CN116626479A - Multi-fault coupling diagnosis method for electromagnetic valve - Google Patents
Multi-fault coupling diagnosis method for electromagnetic valve Download PDFInfo
- Publication number
- CN116626479A CN116626479A CN202310537798.2A CN202310537798A CN116626479A CN 116626479 A CN116626479 A CN 116626479A CN 202310537798 A CN202310537798 A CN 202310537798A CN 116626479 A CN116626479 A CN 116626479A
- Authority
- CN
- China
- Prior art keywords
- electromagnetic valve
- fault
- driving current
- current
- valve
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 238000000034 method Methods 0.000 title claims abstract description 47
- 238000003745 diagnosis Methods 0.000 title claims abstract description 41
- 238000010168 coupling process Methods 0.000 title claims abstract description 38
- 238000005859 coupling reaction Methods 0.000 title claims abstract description 38
- 230000008878 coupling Effects 0.000 title claims abstract description 36
- 238000004422 calculation algorithm Methods 0.000 claims abstract description 67
- 239000013598 vector Substances 0.000 claims abstract description 40
- 230000008569 process Effects 0.000 claims abstract description 10
- 238000002474 experimental method Methods 0.000 claims abstract description 7
- 239000002245 particle Substances 0.000 claims description 61
- 238000000354 decomposition reaction Methods 0.000 claims description 30
- 230000000739 chaotic effect Effects 0.000 claims description 21
- 230000007774 longterm Effects 0.000 claims description 17
- 238000006073 displacement reaction Methods 0.000 claims description 14
- 230000008859 change Effects 0.000 claims description 13
- 238000004458 analytical method Methods 0.000 claims description 12
- 230000006870 function Effects 0.000 claims description 12
- 238000010586 diagram Methods 0.000 claims description 11
- 238000013507 mapping Methods 0.000 claims description 11
- 238000000605 extraction Methods 0.000 claims description 10
- 238000009826 distribution Methods 0.000 claims description 8
- 239000003973 paint Substances 0.000 claims description 8
- 230000000630 rising effect Effects 0.000 claims description 7
- 230000003044 adaptive effect Effects 0.000 claims description 6
- 239000012530 fluid Substances 0.000 claims description 6
- 238000013461 design Methods 0.000 claims description 5
- 239000012535 impurity Substances 0.000 claims description 5
- 239000002184 metal Substances 0.000 claims description 5
- 238000011160 research Methods 0.000 claims description 5
- 238000004804 winding Methods 0.000 claims description 5
- 238000009825 accumulation Methods 0.000 claims description 4
- 238000005259 measurement Methods 0.000 claims description 4
- 239000004576 sand Substances 0.000 claims description 4
- 238000012935 Averaging Methods 0.000 claims description 3
- 230000001133 acceleration Effects 0.000 claims description 3
- 230000009471 action Effects 0.000 claims description 3
- 238000013145 classification model Methods 0.000 claims description 3
- 230000007423 decrease Effects 0.000 claims description 3
- 230000004907 flux Effects 0.000 claims description 3
- 239000000446 fuel Substances 0.000 claims description 3
- 238000010606 normalization Methods 0.000 claims description 3
- 238000012545 processing Methods 0.000 claims description 3
- 230000009467 reduction Effects 0.000 claims description 3
- 230000001105 regulatory effect Effects 0.000 claims description 3
- 238000012360 testing method Methods 0.000 abstract description 5
- 238000001228 spectrum Methods 0.000 abstract description 3
- 238000005299 abrasion Methods 0.000 description 9
- 238000012549 training Methods 0.000 description 7
- 230000008901 benefit Effects 0.000 description 4
- 238000005516 engineering process Methods 0.000 description 4
- 238000012423 maintenance Methods 0.000 description 3
- 238000005457 optimization Methods 0.000 description 3
- 238000012706 support-vector machine Methods 0.000 description 3
- 230000015556 catabolic process Effects 0.000 description 2
- 238000006731 degradation reaction Methods 0.000 description 2
- 230000002159 abnormal effect Effects 0.000 description 1
- 238000013528 artificial neural network Methods 0.000 description 1
- 238000004364 calculation method Methods 0.000 description 1
- 238000007635 classification algorithm Methods 0.000 description 1
- 239000002131 composite material Substances 0.000 description 1
- 238000005260 corrosion Methods 0.000 description 1
- 230000007797 corrosion Effects 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 238000011156 evaluation Methods 0.000 description 1
- 239000000284 extract Substances 0.000 description 1
- 239000002737 fuel gas Substances 0.000 description 1
- 239000000295 fuel oil Substances 0.000 description 1
- 239000011810 insulating material Substances 0.000 description 1
- 230000003993 interaction Effects 0.000 description 1
- 238000012843 least square support vector machine Methods 0.000 description 1
- 238000004519 manufacturing process Methods 0.000 description 1
- 239000000463 material Substances 0.000 description 1
- 230000001537 neural effect Effects 0.000 description 1
- 230000004044 response Effects 0.000 description 1
- 238000012216 screening Methods 0.000 description 1
Classifications
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01R—MEASURING ELECTRIC VARIABLES; MEASURING MAGNETIC VARIABLES
- G01R31/00—Arrangements for testing electric properties; Arrangements for locating electric faults; Arrangements for electrical testing characterised by what is being tested not provided for elsewhere
- G01R31/327—Testing of circuit interrupters, switches or circuit-breakers
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
- G06F18/2411—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on the proximity to a decision surface, e.g. support vector machines
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/004—Artificial life, i.e. computing arrangements simulating life
- G06N3/006—Artificial life, i.e. computing arrangements simulating life based on simulated virtual individual or collective life forms, e.g. social simulations or particle swarm optimisation [PSO]
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- General Physics & Mathematics (AREA)
- Evolutionary Computation (AREA)
- General Engineering & Computer Science (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- Mathematical Physics (AREA)
- Computational Linguistics (AREA)
- General Health & Medical Sciences (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Software Systems (AREA)
- Bioinformatics & Computational Biology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Health & Medical Sciences (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Biology (AREA)
- Magnetically Actuated Valves (AREA)
Abstract
In order to ensure that a certain power system can normally operate, the invention provides a multi-fault coupling diagnosis method for an electromagnetic valve; firstly, analyzing the feasibility of a driving current signal serving as a key fault characteristic of the electromagnetic valve, and acquiring a driving end current signal of the electromagnetic valve with multi-fault coupling through experiments; aiming at the nonlinearity, complex frequency spectrum and doping of a large amount of noise of a current signal, CEEMDAN is adopted to decompose the current signal, and energy entropy of different IMF components is extracted as fault feature vector information; finally, the IPSO-LSSVM algorithm is utilized to realize the multi-fault coupling diagnosis of the electromagnetic valve. The invention has better fault recognition rate, is a necessary link in the test performance process of the power system, can effectively screen out the fault electromagnetic valve, and avoids the influence on the whole system.
Description
Technical Field
The invention relates to the field of feature extraction and multi-fault coupling diagnosis, in particular to a multi-fault coupling diagnosis method for a solenoid valve of a power system.
Background
The electromagnetic valve is a comprehensive product integrating mechanical, electrical, hydraulic and other technologies, and has the advantages of high response speed, easy control and the like, so that the electromagnetic valve is widely applied to the fields of aerospace, transportation and the like. As a key component of a power system, the power system is charged with important tasks of supplying fuel oil and fuel gas, and if the power system fails, the power system will be under-powered, so that the experiment will fail.
However, in the use process of the electromagnetic valve, the electromagnetic valve is in a high-frequency operation and complex and changeable environment for a long time, the material, the structure and the performance of the electromagnetic valve are continuously degraded, the electromagnetic valve is shown as device degradation failure in the initial stage of performance degradation, and once the reliability threshold value of the device exceeds the limit, the electromagnetic valve is failed. If the temperature exceeds the threshold value of the insulating material, the winding wire will be burnt out, and the coil is abnormal; the electromagnetic valve acts on the fluid medium, and will produce corrosion phenomenon under the long-term action, cause the spring stiffness coefficient to change; the electromagnetic valve spring generates metal fatigue under the high-frequency actuation state; the fluid medium has certain impurities, so that the valve core is blocked from moving for a long time, the valve core is blocked, even blocked, and the valve core cannot normally act. Once the solenoid valve fails, a significant loss will be incurred.
