CN108804720B - Oil pumping machine fault diagnosis method based on improved traceless Kalman filtering and RBF neural network - Google Patents
Oil pumping machine fault diagnosis method based on improved traceless Kalman filtering and RBF neural network Download PDFInfo
- Publication number
- CN108804720B CN108804720B CN201710282712.0A CN201710282712A CN108804720B CN 108804720 B CN108804720 B CN 108804720B CN 201710282712 A CN201710282712 A CN 201710282712A CN 108804720 B CN108804720 B CN 108804720B
- Authority
- CN
- China
- Prior art keywords
- pumping unit
- neural network
- improved
- parameters
- indicator diagram
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
- 238000005086 pumping Methods 0.000 title claims abstract description 66
- 238000003745 diagnosis Methods 0.000 title claims abstract description 45
- 238000013528 artificial neural network Methods 0.000 title claims abstract description 36
- 238000000034 method Methods 0.000 title claims abstract description 33
- 238000001914 filtration Methods 0.000 title claims abstract description 17
- 238000004422 calculation algorithm Methods 0.000 claims abstract description 34
- 238000003062 neural network model Methods 0.000 claims abstract description 8
- 239000011159 matrix material Substances 0.000 claims description 36
- 238000010586 diagram Methods 0.000 claims description 30
- 210000002569 neuron Anatomy 0.000 claims description 19
- 238000005070 sampling Methods 0.000 claims description 11
- 238000012549 training Methods 0.000 claims description 11
- 238000004519 manufacturing process Methods 0.000 claims description 9
- 150000001875 compounds Chemical class 0.000 claims description 6
- 230000009466 transformation Effects 0.000 claims description 6
- 239000007788 liquid Substances 0.000 claims description 5
- 239000003129 oil well Substances 0.000 claims description 5
- 238000009826 distribution Methods 0.000 claims description 3
- 239000003960 organic solvent Substances 0.000 claims description 3
- 239000002245 particle Substances 0.000 claims description 3
- 239000004576 sand Substances 0.000 claims description 3
- 238000001228 spectrum Methods 0.000 claims description 3
- 238000005507 spraying Methods 0.000 claims description 3
- 230000004913 activation Effects 0.000 claims description 2
- 230000008021 deposition Effects 0.000 claims description 2
- 238000007667 floating Methods 0.000 claims description 2
- 238000000926 separation method Methods 0.000 claims description 2
- 230000000694 effects Effects 0.000 abstract description 8
- 238000011897 real-time detection Methods 0.000 abstract 1
- 230000008569 process Effects 0.000 description 9
- 230000003068 static effect Effects 0.000 description 6
- 238000004458 analytical method Methods 0.000 description 3
- 230000008901 benefit Effects 0.000 description 3
- 238000013507 mapping Methods 0.000 description 3
- 238000013473 artificial intelligence Methods 0.000 description 2
- 238000006243 chemical reaction Methods 0.000 description 2
- 238000013461 design Methods 0.000 description 2
- 238000005516 engineering process Methods 0.000 description 2
- 238000000605 extraction Methods 0.000 description 2
- 230000006870 function Effects 0.000 description 2
- 238000005312 nonlinear dynamic Methods 0.000 description 2
- 238000005457 optimization Methods 0.000 description 2
- 238000004088 simulation Methods 0.000 description 2
- 238000007792 addition Methods 0.000 description 1
- 238000004364 calculation method Methods 0.000 description 1
- 238000005094 computer simulation Methods 0.000 description 1
- 238000012937 correction Methods 0.000 description 1
- 230000008878 coupling Effects 0.000 description 1
- 238000010168 coupling process Methods 0.000 description 1
- 238000005859 coupling reaction Methods 0.000 description 1
- 230000007812 deficiency Effects 0.000 description 1
- 238000001514 detection method Methods 0.000 description 1
- 230000036541 health Effects 0.000 description 1
- 238000012423 maintenance Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 230000001537 neural effect Effects 0.000 description 1
- 238000012545 processing Methods 0.000 description 1
- 238000004451 qualitative analysis Methods 0.000 description 1
- 239000009671 shengli Substances 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
- 238000012360 testing method Methods 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F30/00—Computer-aided design [CAD]
- G06F30/20—Design optimisation, verification or simulation
-
- E—FIXED CONSTRUCTIONS
- E21—EARTH OR ROCK DRILLING; MINING
- E21B—EARTH OR ROCK DRILLING; OBTAINING OIL, GAS, WATER, SOLUBLE OR MELTABLE MATERIALS OR A SLURRY OF MINERALS FROM WELLS
- E21B47/00—Survey of boreholes or wells
- E21B47/008—Monitoring of down-hole pump systems, e.g. for the detection of "pumped-off" conditions
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2218/00—Aspects of pattern recognition specially adapted for signal processing
- G06F2218/02—Preprocessing
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Life Sciences & Earth Sciences (AREA)
- General Physics & Mathematics (AREA)
- Evolutionary Computation (AREA)
- Geology (AREA)
- Mining & Mineral Resources (AREA)
- General Engineering & Computer Science (AREA)
- Health & Medical Sciences (AREA)
- Molecular Biology (AREA)
- Artificial Intelligence (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- Data Mining & Analysis (AREA)
- Geochemistry & Mineralogy (AREA)
- General Health & Medical Sciences (AREA)
- Fluid Mechanics (AREA)
- Computing Systems (AREA)
- General Life Sciences & Earth Sciences (AREA)
- Environmental & Geological Engineering (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Geophysics (AREA)
- Computer Hardware Design (AREA)
- Geometry (AREA)
- Testing Of Devices, Machine Parts, Or Other Structures Thereof (AREA)
Abstract
The invention provides an improved traceless Kalman filtering and RBF neural network-based oil pumping unit fault diagnosis method. Firstly, modeling decision parameters by using an RBF neural network, then updating the weight, center and width of a hidden layer of the neural network model in real time by using an improved unscented Kalman filtering algorithm to obtain the optimal parameters of the neural network, and establishing the oil pumping unit fault diagnosis method based on the combination of the improved unscented Kalman filtering and the RBF neural network. The invention has the following remarkable effects: the accuracy rate of fault diagnosis is improved, and the purpose of real-time detection of the operation condition of the pumping unit is really achieved.
