2018-12-29 | 2018-12-29 | Reciprocating compressor fault diagnosis system and method based on neural network algorithm |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN109779894A true CN109779894A (en) | 2019-05-21 |
| CN109779894B CN109779894B (en) | 2020-01-10 |
Family
ID=66499520
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201811643547.8A Active CN109779894B (en) | 2018-12-29 | 2018-12-29 | Reciprocating compressor fault diagnosis system and method based on neural network algorithm |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109779894B (en) |
Cited By (24)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110318993A (en) * | 2019-08-01 | 2019-10-11 | 江西资生科技有限公司 | A kind of reciprocating compressor connecting rod slideway remote failure diagnosis system and device |
| CN110360091A (en) * | 2019-06-05 | 2019-10-22 | 合肥通用机械研究院有限公司 | A kind of refrigeration compressor TT&C system neural network based and investigating method |
| CN111027260A (en) * | 2019-12-24 | 2020-04-17 | 大连圣力来监测技术有限公司 | Reciprocating compressor fault diagnosis method based on neural network |
| CN111275007A (en) * | 2020-02-24 | 2020-06-12 | 山东大学 | Bearing fault diagnosis method and system based on multi-scale information fusion |
| CN111272412A (en) * | 2020-03-23 | 2020-06-12 | 北京昊鹏智能技术有限公司 | Fault detection method, device, equipment and storage medium of electric control pneumatic equipment |
| CN111980902A (en) * | 2019-05-23 | 2020-11-24 | 睿捷国际股份有限公司 | Monitoring system and monitoring method for reciprocating compressor |
| CN112049836A (en) * | 2019-06-05 | 2020-12-08 | 大连中和聚能自动控制系统有限公司 | Visual system of pneumatic actuating mechanism |
| CN112051039A (en) * | 2019-06-05 | 2020-12-08 | 大连中和聚能自动控制系统有限公司 | Experiment table system for fault simulation of pneumatic actuating mechanism |
| EP3786746A1 (en) * | 2019-08-28 | 2021-03-03 | Siemens Aktiengesellschaft | Efficient error analysis by means of simulated errors in a digital twin |
| CN112443479A (en) * | 2019-08-27 | 2021-03-05 | 深圳中集智能科技有限公司 | Compressor fault diagnosis method and device, storage medium and electronic equipment |
| CN112610420A (en) * | 2020-12-17 | 2021-04-06 | 华能辽宁清洁能源有限责任公司 | Novel wind generating set safety chain simulation maintenance method |
| CN112926257A (en) * | 2020-09-25 | 2021-06-08 | 中国石油天然气集团有限公司 | Reciprocating natural gas compressor fault diagnosis system and diagnosis method |
| CN112945535A (en) * | 2021-02-20 | 2021-06-11 | 广东石油化工学院 | Rotating machinery fault detection method and device based on numerical simulation |
| CN113011055A (en) * | 2021-02-03 | 2021-06-22 | 中国船级社 | Motor high-temperature fault simulation method and system based on finite element technology |
| CN113435111A (en) * | 2021-06-08 | 2021-09-24 | 西安交通大学 | Fault diagnosis method and system for reciprocating compressor |
| CN113494527A (en) * | 2021-07-30 | 2021-10-12 | 哈尔滨工业大学 | Constant force control method based on electromagnetic auxiliary type constant force spring support |
| CN113587363A (en) * | 2021-08-04 | 2021-11-02 | 广东美的暖通设备有限公司 | Compressor fault detection method and device, computing equipment and storage medium |
| CN113656989A (en) * | 2021-06-29 | 2021-11-16 | 东方电气集团东方电机有限公司 | Bolt fault diagnosis method and device |
| CN113757093A (en) * | 2021-10-14 | 2021-12-07 | 中国海洋石油集团有限公司 | Fault diagnosis method for flash steam compressor unit |
| CN114184883A (en) * | 2021-11-22 | 2022-03-15 | 国网河南省电力公司漯河供电公司 | Distribution network fault detection precision calculation method based on distribution network fault simulation |
| CN115596654A (en) * | 2022-09-21 | 2023-01-13 | 西安交通大学（Cn） | Reciprocating compressor fault diagnosis method and system based on state parameter learning |
| CN116538092A (en) * | 2023-07-06 | 2023-08-04 | 中国科学院理化技术研究所 | Compressor on-line monitoring and diagnosing method, device, equipment and storage medium |
| CN117370824A (en) * | 2023-12-07 | 2024-01-09 | 深圳市伟昊净化设备有限公司 | Intelligent monitoring method and system for air inlet state of air compressor |
| US11906112B2 (en) | 2022-12-19 | 2024-02-20 | Chengdu Qinchuan Iot Technology Co., Ltd | Methods for safety management of compressors in smart gas pipeline network and internet of things systems thereof |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US5919267A (en) * | 1997-04-09 | 1999-07-06 | Mcdonnell Douglas Corporation | Neural network fault diagnostics systems and related method |
| CN104750742A (en) * | 2013-12-31 | 2015-07-01 | 南京理工大学常熟研究院有限公司 | Fault diagnosis method and system for heading machine hydraulic system |