Along with the extension of the working period and the long-term complex and changeable working environment, the electromagnetic valve has multiple fault coupling, such as fault superposition of spring fracture, valve core clamping stagnation, coil abrasion and the like. However, compared with the electromagnetic valve in a single fault state, the multi-fault coupling state is complex in fault condition, and the fault data collected on site has the characteristics of nonlinearity, complex frequency spectrum, large noise doping, small distribution difference among the fault data and the like, so that the fault diagnosis of the electromagnetic valve is greatly challenged. How does the solenoid valve multi-fault-coupled raw data thus be processed for more significant fault signatures? What kind of fault diagnosis method is adopted for the processed solenoid valve multi-fault coupling data to have higher accuracy? These two problems have become key difficulties in the use of solenoid valves.
Disclosure of Invention
Aiming at the defects of the existing method, the invention provides a method for establishing a multi-fault coupling feature vector of an electromagnetic valve, which comprises the following steps:
step1: feasibility analysis of drive current signal as solenoid valve multi-fault coupling state representation
The solenoid valve has 5 stages in its operation: the actuation touch, actuation motion, power-on hold, release touch and release motion stages; four stages in the working process of the electromagnetic valve: the method comprises the steps of carrying out analysis and research on the phases of actuation touch, actuation movement, power-on holding, release touch and the like, and checking whether a driving current signal is related to a fault state of the electromagnetic valve;
before the actuation and touch stage, the electromagnetic valve is not electrified, the driving current is 0A, and the electromagnetic valve does not work;
when the electromagnetic valve is electrified, the electromagnetic valve is in a suction touch stage, rising current is generated in the loop, and the magnetic flux of the coil is changed due to the self-inductance phenomenon of the coil in the electromagnetic circuit, so that electromotive force is generated, and abrupt change of current cannot be prevented; therefore, the current cannot directly reach the steady-state current, but rises at a higher speed, and at the moment, the electromagnetic force generated by the electromotive force cannot overcome the friction force and the spring tension, and the valve core does not displace;
When the driving current is increased to a first threshold value, the electromotive force generated by the coil causes the electromagnetic force to be larger than the friction force between the valve core and fuel and the tensile force of the spring, and the valve core starts to move towards the maximum displacement under the action of the combined force; the motion of the valve core generates dynamic electromotive force, so that the driving current is continuously reduced, when the driving current is reduced to a second threshold value, the valve core of the electromagnetic valve reaches the maximum displacement position and no displacement is generated, and the dynamic electromotive force at the moment disappears;
when the electromotive force disappears, the current increases at the current increasing speed at the moment of actuation touch, but the increasing speed of the driving current is slowed down along with the change of the induced electromotive force, but the loop current is still increased, and finally the stable value is reached and does not change any more;
the release stage: when the electromagnetic valve is in a de-energized state, the current in the loop is rapidly reduced, meanwhile, the valve core of the electromagnetic valve is reset due to the fact that the electromagnetic force at the moment of turn-off is smaller than the friction force and the spring tension force, the generated dynamic electromotive force reduces the reduction speed of the driving current, but as the valve core reaches the original position, the dynamic electromotive force disappears, and at the moment, the driving current is rapidly reduced; thus, during the release phase, the drive current decreases at a rate of change from small to large;
Step2: designing a solenoid valve driving current signal data acquisition system, setting a solenoid valve fault mode, and carrying out data acquisition
Based on the electromagnetic valve circuit, electromagnetic and mechanical analysis, the driving current is closely related to the displacement of the valve core of the electromagnetic valve; in order to verify that the electromagnetic valve driving current contains fault information, 8 multi-fault modes of an electromagnetic valve of a certain power system are simulated, and an electromagnetic valve driving current data acquisition system is designed; the method comprises the following steps:
the first step: design data acquisition system
The experimental equipment comprises a computer, an electromagnetic valve, a data acquisition card PCI-9114, a current sensor LTS 25-NP, a diode FR307, a 5V voltage-stabilizing direct current power supply GPS-3303C, a 24V voltage-stabilizing direct current power supply RIGOL-DP712 and the like;
and a second step of: connecting experimental equipment according to a circuit diagram
The computer is internally provided with a data acquisition card PCI-9114; the 37 port of the data acquisition card PCI-9114 is connected with the port 7 of the current sensor LTS 25-NP; the port 19 of the data acquisition card PCI-9114 is connected with the port 8 of the current sensor LTS 25-NP; a 5V stabilized direct current power supply is connected between the 19 port of the data acquisition card PCI-9114 and the 9 port of the current sensor LTS 25-NP, the positive electrode of the 5V stabilized direct current power supply is connected with the 9 port of the current sensor LTS 25-NP, and the negative electrode of the 5V stabilized direct current power supply is connected with the 19 port of the data acquisition card PCI-9114; ports 2 and 6 of the current sensor LTS 25-NP are directly connected, and ports 3 and 5 are directly connected; the electromagnetic valve is connected with a rectifying diode FR307 in parallel, the positive end of the rectifying diode FR307 is connected with a port 1 of a current sensor LTS 25-NP, and the negative end of the rectifying diode FR307 is connected with a switch model through a toggle switch: KN1-202 is connected with the positive electrode of 24V stabilized DC power supply RIGOL-DP712, and the negative electrode of 24V stabilized DC power supply RIGOL-DP712 is connected with port 4 of current sensor LTS 25-NP;
And a third step of: setting different fault modes for electromagnetic valves
In order to verify that the driving current of the electromagnetic valve can be used for performing fault diagnosis of the electromagnetic valve, the electromagnetic valve fault caused by long-term high-frequency actuation and external environment is simulated through setting a fault expression form, and real driving current data are acquired through a data acquisition system; setting 8 multi-fault modes: normal, coil wear, spool sticking, spring break, coil wear + spool sticking, spring break + spool sticking, coil wear + spring break + spool sticking; wherein, the normal state selects the electromagnetic valve which can work normally;
in the experiment, the voltage-stabilizing direct-current power supply respectively provides direct-current voltage for the electromagnetic valve and the current sensor; the electromagnetic valve is determined to be opened and closed by a control switch; the data acquisition software in the computer is used for interacting with the data acquisition card, acquiring a driving current signal, and carrying out maximum and minimum standardization processing on the current signal;
fourth step: driving current signal acquisition using data acquisition software
The data acquisition software is used for realizing the acquisition of the driving current signals and storing the driving current signals into a folder;
step3: feature extraction is carried out on electromagnetic valve driving current signals by using CEEMDAN algorithm, energy entropy is calculated, and a fault feature vector table of the electromagnetic valve is constructed
Decomposing the driving current signal by CEEMDAN algorithm (Torres M E, colominas M A, schlotthauer G, et al A complete ensemble empirical mode decomposition with adaptive noise [ C ]// IEEE international conference on acoustics, speech and signalprocessing (ICASSP). IEEE, 2011:4144-4147.) to obtain natural modal components of different frequency bands, wherein it is noted that the extreme points of the final residual signal are not more than two; on the basis, energy of different frequency bands of the driving current is extracted as fault characteristic information of the driving current, and a multi-fault coupling characteristic vector table of the electromagnetic valve is established;
1) Normalizing raw drive current data
In order to eliminate the influence of the measurement unit and the magnitude thereof among the characteristic data, the maximum and minimum normalization is adopted to process the original driving current data, as shown in the formula 1:
wherein z is normalized data, x is a sample, x max For the maximum value of the sample, x min Is the sample minimum;
2) Decomposing the normalized data by CEEMDAN algorithm to obtain 7 natural modal components and a residual component
The CEEMDAN was calculated as follows:
first, with a driving current signal as an original signal, an adaptive noise ζ compliant with a normal distribution is added to the original signal x (t) i (t) obtaining a mixed signal x i (t), wherein i represents the signal of the ith dimension, for x i (t) EMD decomposition was performed N times and N IMFs were subjected to 1 i (t) averaging to obtain a first modal componentWherein N represents the empirical mode decomposition number, IMF 1 i (t) calculating a first residual signal R as an eigenmode component 1 (t)；
For R 1 (t)+ξ i (t)E 1 (ω i (t)) to obtain second-order modal components wherein ωi (t) white noise following the N (0, 1) distribution, E 1 The (-) function is the IMF component of the 1 st order obtained by EMD decomposition, E 1 (ω i (t)) is an IMF component of the 1 st order obtained by EMD decomposition of white noise, and calculates a second residual signal R 2 (t)；
Repeating the steps until the extreme points of the residual signals are not more than two, obtaining k-order modal components, and decomposing the original signals into:
3) Calculating the energy entropy of a sample, and constructing an electromagnetic valve fault characteristic vector
After CEEMDAN decomposition, eigen-mode function IMF components of different frequency bands are obtained, the fault state of the electromagnetic valve is reflected more carefully, and the electromagnetic valve can be effectively subjected to fault diagnosis by extracting energy of different frequency bands of the driving current as fault characteristic information;
the total energy of each IMF of the solenoid valve is:
wherein ,Ek Represents the energy of the kth IMF, |x ki The I is the amplitude of the kth IMF component discrete point;
The feature vectors are expressed as:
X＝[E 1 ，E 2 ，…E m ] (8)
wherein X is a feature vector, E is a two-norm of the feature vector,the m is the number of samples and is the normalized fault characteristic vector; thereby obtaining a normalized fault feature vector table; obtaining 7 modal components and 1 residual signal by CEEMDAN, R k (t) is the final residual signal, R 1 (t) corresponds to R 2 (t) raw signal, R 2 (t) is the residual signal and,using the LSSVM algorithm (Suykens J A K, aandewalle J.least squares support vector machine classifiers [ J)]Neural processing letters,1999,9: 293-300.) to classify the sample data.