Description
Technical Field
The invention relates to a fault diagnosis technology of an oil pumping unit, in particular to a fault diagnosis method of an oil pumping unit based on improved traceless Kalman filtering and RBF neural network.
Background
At present, people mainly judge the fault of the pumping unit manually according to an indicator diagram, and can only carry out qualitative analysis, the diagnosis result is influenced by the aspects of expert experience, technology and the like, and the diagnosis has certain hysteresis, so that the real-time accurate diagnosis cannot be achieved. The operation process of the oil pumping unit has the characteristics of nonlinearity and strong coupling, and great difficulty is brought to fault diagnosis. The RBF neural network has strong nonlinear mapping capability, is suitable for solving the problem of nonlinear system modeling, and provides a new idea for process modeling of the scheme. According to the method, an artificial intelligence method is adopted, Fourier transform processing is carried out on indicator diagram parameters collected by the oil pumping unit, the current parameters of the oil pumping unit are combined, an RBF neural network is used for building an oil pumping unit fault diagnosis model, a UKF algorithm is used for optimizing the built diagnosis model, and optimal model parameters are obtained. In practical application, after collected data are preprocessed, output obtained after RBF neural network mapping is compared with model output, and then the fault type of the oil pumping unit can be judged. The method solves the problem that the judgment error is possibly caused by only intuitively judging the health for a long time, improves the accuracy and efficiency of fault diagnosis, reduces the randomness and uncertainty, really achieves the purpose of real-time diagnosis of the pumping unit, provides a new idea for solving similar problems, and embodies the powerful use of an artificial intelligence algorithm in the industry.
Disclosure of Invention
The application provides a fault diagnosis method of an oil pumping unit based on improved traceless Kalman filtering and RBF neural network, so as to solve the technical problem that in the prior art, when a fault occurs in the operation process of the oil pumping unit, the fault of the oil pumping unit cannot be detected in time, so that the optimal maintenance period is missed.
In order to solve the technical problems, the application adopts the following technical scheme:
a fault diagnosis method of an oil pumping unit based on improved Kalman filtering and RBF neural network is characterized by comprising the following steps:
s1: selecting a complete indicator diagram in one stroke of the pumping unit, carrying out Fourier transform on the indicator diagram, selecting the first f indicator diagram coordinate parameters of a low-frequency part, and combining three-phase current parameters b1, b2 and b3 of the pumping unit to form a decision variable X (a 1, a2, …, af, b1, b2 and b 3), wherein f is the number of the selected indicator diagram coordinate parameters;
s2: at the production site of the pumping unit, at least one group of decision variables X ═ a1, a2, …, af, b1, b2 and b3 is selected]As sample data, the decision variable X or X is output1～XiCorresponding fault type Y or Y1～Yl；
Using RBF neural network to determine the collected decision variable X or X1～XiTraining and checking are carried out, so that a fault diagnosis model of the oil pumping unit is established;
s3: improving the traditional unscented Kalman algorithm by utilizing spherical unscented transformation, and establishing an improved unscented Kalman algorithm, namely a CUKF algorithm;
for the traditional traceless Kalman algorithm, 2n +1 Sigma points are calculated when UT transformation is carried out, n refers to the dimension of a state, in the improved simplex sampling, the number of the Sigma points is n +2, the Sigma points are in an improved state on the spatial distribution, and the determination mode of the Sigma points is as follows:
(1) randomly selecting omega with 0 ≤0≤1；
(2) Calculating the weight value of the corresponding Sigma point as:
(3) when the state is 1-dimensional, the initialization vector sequence is:
(4) when the input dimension j is 2,3, …, n, the iterative formula is:
in the formula, the first and second organic solvents are,
is the ith particle point of the jth dimension;
(5) the mean and covariance of the system state v added to the generated Sigma points are:
wherein
Is the mean value of the state variable, PxIs the covariance matrix of the state vector, it can be seen from the above sampling algorithm that other sampling points except the origin have the same weight and are all located in the improved half radius
The above step (1);
the number of Sigma sampling points during UT conversion can be reduced by utilizing spherical conversion, the running time of the traditional traceless Kalman algorithm is shortened, and the speed of fault diagnosis is increased;
s4: optimizing the RBF neural network model obtained in the step S2 by using the CUKF algorithm in the step S3 to obtain a group of optimal parameters;
the CUKF algorithm is used for optimizing the static RBF neural network model, so that the influence of RBF initial random value on the model precision can be avoided, dynamic evolution modeling is realized, and the real-time diagnosis of the oil pumping unit is carried out;
s5: and modeling and diagnosing the fault of the pumping unit selected in the step S2 according to the optimal model obtained in the step S4, so that the fault diagnosis purpose is achieved.
