CN110175698B - Water pump equipment state prediction method based on improved particle swarm optimization BP neural network - Google Patents

Water pump equipment state prediction method based on improved particle swarm optimization BP neural network Download PDF

Info

Publication number
CN110175698B
CN110175698B CN201910342790.4A CN201910342790A CN110175698B CN 110175698 B CN110175698 B CN 110175698B CN 201910342790 A CN201910342790 A CN 201910342790A CN 110175698 B CN110175698 B CN 110175698B
Authority
CN
China
Prior art keywords
vibration intensity
water pump
class
pump equipment
max
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Active
Application number
CN201910342790.4A
Other languages
Chinese (zh)
Other versions
CN110175698A (en
Inventor
潘建
吴攀峰
赵焕东
汤绍雄
奚家字
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Zhejiang University of Technology ZJUT
Original Assignee
Zhejiang University of Technology ZJUT
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Zhejiang University of Technology ZJUT filed Critical Zhejiang University of Technology ZJUT
Priority to CN201910342790.4A priority Critical patent/CN110175698B/en
Publication of CN110175698A publication Critical patent/CN110175698A/en
Application granted granted Critical
Publication of CN110175698B publication Critical patent/CN110175698B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/08—Learning methods
    • G06N3/084—Backpropagation, e.g. using gradient descent
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q10/00—Administration; Management
    • G06Q10/04—Forecasting or optimisation specially adapted for administrative or management purposes, e.g. linear programming or "cutting stock problem"

Landscapes

  • Engineering & Computer Science (AREA)
  • Business, Economics & Management (AREA)
  • Theoretical Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Strategic Management (AREA)
  • Human Resources & Organizations (AREA)
  • Economics (AREA)
  • Computational Linguistics (AREA)
  • Development Economics (AREA)
  • General Health & Medical Sciences (AREA)
  • Molecular Biology (AREA)
  • Computing Systems (AREA)
  • General Engineering & Computer Science (AREA)
  • Data Mining & Analysis (AREA)
  • Mathematical Physics (AREA)
  • Software Systems (AREA)
  • Biophysics (AREA)
  • Biomedical Technology (AREA)
  • Evolutionary Computation (AREA)
  • Artificial Intelligence (AREA)
  • Game Theory and Decision Science (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Health & Medical Sciences (AREA)
  • Entrepreneurship & Innovation (AREA)
  • Marketing (AREA)
  • Operations Research (AREA)
  • Quality & Reliability (AREA)
  • Tourism & Hospitality (AREA)
  • General Business, Economics & Management (AREA)
  • Feedback Control In General (AREA)

Abstract

A water pump equipment state prediction method based on an improved particle swarm optimization BP neural network comprises the following steps: step 1, carrying out noise reduction treatment on the vibration intensity data of a bearing at a driving end of water pump equipment by using a five-point three-time smoothing method; step 2, calculating delay time and embedding dimension of the vibration intensity by using a Cao method of a mutual information method, and performing phase space reconstruction on the vibration intensity data; step 3, dividing an input set and an output set to be used as a prediction model training sample set; step 4, optimizing the weight in the three-layer BP neural network by using an improved particle swarm algorithm; step 5, training to obtain a BP neural network model function; step 6, predicting the vibration intensity data of the drive end bearing in a period of time in the future by using a BP neural network model function to obtain a prediction set; and 7, analyzing the predicted concentrated vibration intensity to obtain the running state of the water pump equipment in a future period of time. The invention ensures the safe and reliable operation of the water pump equipment to the maximum extent.