In one embodiment of the present invention, a 2W160Y32B-DC24V type direct-acting solenoid valve is selected as the subject, a first threshold value=0.58A, a second threshold value=0.25A, and a steady value=0.85A.
In another embodiment of the present invention, the 8 multiple failure modes are specified as follows,
coil wear: using 30700to polish part of windings of the solenoid valve coil, damaging insulating paint, simulating the falling off of the insulating paint caused by the rising of the external environment and the temperature of the solenoid valve, and placing the polished coil in a normal solenoid valve;
spring breakage: the sheared spring is lifted to replace the spring of the normal valve, so that the spring fracture caused by metal fatigue of the spring due to long-term high-frequency actuation of the electromagnetic valve is simulated;
Valve core clamping stagnation: a certain amount of sand is filled into a normal electromagnetic valve to simulate valve core clamping stagnation caused by long-term accumulation of a fluid medium containing impurities;
coil wear + spring break, coil wear + spool clamping, spring break + spool clamping and coil wear + spring break + spool clamping are all set up based on above-mentioned single fault condition, simulate the many fault conditions that the solenoid valve probably appears in long-term use respectively.
In order to optimize and obtain the optimal parameters of the LSSVM algorithm, the invention also provides a multi-fault coupling diagnosis method for the electromagnetic valve, which comprises the following steps:
the first step: improved particle swarm algorithm by adopting self-adaptive dynamic inertia weight coefficient and Sine-Tent-Cosine chaotic mapping method
PSO algorithm (Kennedy J, eberhart R.particle swarm optimization [ C ]// Proceedings of ICVV 95-international conference on neural networks, IEEE,1995, 4:1942-1948.) is implemented as follows:
assuming that the number of particle populations is N' dimensional, each population has M feasible solutions, and the position of particle j is represented as X (j) = (X) j1 ，x j2 ，…，x jN′), wherein xj1 ，x j2 ，…，x jN′ The position information of the particle j in different dimensions is respectively shown, and the velocity of the particle j is expressed as V (j) = (V) j1 ，v j2 ，…v jN′), wherein vj1 ，v j2 ，…v jN′ The individual optimal position Pbest is denoted pbest= (P) for velocity information of the particle j in different dimensions, respectively j1 ，P j2 ，…P jN′), wherein Pj1 ，P j2 ，…P jN′ The individual optimal position information of the particle j in different dimensions is respectively shown as a global optimal position gbest= (g) j1 ，g j2 ，…g jN′), wherein gj1 ，g j2 ，…g jN′ Global optimal position information of the particles j in different dimensions is respectively obtained; the position and velocity update formula of particle j in the N' th dimension is:
wherein ,the speed of the particle j in the nth dimension of the t generation is given, and t is the current iteration number; w is a weight; c 1 、c 2 Are acceleration factors and are responsible for regulating the learning rate; r is (r) 1 、r 2 Are random values within (0, 1); pbest is the best position of the individualPlacing; gbest is the optimal position of the population; />Position information of the particle j in the nth dimension of the t-1 generation;
an iterative mode of improving particles by adopting an adaptive dynamic inertia weight coefficient is adopted, and the expression is as follows:
wherein ,wmax For maximum weight, w min Is the minimum value of the weight; f. f (f) min 、f avg The algorithm fitness function, the algorithm fitness minimum value and the algorithm fitness average value are respectively;
initializing a particle swarm by adopting a chaotic mapping method, generating uniformly distributed particles, and improving the quality of the particles; the Sine-Tent-Cosine chaotic mapping expression is as follows:
Wherein r epsilon (0, 1) is a random number, x (z) represents a z-th chaotic variable, x (1) epsilon (0, 1) is an initial variable and is a random number, x (z) is continuously updated according to a formula, and finally, the generated chaotic sequence is mapped into a solution space;
and a second step of: classifying solenoid valve fault characteristic samples by using IPSO-LSSVM algorithm to realize solenoid valve multi-fault coupling diagnosis
The found optimal solution is input into an LSSVM algorithm, 8 fault states of the electromagnetic valve are classified by utilizing a classification model in the LSSVM algorithm, and the mean square error and the root mean square error of the true value and the classification value are calculated.
In one embodiment of the present invention, the 8 multiple failure modes are: normal, coil wear, spool sticking, spring break, coil wear + spool sticking, spring break + spool sticking, coil wear + spring break + spool sticking; wherein, normal state selects the solenoid valve that can normally work.
In another embodiment of the present invention, the 8 multiple failure modes are specified as follows,
coil wear: using 30700to polish part of windings of the solenoid valve coil, damaging insulating paint, simulating the falling off of the insulating paint caused by the rising of the external environment and the temperature of the solenoid valve, and placing the polished coil in a normal solenoid valve;
Spring breakage: the sheared spring is lifted to replace the spring of the normal valve, so that the spring fracture caused by metal fatigue of the spring due to long-term high-frequency actuation of the electromagnetic valve is simulated;
valve core clamping stagnation: a certain amount of sand is filled into a normal electromagnetic valve to simulate valve core clamping stagnation caused by long-term accumulation of a fluid medium containing impurities;
coil wear + spring break, coil wear + spool clamping, spring break + spool clamping and coil wear + spring break + spool clamping are all set up based on above-mentioned single fault condition, simulate the many fault conditions that the solenoid valve probably appears in long-term use respectively.
The invention extracts the characteristics of the collected characteristic data through complete set empirical mode decomposition (CEEMDAN), and then classifies fault characteristic vectors by utilizing the improved IPSO-LSSVM, thereby realizing the multi-fault coupling diagnosis of the electromagnetic valve.
Specifically:
1. aiming at the characteristics of nonlinearity, complex frequency spectrum, large noise doping, small distribution difference among fault data and the like of the fault data collected on site, the invention utilizes a CECEMDAN method to extract the characteristics of the multi-fault coupling data of the electromagnetic valve and establishes a fault characteristic vector table.
2. The method is characterized in that an improved Particle Swarm Optimization (PSO) algorithm is provided for intelligently selecting the super parameters of a Least Squares Support Vector Machine (LSSVM) algorithm, and the acquired and processed fault characteristic parameters are classified by utilizing the improved LSSVM so as to realize the multi-fault coupling diagnosis of the electromagnetic valve.
The invention mainly researches a multi-fault coupling diagnosis technology of an electromagnetic valve, and aims to provide an effective and reliable multi-fault diagnosis method of the electromagnetic valve, which is used for screening and testing the electromagnetic valve of a system with higher reliability requirements.
Compared with the prior art, the invention has the advantages that the CEEMDAN method solves the problem of difficult extraction of the fault characteristics of the electromagnetic valve which is non-stable and nonlinear and contains a large amount of measurement noise, and the improved particle swarm algorithm (IPSO-LSSVM) algorithm is utilized to solve the problem of difficult parameter selection of the LSSVM, so that the multi-fault coupling diagnosis of the electromagnetic valve can be reliably realized, technical support is provided for the production design, assembly and experiment of a certain power system, the maintenance efficiency of maintenance personnel is improved, and a large amount of test and maintenance time and cost are saved.