When one decision variable is selected in step S2: selecting a complete indicator diagram in one stroke of the pumping unit, carrying out Fourier transform on the indicator diagram, selecting the first 8 indicator diagram coordinate parameters of a low-frequency part, and combining three-phase current parameters b1, b2 and b3 of the pumping unit to form decision variables X (a 1, a2, …, a8, b1, b2 and b 3), inputting 1 group of decision variables X, and outputting a fault type Y corresponding to the group of decision variables X;
in step S2, when 12 decision variables are selected: in the pumping unitSelecting 12 decision variables X in production site1,X2,...,X12And the corresponding failure types of insufficient liquid supply, oil well sand production, gas influence, airlock, fixed valve leakage, traveling valve leakage, double valve leakage, broken and separated sucker rod, bump on pump, bump under pump and continuous pumping and strip spraying are used as data samples and input into n groups of decision variables X1～XiThe output is n groups of decision variables X1～XiCorresponding fault type Y1～Yl；1<n is less than or equal to 12, i is 12;
the RBF neural network in the step S2 is composed of an input layer, a hidden layer and an output layer;
for the fault diagnosis model of the pumping unit, the network structure is that A-B-C is an input layer, B is a hidden layer, C is an output layer, the activation function adopts a Sigmod function, and the iteration number during sample training is 800.
In the spherical traceless transform in step S3, the UT transform samples are n + 2.
The CUKF-RBF algorithm in step S5 includes the steps of:
wherein, the CUKF algorithm part is as follows:
s511: initializing system parameters;
s512: calculating a Sigma point state vector;
s513: performing one-step prediction of system state and covariance matrix;
s514: calculating system observation and covariance matrixes;
s515: calculating a Kalman gain;
s516: updating a system state estimation matrix and a covariance matrix;
in the formula (I), the compound is shown in the specification,
the matrix is estimated for the system state at time k-1,
is a Kalman gain matrix, Y (k | k-1) is a system observation matrix at the moment k-1,
a prediction matrix is observed for the system at the time k-1;
in the formula (I), the compound is shown in the specification,
a matrix covariance matrix is estimated for the system at time k-1,
a covariance matrix of a system observation matrix at the moment of k-1;
the RBF algorithm part is as follows:
s521: updating the output of the RBF hidden layer:
wherein m is hidden layer neuron, total J,
is hidden layer neuron output, cmIs the center of the hidden layer neuron, σmWidth of hidden layer neurons;
s522: computing RBF output layer outputs
In the formula, ylFor network output layer output, omegam,lConnecting the weight from the hidden layer to the output layer after updating;
compared with the prior art, the technical scheme that this application provided, the technological effect or advantage that have are: when the oil pumping unit has a fault, the method can quickly diagnose and identify the fault, implement the fault diagnosis and improve the oil extraction efficiency.
Drawings
FIG. 1 is a flow chart of the present invention;
FIG. 2 is a frequency spectrum diagram of the present invention after Fourier transform;
FIG. 3 is a graph of the output of a static RBF diagnostic model;
FIG. 4 is a graph of static RBF diagnostic error;
FIG. 5 is a graph of the output of the CUKF-RBF diagnostic model;
FIG. 6 is a CUKF-RBF diagnostic error plot;
FIG. 7 is a simulation diagram of RBFNN in multi-fault diagnosis, wherein the output 1 is a normal sample, 2 is insufficient liquid supply, 3 is gas influence, and 4 is fixed valve leakage;
FIG. 8 is a simulation diagram of CUKF-RBF in multi-fault diagnosis, with output 1 being a normal sample, 2 being a liquid supply deficiency, 3 being a gas influence, and 4 being a fixed valve leak.
Detailed Description
The embodiment of the application provides a method for diagnosing the fault of the oil pumping unit based on the improved unscented Kalman filtering and RBF neural network, and by referring to the prior art means, the technical scheme provided by the application has the following technical effects or advantages: the method adopts an intelligent algorithm for fault diagnosis of the oil pumping unit, effectively improves the diagnosis efficiency and really achieves the purpose of implementing the fault diagnosis of the oil pumping unit.