Description

Water pump equipment state prediction method based on improved particle swarm optimization BP neural network
Technical Field
The invention relates to a water pump equipment state prediction method based on an improved particle swarm optimization BP neural network.
Technical Field
With the development of industry and the dramatic increase in population, water utilities are facing unprecedented water supply pressures. In order to meet the demand of ultra-large water supply, many water utilities have to operate the water pump for a longer time without stopping, which increases the failure rate of the water supply equipment. If the water pump equipment fails frequently, the production of related water enterprises can be influenced, economic loss is brought, and the life quality of residents can be seriously influenced. Therefore, how to predict the running state of the water pump equipment in advance and make a corresponding maintenance plan in time becomes a common problem for many water supply enterprises.
Disclosure of Invention
In order to accurately predict the running state of water pump equipment and reduce the loss caused by the fault of the water pump equipment, the invention provides a water pump equipment state prediction method based on an improved particle swarm optimization BP neural network (PSO-BP).
The technical scheme adopted by the invention is as follows:
a water pump equipment state prediction method based on an improved particle swarm optimization BP neural network comprises the following steps:
step 1, sampling n groups of water pump equipment driving end bearing vibration intensity at equal time intervals
Figure BDA0002041298280000011
Figure BDA0002041298280000021
Performing noise reduction treatment by using a five-point triple smoothing method to obtain preprocessed driving end bearing vibration intensity S { (t)1，x1)，(t2，x2)，...，(ti，xi)，...，(tn，xn) Therein of
Figure BDA0002041298280000022
Is tiOriginal vibration intensity value at time, xiIs tiConstantly reducing the noise and then vibrating the intensity value;
step 2, calculating the delay time and the embedding dimension of the noise-reduced vibration intensity data of the drive-end bearing by using a mutual information method and a Cao method to obtain the delay time tau and the embedding dimension m, performing phase space reconstruction on the vibration intensity of the drive-end bearing by using the delay time and the embedding dimension obtained by calculation to obtain a reconstructed vibration intensity set D of the drive-end bearing { D ═ D }1，d2，...，di，...，dn-(m-1)τWhere di is the ith phase point in the D set, Di＝{xi，xi+τ，xi+2τ，...，xi+(m-1)τ}；
Step 3, dividing an input set and an output set from the reconstructed vibration intensity set D of the drive-end bearing, and taking the first (m-1) vibration intensity values of each phase point as the input set, namely the input set Din＝{d1，in，d2，in，...，di，in，...，dn-(m-1)τ，inIn which d isi，in＝{xi，xi+τ，xi+2τ，...，xi+(m-2)τ}; using the m-th vibration intensity value of each phase point as an output set, namely an output set D out＝{x1+(m-1)τ，x2+(m-1)τ，...，xi+(m-1)τ，...，xn}；
Step 4, combining an input set D by using an improved particle swarm algorithminOutput set DoutOptimizing the hidden layer node weight W1, the hidden layer node threshold A1, the output layer node weight W2 and the output layer node threshold A2 of the three-layer BP neural network to obtain the optimal hidden layer node weight Best W1, the optimal hidden layer node threshold Best A1, the optimal output layer node weight Best W2 and the optimal output layer node threshold Best A2;
step 5, using the input set DinOutput set DoutAnd Best W1, Best A1, Best W2 and Best A2, and the BP neural network is trained to obtain a BP neural network model function y ═ f (d)in)；
Step 6, use f (d)in) Predicting tnThe vibration intensity of k drive-end bearings after a moment, i.e. Spre＝{(tn+1，xn+1)，(tn+2，xn+2)，...，(tn+k，xn+k)}；
Step 7, measuring and evaluating S according to the vibration measurement and evaluation analysis method (GB/T29531-2013) of the pumppreAnalyzing and judging SpreAnd finally, the running state of the water pump equipment is given according to the vibration level of the vibration intensity of the bearing at the middle driving end.
Further, the judging process in step 7 is as follows:
(7.1) pumps are classified as follows, based on their height H (mm) and speed R (rpm):
when H is less than or equal to 225, if R is less than or equal to 1800, the pump belongs to the first class; if R is more than 1800 and less than or equal to 4500, the pump belongs to the second class; if R is more than 4500 and less than or equal to 12000, the pump belongs to the third class;
When H is more than 225 and less than or equal to 225, if R is less than or equal to 1000, the pump belongs to the first class; if R is more than 1000 and less than or equal to 1800, the pump belongs to the second class; if R is more than 1800 and less than or equal to 4500, the pump belongs to the third class; if R is more than 4500 and less than or equal to 12000, the pump belongs to the fourth class;
when H is more than 550, if R is more than 600 and less than or equal to 1500, the pump belongs to the second class; if R is more than 1500 and less than or equal to 3600, the pump belongs to the third class; if R is more than 3600 and less than or equal to 12000, the pump belongs to the fourth class;
(7.2) judging the class of the water pump equipment according to the step 7.1, and determining the maximum effective vibration intensity range corresponding to the vibration intensity of the bearing at the driving end of the water pump equipment by comparing with a vibration level evaluation card of the pump. The four states of the water pump equipment are respectively as follows:
a: the water pump equipment has excellent state and can safely and continuously operate;
b: the water pump equipment has good state and can continuously operate within an allowable working range;
c: the water pump equipment is in a medium state, and the working range and the running time length need to be limited;
d: the water pump equipment has poor state, is not allowed to run and has damage risk;
the maximum effective vibration intensity ranges corresponding to the four states are respectively as follows:
a: the upper and lower limit values are:
Figure BDA0002041298280000031
j ∈ [ first class, second class, third class, fourth class]；
B: the upper and lower limit values are:
Figure BDA0002041298280000032
j ∈ [ first class, second class, third class, fourth class ]；
C: the upper and lower limit values are:
Figure BDA0002041298280000033
j ∈ [ first class, second class, third class, fourth class]；
D: the upper and lower limit values are:
Figure BDA0002041298280000034
j ∈ [ first class, second class, third class, fourth class]；
(7.3) adding SpreComparing the vibration intensity of the bearing at the middle driving end with the four maximum effective vibration intensity ranges in the step 7.2, and if the vibration intensity exceeds r1% vibration intensity greater than
Figure BDA0002041298280000035
The maximum effective vibration intensity range of the water pump device is considered to be within
Figure BDA0002041298280000036
Internal; if not exceeding r1% vibration intensity greater than
Figure BDA0002041298280000037
But have more than r2% vibration intensity greater than
Figure BDA0002041298280000041
The maximum effective vibration intensity range of the water pump device is considered to be within
Figure BDA0002041298280000042
Internal; if not exceeding r2% vibration intensity greater than
Figure BDA0002041298280000043
But have more than r3% vibration intensity greater than
Figure BDA0002041298280000044
The maximum effective vibration intensity range of the water pump device is considered to be within
Figure BDA0002041298280000045
Internal; if not exceeding r3% vibration intensity greater than
Figure BDA0002041298280000046
The maximum effective vibration intensity range of the water pump device is considered to be within
Figure BDA0002041298280000047
Internal;
and (7.4) judging the state of the water pump equipment according to the maximum effective vibration intensity range of the water pump equipment obtained in the step (7.3).
Further, in step 4, the hidden layer node weight, the hidden layer node threshold, the output layer node weight, and the output layer node threshold in the three-layer BP neural network are optimized by using an improved particle swarm optimization, and the specific process is as follows:
(4.1) initializing relevant parameters: the maximum iteration number of the particle combination group is T, the number of the particles is ms, and a learning factor c1The value range is [ c ]1，start，c1，end]Learning factor c2The value range is [ c ]2，start，c2，end]The value range of the inertia weight w is [ wstart，wend]All the hidden layer node weight value ranges are [ W1 ]min，W1max]All the hidden layer node threshold value ranges are [ A1 ]min，A1max]All output layer node weight value ranges are [ W2 ]min，W2max]All output layer node threshold value ranges are [ A2 ]min，A2max]The number of nodes of an input layer of the network is embedded dimension m-1, the number of nodes of an implicit layer is h, and the number of nodes of an output layer is 1;
(4.2) initializing particle swarm individuals: setting an iterative control variable t to 1, and obtaining a search space dimension R to h (m-1) +2h +1 and a particle swarm P to { P ═ according to the initialization parameters in step 4.11，p2，...，pi...，pmsH, then p isiThe random initial position of the particle is
Figure BDA0002041298280000048
P thiThe random initial velocity of the particles is
Figure BDA0002041298280000049
Where R ∈ R, piThe initial individual optimal position of the particle is
Figure BDA00020412982800000410
The global optimum position of the particle is
Figure BDA00020412982800000411
(4.3) calculating a particle fitness value: calculating the fitness value of each particle by using a root mean square error formula, wherein the root mean square error formula is as follows:
Figure BDA0002041298280000051
wherein x isiIs a value of the actual vibration intensity,
Figure BDA0002041298280000052
a vibration severity value is predicted for the model. The smaller the root mean square error, the better the fitness.
(4.4) calculating the current fitness RMSE of the particlestRMSE (fitness to original)t-1Making a comparison if RMSEtIs superior to RMSEt-1Then the particle individual optimum position
Figure BDA0002041298280000053
Equal to RMSEtAnd (4) the corresponding particle position, otherwise, the optimal position of the particle individual keeps the original position unchanged. Global optimum position of particle GBtThe position of the particle corresponding to the best fitness in all particle fitness values is set.
(4.5) use of learning factor c1、c2And c, calculating and updating a calculation formula of the inertia weight w1、c2And w. The calculation formula is as follows:
Figure BDA0002041298280000054
(4.6) iterating according to the position and speed iterative formula of the particle swarm, and updating the position and speed of each particle;
Figure BDA0002041298280000055
wherein r is1、r2Is [0, 1 ]]A random number;
(4.7) judging whether the condition of finishing the optimization searching is achieved: if the optimization result reaches the requirement or T is T, the optimization is finished, and BestW1, BestA1, BestW2 and BestA2 respectively take GBtOf the corresponding portion, i.e. of
Figure BDA0002041298280000056
Figure BDA0002041298280000057
Otherwise, return to step 4.3.
Still further, the prediction process of step 6 is:
(6.1) initializing the set
Figure BDA0002041298280000061
An iteration control variable j is 1;
(6.2) adding a vibration intensity value (t) to the last position in the set Sn+j0), where S has a length of (n + j);
(6.3) carrying out phase space reconstruction on the vibration intensity data in the S by using the tau and the m obtained by calculation in the step 2 to obtain a reconstructed vibration intensity data set D of the drive end bearing t；
(6.4) treating D using the method of step 3tDivision into input sets Dt，inAnd output set Dt，out；
(6.5) mixing Dt，inD in (1)n+j-(m-1)τ，inCarry in f (d)in) To obtain the corresponding output y, where y is t predicted by the prediction modeln+jMoment-by-moment drive-end bearing vibration intensity value, i.e. xn+jReplacing t in S with yn+jThe vibration intensity at the time will be (t)n+jY) substituting into the set SprePerforming the following steps;
(6.6) judging whether j is equal to k, if so, executing the step 7, if not, increasing j by 1, and then returning to the step 6.2.
In the step 1, a five-point cubic smoothing method has a calculation formula as follows:
Figure BDA0002041298280000062
in the above formula, i is required to satisfy that i is more than or equal to 3 and less than or equal to n-2.
In step 3, the mutual information method has the following calculation formula:
Figure BDA0002041298280000063
wherein s ═ xi，q＝xi+τ，Psq(si，qj) Is s isiAnd q isjOf joint distribution probability, Px(xi) And Pq(qj) Is S and S in Q, respectivelyiAnd q isjThe Cao method has the following calculation formula:
Figure BDA0002041298280000064
Figure BDA0002041298280000065
wherein
Figure BDA0002041298280000071
Is the nearest neighbor point in the m-dimension,
Figure BDA0002041298280000072
nearest neighbor point of dimension m +1
In the step 1, the vibration intensity refers to the root mean square value of the vibration speed of the water pump equipment. At present, three standards of displacement, speed and acceleration are often used to measure the magnitude of the vibration intensity of an object, and in general, the magnitude of the vibration intensity is measured by using the vibration intensity, which reflects the magnitude of the total vibration energy containing each harmonic energy.
In the step 7, the vibration measurement and evaluation analysis method (GB/T29531-2013) of the pump is a recommended national standard for evaluating the vibration intensity of the water pump equipment, and the standard is suitable for various water pumps with the rotating speed of 600rpm to 12000rpm, including a water pump unit with a motor and a pump coaxial.
The invention has the beneficial effects that: the vibration intensity data of the bearing at the driving end of the water pump equipment can be predicted, the running state of the water pump equipment in a period of time in the future is judged according to the predicted data, and the safe and reliable running of the water pump equipment is guaranteed to the maximum extent.
Drawings
FIG. 1 is an overall flow chart of the method of the present invention:
FIG. 2 is a schematic diagram of a three-layer BP neural network structure
FIG. 3 is a schematic diagram of the vibration intensity raw data of the drive end bearing:
FIG. 4 is a schematic diagram of data after noise reduction of vibration intensity of a drive end bearing:
FIG. 5 is a vibration intensity rating scale evaluation card:
fig. 6 is a graph of vibration intensity prediction effect.
Detailed Description
The invention is further described below with reference to the accompanying drawings.
Referring to fig. 1 to 6, a water pump equipment state prediction method based on an improved particle swarm optimization BP neural network comprises the following steps:
step 1, sampling n groups of drive end bearing vibration intensity at equal time intervals
Figure BDA0002041298280000073
Performing noise reduction treatment by using a five-point triple smoothing method to obtain preprocessed driving end bearing vibration intensity S { (t)1，x1)，(t2，x2)，...，(ti，xi)，...，(tn，xn) Therein of
Figure BDA0002041298280000081
Is tiOriginal vibration intensity value at time, xiIs tiAnd (3) the vibration intensity value after noise reduction at the moment is calculated by a five-point cubic smoothing method according to the formula:
Figure BDA0002041298280000082
in the above formula, i needs to satisfy i is more than or equal to 3 and less than or equal to n-2;
step 2, calculating the delay time and the embedding dimension of the denoised vibration intensity data of the drive-end bearing by using a mutual information method and a Cao method to obtain the delay time tau and the embedding dimension m, performing phase space reconstruction on the vibration intensity of the drive-end bearing by using the delay time and the embedding dimension obtained by calculation to obtain a reconstructed vibration intensity set D of the drive-end bearing { D ═ D1，d2，...，di，...，dn-(m-1)τIn which d isiIs the ith phase point, di＝{xi，xi+τ，xi+2τ，...，xi+(m-1)τThe formula of the mutual information method is as follows:
Figure BDA0002041298280000083
wherein s ═ xi，q＝xi+τ，Psq(si，qj) Is s isiAnd q isjOf joint distribution probability, Px(xi) And Pq(qj) Is S and S in Q, respectivelyiAnd q isjThe Cao method has the following calculation formula:
Figure BDA0002041298280000084
Figure BDA0002041298280000085
wherein
Figure BDA0002041298280000086
Is the nearest neighbor point in the m-dimension,
Figure BDA0002041298280000087
nearest neighbors in dimension m + 1;
step 3, dividing an input set and an output set from the reconstructed drive end bearing vibration intensity set D, and taking the front (i + (m-2) tau) vibration intensity values of each phase point as the input set, namely the input set Din＝{d1，in，d2，in，...，di，in，...，dn-(m-1)τ，inIn which d is i，in＝{xi，xi+τ，xi+2τ，...，xi+(m-2)τ}; taking the (i + (m-1) tau) th vibration intensity value of each phase point as an output set, namely an output set Dout＝{x1+(m-1)τ，x2+(m-1)τ，...，xi+(m-1)τ，...，xn}；