Drawings
FIG. 1 shows a physical diagram of a 2W160Y32B-DC24V type electromagnetic valve for experiments;
FIG. 2 shows a graph of drive current and displacement versus time for various stages of a solenoid valve of the present invention;
FIG. 3 shows a physical diagram of the electromagnetic valve data acquisition system of the invention;
FIG. 4 shows a circuit diagram of the solenoid valve data acquisition system of the present invention;
FIG. 5 is a graph showing the current profile at the drive end for each state of the solenoid valve in accordance with the present invention
Fig. 6 shows a diagram of the extracted CEEMDAN characteristics in the present invention, in which fig. 6 (a) shows the CEEMDAN decomposition result of the driving current at the normal valve actuation stage and fig. 6 (b) shows the CEEMDAN decomposition result of the driving current at the coil wear actuation stage;
fig. 7 shows a CEEMDAN feature extraction result diagram in the present invention, in which fig. 7 (a) shows a CEEMDAN decomposition result of a driving current in a clamping and actuation stage of a valve core, and fig. 7 (b) shows a CEEMDAN decomposition result of a driving current in a breaking and actuation stage of a spring;
fig. 8 shows a CEEMDAN feature extraction result diagram in the present invention, in which fig. 8 (a) shows a coil wear + spring break pull-in stage driving current CEEMDAN decomposition result, and fig. 8 (b) shows a coil wear + spool clamping pull-in stage driving current CEEMDAN decomposition result;
fig. 9 shows a diagram of the extracted result of the CEEMDAN feature in the present invention, wherein fig. 9 (a) shows the decomposition result of the driving current CEEMDAN at the spring break + spool clamping stage, and fig. 9 (b) shows the decomposition result of the driving current CEEMDAN at the coil wear + spring break + spool clamping stage;
FIG. 10 shows a flow chart for optimizing LSSVM model parameters by the IPSO algorithm of the present invention;
fig. 11 shows a comparison result graph of a solenoid valve multi-fault diagnosis model in the present invention, in which fig. 11 (a) shows an SVM algorithm training result, fig. 11 (b) shows an LSSVM algorithm training result, in which fig. 11 (c) shows a GWO-LSSVM algorithm training result, fig. 11 (d) shows a VGWO-LSSVM algorithm training result, in which fig. 11 (e) shows a PSO-LSSVM algorithm training result, and fig. 11 (f) shows an IPSO-LSSVM algorithm training result.
Detailed Description
The invention provides a multi-fault coupling diagnosis method for an electromagnetic valve, which mainly comprises the steps of electromagnetic valve driving current characteristic extraction and electromagnetic valve multi-fault coupling diagnosis.
Carrying out feasibility analysis on the driving current signal serving as a multi-fault coupling state representation of the electromagnetic valve;
designing a solenoid valve driving current signal data acquisition system, setting a solenoid valve fault mode, and carrying out data acquisition;
extracting characteristics of a solenoid valve driving current signal by using a CEMMDAN algorithm, calculating energy entropy of the solenoid valve driving current signal, and constructing a fault characteristic vector table of the solenoid valve;
and optimizing the parameter selection of the LSSVM by using an IPSO algorithm, and performing fault diagnosis on the fault feature vector of the electromagnetic valve.
The technical solutions of the embodiments of the present invention will be described below with reference to specific embodiments of the present invention and the above-described drawings.
In a first aspect, the invention provides a method for establishing a multi-fault coupling feature vector of an electromagnetic valve, which comprises a feasibility analysis method, a data acquisition system design and a feature extraction method.
In order to analyze whether the driving current signal can be used as a key fault characteristic signal of the electromagnetic valve, firstly, in the working process of the electromagnetic valve, the connection of the driving current signal and the electromagnetic valve fault state is obtained by analyzing the five stages of the suction touch, the suction motion, the power-on holding, the release touch and the release motion of the electromagnetic valve. Thereby proving the feasibility of using the drive current signal as a key fault signature for the solenoid valve.
In order to achieve the above purpose, the technical scheme adopted by the invention is as follows:
step1: the drive current signal is analyzed for feasibility as a solenoid valve multi-fault coupling state representation.
Solenoid valve drive current is key characteristic information of solenoid valves, which have 5 stages in their operation: the actuation touch, actuation movement, power-on hold, release touch and release movement stages. Four stages in the working process of the electromagnetic valve: the phases of actuation touch, actuation movement, power-on holding, release touch and the like are analyzed and researched to examine whether the driving current signal is related to the fault state of the electromagnetic valve.
The invention selects a direct-acting electromagnetic valve as a research object, and a 2W160Y32B-DC24V electromagnetic valve is shown in figure 1. The first four stages in the working process of the electromagnetic valve are mainly: the electromagnetic valve is characterized in that the electromagnetic valve is provided with a valve core, the valve core is provided with a valve seat, the valve seat is provided with a valve core, the valve core is provided with a valve core, and the valve core is connected with a valve seat. Therefore, in order to study the multi-fault coupling state of the solenoid valve, a driving current signal which is easy to perform data acquisition is selected as a characteristic parameter of the solenoid valve. The four phases of the solenoid valve were analyzed in conjunction with the electrical, magnetic and mechanical equations described below.
FIG. 2 is a graph showing the changes of driving current and displacement with time at each stage, and the specific analysis steps are as follows:
before the actuation and touch stage, the electromagnetic valve is not electrified, the driving current is 0A, and the electromagnetic valve does not work;
when the electromagnetic valve starts to be electrified (24V), the electromagnetic valve is in a suction touch stage, rising current is generated in the loop, and the magnetic flux of the coil is changed due to the self-inductance phenomenon of the coil in the electromagnetic circuit, so that electromotive force is generated, and abrupt change of current cannot be prevented. Therefore, the current cannot directly reach the steady-state current, but rises at a relatively high speed, and at this time, the electromagnetic force generated by the electromotive force cannot overcome the friction force and the spring tension, and the valve core is not displaced.
When the driving current is increased to 0.58A, the electromotive force generated by the coil causes the electromagnetic force to be larger than the friction force between the valve core and the fuel and the tensile force of the spring, and the valve core starts to move towards the maximum displacement under the action of the combined force. The motion of the valve core generates dynamic electromotive force, so that the driving current is continuously reduced, when the driving current is reduced to 0.25A, the valve core of the electromagnetic valve reaches the maximum displacement position, no displacement is generated, and the dynamic electromotive force at the moment disappears.
When the electromotive force disappears, the current increases at the current increasing speed at the time of actuation touch, but the increasing speed of the driving current is slowed down along with the change of the induced electromotive force, but the loop current is still increased, and finally the stable value of 0.85A is reached and does not change any more.
The release stage: when the electromagnetic valve is released from the electrified state, the current in the loop is rapidly reduced, meanwhile, the valve core of the electromagnetic valve is reset due to the fact that the electromagnetic force is smaller than the friction force and the spring tension force at the moment of switching off, the generated dynamic electromotive force reduces the reduction speed of the driving current, but as the valve core reaches the original position, the dynamic electromotive force disappears, and at the moment, the driving current is rapidly reduced. Thus, during the release phase, the drive current decreases at a rate of change from small to large.
Step2: and designing a solenoid valve driving current signal data acquisition system, setting a solenoid valve fault mode, and carrying out data acquisition.
And (3) electromagnetic valve driving current data acquisition: based on the above-described solenoid circuit, electromagnetic and mechanical analysis, it is known that the drive current is closely related to solenoid valve spool displacement. In order to verify that the electromagnetic valve driving current contains fault information, the invention simulates 8 multi-fault modes of an electromagnetic valve of a power system, designs an electromagnetic valve driving current data acquisition system, selects a 2W160Y32B-DC24V type electromagnetic valve to carry out experimental study, the electromagnetic valve data acquisition system is shown in a figure 3, and a circuit diagram of the electromagnetic valve data acquisition system is shown in a figure 4.
The first step: designing a data acquisition system.
The experimental equipment mainly comprises a computer, an electromagnetic valve (2W 160Y32B-DC 24V), a data acquisition card (PCI-9114), a current sensor (LTS 25-NP), a diode (FR 307), a 5V voltage-stabilizing direct-current power supply (GPS-3303C), a 24V voltage-stabilizing direct-current power supply (RIGOL-DP 712) and the like.