In order to better understand the technical solutions, the technical solutions will be described in detail below with reference to the drawings and specific embodiments.
As shown in fig. 1, a method for diagnosing a fault of a pumping unit based on an improved traceless Kalman filter and an RBF neural network includes the following steps:
as shown in fig. 2,3, 4, 5, and 6, S1: when a set of decision variables is selected: selecting a complete indicator diagram in one stroke of the pumping unit, carrying out Fourier transform on the indicator diagram, selecting the first 8 indicator diagram coordinate parameters of a low-frequency part, and combining three-phase current parameters b1, b2 and b3 of the pumping unit to form decision variables X (a 1, a2, …, a8, b1, b2 and b 3), inputting 1 group of decision variables X, and outputting a fault type Y corresponding to the group of decision variables X;
when 12 sets of decision variables are selected: in the production field of the oil pumping unit, 12 groups of decision variables X are selected1,X2,...,X12And the corresponding failure types of insufficient liquid supply, sand production of an oil well, wax deposition of the oil well, gas influence, air lock, fixed valve loss, floating valve loss, double valve loss, breakage and separation of a sucker rod, collision on a pump, collision under the pump and continuous pumping and strip spraying are used as data samples, and n groups of decision variables X are input into the data samples1～XiThe output is n groups of decision variables X1～XiCorresponding fault type Y1～Yl；1<n is less than or equal to 12, i is 12;
s2: using RBF neural network to determine the collected decision variable X or X1～XiTraining and checking are carried out, so that a fault diagnosis model of the oil pumping unit is established;
in this embodiment, 800 groups of data of the operation of the pumping unit in the Dongxin oil production plant in the Shengli oil field are collected, wherein 740 groups of data are used as a modeling training sample, 60 groups of data are used as a test sample, the output of the pumping unit in normal operation is 1, the output of the pumping unit in failure is 2, and the failure is detected when the prediction error is greater than 0.5.
Data samples are shown in table 1 below;
TABLE 1 data samples
In the design of the neural network, the number of hidden layer nodes is the key for determining the quality of a neural network model and is also a difficult point in the design of the neural network, and the number of hidden layer nodes is determined by adopting a trial and error method;
in the formula, h is the number of hidden layer neuron nodes, q is the number of input layer neurons, e is the number of output layer neurons, r is a constant between 1 and 10, and the setting parameters of the RBF neural network in the embodiment are shown in the following table 2;
TABLE 2 RBF neural setup parameters
The training process of the neural network is mainly carried out according to the following steps:
set up Xk＝[xk1,xk2,…,xkM](k-1, 2, …, T) is the input vector, T is the number of training samples,
is a weight vector between the hidden layer M and the output I at the g-th iteration, yn(l)＝[yk1(l),yk2(l),…,ykP(l)](k-1, 2, …, T) is the actual output of the network at the g-th iteration, dk＝[dk1,dk2,…,dkP](k ═ 1,2, …, T) is the desired output;
the step S2 of establishing the model of the fault diagnosis of the pumping unit specifically includes the following steps:
s21: initializing, setting the initial value of the iteration times g as 0, and assigning WMI(0) A random value in the interval (0, 1);
s22: random input sample Xk；
S23: for input sample XkCalculating an input signal and an output signal of each layer of neuron of the RBF neural network in a forward direction;
s24: output d according to desirekAnd the actual output Yk(l) Calculating error E (l);
s25: judging whether the error E (l) meets the requirement, if not, entering the step S26, and if so, entering the step S29;
s26: judging whether the iteration number g +1 is greater than the maximum iteration number, if so, entering a step S29, otherwise, entering a step S27;
s27: for input sample XkCalculating each layer backwardsLocal gradient δ of neurons;
s28: calculating the weight correction quantity delta W and correcting the weight, wherein the calculation formula is as follows:
in the formula, eta is learning efficiency; let g be g +1, go to step S23;
s29: judging whether all training samples are finished, if so, finishing modeling, and otherwise, continuing to step S22;
through the above process, the prediction effect of the RBF neural network can be obtained as shown in fig. 3, and the error is shown in fig. 4. As can be seen from the analysis of fig. 3 and 4, most of the static prediction models established by the traditional RBF neural network training satisfy the requirements of fault detection, preliminary modeling requirements, and optimization adjustment;
s3: improving the traditional unscented Kalman algorithm by utilizing spherical unscented transformation, and establishing an improved unscented Kalman algorithm, namely a CUKF algorithm;
the spherical unscented transform in step S3:
for the traditional traceless Kalman algorithm, 2n +1 Sigma points are calculated when UT transformation is carried out, n refers to the dimension of a state, namely the number of RBF parameters to be optimized, in the improved simplex sampling, the number of the Sigma points is n +2, the Sigma points are in an improved shape in spatial distribution, and the determination mode of the Sigma points is as follows:
s311 randomly selecting 0 to be not less than omega0≤1；
S312, calculating the weight value of the corresponding Sigma point as:
s313 when the input is 1-dimensional, the initialization vector sequence is:
when the input dimension j is 2,3, …, n, the iterative formula is:
in the formula, the first and second organic solvents are,
is the ith particle point of the jth dimension;
s321, adding the mean and covariance of the system state v to the generated Sigma point:
wherein
Is the mean value of the state variable, PxIs the covariance matrix of the state vector, it can be seen from the above sampling algorithm that other sampling points except the origin have the same weight and are all located in the improved half radius
The above step (1);
s4: optimizing the RBF neural network model obtained in the step S2 by using the CUKF algorithm in the step S3 to obtain a group of optimal parameters
S5: modeling and diagnosing the fault of the pumping unit selected in the step S2 according to the optimal model obtained in the step S4, so that the fault diagnosis purpose is achieved, and the method specifically comprises the following steps:
the CUKF algorithm part is as follows:
s511: initializing system parameters;
s512: calculating a Sigma point state vector;
s513: performing one-step prediction of system state and covariance matrix;
s514: calculating system observation and covariance matrixes;
s515: calculating a Kalman gain;
s516: updating a system state estimation matrix and a covariance matrix;
in the formula (I), the compound is shown in the specification,
the matrix is estimated for the system state at time k-1,
is a Kalman gain matrix, Y (k | k-1) is a system observation matrix at the moment k-1,
a prediction matrix is observed for the system at the time k-1;
in the formula (I), the compound is shown in the specification,
a matrix covariance matrix is estimated for the system at time k-1,
a covariance matrix of a system observation matrix at the moment of k-1;
the RBF algorithm part is as follows:
s521: updating the output of the RBF hidden layer:
wherein m is hidden layer neuron, total J,
is hidden layer neuron output, cmIs the center of the hidden layer neuron, σmWidth of hidden layer neurons;
s522: computing RBF output layer outputs
Wherein l is output layer neuron, and total number of M, ylFor network output layer output, omegam,lConnecting the weight from the hidden layer to the output layer after updating;
through the process, the prediction effect of the CUKF-RBF neural network is shown in FIG. 5, the error is shown in FIG. 6, and the analysis on FIG. 5 and FIG. 6 shows that the CUKF-RBF diagnosis model established by the optimization model can accurately detect the fault and meet the requirement of modeling precision, and the analysis on FIG. 7 and FIG. 8 shows that the CUKF-RBF has better classification effect in multi-target diagnosis;
compared with the prior art, the technical scheme that this application provided, the technological effect or advantage that have are: when the oil pumping unit has a fault, the method can quickly diagnose and identify the fault, implement the fault diagnosis and improve the oil extraction efficiency;
s5: and modeling and diagnosing the fault of the pumping unit selected in the step S2 according to the optimal model obtained in the step S4, so that the fault diagnosis purpose is achieved.
The invention provides an improved traceless Kalman filtering and RBF neural network-based oil pumping unit fault diagnosis method. Firstly, modeling decision parameters by using an RBF neural network, then updating the weight, center and width of a hidden layer of the neural network model in real time by using an improved unscented Kalman filtering algorithm to obtain the optimal parameters of the neural network, and establishing the oil pumping unit fault diagnosis method based on the combination of the improved unscented Kalman filtering and the RBF neural network. The operation process of the pumping unit is often a complex nonlinear dynamic system operation process, the neural network model describing the operation process is often a static mapping, and the influence of external conditions such as specific working conditions in a well on the pumping unit pump is not considered, so that the technical decision effect depending on the static model is unstable, the unscented Kalman filtering has real-time updating performance, the nonlinear dynamic modeling of the RBF neural network is realized, and the unscented Kalman filtering algorithm is improved aiming at the problems of complex iteration of the UKF algorithm, long operation time and the like. The method improves the accuracy rate of fault diagnosis and really achieves the purpose of detecting the operation condition of the oil pumping unit in real time.
It should be noted that the above description is not intended to limit the present invention, and the present invention is not limited to the above examples, and those skilled in the art may make variations, modifications, additions or substitutions within the spirit and scope of the present invention.