Step 4, combining an input set D by using an improved particle swarm algorithminOutput set DoutOptimizing the hidden layer node weight W1, the hidden layer node threshold A1, the output layer node weight W2 and the output layer node threshold A2 in the BP neural network to obtain the optimal hidden layer node weight Best W1, the optimal hidden layer node threshold Best A1, the optimal output layer node weight Best W2 and the optimal output layer node threshold Best A2;
the BP neural network is an Artificial Neural Network (ANN) -based error back propagation algorithm, and a three-layer network structure diagram of the BP neural network is shown in fig. 2, where x ═ x (x ═ x)1，...，xi，...，xc)TTo input data, then the input layer section is knownThe number of the points is c; y ═ Y1，...，yj，...，yv)TTo output data, it is then known that the number of output layer nodes is v. Assuming that the number of hidden layer nodes is h, the activation function of the hidden layer nodes is
Figure BDA0002041298280000094
The output layer node activation function is psi, the weight of the ith input layer node and the tth hidden layer node is W1itThe weight of the t hidden layer node and the j output layer node is W2tjThe threshold of the tth hidden layer node is A1tThe threshold value of the jth output layer node is A2 j；
The BP neural network is generally divided into two phases in the training process:
1) forward propagation phase
The data progression from the input layer to the output layer is referred to as the forward propagation phase. The input of the t-th hidden layer node is recorded as innettThe calculation formula is as follows:
Figure BDA0002041298280000091
the output of the tth hidden layer node is recorded as outnettThe calculation formula is as follows:
Figure BDA0002041298280000092
the input of the jth output layer node is recorded as injThe calculation formula is as follows:
Figure BDA0002041298280000093
thus the output out of the jth output layer nodejComprises the following steps:
Figure BDA0002041298280000101
2) phase of back propagation
The data progression from the output layer to the input layer is referred to as the back propagation phase. And in the back propagation stage, continuously adjusting the parameters of each network layer node by using a gradient descent method until the error between the output value and the expected value meets the requirement, and finishing the adjustment.
Selecting a Sigmoid function as an activation function of the BP neural network, wherein the Sigmoid function formula is as follows:
Figure BDA0002041298280000102
the Particle Swarm Optimization (PSO) core idea is to continuously adjust the self flying speed and position through cooperation and information sharing among individuals in a group to find an optimal solution, wherein the speed and position updating formula is as follows:
Figure BDA0002041298280000103
wherein c is1、c2As a learning factor, w is the inertial weight, r1、r2Is [0, 1 ]]A random number within the range of the random number,
Figure BDA0002041298280000107
the individual optimal position of the ith particle at the t iteration is taken as the position of the ith particle; GB tThe optimal position of the population at the t-th iteration is obtained.
For c1、c2And w value, the invention provides a Group Optimization Strategy (GOS) based on a cosine function for optimizing c in PSO1、c2And the value of w, i.e. c will be taken in each iteration of the PSO1、c2And w are divided into a group, and three of them are optimized at the same time, and the method can be described as follows:
Figure BDA0002041298280000104
Figure BDA0002041298280000105
Figure BDA0002041298280000106
wherein c is1，start、c1，endIs c1A value range of (c)2，start、c2，endIs c2A value range of (d), wstart、wendTo obtain the value range of w, the process of step 4 is as follows:
(4.1) initializing particle swarm related parameters: the maximum iteration number of the particle combination group is T, the number of the particles is ms, and a learning factor c1The value range is [ c ]1，start，c1，end]Learning factor c2The value range is [ c ]2，start，c2，end]The value range of the inertia weight w is [ wstart，wend]All the hidden layer node weight value ranges are [ W1 ]min，W1max]All the hidden layer node threshold value ranges are [ A1 ]min，A1max]All output layer node weight value ranges are [ W2 ]min，W2max]All output layer node threshold value ranges are [ A2 ]min，A2max]The number of input layer nodes of the BP neural network is m-1, the number of hidden layer nodes is h, and the number of output layer nodes is 1;
(4.2) initializing particle swarm individuals: setting an iterative control variable t to 1, and obtaining a search space dimension R to h (m-1) +2h +1 and a particle group P to { P ═ according to the initialization parameters in step 4.1 1，p2，...，pi...，pmsH, then p isiThe random initial position of the particle is
Figure BDA0002041298280000111
P thiThe random initial velocity of the particles is
Figure BDA0002041298280000112
Where R ∈ R, piThe initial individual optimum value of the particle is
Figure BDA0002041298280000113
The global optimum position of the particle is
Figure BDA0002041298280000114
(4.3) calculating a particle fitness value: calculating the fitness value of each particle by using a root mean square error function, wherein the root mean square error function is calculated by the formula:
Figure BDA0002041298280000115
wherein x isiIs a value of the actual vibration intensity,
Figure BDA0002041298280000116
predicting a vibration intensity value for the model, wherein the smaller the root mean square error is, the better the fitness is;
(4.4) calculating the current fitness RMSE of the particlestRMSE (fitness to original)t-1Making a comparison if RMSEtIs superior to RMSEt-1Then the particle individual optimum position
Figure BDA0002041298280000117
Equal to RMSEtAnd (4) the corresponding particle position, otherwise, the optimal position of the particle individual keeps the original position unchanged. Global optimum position of particle GBtSetting the position of the particle corresponding to the best fitness in all particle fitness values;
(4.5) separately calculating the learning factor c by using the above-mentioned cosine function-based grouping optimization strategy1、c2And the value of the inertial weight w, update c1、c2、w；
(4.6) iterating according to the position and speed iterative formula of the particle swarm, and updating the position and speed of each particle;
(4.7) judgmentWhether the interruption reaches the condition of finishing the optimization is as follows: if the optimization result reaches the requirement or T is T, the optimization is finished, and BestW1, BestA1, BestW2 and BestA2 respectively take GB tOf the corresponding portion, i.e. of
Figure BDA0002041298280000121
Figure BDA0002041298280000122
Otherwise, return to step 4.3.
Step 5, using the input set DinOutput set DoutAnd Best W1, Best A1, Best W2 and Best A2, and the BP neural network is trained to obtain a BP neural network model function y ═ f (d)in)；
Step 6, use f (d)in) Predicting tnThe vibration intensity of k drive-end bearings after a moment, i.e. Spre＝{(tn+1，xn+1)，(tn+2，xn+2)，...，(tn+k，xn+k) The prediction process is as follows:
(6.1) initializing the set.
Figure BDA0002041298280000123
An iteration control variable j is 1;
(6.2) adding a vibration intensity value (t) to the last position in the set Sn+j0), where S has a length of (n + j);
(6.3) carrying out phase space reconstruction on the vibration intensity data in the S by using the tau and the m obtained by calculation in the step 2 to obtain a reconstructed vibration intensity data set D of the drive end bearingt；
(6.4) treating D using the method of step 3tDivision into input sets Dt，inAnd output set Dt，out；
(6.5) mixing Dt，inD in (1)n+j-(m-1)τ，inCarry in f (d)in) To obtain the corresponding output y, where y is t predicted by the prediction modeln+jMoment value of vibration intensity of bearing at drive end, i.e. xn+jReplacing t in S with yn+jThe vibration intensity value at the moment of time is(tn+jY) substituting into the set SprePerforming the following steps;
(6.6) judging whether j is equal to k, if so, executing the step 7, if not, increasing j by 1, and then returning to the step 6.2.
Step 7, measuring and evaluating S according to the vibration measurement and evaluation analysis method (GB/T29531-2013) of the pumppreAnalyzing and judging SpreThe vibration intensity of the bearing at the middle driving end belongs to what vibration level, the running state of the water pump equipment is finally given, and the judgment process is as follows:
(7.1) pumps are classified as follows, based on their height H (mm) and speed R (rpm):
when H is less than or equal to 225, if R is less than or equal to 1800, the pump belongs to the first class; if R is more than 1800 and less than or equal to 4500, the pump belongs to the second class; if R is more than 4500 and less than or equal to 12000, the pump belongs to the third class;