And a second step of: the experimental equipment was connected according to the circuit diagram.
The computer has built-in data acquisition card (PCI-9114). The 37 port of the data acquisition card (PCI-9114) is connected with the port 7 of the current sensor (LTS 25-NP); the 19 port of the data acquisition card (PCI-9114) is connected with the port 8 of the current sensor (LTS 25-NP); the port 19 of the data acquisition card (PCI-9114) is connected with the port 9 of the current sensor (LTS 25-NP) by a 5V voltage-stabilizing direct current power supply, the positive electrode of the 5V voltage-stabilizing direct current power supply is connected with the port 9 of the current sensor (LTS 25-NP), and the negative electrode of the 5V voltage-stabilizing direct current power supply is connected with the port 19 of the data acquisition card (PCI-9114). Ports 2 and 6 of the current sensor (LTS 25-NP) are directly connected, and ports 3 and 5 are directly connected; the solenoid valve is connected with a rectifier diode (FR 307) in parallel, the positive end of the rectifier diode (FR 307) is connected with a port 1 of a current sensor (LTS 25-NP), and the negative end of the rectifier diode (FR 307) is connected with a toggle switch (switch model: KN 1-202) is connected with the positive pole of a 24V stabilized direct current power supply (RIGOL-DP 712), and the negative pole of the 24V stabilized direct current power supply (RIGOL-DP 712) is connected with the port 4 of a current sensor (LTS 25-NP).
And a third step of: different fault modes are respectively set for the solenoid valves.
In order to verify that the driving current of the electromagnetic valve can be used for performing fault diagnosis of the electromagnetic valve, the electromagnetic valve fault caused by long-term high-frequency actuation and external environment is simulated through setting a fault expression form, and real driving current data are acquired through a data acquisition system. From statistics, common faults are classified into coil wear, spool sticking and spring breakage. However, in complex environments, solenoid valve failure does not occur in a single form, but rather tends to occur in multiple failure forms. For possible multiple fault conditions, 8 multiple fault modes are set: normal, coil wear, spool sticking, spring break, coil wear + spool sticking, spring break + spool sticking, coil wear + spring break + spool sticking. Wherein, the normal state selects the electromagnetic valve which can work normally;
coil wear: using 30700to polish part of windings of the solenoid valve coil, damaging insulating paint, simulating the falling off of the insulating paint caused by the rising of the external environment and the temperature of the solenoid valve, and placing the polished coil in a normal solenoid valve;
spring breakage: the sheared spring is lifted to replace the spring of the normal valve, so that the spring fracture caused by metal fatigue of the spring due to long-term high-frequency actuation of the electromagnetic valve is simulated;
Valve core clamping stagnation: and (3) filling a certain amount of sand into the normal electromagnetic valve to simulate the valve core clamping stagnation caused by long-term accumulation of the fluid medium containing impurities.
Coil wear + spring break, coil wear + spool clamping, spring break + spool clamping and coil wear + spring break + spool clamping are all set up based on above-mentioned single fault condition, simulate the many fault conditions that the solenoid valve probably appears in long-term use respectively.
In the experiment, a voltage-stabilizing direct-current power supply respectively provides 24V direct-current voltage and 5V direct-current voltage for the electromagnetic valve and the current sensor; the electromagnetic valve is determined to be opened and closed by a control switch; and then the data acquisition software in the computer is used for interacting with the data acquisition card, and the driving current signal is acquired, and the current signal is standardized by the maximum and minimum, as shown in fig. 5. Wherein, the Sample packet (Sample) contains 160 samples, and 20 samples are sequentially set as a group of states, and the 8 states are respectively: normal (1-20), coil wear (21-40), spool sticking (41-60), spring break (61-80), coil wear + spring break (81-100), coil wear + spool sticking (101-120), spring break + spool sticking (121-140), coil wear + spring break + spool sticking (141-160).
Fourth step: and the driving current signal acquisition is realized by utilizing data acquisition software.
And (3) realizing driving current signal acquisition by using AD-log data acquisition software (Ling Hua technology) and storing the driving current signal into a folder.
Step3: and extracting the characteristics of the electromagnetic valve driving current signal by using a CEMMDAN algorithm, calculating the energy entropy of the electromagnetic valve driving current signal, and constructing a fault characteristic vector table of the electromagnetic valve.
Fault feature information extraction: the driving current signal is decomposed by using a CEEMDAN algorithm to obtain natural modal components of different frequency bands (note that the extreme points of the final residual signal are not more than two). On the basis, energy of different frequency bands of the driving current is extracted as fault characteristic information of the driving current, and a multi-fault coupling characteristic vector table of the electromagnetic valve is established.
1) The raw drive current data is normalized.
In order to eliminate the influence of the measurement unit and the magnitude thereof among the characteristic data, the maximum and minimum normalization is adopted to process the original driving current data, as shown in the formula 1:
wherein z is normalized data, x is a sample, x max For the maximum value of the sample, x min Is the sample minimum.
2) And decomposing the normalized data by using a CEEMDAN algorithm to obtain 7 natural modal components and a residual component.
The CEEMDAN was calculated as follows:
first, with a driving current signal as an original signal, an adaptive noise ζ compliant with a normal distribution is added to the original signal x (t) i (t) obtaining a mixed signal x i (t), wherein i represents the signal of the ith dimension, for x i (t) performing Empirical Mode Decomposition (EMD) decomposition N times and performing the same on N IMFs 1 i (t) averaging to obtain a first modal componentWherein N represents the empirical mode decomposition number, IMF 1 i (t) calculating a first residual signal R as an eigenmode component 1 (t)。
For R 1 (t)+ξ i (t)E 1 (ω i (t)) to obtain a second-order modal component wherein ωi (t) white noise following the N (0, 1) distribution, E 1 The (-) function is the IMF component of the 1 st order obtained by EMD decomposition, E 1 (ω i (t)) is an IMF component of the 1 st order obtained by EMD decomposition of white noise, and calculates a second residual signal R 2 (t)。
Repeating the steps until the extreme points of the residual signal are not more than two, and finally obtaining k-order modal components, wherein the original signal can be decomposed into:
among the results of the CEEMDAN decomposition of the drive current signal, the results of the CEEMDAN decomposition of the normal valve and the coil wear drive current are shown in fig. 6, the results of the CEEMDAN decomposition of the coil wear + the spring break and the coil wear + the coil wear drive current are shown in fig. 7, and the results of the CEEMDAN decomposition of the spring break + the coil wear + the spring break + the coil wear drive current are shown in fig. 8.
3) And calculating the energy entropy of the sample, and constructing an electromagnetic valve fault feature vector.
And (3) energy entropy calculation: after CEEMDAN decomposition, eigen-mode function (IMF) components of different frequency bands are obtained, the fault state of the electromagnetic valve is reflected more carefully, and the electromagnetic valve can be effectively diagnosed by extracting energy of different frequency bands of the driving current as fault characteristic information.
The total energy of each IMF of the solenoid valve is:
wherein ,Ek Represents the energy of the kth IMF, |x ki And I is the amplitude of the kth IMF component discrete point.
The feature vectors are expressed as:
X＝[E 1 ，E 2 ，…E m ] (8)
wherein X is a feature vector, E is a two-norm of the feature vector,and m is the number of samples, which is the normalized fault characteristic vector.
Table 1 shows normalized fault signature vectors, 7 modal components and 1 residual signal obtained by CEEMDAN, R k (t) is the final residual signal, also denoted RES, R 1 (t) corresponds to R 2 (t) raw signal, R 2 (t) is the residual signal and,the fault feature vector is then used for diagnostic purposes.
TABLE 1 energy entropy for each fault state
The LSSVM can better classify small samples, however, different parameters of the LSSVM have great influence on classification accuracy, parameter selection is a typical optimizing problem, and research on parameter optimizing technology is conducted below.
In a second aspect, the invention provides a multi-fault coupling diagnosis method for electromagnetic valves, which optimizes the parameter selection of LSSVM by using IPSO algorithm and performs fault diagnosis on fault feature vectors of the electromagnetic valves. The method comprises the following steps:
the first step: and improving a particle swarm algorithm by adopting a self-adaptive dynamic inertia weight coefficient and a Sine-Tent-Cosine chaotic mapping method.