Claims (4)
1. A fault diagnosis method of an oil pumping unit based on improved traceless Kalman filtering and RBF neural network is characterized by comprising the following steps:
s1: selecting a complete indicator diagram in one stroke of the pumping unit, carrying out Fourier transform on the indicator diagram, selecting the first f indicator diagram coordinate parameters of a low-frequency part, and combining three-phase current parameters b1, b2 and b3 of the pumping unit to form a decision variable X (a 1, a2, …, af, b1, b2 and b 3), wherein f is the number of the selected indicator diagram coordinate parameters;
s2: at the production site of the pumping unit, at least one group of decision variables X ═ a1, a2, …, af, b1, b2 and b3 is selected]As sample data, the decision variable X or X is output1～XiCorresponding fault type Y or Y1～Yl；
Using RBF neural network to determine the collected decision variable X or X1～XiTraining and checking are carried out, so that a fault diagnosis model of the oil pumping unit is established;
in step S1, when a set of decision variables is selected: selecting a complete indicator diagram in one stroke of the pumping unit, carrying out Fourier transform on the indicator diagram, selecting the first 8 indicator diagram coordinate parameters of a low-frequency part, and combining three-phase current parameters b1, b2 and b3 of the pumping unit to form decision variables X (a 1, a2, …, a8, b1, b2 and b 3), inputting 1 group of decision variables X, and outputting a fault type Y corresponding to the group of decision variables X;
in step S1, when 12 sets of decision variables are selected: in the production field of the oil pumping unit, 12 groups of decision variables X are selected1,X2,...,X12And the corresponding failure types of insufficient liquid supply, sand production of an oil well, wax deposition of the oil well, gas influence, air lock, fixed valve loss, floating valve loss, double valve loss, breakage and separation of a sucker rod, collision on a pump, collision under the pump and continuous pumping and strip spraying are used as data samples, and n groups of decision variables X are input into the data samples1～XiThe output is n groups of decision variables X1～XiCorresponding fault type Y1～Yl；1<n is less than or equal to 12, i is 12;
training and checking the acquired sample data by using an RBF neural network, thereby establishing a fault diagnosis model of the oil pumping unit;
s3: improving the traditional unscented Kalman algorithm by utilizing spherical unscented transformation, and establishing an improved unscented Kalman algorithm, namely a CUKF algorithm; for the traditional traceless Kalman algorithm, 2n +1 Sigma points are calculated when UT transformation is carried out, n refers to the dimension of a state, namely the number of RBF parameters to be optimized, in the improved simplex sampling, the number of the Sigma points is n +2, the Sigma points are in an improved shape in spatial distribution, and the determination mode of the Sigma points is as follows:
(1) when the state is 1-dimensional, the initialization vector sequence is:
(2) when the input dimension j is 2,3, …, n, the iterative formula is:
in the formula, the first and second organic solvents are,
the ith particle point of the jth dimension has n dimensions;
(3) the mean and covariance of the system state v added to the generated Sigma points are:
wherein
Is the mean value of the state variable, PxIs the covariance matrix of the state vector, it can be seen from the above sampling algorithm that other sampling points except the origin have the same weight and are all located in the improved half radius
The above step (1);
s4: optimizing the RBF neural network model obtained in the step S2 by using the CUKF algorithm in the step S3 to obtain a group of optimal parameters;
s5: constructing an optimal model according to the parameters obtained in the step S4 to perform modeling diagnosis on the fault of the pumping unit selected in the step S2, so that the fault diagnosis purpose is achieved;
wherein, the CUKF algorithm part is as follows:
s511: initializing system parameters;
s512: calculating a Sigma point state vector;
s513: performing one-step prediction of system state and covariance matrix;
s514: calculating system observation and covariance matrixes;
s515: calculating a Kalman gain;
s516: updating a system state estimation matrix and a covariance matrix;
in the formula (I), the compound is shown in the specification,
the matrix is estimated for the system state at time k-1,
is a Kalman gain matrix, Y (k | k-1) is a system observation matrix at the moment k-1,
a prediction matrix is observed for the system at the time k-1;
in the formula (I), the compound is shown in the specification,
a matrix covariance matrix is estimated for the system at time k-1,
a covariance matrix of a system observation matrix at the moment of k-1;
the RBF algorithm part is as follows:
s521: updating the output of the RBF hidden layer:
wherein m is hidden layer neuron, total J,
is hidden layer neuron output, cmIs the center of the hidden layer neuron, σmWidth of hidden layer neurons;
s522: computing RBF output layer outputs
Wherein l is output layer neuron, and total number of M, ylFor network output layer output, omegam,lAnd connecting the weights from the hidden layer to the output layer after updating.
2. The method of claim 1, wherein the fault diagnosis of the pumping unit based on the improved unscented Kalman filtering and RBF neural network is performed,
in step S1, two major types of parameters are selected to form a decision variable X, the first major type is that a fourier transform is applied to the indicator diagram coordinate parameters a1, a2, and … a8 after fourier transform, fourier transform is performed on each fault indicator diagram to obtain an indicator diagram frequency spectrum diagram, the first 8 points of the low-frequency part of the indicator diagram frequency spectrum diagram represent indicator diagram graphic features, fourier transform is performed on each fault respectively, the indicator diagram coordinate parameters a1, a2, and … a8 of the first 8 low-frequency parts after transform are selected, and the second major type is pumping unit current parameters b1, b2, and b 3.
3. The method for diagnosing the fault of the pumping unit based on the improved traceless Kalman filtering and the RBF neural network as claimed in claim 1, wherein the RBF neural network in the step S2 is composed of an input layer, a hidden layer and an output layer;
for the fault diagnosis model of the pumping unit, the network structure is A-B-C, A is an input layer, B is a hidden layer, C is an output layer, the activation function adopts a Sigmod function, and the iteration times during sample training are 800 times.