when H is more than 225 and less than or equal to 225, if R is less than or equal to 1000, the pump belongs to the first class; if R is more than 1000 and less than or equal to 1800, the pump belongs to the second class; if R is more than 1800 and less than or equal to 4500, the pump belongs to the third class; if R is more than 4500 and less than or equal to 12000, the pump belongs to the fourth category:
when H is more than 550, if R is more than 600 and less than or equal to 1500, the pump belongs to the second class; if R is more than 1500 and less than or equal to 3600, the pump belongs to the third class; if R is more than 3600 and less than or equal to 12000, the pump belongs to the fourth class;
(7.2) judging the class of the water pump equipment according to the step 7.1, and determining the maximum effective vibration intensity range corresponding to the vibration intensity of the bearing at the driving end of the water pump equipment by comparing with a vibration level evaluation card of the pump. The four states of the water pump equipment are respectively as follows:
a: the water pump equipment has excellent state and can safely and continuously operate;
B: the water pump equipment has good state and can continuously operate within an allowable working range;
c: the water pump equipment is in a medium state, and the working range and the running time length need to be limited;
d: the water pump equipment has poor state, is not allowed to run and has damage risk;
the maximum effective vibration intensity ranges corresponding to the four states are respectively as follows:
a: the upper and lower limit values are:
Figure BDA0002041298280000131
j ∈ [ first class, second class, third class, fourth class]；
B: the upper and lower limit values are:
Figure BDA0002041298280000132
j ∈ [ first class, second class, third class, fourth class]；
C: the upper and lower limit values are:
Figure BDA0002041298280000133
j ∈ [ first class, second class, third class, fourth class]；
D: the upper and lower limit values are:
Figure BDA0002041298280000134
j ∈ [ first class, second class, third class, fourth class]；
(7.3) adding SpreComparing the vibration intensity of the bearing at the middle driving end with the four maximum effective vibration intensity ranges in the step 7.2, and if the vibration intensity of more than 20 percent is larger than the vibration intensity range of the bearing at the middle driving end
Figure BDA0002041298280000135
The maximum effective vibration intensity range of the water pump device is considered to be within
Figure BDA0002041298280000136
Internal; if not more than 20% of the vibration intensity is greater than
Figure BDA0002041298280000137
But more than 20% of the vibration intensity is greater than
Figure BDA0002041298280000138
The maximum effective vibration intensity range of the water pump device is considered to be within
Figure BDA0002041298280000139
Internal; if not more than 20% of the vibration intensity is greater than
Figure BDA00020412982800001310
But more than 20% of the vibration intensity is greater than
Figure BDA0002041298280000141
The maximum effective vibration intensity range of the water pump device is considered to be within
Figure BDA0002041298280000142
Internal; if not more than 20% of the vibration intensity is greater than
Figure BDA0002041298280000143
The maximum effective vibration intensity range of the water pump device is considered to be within
Figure BDA0002041298280000144
Internal:
and (7.4) judging the state of the water pump equipment according to the maximum effective vibration intensity range of the water pump equipment obtained in the step (7.3).
The present invention is further illustrated by the following examples, which include, but are not limited to, the following examples.
Taking the vibration intensity data of certain water pump equipment as an example, the rated power of the water pump equipment is 20KW, the rated rotating speed is 1750rpm, and the height of the center of the pump is 243mm, so that the water pump equipment can be judged to belong to the second class. Collecting the vibration intensity of the bearing at the driving end of the water pump equipment at each time of 10 minutes, finally obtaining the vibration intensity data of 805 bearings at the driving end and storing the data into a set S*In pair S*Performing noise reduction treatment, storing the processed vibration intensity into a set S, and comparing partial vibration intensity before and after the noise reduction treatment with a ratio shown in Table 1, wherein S is*The graphs with S data are shown in fig. 3 and 4. Table 1 shows comparison before and after noise reduction treatment of partial vibration intensity;
Figure BDA0002041298280000145
TABLE 1
After the vibration intensity of the bearing at the driving end is denoised, the delay time and the embedding dimension of the vibration intensity in the set S are calculated by using a mutual information method and a Cao method, the delay time tau is 2 and the embedding dimension m is 8 through calculation. Performing phase space reconstruction on the vibration intensity in the S by using the delay time and the embedding dimension to obtain 791 groups of reconstructed phase points, and dividing the first 7 vibration intensity data of each phase point into an input set D inThe last data is divided into output sets DoutFinally, the input set and output set for training are obtained. The partial phase points are shown in table 2.
Figure BDA0002041298280000151
TABLE 2
After the sample input set and the output set are prepared, optimizing the hidden layer node weight, the hidden layer node threshold value, the output layer node weight and the output layer node threshold value in the BP neural network by using an improved particle swarm algorithm and combining the input set and the output set, wherein the values of relevant parameters are set as follows: t100, ms 25, [ c1，start，c1，end]＝[2.75，1.25]，[c2，start，c2，end]＝[0.5，2.25]，[wstart，wend]＝[0.9，0.4]All hidden layer nodes have weight of-10, 10]All hidden layer node thresholds are [ -10, 10 [)]All output layer nodes have weight of-10, 10]All output layer node thresholds are [ -10, 10 [)]. The final product BestW1 ═ (-3.1732, 0.1279, 3.3352, -3.5211, 3.9409, 0.6858, 1.0976, -1.4948, 0.6220, -2.0482, 1.0782, -2.3054, -2.6585, -0.7521, 0.3187, 4.1675, 2.6741, -1.2950, 1.9743, 0.8797, 0.7823, -4.0298, -0.9680, -2.6340, 0.3249, 1.8298, 0.3008, -1.4996, 0.0810, -1.8230, 2.6539, 2.9508, 0.68861.4176, 0.5825), BestA1 ═ (1.0941, 2.6565, -2.6752, -1.1872, -3.9007), BestW2 ═ (0.1383, 0.1864, 0.1209, -0.0715, -0.2347), BestA2 ═ 0.4026.
Training the BP neural network by using the input set, the output set, BestW1, BestA1, BestW2 and BestA2 to obtain BPThe neural network model function y ═ f (d)in) In turn, 35 sets of vibration intensity values were predicted to be generated, as shown in table 3.
Figure BDA0002041298280000152
Figure BDA0002041298280000161
TABLE 3
The water pump equipment is judged to belong to the second category, and the vibration intensity grade evaluation card of the pump in comparison with the figure 5 shows that all the 30 vibration intensity data are in the range of 0.28 and belong to the state B, and finally the operation state of the fan in the future 20 minutes is a satisfactory state.
Further, in this example, after 805 drive-end bearing vibration severity values were collected, 35 vibration severity values were collected, as shown in table 4. The 35 predicted vibration intensity values were compared with the actual vibration intensity values in table 4, and the prediction results substantially matched as shown in fig. 6.
ti 806 807 808 809 810 811 812
xi 0.135 0.140 0.137 0.128 0.115 0.106 0.106
ti 813 814 815 816 817 818 819
xi 0.113 0.121 0.127 0.127 0.126 0.128 0.136
ti 820 821 822 823 824 825 826
xi 0.144 0.144 0.138 0.137 0.143 0.151 0.155
ti 827 828 829 830 831 832 833
xi 0.155 0.150 0.140 0.128 0.123 0.128 0.136
ti 834 835 836 837 838 839 840
xi 0.139 0.136 0.133 0.133 0.133 0.131 0.131
TABLE 4
The water pump equipment state prediction method based on the improved particle swarm optimization BP neural network predicts the vibration intensity data of the drive end bearing of the water pump equipment, judges the running state of the water pump equipment in a period of time in the future by utilizing the predicted vibration intensity, and ensures the safe and reliable running of a water supply system to the maximum extent.
It will be appreciated by persons skilled in the art that the foregoing is illustrative only and is not to be construed as limiting the invention, as variations and modifications of the foregoing examples are within the spirit and scope of the invention as defined by the appended claims.