Normalizing the feature vector to obtain a normalized fault feature vector with the amplitude between (0, 1). Firstly, an iteration mode of improving particles by adopting a self-adaptive dynamic inertia weight coefficient is adopted, and secondly, a fine-tone-Tent-Cosine chaotic mapping method is used for initializing a particle swarm to generate uniformly distributed particles.
The PSO algorithm firstly obtains a group of particles through population initialization, each particle takes the position and the speed as the characteristics thereof, the adaptability is taken as the feasible solution thereof, and the speed and the position of each particle are limited in a certain range. Through information interaction among particles, an individual optimal position Pbest can be obtained, and then a global optimal position Gbest can be obtained through updating the population optimal position by the Pbest. The specific PSO implementation steps are as follows:
assuming that the number of particle populations is N' dimensional, each population has M possible solutions, the position of particle j can be expressed as X (j) = (X) j1 ，x j2 ，…，x jN′), wherein xj1 ，x j2 ，…，x jN′ The velocity of particle j can be expressed as V (j) = (V) for the position information of particle j in different dimensions j1 ，v j2 ，…v jN′), wherein vj1 ，v j2 ，…v jN′ The individual optimal position Pbest may be expressed as pbest= (P) for velocity information of the particle j in different dimensions, respectively j1 ，P j2 ，…P jN′), wherein Pj1 ，P j2 ，…P jN′ The individual optimal position information of the particle j in different dimensions is respectively shown as a global optimal position gbest= (g) j1 ，g j2 ，…g jN′), wherein gj1 ，g j2 ，…g jN′ And the global optimal position information of the particles j in different dimensions is respectively obtained. The position and velocity update formula of particle j in the N' th dimension is:
wherein ,the speed of the particle j in the nth dimension of the t generation is given, and t is the current iteration number; w is a weight; c 1 、c 2 Are acceleration factors and are responsible for regulating the learning rate; r is (r) 1 、r 2 Are random values within (0, 1); pbest is the optimal position of the individual; gbest is the optimal position of the population; />Is the position information of the particle j in the nth dimension of the t-1 generation.
Simulating parameters to be solved into particles through a PSO algorithm, initializing a population to obtain different parameters which are randomly distributed, substituting the parameters into a loss function, reducing the loss function through continuous iteration, finding a local optimal solution, and finding a global optimal solution through repeated loop iteration. The PSO algorithm is applied to the least square support vector machine or the support vector machine, so that higher fitness can be obtained, the network structure is updated by using the optimized parameters, and the fault diagnosis rate of the classification algorithm is improved. The use of PSO algorithms to optimize parameters is well known to those skilled in the art and will not be described in detail. However, the PSO algorithm is easily trapped in a local optimum, resulting in a low accuracy of the fault diagnosis algorithm. The invention improves the inertia weight factor iteration mode in the PSO algorithm.
The inertia weight factor is used as an important index for influencing the PSO global searching capability and the local searching capability, and when the value of the inertia weight factor is large, the inertia weight factor is helpful to jump out of the local searching, but particles can not be converged to a smaller value; and when the inertia weight factor is smaller, quick convergence can be realized, but the global optimal target can not be converged. Therefore, the choice of inertial weighting factors will affect the performance of the algorithm to a large extent. Aiming at the problem that the weights of the local optimal solution and the global optimal solution of the particles in different populations are different, the invention adopts an iteration mode of improving the particles by adopting a self-adaptive dynamic inertia weight coefficient, and the expression is as follows:
wherein ,wmax For maximum weight, w min Is the minimum value of the weight; f. f (f) min 、f avg The algorithm fitness function, the algorithm fitness minimum value and the algorithm fitness average value are respectively adopted.
On the other hand, the quality of the initial population in the PSO algorithm is also an important factor affecting its performance. The chaotic system is a random, unpredictable and nonlinear system, and compared with a single low-dimensional chaotic algorithm, the composite chaotic algorithm has the advantages of low complexity and better uniformity, and the global search function can be realized by using the chaotic algorithm. The invention provides a chaotic mapping method (Sine-Tent-Cosine) for initializing a particle swarm to generate uniformly distributed particles and improving the quality of the particles. The Sine-Tent-Cosine chaotic mapping expression is as follows:
Wherein r epsilon (0, 1) is a random number, x (z) represents a z-th chaotic variable, x (1) epsilon (0, 1) is an initial variable and is a random number, and x (z) is continuously updated according to a formula, and finally the generated chaotic sequence is mapped into a solution space.
The invention combines the advantages of the above IPSO in convergence and global, so the invention applies the IPSO algorithm to the LSSVM to carry out fault diagnosis on the multi-fault coupling state of the electromagnetic valve.
And a second step of: and classifying the electromagnetic valve fault characteristic samples by using an IPSO-LSSVM algorithm to realize the multi-fault coupling diagnosis of the electromagnetic valve.
The found optimal solution is input into an LSSVM algorithm, 8 fault states of the electromagnetic valve are classified by utilizing a classification model in the LSSVM algorithm, and the mean square error and the root mean square error of a true value and a classification value are calculated.
The core idea of the invention is that: initializing characteristic parameters of particles in the population by using a Sine-Tent-Cosine chaotic mapping algorithm to obtain particles with higher quality; meanwhile, a self-adaptive dynamic inertia weight mode is adopted, so that the problem that the weight cannot be converged to the minimum value and falls into a locally optimal state easily caused by improper weight setting is solved; finally, optimizing key parameters affecting the classification performance of the LSSVM by utilizing an improved PSO algorithm, substituting the optimized key parameters into a network structure of the LSSVM to realize the optimal fault classification effect, and providing an IPSO-LSSVM fault diagnosis algorithm, wherein a flow chart of the algorithm is shown in figure 10.
Setting experimental parameters: the [160 multiplied by 8 ] is obtained by the data acquisition system and the characteristic extraction mode]The fault data of the electromagnetic valve are 8 kinds of electromagnetic valve characteristic vectors in different states, fault labels of the electromagnetic valve characteristic vectors are respectively set, and the data are randomly disturbed in a shuffling mode. Wherein the training set sample is [120×8 ]]Test set sample is [40×8 ]]. In parameter optimization, the number of particle groups is 100; learning factor C 1 ＝C 2 =2; the maximum value of the inertia weight is 0.9, and the minimum value of the inertia weight is 0.4; the iteration number is 300; regular parameter C max ＝300，C min =0.01; RBF kernel parameter sigma max ＝200，σ min ＝0.01；V＝25。
Evaluation of results: to evaluate the effectiveness of the methods presented herein, the present invention employs 5 fault diagnosis algorithms and the proposed methods for fault diagnosis of solenoid valve multiple fault coupling states, namely SVM, LSSVM, GWO-LSSVM, VGWO-LSSVM, PSO-LSSVM. The labels of various fault states such as normal state, coil abrasion state, valve core clamping stagnation state, spring breakage state, coil abrasion state, valve core clamping stagnation state, spring breakage state, valve core clamping stagnation state, coil abrasion state, spring breakage state, valve core clamping stagnation state and the like are respectively 1, 2, 3, 4, 5, 6, 7 and 8 numbers.
Table 2 shows that the classification accuracy of the IPSO-LSSVM is highest by analyzing the classification of the test samples by SVM, LSSVM, GWO-LSSVM, VGWO-LSSVM, PSO-LSSVM and IPSO-LSSVM. Compared with the SVM, the LSSVM has higher diagnosis precision through the comparison of the parameter setting, and the superiority of using the LSSVM for fault diagnosis is verified. From the specific classification, the SVM is superior to the SVM in terms of coil abrasion, spring breakage and valve core clamping stagnation diagnosis errors, but the LSSVM-based method can accurately diagnose the valve core clamping stagnation fault state, and only the coil abrasion and spring breakage diagnosis errors. . Meanwhile, compared with other methods, the method based on the IPSO-LSSVM has fewer coil abrasion and spring breakage diagnosis errors. According to the analysis of the data characteristics, the coil abrasion and spring fracture fault data sets are high in similarity, so that diagnosis errors are easy to occur, key hidden characteristic parameters of the coil abrasion and spring fracture fault data sets are obtained through deep analysis in subsequent work, and the comprehensive fault characteristics are utilized to diagnose the fault state of the electromagnetic valve.