4. The method as claimed in claim 1, wherein the number of UT transform sampling points in the spherical traceless transform in step S3 is n + 2.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710282712.0A CN108804720B (en) | 2017-04-26 | 2017-04-26 | Oil pumping machine fault diagnosis method based on improved traceless Kalman filtering and RBF neural network |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710282712.0A CN108804720B (en) | 2017-04-26 | 2017-04-26 | Oil pumping machine fault diagnosis method based on improved traceless Kalman filtering and RBF neural network |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN108804720A CN108804720A (en) | 2018-11-13 |
| CN108804720B true CN108804720B (en) | 2021-09-14 |
Family
ID=64069291
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201710282712.0A Active CN108804720B (en) | 2017-04-26 | 2017-04-26 | Oil pumping machine fault diagnosis method based on improved traceless Kalman filtering and RBF neural network |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN108804720B (en) |
Families Citing this family (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109739210B (en) * | 2018-12-25 | 2021-02-19 | 中车工业研究院有限公司 | Method and device for evaluating health state of equipment component |
| CN109920514A (en) * | 2019-03-11 | 2019-06-21 | 重庆科技学院 | A kind of self-closing disease based on Kalman filtering neural network embraces body and tests evaluation method and system |
| CN110439537B (en) * | 2019-08-13 | 2020-04-28 | 北京助创科技有限公司 | Method for inverting ground indicator diagram by using electric indicator diagram of pumping unit |
| CN111537679B (en) * | 2020-05-25 | 2021-06-08 | 北京化工大学 | Soft measurement method for concentration of dividing wall rectifying tower based on nonlinear filtering |
| CN115434690B (en) * | 2021-06-04 | 2024-05-14 | 中国科学院沈阳自动化研究所 | Bayesian-based oil pumping unit unsupervised online mutation point detection and fusion method |
| WO2023136856A1 (en) * | 2022-01-17 | 2023-07-20 | Schlumberger Technology Corporation | Field pump equipment system |
Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101826856A (en) * | 2010-03-11 | 2010-09-08 | 哈尔滨工程大学 | Particle filtering method based on spherical simplex unscented Kalman filter |
| CN104680025A (en) * | 2015-03-12 | 2015-06-03 | 重庆科技学院 | Oil pumping unit parameter optimization method on basis of genetic algorithm extreme learning machine |
| CN105045941A (en) * | 2015-03-13 | 2015-11-11 | 重庆科技学院 | Oil pumping unit parameter optimization method based on traceless Kalman filtering |
| WO2016059126A1 (en) * | 2014-10-17 | 2016-04-21 | Jaguar Land Rover Limited | Battery condition monitoring |
| CN105672988A (en) * | 2015-12-30 | 2016-06-15 | 中国石油天然气股份有限公司 | Oil pumping unit indicator diagram diagnosis system and method |
| CN105891724A (en) * | 2016-05-05 | 2016-08-24 | 南京航空航天大学 | On-line estimation method for state of charge of lithium ion battery based on extended single particle model |
| CN106093783A (en) * | 2016-06-03 | 2016-11-09 | 哈尔滨工业大学 | The battery SOC method of estimation that Kalman filtering merges with data-driven |
-
2017
- 2017-04-26 CN CN201710282712.0A patent/CN108804720B/en active Active
Patent Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101826856A (en) * | 2010-03-11 | 2010-09-08 | 哈尔滨工程大学 | Particle filtering method based on spherical simplex unscented Kalman filter |
| WO2016059126A1 (en) * | 2014-10-17 | 2016-04-21 | Jaguar Land Rover Limited | Battery condition monitoring |
| CN104680025A (en) * | 2015-03-12 | 2015-06-03 | 重庆科技学院 | Oil pumping unit parameter optimization method on basis of genetic algorithm extreme learning machine |
| CN105045941A (en) * | 2015-03-13 | 2015-11-11 | 重庆科技学院 | Oil pumping unit parameter optimization method based on traceless Kalman filtering |
| CN105672988A (en) * | 2015-12-30 | 2016-06-15 | 中国石油天然气股份有限公司 | Oil pumping unit indicator diagram diagnosis system and method |
| CN105891724A (en) * | 2016-05-05 | 2016-08-24 | 南京航空航天大学 | On-line estimation method for state of charge of lithium ion battery based on extended single particle model |
| CN106093783A (en) * | 2016-06-03 | 2016-11-09 | 哈尔滨工业大学 | The battery SOC method of estimation that Kalman filtering merges with data-driven |
Non-Patent Citations (2)
| Title |
|---|
| 基于RBF网络的抽油机减速器故障诊断;王江萍;《石油机械》;20071231;第28-31页 * |
| 改进自适应UKF在组合导航系统中的应用研究;王明建;《无线电工程》;20161231;第76-78页 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN108804720A (en) | 2018-11-13 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108804720B (en) | Oil pumping machine fault diagnosis method based on improved traceless Kalman filtering and RBF neural network | |
| CN110029986B (en) | Beam-pumping unit working fluid level prediction method based on particle swarm extreme learning machine | |
| CN107941537B (en) | A kind of mechanical equipment health state evaluation method | |
| CN108897286B (en) | Fault detection method based on distributed nonlinear dynamic relation model | |