Claims (6)

1. A water pump equipment state prediction method based on an improved particle swarm optimization BP neural network is characterized by comprising the following steps:
step 1, sampling n groups of water pump equipment driving end bearing vibration intensity S at equal time intervals*＝{(t1,x1 *),(t2,x2 *),…,(ti,xi *),…,(tn,xn *) And performing noise reduction treatment by using a five-point three-time smoothing method to obtain preprocessed driving-end bearing vibration intensity S { (t)1,x1),(t2,x2),…,(ti,xi),…,(tn,xn) In which xi *Is tiOriginal vibration intensity value at time, xiIs tiConstantly reducing the noise and then vibrating the intensity value;
step 2, calculating the delay time and the embedding dimension of the noise-reduced vibration intensity data of the drive-end bearing by using a mutual information method and a Cao method to obtain the delay time tau and the embedding dimension m, performing phase space reconstruction on the vibration intensity of the drive-end bearing by using the delay time and the embedding dimension obtained by calculation to obtain a reconstructed vibration intensity set D of the drive-end bearing { D ═ D }1,d2,…,di,…,dn-(m-1)τIn which d isiIs the ith phase point in the set D, Di＝{xi，xi+τ，xi+2τ，...，xi+(m-1)τ}；
Step 3, dividing an input set and an output set from the reconstructed vibration intensity set D of the drive-end bearing, and taking the first (m-1) vibration intensity values of each phase point as the input set, namely the input set Din＝{d1,in,d2,in,…,di,in,…,dn-(m-1)τ,inIn which d isi,in＝{xi,xi+τ,xi+2τ,…,xi+(m-2)τThe m-th vibration intensity value of each phase point is taken as an output set, namely an output set D out＝{x1+(m-1)τ,x2+(m-1)τ,…,xi+(m-1)τ…,xn}；
Step 4, combining an input set D by using an improved particle swarm algorithminOutput set DoutOptimizing the hidden layer node weight W1, the hidden layer node threshold A1, the output layer node weight W2 and the output layer node threshold A2 of the three-layer BP neural network to obtain the optimal hidden layer node weight Best W1, the optimal hidden layer node threshold Best A1, the optimal output layer node weight Best W2 and the optimal output layer node threshold Best A2;
step 5, using the input set DinOutput set DoutAnd Best W1, Best A1, Best W2 and Best A2, and the BP neural network is trained to obtain a BP neural network model function y ═ f (d)in)；
Step 6, use f (d)in) Predicting tnThe vibration intensity of k drive-end bearings after a moment, i.e. Spre＝{(tn+1,xn+1),(tn+2,xn+2),…,(tn+k,xn+k)}；
Step 7, measuring and evaluating S according to the vibration measurement and evaluation analysis method (GB/T29531-2013) of the pumppreAnalyzing and determining SpreThe vibration level of the vibration intensity of the bearing at the middle driving end is used for finally giving the running state of the water pump equipment;
in the step 7, the step S is carried out according to the vibration measurement and evaluation analysis method (GB/T29531-2013) of the pumppreAnalyzing and judging SpreThe vibration level of the vibration intensity of the bearing at the middle driving end is finally given out the running state of the water pump equipment, and the judging process is as follows:
(7.1) pumps are classified as follows, based on their height H (mm) and speed R (rpm):
when H is less than or equal to 225, if R is less than or equal to 1800, the pump belongs to the first class; if 1800< R.ltoreq.4500, the pump belongs to the second class; if 4500< R < 12000, the pump belongs to the third category;
when 225< H.ltoreq.550, if R.ltoreq.1000, the pump belongs to the first class; if 1000< R.ltoreq.1800, the pump belongs to the second class; if 1800< R.ltoreq.4500, the pump belongs to the third class; if 4500< R < 12000, the pump belongs to the fourth class;
when H >550, if 600< R ≦ 1500, the pump belongs to the second class; if 1500< R is less than or equal to 3600, the pump belongs to the third class; if 3600< R < 12000, the pump belongs to the fourth class;
(7.2) judging the category of the water pump equipment according to the step (7.1), determining the maximum effective vibration intensity range corresponding to the vibration intensity of the bearing at the drive end of the water pump equipment by comparing with a vibration level evaluation card of the pump, wherein the four states of the water pump equipment are respectively as follows:
a: the water pump equipment has excellent state and can safely and continuously operate;
b: the water pump equipment has good state and can continuously operate within an allowable working range;
c: the water pump equipment is in a medium state, and the working range and the running time length need to be limited;
d: the water pump equipment has poor state, is not allowed to run and has damage risk;
The maximum effective vibration intensity ranges corresponding to the four states are respectively as follows:
a: the upper and lower limit values are: [0, A ]max j]J ∈ [ first class, second class, third class, fourth class)]；
B: the upper and lower limit values are: (A)max j,Bmax j]J ∈ [ first class, second class, third class, fourth class)]；
C: the upper and lower limit values are: (B)max j,Cmax j]J ∈ [ first class, second class, third class, fourth class)]；
D: the upper and lower limit values are: (C)max j, + infinity), j ∈ [ first class, second class, third class, fourth class]；
(7.3) adding SpreThe vibration intensity of the bearing at the middle driving end is compared with the maximum effective vibration intensity range of the four states in (7.2), and if the vibration intensity exceeds r1% vibration intensity greater than Cmax jAnd then the maximum effective vibration intensity range of the water pump equipment is considered to be (C)max jWithin, + ∞); if not exceeding r1% vibration intensity greater than Cmax jBut has more than r2% vibration intensity greater than Bmax jAnd then the maximum effective vibration intensity range of the water pump equipment is considered to be (B)max j,Cmax j]Internal; if not exceeding r2% vibration intensity greater than Bmax jBut has more than r3% vibration intensity greater than Amax jAnd then the maximum effective vibration intensity range of the water pump equipment is considered to be (A)max j,Bmax j]Internal; if not exceeding r3% vibration intensity greater than Amax jThen the maximum effective vibration intensity range of the water pump equipment is considered to be [0, A ] max j]Internal;
(7.4) judging the state of the water pump equipment according to the maximum effective vibration intensity range of the water pump equipment obtained in the step (7.3);