| CN106124982A (en) * | 2016-06-14 | 2016-11-16 | 都城绿色能源有限公司 | Automatic expert's resultant fault diagnostic system of a kind of Wind turbines and diagnostic method |
-
2018
- 2018-12-29 CN CN201811643547.8A patent/CN109779894B/en active Active
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US5919267A (en) * | 1997-04-09 | 1999-07-06 | Mcdonnell Douglas Corporation | Neural network fault diagnostics systems and related method |
| CN104750742A (en) * | 2013-12-31 | 2015-07-01 | 南京理工大学常熟研究院有限公司 | Fault diagnosis method and system for heading machine hydraulic system |
| CN106124982A (en) * | 2016-06-14 | 2016-11-16 | 都城绿色能源有限公司 | Automatic expert's resultant fault diagnostic system of a kind of Wind turbines and diagnostic method |
Non-Patent Citations (1)
| Title |
|---|
| 张华: "基于LM-BP神经网络的多级往复式压缩机气阀故障诊断研究", 《中国优秀硕士学位论文全文数据库 信息科技辑》 * |
Cited By (37)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111980902B (en) * | 2019-05-23 | 2022-08-12 | 睿捷国际股份有限公司 | Monitoring system and monitoring method for reciprocating compressor |
| CN111980902A (en) * | 2019-05-23 | 2020-11-24 | 睿捷国际股份有限公司 | Monitoring system and monitoring method for reciprocating compressor |
| CN112049836A (en) * | 2019-06-05 | 2020-12-08 | 大连中和聚能自动控制系统有限公司 | Visual system of pneumatic actuating mechanism |
| CN110360091A (en) * | 2019-06-05 | 2019-10-22 | 合肥通用机械研究院有限公司 | A kind of refrigeration compressor TT&C system neural network based and investigating method |
| CN112051039A (en) * | 2019-06-05 | 2020-12-08 | 大连中和聚能自动控制系统有限公司 | Experiment table system for fault simulation of pneumatic actuating mechanism |
| CN110318993A (en) * | 2019-08-01 | 2019-10-11 | 江西资生科技有限公司 | A kind of reciprocating compressor connecting rod slideway remote failure diagnosis system and device |
| CN112443479A (en) * | 2019-08-27 | 2021-03-05 | 深圳中集智能科技有限公司 | Compressor fault diagnosis method and device, storage medium and electronic equipment |
| US11567489B2 (en) | 2019-08-28 | 2023-01-31 | Siemens Aktiengesellschaft | Method and device for efficient fault analysis through simulated faults in a digital twin |
| EP3786746A1 (en) * | 2019-08-28 | 2021-03-03 | Siemens Aktiengesellschaft | Efficient error analysis by means of simulated errors in a digital twin |
| WO2021037498A1 (en) * | 2019-08-28 | 2021-03-04 | Siemens Aktiengesellschaft | Efficient fault analysis through simulated faults in a digital twin |
| CN114341814A (en) * | 2019-08-28 | 2022-04-12 | 西门子股份公司 | Efficient fault analysis by simulating faults in digital twinning |
| CN111027260A (en) * | 2019-12-24 | 2020-04-17 | 大连圣力来监测技术有限公司 | Reciprocating compressor fault diagnosis method based on neural network |
| CN111275007A (en) * | 2020-02-24 | 2020-06-12 | 山东大学 | Bearing fault diagnosis method and system based on multi-scale information fusion |
| CN111275007B (en) * | 2020-02-24 | 2021-10-22 | 山东大学 | Bearing fault diagnosis method and system based on multi-scale information fusion |
| CN111272412B (en) * | 2020-03-23 | 2022-02-22 | 北京昊鹏智能技术有限公司 | Fault detection method, device, equipment and storage medium of electric control pneumatic equipment |
| CN111272412A (en) * | 2020-03-23 | 2020-06-12 | 北京昊鹏智能技术有限公司 | Fault detection method, device, equipment and storage medium of electric control pneumatic equipment |
| CN112926257A (en) * | 2020-09-25 | 2021-06-08 | 中国石油天然气集团有限公司 | Reciprocating natural gas compressor fault diagnosis system and diagnosis method |
| CN112610420B (en) * | 2020-12-17 | 2024-04-02 | 华能辽宁清洁能源有限责任公司 | Novel wind generating set safety chain simulation maintenance method |
| CN112610420A (en) * | 2020-12-17 | 2021-04-06 | 华能辽宁清洁能源有限责任公司 | Novel wind generating set safety chain simulation maintenance method |
| CN113011055A (en) * | 2021-02-03 | 2021-06-22 | 中国船级社 | Motor high-temperature fault simulation method and system based on finite element technology |
| CN112945535A (en) * | 2021-02-20 | 2021-06-11 | 广东石油化工学院 | Rotating machinery fault detection method and device based on numerical simulation |
| CN113435111B (en) * | 2021-06-08 | 2022-10-25 | 西安交通大学 | Fault diagnosis method and system for reciprocating compressor |
| CN113435111A (en) * | 2021-06-08 | 2021-09-24 | 西安交通大学 | Fault diagnosis method and system for reciprocating compressor |
| CN113656989A (en) * | 2021-06-29 | 2021-11-16 | 东方电气集团东方电机有限公司 | Bolt fault diagnosis method and device |
| CN113494527B (en) * | 2021-07-30 | 2022-06-24 | 哈尔滨工业大学 | Constant force control method based on electromagnetic auxiliary type constant force spring support |