| CN111539132B (en) | Dynamic load time domain identification method based on convolutional neural network | |
| CN109740687B (en) | Fermentation process fault monitoring method based on DLAE | |
| CN105981025A (en) | Actually-measured marine environment data assimilation method based on sequence recursive filtering three-dimensional variation | |
| CN108804721B (en) | Oil pumping machine fault diagnosis method based on self-adaptive unscented Kalman filtering and RBF neural network | |
| CN111459144A (en) | Airplane flight control system fault prediction method based on deep cycle neural network | |
| CN115130495A (en) | Rolling bearing fault prediction method and system | |
| CN111324110A (en) | Fermentation process fault monitoring method based on multiple shrinkage automatic encoders | |
| CN106934242A (en) | The health degree appraisal procedure and system of equipment under multi-mode based on Cross-Entropy Method | |
| CN102749584B (en) | Prediction method for residual service life of turbine generator based on ESN (echo state network) of Kalman filtering | |
| CN106301610A (en) | The adaptive failure detection of a kind of superhet and diagnostic method and device | |
| CN110222825B (en) | Cement product specific surface area prediction method and system | |
| CN106682312A (en) | Industrial process soft-measurement modeling method of local weighing extreme learning machine model | |
| CN106503456B (en) | Ensemble Kalman Filter Reservoir behavior history-matching method based on suprasphere transformation | |
| CN111539157B (en) | Dynamic load time domain identification method based on multilayer perceptron | |
| CN111241749B (en) | Permanent magnet synchronous motor chaos prediction method based on reserve pool calculation | |
| CN117473260A (en) | Method for predicting residual life of space bearing by digital-analog linkage | |
| CN105652795B (en) | 3PTT-2R series-parallel numerical control machine tool servo system fault prediction device and method based on residual observer | |
| CN111931368A (en) | UUV target state estimation method based on GRU particle filter | |
| CN115327504B (en) | Sea clutter amplitude distribution non-typed prediction method based on measurement condition parameters | |
| CN102830341A (en) | Online intelligent fault prediction method for power electronic circuit based on RS-CMAC (rough sets and cerebellar model articulation controller) | |
| CN108446506A (en) | A kind of uncertain system modeling method based on section Feedback Neural Network |
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 | ||
| EE01 | Entry into force of recordation of patent licensing contract | ||
| EE01 | Entry into force of recordation of patent licensing contract |
Application publication date: 20181113 Assignee: Guangxi Chunmeng Intelligent Technology Co.,Ltd. Assignor: Chongqing University of Science & Technology Contract record no.: X2023980053984 Denomination of invention: A Fault Diagnosis Method for Pumping Units Based on Improved Unscented Kalman Filter and RBF Neural Network Granted publication date: 20210914 License type: Common License Record date: 20231227 |
|
| EE01 | Entry into force of recordation of patent licensing contract | ||
| EE01 | Entry into force of recordation of patent licensing contract |
Application publication date: 20181113 Assignee: Foshan shangxiaoyun Technology Co.,Ltd. Assignor: Chongqing University of Science & Technology Contract record no.: X2024980003005 Denomination of invention: A Fault Diagnosis Method for Pumping Units Based on Improved Unscented Kalman Filter and RBF Neural Network Granted publication date: 20210914 License type: Common License Record date: 20240322 Application publication date: 20181113 Assignee: FOSHAN YAOYE TECHNOLOGY Co.,Ltd. Assignor: Chongqing University of Science & Technology Contract record no.: X2024980003003 Denomination of invention: A Fault Diagnosis Method for Pumping Units Based on Improved Unscented Kalman Filter and RBF Neural Network Granted publication date: 20210914 License type: Common License Record date: 20240322 Application publication date: 20181113 Assignee: Foshan helixing Technology Co.,Ltd. Assignor: Chongqing University of Science & Technology Contract record no.: X2024980003002 Denomination of invention: A Fault Diagnosis Method for Pumping Units Based on Improved Unscented Kalman Filter and RBF Neural Network Granted publication date: 20210914 License type: Common License Record date: 20240322 |
|
| EE01 | Entry into force of recordation of patent licensing contract | ||
| EE01 | Entry into force of recordation of patent licensing contract |
Application publication date: 20181113 Assignee: Foshan qianshun Technology Co.,Ltd. Assignor: Chongqing University of Science & Technology Contract record no.: X2024980004523 Denomination of invention: A Fault Diagnosis Method for Pumping Units Based on Improved Unscented Kalman Filter and RBF Neural Network Granted publication date: 20210914 License type: Common License Record date: 20240419 |