Table 2 comparison of multiple fault diagnostic models for solenoid valves
/>
Claims (5)
1. The method for establishing the multi-fault coupling characteristic vector of the electromagnetic valve is characterized by comprising the following steps of:
step1: feasibility analysis of drive current signal as solenoid valve multi-fault coupling state representation
The solenoid valve has 5 stages in its operation: the actuation touch, actuation motion, power-on hold, release touch and release motion stages; four stages in the working process of the electromagnetic valve: the method comprises the steps of carrying out analysis and research on the phases of actuation touch, actuation movement, power-on holding, release touch and the like, and checking whether a driving current signal is related to a fault state of the electromagnetic valve;
before the actuation and touch stage, the electromagnetic valve is not electrified, the driving current is 0A, and the electromagnetic valve does not work;
when the electromagnetic valve is electrified, the electromagnetic valve is in a suction touch stage, rising current is generated in the loop, and the magnetic flux of the coil is changed due to the self-inductance phenomenon of the coil in the electromagnetic circuit, so that electromotive force is generated, and abrupt change of current cannot be prevented; therefore, the current cannot directly reach the steady-state current, but rises at a higher speed, and at the moment, the electromagnetic force generated by the electromotive force cannot overcome the friction force and the spring tension, and the valve core does not displace;
When the driving current is increased to a first threshold value, the electromotive force generated by the coil causes the electromagnetic force to be larger than the friction force between the valve core and fuel and the tensile force of the spring, and the valve core starts to move towards the maximum displacement under the action of the combined force; the motion of the valve core generates dynamic electromotive force, so that the driving current is continuously reduced, when the driving current is reduced to a second threshold value, the valve core of the electromagnetic valve reaches the maximum displacement position and no displacement is generated, and the dynamic electromotive force at the moment disappears;
when the electromotive force disappears, the current increases at the current increasing speed at the moment of actuation touch, but the increasing speed of the driving current is slowed down along with the change of the induced electromotive force, but the loop current is still increased, and finally the stable value is reached and does not change any more;
the release stage: when the electromagnetic valve is in a de-energized state, the current in the loop is rapidly reduced, meanwhile, the valve core of the electromagnetic valve is reset due to the fact that the electromagnetic force at the moment of turn-off is smaller than the friction force and the spring tension force, the generated dynamic electromotive force reduces the reduction speed of the driving current, but as the valve core reaches the original position, the dynamic electromotive force disappears, and at the moment, the driving current is rapidly reduced; thus, during the release phase, the drive current decreases at a rate of change from small to large;
Step2: designing a solenoid valve driving current signal data acquisition system, setting a solenoid valve fault mode, and carrying out data acquisition
Based on the electromagnetic valve circuit, electromagnetic and mechanical analysis, the driving current is closely related to the displacement of the valve core of the electromagnetic valve; in order to verify that the electromagnetic valve driving current contains fault information, 8 multi-fault modes of an electromagnetic valve of a certain power system are simulated, and an electromagnetic valve driving current data acquisition system is designed; the method comprises the following steps:
the first step: design data acquisition system
The experimental equipment comprises a computer, an electromagnetic valve, a data acquisition card PCI-9114, a current sensor LTS 25-NP, a diode FR307 and a 5V voltage-stabilizing direct current power supply GPS-3303C, and a 24V voltage-stabilizing direct current power supply RIGOL-DP712;
and a second step of: connecting experimental equipment according to a circuit diagram
The computer is internally provided with a data acquisition card PCI-9114; the 37 port of the data acquisition card PCI-9114 is connected with the port 7 of the current sensor LTS 25-NP; the port 19 of the data acquisition card PCI-9114 is connected with the port 8 of the current sensor LTS 25-NP; a 5V stabilized direct current power supply is connected between the 19 port of the data acquisition card PCI-9114 and the 9 port of the current sensor LTS 25-NP, the positive electrode of the 5V stabilized direct current power supply is connected with the 9 port of the current sensor LTS 25-NP, and the negative electrode of the 5V stabilized direct current power supply is connected with the 19 port of the data acquisition card PCI-9114; ports 2 and 6 of the current sensor LTS 25-NP are directly connected, and ports 3 and 5 are directly connected; the electromagnetic valve is connected with a rectifying diode FR307 in parallel, the positive end of the rectifying diode FR307 is connected with a port 1 of a current sensor LTS 25-NP, and the negative end of the rectifying diode FR307 is connected with a switch model through a toggle switch: KN1-202 is connected with the positive electrode of 24V stabilized DC power supply RIGOL-DP712, and the negative electrode of 24V stabilized DC power supply RIGOL-DP712 is connected with port 4 of current sensor LTS 25-NP;
And a third step of: setting different fault modes for electromagnetic valves
In order to verify that the driving current of the electromagnetic valve can be used for performing fault diagnosis of the electromagnetic valve, the electromagnetic valve fault caused by long-term high-frequency actuation and external environment is simulated through setting a fault expression form, and real driving current data are acquired through a data acquisition system; setting 8 multi-fault modes: normal, coil wear, spool sticking, spring break, coil wear + spool sticking, spring break + spool sticking, coil wear + spring break + spool sticking; wherein, the normal state selects the electromagnetic valve which can work normally;
in the experiment, the voltage-stabilizing direct-current power supply respectively provides direct-current voltage for the electromagnetic valve and the current sensor; the electromagnetic valve is determined to be opened and closed by a control switch; the data acquisition software in the computer is used for interacting with the data acquisition card, acquiring a driving current signal, and carrying out maximum and minimum standardization processing on the current signal;
fourth step: driving current signal acquisition using data acquisition software
The data acquisition software is used for realizing the acquisition of the driving current signals and storing the driving current signals into a folder;
step3: feature extraction is carried out on electromagnetic valve driving current signals by using CEMMDAN algorithm, energy entropy is calculated, and a fault feature vector table of the electromagnetic valve is constructed
Decomposing the driving current signal through CEEMDAN algorithm to obtain natural modal components of different frequency bands, wherein the extreme points of the final residual signal are not more than two; on the basis, energy of different frequency bands of the driving current is extracted as fault characteristic information of the driving current, and a multi-fault coupling characteristic vector table of the electromagnetic valve is established;
1) Normalizing raw drive current data
In order to eliminate the influence of the measurement unit and the magnitude thereof among the characteristic data, the maximum and minimum normalization is adopted to process the original driving current data, as shown in the formula 1:
wherein z is normalized data, x is a sample, x max For the maximum value of the sample, x min Is the sample minimum;
2) Decomposing the normalized data by CEEMDAN algorithm to obtain 7 natural modal components and a residual component
The CEEMDAN was calculated as follows:
first, with a driving current signal as an original signal, an adaptive noise ζ compliant with a normal distribution is added to the original signal x (t) i (t) obtaining a mixed signal x i (t), wherein i represents the signal of the ith dimension, for x i (t) EMD decomposition was performed N times and N IMFs were subjected to 1 i (t) averaging to obtain a first modal componentWherein N represents the empirical mode decomposition number, IMF 1 i (t) calculating a first residual signal R as an eigenmode component 1 (t)；
For R 1 (t)+ξ i (t)E 1 (ω i (t)) to obtain second-order modal components wherein ωi (t) white noise following the N (0, 1) distribution, E 1 The (-) function is the IMF component of the 1 st order obtained by EMD decomposition, E 1 (ω i (t)) is the 1 st order obtained by EMD decomposition of white noiseAn IMF component and calculates a second residual signal R 2 (t)；
Repeating the steps until the extreme points of the residual signals are not more than two, obtaining k-order modal components, and decomposing the original signals into:
3) Calculating the energy entropy of a sample, and constructing an electromagnetic valve fault characteristic vector
After CEEMDAN decomposition, eigen-mode function IMF components of different frequency bands are obtained, the fault state of the electromagnetic valve is reflected more carefully, and the electromagnetic valve can be effectively subjected to fault diagnosis by extracting energy of different frequency bands of the driving current as fault characteristic information;
the total energy of each IMF of the solenoid valve is:
wherein ,Ek Represents the energy of the kth IMF, |x ki The I is the amplitude of the kth IMF component discrete point;
the feature vectors are expressed as:
X＝[E 1 ，E 2 ，…E m ] (8)
wherein X is a feature vector, E is a two-norm of the feature vector,the m is the number of samples and is the normalized fault characteristic vector; thereby obtaining a normalized fault feature vector table; obtaining 7 modal components and 1 residual signal by CEEMDAN, R k (t) is the final residual signal, R 1 (t) corresponds to R 2 (t) raw signal, R 2 (t) is the residual signal and,
2. the method for establishing the multi-fault-coupling feature vector of the electromagnetic valve according to claim 1, wherein the 2W160Y32B-DC24V type direct-acting electromagnetic valve is selected as a study object, wherein the first threshold value=0.58A, the second threshold value=0.25A, and the stable value=0.85A.