in the step 4, the hidden layer node weight W1, the hidden layer node threshold a1, the output layer node weight W2 and the output layer node threshold a2 of the three-layer BP neural network are optimized by using an improved particle swarm optimization, and the process is as follows:
(4.1) initializing relevant parameters: let the maximum iteration number of the particle swarm be T, the number of the particles be ms, and learn the factor c1The value range is [ c ]1,start,c1,end]Learning factor c2The value range is [ c ]2,start,c2,end]The value range of the inertia weight w is [ wstart,wend]All the hidden layer node weight value ranges are [ W1 ]min,W1max]All the hidden layer node threshold value ranges are [ A1 ]min,A1max]All output layer node weight value ranges are [ W2 ]min,W2max]All output layer node threshold value ranges are [ A2 ]min,A2max]The number of nodes of an input layer of the network is embedded dimension m-1, the number of nodes of an implicit layer is h, and the number of nodes of an output layer is 1;
(4.2) initializing particle swarm individuals: setting an iteration control variable t as 1, obtaining a search space dimension R as h (m-1) +2h +1 and a particle swarm P as { P) according to the initialization parameters in the step (1)1,p2,…,pi…,pmsH, then p isiThe random initial position of the particle is X i t＝(qi,1 t,qi,2 t,…,qi,r t,…,qi,R t) P thiThe random initial velocity of the particles is Vi t＝(vi,1 t,vi,2 t,…,vi,r t,…,vi,R t) Where R ∈ R, piThe initial individual optimal position of the particle is PBi t＝Xi tThe global optimum position of the particle is GBt＝X1 t；
(4.3) calculating a particle fitness value: calculating the fitness value of each particle by using a root mean square error formula;
(4.4) comparing particle fitness values: the calculated current fitness RMSE of the particlestRMSE (fitness to original)t-1Making a comparison if RMSEtIs superior to RMSEt-1Then the individual optimum position PB of the particlei tEqual to RMSEtThe corresponding particle position, otherwise, the individual optimal position of the particle keeps the original position unchanged, and the global optimal position GB of the particletSetting the position of the particle corresponding to the best fitness in all particle fitness values;
(4.5) calculating and updating c1、c2W: calculating and updating a learning factor c in the particle swarm algorithm according to a group optimization strategy calculation formula based on a cosine function1、c2And the value of the inertial weight w;
(4.6) calculating and updating particle position and velocity: iterating according to the position and speed iterative formula of the particle swarm, and updating the position and speed of each particle;
(4.7) judging whether the condition of finishing the optimization searching is achieved: if the optimization result reaches the requirement or T is T, finishing the optimization, and respectively taking GB from BestW1, BestA1, BestW2 and BestA2 tThe value of the corresponding portion of (q), BestW1 ═ q1 t,q2 t,…,qh(m-1) t)，BestA1＝(q(h(m-1)+1) t,q(h(m-1)+2) t,…,q(h(m-1)+h) t)，BestW2＝(q(h(m-1)+h+1) t,q(h(m-1)+h+2) t,…,q(h(m-1)+2h) t)，BestA2＝q(h(m-1)+2h+1) tOtherwise, return to (4.3).
2. The method for predicting the state of the water pump equipment based on the improved particle swarm optimization BP neural network according to claim 1, wherein the prediction process in the step 6 is as follows:
(6.1) initializing the set
Figure FDA0003101568610000041
An iteration control variable j is 1;
(6.2) adding a vibration intensity value (t) to the last position in the set Sn+j0), where S has a length of (n + j);
(6.3) carrying out phase space reconstruction on the vibration intensity data in the S by using the tau and the m obtained by calculation in the step 2 to obtain a reconstructed vibration intensity data set D of the drive end bearingt；
(6.4) treating D using the method of step 3tDivision into input sets Dt，inAnd output set Dt，out；
(6.5) mixing Dt，inD in (1)n+j-(m-1)τ，inCarry in f (d)in) To obtain the corresponding output y, where y is t predicted by the prediction modeln+jMoment-by-moment drive-end bearing vibration intensity value, i.e. xn+jReplacing t in S with yn+jThe vibration intensity at the time will be (t)n+jY) substituting into the set SprePerforming the following steps;
(6.6) judging whether j is equal to k, if so, executing the step 7, if not, increasing j by 1, and then returning to the step 6.2.
3. The method for predicting the state of the water pump equipment based on the improved particle swarm optimization BP neural network according to claim 1, wherein in the step 1, a five-point cubic smoothing method has a calculation formula as follows:
Figure FDA0003101568610000042
In the above formula, i is required to satisfy that i is more than or equal to 3 and less than or equal to n-2.
4. The method for predicting the state of the water pump equipment based on the improved particle swarm optimization BP neural network according to claim 1, wherein in the step 2, a mutual information method calculation formula is as follows:
Figure FDA0003101568610000043
wherein s isi＝xi，qj＝xi+τ，Psq(si，qj) Is s isiAnd q isjOf joint distribution probability, Ps(si) And Pq(qj) Is S and S in Q, respectivelyiAnd q isjThe Cao method has the following calculation formula:
Figure FDA0003101568610000044
Figure FDA0003101568610000051
wherein
Figure FDA0003101568610000052
Is the nearest neighbor point in the m-dimension,
Figure FDA0003101568610000053
the nearest neighbor in dimension m + 1.
5. The method for predicting the state of the water pump equipment based on the improved particle swarm optimization BP neural network according to claim 1, wherein in the step 1, the vibration intensity refers to a root mean square value of a vibration speed of the water pump equipment.
6. The method as claimed in claim 1, wherein in the step 7, the method for measuring and evaluating the vibration of the pump (GB/T29531-2013) is a recommended national standard for evaluating the vibration intensity of the water pump equipment, and the standard is suitable for various types of water pumps with the rotation speed of 600rpm to 12000rpm, including a water pump unit with a motor and a pump coaxial.
CN201910342790.4A 2019-04-26 2019-04-26 Water pump equipment state prediction method based on improved particle swarm optimization BP neural network Active CN110175698B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201910342790.4A CN110175698B (en) 2019-04-26 2019-04-26 Water pump equipment state prediction method based on improved particle swarm optimization BP neural network