| CN113494527A (en) * | 2021-07-30 | 2021-10-12 | 哈尔滨工业大学 | Constant force control method based on electromagnetic auxiliary type constant force spring support |
| CN113587363B (en) * | 2021-08-04 | 2022-07-26 | 广东美的暖通设备有限公司 | Compressor fault detection method and device, computing equipment and storage medium |
| CN113587363A (en) * | 2021-08-04 | 2021-11-02 | 广东美的暖通设备有限公司 | Compressor fault detection method and device, computing equipment and storage medium |
| CN113757093A (en) * | 2021-10-14 | 2021-12-07 | 中国海洋石油集团有限公司 | Fault diagnosis method for flash steam compressor unit |
| CN114184883A (en) * | 2021-11-22 | 2022-03-15 | 国网河南省电力公司漯河供电公司 | Distribution network fault detection precision calculation method based on distribution network fault simulation |
| CN115596654A (en) * | 2022-09-21 | 2023-01-13 | 西安交通大学（Cn） | Reciprocating compressor fault diagnosis method and system based on state parameter learning |
| CN115596654B (en) * | 2022-09-21 | 2023-12-22 | 西安交通大学 | Reciprocating compressor fault diagnosis method and system based on state parameter learning |
| US11906112B2 (en) | 2022-12-19 | 2024-02-20 | Chengdu Qinchuan Iot Technology Co., Ltd | Methods for safety management of compressors in smart gas pipeline network and internet of things systems thereof |
| CN116538092B (en) * | 2023-07-06 | 2023-11-14 | 中国科学院理化技术研究所 | Compressor on-line monitoring and diagnosing method, device, equipment and storage medium |
| CN116538092A (en) * | 2023-07-06 | 2023-08-04 | 中国科学院理化技术研究所 | Compressor on-line monitoring and diagnosing method, device, equipment and storage medium |
| CN117370824A (en) * | 2023-12-07 | 2024-01-09 | 深圳市伟昊净化设备有限公司 | Intelligent monitoring method and system for air inlet state of air compressor |
| CN117370824B (en) * | 2023-12-07 | 2024-03-22 | 深圳市伟昊净化设备有限公司 | Intelligent monitoring method and system for air inlet state of air compressor |
Also Published As
| Publication number | Publication date |
|---|---|
| CN109779894B (en) | 2020-01-10 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109779894A (en) | A kind of reciprocating compressor fault diagnosis system and method based on neural network algorithm | |
| US6556939B1 (en) | Inferential signal generator for instrumented equipment and processes | |
| CN106404403B (en) | Method and system for analysis of a turbomachine | |
| US6092016A (en) | Apparatus and method for diagnosing an engine using an exhaust temperature model | |
| CN109033930B (en) | Equipment fault diagnosis method based on fault mechanism and statistical model online learning | |
| CN114912371B (en) | Digital twin body modeling system and method for running stability detection of tubular pump | |
| CN111441864A (en) | Engine health diagnosis method and engine diagnosis system | |
| CN111581763B (en) | Method for evaluating air path fault diagnosis result of aero-engine | |
| JP6444885B2 (en) | Exercise equipment condition monitoring system | |
| Centomo et al. | The design of a digital-twin for predictive maintenance | |
| CN112861425A (en) | Method for detecting performance state of double-shaft gas turbine by combining mechanism and neural network | |
| CN109063246A (en) | Frequency-changeable compressor is piped evaluation of vibration method | |
| CN109826783A (en) | A kind of reciprocating compressor indicator diagram testing device and method for diagnosing faults | |
| CN114154360B (en) | Multi-physical field probability digital twin modeling method for high-pressure internal gear pump | |
| CN115640741A (en) | Method for diagnosing fault based on simulation model driven by model and data | |
| CN110513336B (en) | Method for determining offline water washing time of gas turbine of power station | |
| CN113420465B (en) | Hydraulic support life cycle management method based on digital twin model | |
| CN112926698B (en) | Vibration prediction and assembly evaluation method for large-scale rotating equipment | |
| Hou et al. | Inter-shaft bearing fault diagnosis based on aero-engine system: a benchmarking dataset study | |
| CN114201872A (en) | Digital twin modeling method for diesel engine fuel system | |
| CN114925502A (en) | Method and device for constructing digital verification model | |
| CN108757224B (en) | Fault diagnosis method for impact test of solid rocket engine | |
| Zweiri et al. | A comparison of dynamic models of various complexity for diesel engines | |
| CN116227086B (en) | Aeroengine gas circuit fault simulation method | |
| CN116090266B (en) | Real-time fault diagnosis method for air path performance of power device |
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 |