3. The method for establishing a multi-fault-coupling feature vector of a solenoid valve as claimed in claim 1, wherein among 8 multi-fault modes,
coil wear: using 30700to polish part of windings of the solenoid valve coil, damaging insulating paint, simulating the falling off of the insulating paint caused by the rising of the external environment and the temperature of the solenoid valve, and placing the polished coil in a normal solenoid valve;
spring breakage: the sheared spring is lifted to replace the spring of the normal valve, so that the spring fracture caused by metal fatigue of the spring due to long-term high-frequency actuation of the electromagnetic valve is simulated;
valve core clamping stagnation: a certain amount of sand is filled into a normal electromagnetic valve to simulate valve core clamping stagnation caused by long-term accumulation of a fluid medium containing impurities;
coil wear + spring break, coil wear + spool clamping, spring break + spool clamping and coil wear + spring break + spool clamping are all set up based on above-mentioned single fault condition, simulate the many fault conditions that the solenoid valve probably appears in long-term use respectively.
4. A multi-fault coupling diagnosis method for electromagnetic valves is characterized by comprising the following steps:
the first step: improved particle swarm algorithm by adopting self-adaptive dynamic inertia weight coefficient and Sine-Tent-Cosine chaotic mapping method
The PSO algorithm comprises the following implementation steps:
assuming that the number of particle populations is N' dimensional, each population has M feasible solutions, and the position of particle j is represented as X (j) = (X) j1 ，x j2 ，…，x jN′), wherein xj1 ，x j2 ，…，x jN′ The position information of the particle j in different dimensions is respectively shown, and the velocity of the particle j is expressed as V (j) = (V) j1 ，v j2 ，…v jN′), wherein vj1 ，v j2 ，…v jN′ The individual optimal position Pbest is denoted pbest= (P) for velocity information of the particle j in different dimensions, respectively j1 ，P j2 ，…P jN′), wherein Pj1 ，P j2 ，…P jN′ The individual optimal position information of the particle j in different dimensions is respectively shown as a global optimal position gbest= (g) j1 ，g j2 ，…g jN′), wherein gj1 ，g j2 ，…g jN′ Global optimal position information of the particles j in different dimensions is respectively obtained; the position and velocity update formula of particle j in the N' th dimension is:
wherein ,the speed of the particle j in the nth dimension of the t generation is given, and t is the current iteration number; w isA weight; c 1 、c 2 Are acceleration factors and are responsible for regulating the learning rate; r is (r) 1 、r 2 Are random values within (0, 1); pbest is the optimal position of the individual; gbest is the optimal position of the population; />Position information of the particle j in the nth dimension of the t-1 generation;
An iterative mode of improving particles by adopting an adaptive dynamic inertia weight coefficient is adopted, and the expression is as follows:
wherein ,wmax For maximum weight, w min Is the minimum value of the weight; f. f (f) min 、f avg The algorithm fitness function, the algorithm fitness minimum value and the algorithm fitness average value are respectively adopted.
Initializing a particle swarm by adopting a chaotic mapping method, generating uniformly distributed particles, and improving the quality of the particles; the Sine-Tent-Cosine chaotic mapping expression is as follows:
wherein r epsilon (0, 1) is a random number, x (z) represents a z-th chaotic variable, x (1) epsilon (0, 1) is an initial variable and is a random number, x (z) is continuously updated according to a formula, and finally, the generated chaotic sequence is mapped into a solution space;
and a second step of: classifying solenoid valve fault characteristic samples by using IPSO-LSSVM algorithm to realize solenoid valve multi-fault coupling diagnosis
The found optimal solution is input into an LSSVM algorithm, 8 fault states of the electromagnetic valve are classified by utilizing a classification model in the LSSVM algorithm, and the mean square error and the root mean square error of the true value and the classification value are calculated.
5. The method of claim 4, wherein the 8 multiple failure modes are: normal, coil wear, spool sticking, spring break, coil wear + spool sticking, spring break + spool sticking, coil wear + spring break + spool sticking; wherein, normal state selects the solenoid valve that can normally work.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310537798.2A CN116626479A (en) | 2023-05-12 | 2023-05-12 | Multi-fault coupling diagnosis method for electromagnetic valve |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310537798.2A CN116626479A (en) | 2023-05-12 | 2023-05-12 | Multi-fault coupling diagnosis method for electromagnetic valve |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN116626479A true CN116626479A (en) | 2023-08-22 |
Family
ID=87612593
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202310537798.2A Pending CN116626479A (en) | 2023-05-12 | 2023-05-12 | Multi-fault coupling diagnosis method for electromagnetic valve |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN116626479A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117687394A (en) * | 2024-01-27 | 2024-03-12 | 南京德克威尔自动化有限公司 | Solenoid valve island control signal verification method and system |
-
2023
- 2023-05-12 CN CN202310537798.2A patent/CN116626479A/en active Pending
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117687394A (en) * | 2024-01-27 | 2024-03-12 | 南京德克威尔自动化有限公司 | Solenoid valve island control signal verification method and system |
| CN117687394B (en) * | 2024-01-27 | 2024-04-16 | 南京德克威尔自动化有限公司 | Solenoid valve island control signal verification method and system |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN106980822B (en) | A kind of rotary machinery fault diagnosis method based on selective ensemble study | |
| Xiao et al. | Fault diagnosis of asynchronous motors based on LSTM neural network | |
| CN109187025A (en) | A kind of rolling bearing remaining life prediction technique of integrated KELM | |
| CN116626479A (en) | Multi-fault coupling diagnosis method for electromagnetic valve | |
| CN109284672A (en) | A kind of Mechanical Failure of HV Circuit Breaker diagnostic method based on PSO-Kmeans algorithm | |
| CN110118928B (en) | Breaker fault diagnosis method based on error inverse propagation algorithm | |
| Wang et al. | Few-shot transfer learning with attention mechanism for high-voltage circuit breaker fault diagnosis | |
| CN112345213A (en) | Low-voltage direct-current circuit breaker mechanical fault diagnosis method | |
| CN112800682A (en) | Feedback optimization fan blade fault monitoring method | |
| CN109187022A (en) | A kind of Fault Diagnosis of Roller Bearings based on accidental resonance and autocoder | |
| CN110703077A (en) | HPSO-TSVM-based high-voltage circuit breaker fault diagnosis method | |
| CN115791174B (en) | Rolling bearing abnormality diagnosis method, system, electronic equipment and storage medium | |
| Kuo | Artificial recognition system for defective types of transformers by acoustic emission | |
| CN115600088A (en) | Distribution transformer fault diagnosis method based on vibration signals | |
| CN113255458A (en) | Bearing fault diagnosis method based on multi-view associated feature learning | |
| Wong et al. | Fault diagnosis of induction motors under untrained loads with a feature adaptation and improved broad learning framework | |
| Bessam et al. | A novel method for induction motors stator inter-turn short circuit fault diagnosis based on wavelet energy and neural network | |
| Ji et al. | Typical current modelling and feature extraction of high voltage circuit breaker towards condition analysis and fault diagnosis | |
| CN115113038B (en) | Circuit breaker fault diagnosis method based on current signal phase space reconstruction | |
| CN115456106A (en) | High-voltage circuit breaker fault diagnosis model optimization method | |
| CN105116323A (en) | Motor fault detection method based on RBF | |
| Liu et al. | A fast and accurate piezoelectric actuator modeling method based on truncated least squares support vector regression | |
| CN109116833A (en) | Based on improvement drosophila-bat algorithm mechanical failure diagnostic method | |
| CN115754713A (en) | Motor fault diagnosis method based on frequency diagram and EFA-BP deep learning | |
| Zhang et al. | fault diagnosis of servo drive system of CNC machine based on deep learning |
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 |