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201910342790.4A CN110175698B (en) 2019-04-26 2019-04-26 Water pump equipment state prediction method based on improved particle swarm optimization BP neural network

Publications (2)

Publication Number Publication Date
CN110175698A CN110175698A (en) 2019-08-27
CN110175698B true CN110175698B (en) 2021-08-03

Family

ID=67690194

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201910342790.4A Active CN110175698B (en) 2019-04-26 2019-04-26 Water pump equipment state prediction method based on improved particle swarm optimization BP neural network

Country Status (1)

Country Link
CN (1) CN110175698B (en)

Families Citing this family (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN112331274A (en) * 2020-03-06 2021-02-05 西安工程大学 Method for predicting dissolved gas in transformer oil based on phase reconstruction and neural network
CN113236581B (en) * 2021-06-29 2022-02-22 江苏大学 Intelligent parallel pump system and optimal adjustment method

Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN103544526A (en) * 2013-11-05 2014-01-29 辽宁大学 Improved particle swarm algorithm and application thereof
US9274036B2 (en) * 2013-12-13 2016-03-01 King Fahd University Of Petroleum And Minerals Method and apparatus for characterizing composite materials using an artificial neural network
CN106289777A (en) * 2016-08-01 2017-01-04 北京航空航天大学 A kind of multi-state rolling bearing performance appraisal procedure based on geometry tolerance
CN108376286A (en) * 2018-02-26 2018-08-07 上海理工大学 A kind of equipment fault diagnosis apparatus and system
CN109242215A (en) * 2018-10-26 2019-01-18 浙江工业大学之江学院 It is a kind of based on population-support vector machines rotating machinery operating condition prediction technique

Patent Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN103544526A (en) * 2013-11-05 2014-01-29 辽宁大学 Improved particle swarm algorithm and application thereof
US9274036B2 (en) * 2013-12-13 2016-03-01 King Fahd University Of Petroleum And Minerals Method and apparatus for characterizing composite materials using an artificial neural network
CN106289777A (en) * 2016-08-01 2017-01-04 北京航空航天大学 A kind of multi-state rolling bearing performance appraisal procedure based on geometry tolerance
CN108376286A (en) * 2018-02-26 2018-08-07 上海理工大学 A kind of equipment fault diagnosis apparatus and system
CN109242215A (en) * 2018-10-26 2019-01-18 浙江工业大学之江学院 It is a kind of based on population-support vector machines rotating machinery operating condition prediction technique

Non-Patent Citations (1)

* Cited by examiner, † Cited by third party
Title
基于MATLAB的振动信号平滑处理方法;孙苗钟;《电子测量技术》;20070630;第30卷(第6期);正文第1-2页 *

Also Published As

Publication number Publication date
CN110175698A (en) 2019-08-27

Similar Documents

Publication Publication Date Title
CN108334059B (en) Particle swarm optimization-based BP neural network model fault diagnosis method
CN110175698B (en) Water pump equipment state prediction method based on improved particle swarm optimization BP neural network
CN108181591B (en) Battery SOC value prediction method based on improved BP neural network
CN111080039B (en) Fan cluster fault prediction method and system
CN110909802A (en) Improved PSO (particle swarm optimization) based fault classification method for optimizing PNN (portable network) smoothing factor
CN110794308B (en) Method and device for predicting train battery capacity
CN107168292B (en) Submarine navigation device circuit failure diagnosis method based on ELM algorithm
CN111159063B (en) Cache allocation method for multi-layer Sketch network measurement
CN109740481B (en) CNN and LSTM combined atrial fibrillation signal classification system based on jump connection
CN113239534B (en) Fault and service life prediction method and device of wind generating set
CN111507045B (en) Structural damage identification method based on adaptive weight whale optimization algorithm
CN112288147B (en) Method for predicting insulation state of generator stator by BP-Adaboost strong predictor
CN116010900A (en) Multi-scale feature fusion gearbox fault diagnosis method based on self-attention mechanism
CN115270867A (en) Improved DQN fault diagnosis method and system for gas turbine rotor system
CN112990420A (en) Pruning method for convolutional neural network model
CN115879369A (en) Coal mill fault early warning method based on optimized LightGBM algorithm
CN111695288A (en) Transformer fault diagnosis method based on Apriori-BP algorithm
CN111241749B (en) Permanent magnet synchronous motor chaos prediction method based on reserve pool calculation
JP6538573B2 (en) Machine learning device, motor control device, motor control system, and machine learning method for learning values of resistance regeneration start voltage and resistance regeneration stop voltage
CN106773667B (en) Oil field mechanical production parameter modeling method based on unscented particle filter neural network
CN110348005B (en) Distribution network equipment state data processing method and device, computer equipment and medium
CN116992624B (en) New energy automobile motor heat management method
CN113807005B (en) Bearing residual life prediction method based on improved FPA-DBN
CN113516163B (en) Vehicle classification model compression method, device and storage medium based on network pruning
CN114091782B (en) Medium-long term power load prediction method

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