CN114386312A - Equipment fault diagnosis method - Google Patents
Equipment fault diagnosis method Download PDFInfo
- Publication number
- CN114386312A CN114386312A CN202111491009.3A CN202111491009A CN114386312A CN 114386312 A CN114386312 A CN 114386312A CN 202111491009 A CN202111491009 A CN 202111491009A CN 114386312 A CN114386312 A CN 114386312A
- Authority
- CN
- China
- Prior art keywords
- output
- network
- equipment
- neural network
- weight
- 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.)
- Withdrawn
Links
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
- G06F30/27—Design optimisation, verification or simulation using machine learning, e.g. artificial intelligence, neural networks, support vector machines [SVM] or training a model
-
- 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
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/048—Activation functions
-
- 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
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
- G06N3/088—Non-supervised learning, e.g. competitive learning
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2119/00—Details relating to the type or aim of the analysis or the optimisation
- G06F2119/06—Power analysis or power optimisation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2119/00—Details relating to the type or aim of the analysis or the optimisation
- G06F2119/08—Thermal analysis or thermal optimisation
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Evolutionary Computation (AREA)
- Software Systems (AREA)
- Artificial Intelligence (AREA)
- General Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- General Health & Medical Sciences (AREA)
- Mathematical Physics (AREA)
- Computational Linguistics (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- Biophysics (AREA)
- Biomedical Technology (AREA)
- Data Mining & Analysis (AREA)
- Life Sciences & Earth Sciences (AREA)
- Health & Medical Sciences (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Medical Informatics (AREA)
- Computer Hardware Design (AREA)
- Geometry (AREA)
- Testing And Monitoring For Control Systems (AREA)
Abstract
The invention belongs to the technical field of equipment operation guarantee, and particularly relates to an equipment fault diagnosis method; it comprises the following steps: determining the category, rated parameters and initial life cycle of research equipment; establishing an expert system model rule base aiming at the object category characteristics; establishing an object historical database; introducing an artificial neural network to model a high-dimensional nonlinear problem; realizing fault diagnosis by using a BP neural network; according to the equipment fault diagnosis method, technologies such as big data and artificial intelligence algorithms are introduced to carry out datamation, modeling and standardization on equipment rated parameters, equipment operation conditions, equipment operation and maintenance history and experience, intelligent learning and alternation capabilities are given to a diagnosis system, the debugging parameters are fed back and corrected according to result output, and finally the whole life cycle state of the equipment is displayed and reflected through a visual platform, so that data support is provided for fault early warning, diagnosis and analysis and fault maintenance decision, and the equipment fault rate is reduced.
Description
Technical Field
The invention belongs to the technical field of equipment operation guarantee, and particularly relates to an equipment fault diagnosis method.
Background
At present, the early warning to operation equipment trouble, the definite time is examined the diagnosis and is judged based on operation and maintenance personnel's experience completely, but because operation and maintenance personnel work experience, operating condition and individual difference, thereby can lead to the diagnosis to the trouble to have the deviation, personnel flow simultaneously, the succession that divides the labour change to lead to operation and maintenance personnel experience value can't continue effectively, trial-and-error cost and opportunity cost are in the wave fluctuation state, the equipment is numerous, operation and maintenance personnel's energy can't be concentrated completely also can lead to judging the emergence of the condition such as lagged behind, omit, can't accomplish early warning in advance and effectual definite time decision-making.
Disclosure of Invention
The invention aims to overcome the defects of the prior art, and provides a scheme capable of effectively diagnosing equipment faults and an equipment fault diagnosis method for providing data support for equipment overhaul expiration reminding, fault diagnosis and decision making.
The invention aims to realize the method, and the method for diagnosing the equipment fault comprises the following steps:
step 1: determining the category, rated parameters and initial life cycle of research equipment;
step 2: establishing an expert system model rule base aiming at the object category characteristics;
and step 3: establishing an object historical database;
and 4, step 4: introducing an artificial neural network to model a high-dimensional nonlinear problem;
and 5: and (4) utilizing the BP neural network to realize fault diagnosis.
The equipment types in the step 1 comprise machinery, a motor and a valve, and the rated parameters of the equipment comprise power, voltage, frequency, rotating speed and temperature;
the expert system model rule base in the step 2 comprises a fault database, a rule base and a knowledge base; the RBF network is a three-layer neural network which comprises an input layer, a hidden layer and an output layer, wherein the transformation from an input space to the hidden layer space is nonlinear, and the transformation from the hidden layer space to the output layer space is linear;
the basic principle of the RBF network is: an RBF is used as a base of a hidden unit to form a hidden layer space, so that an input vector can be directly mapped to the hidden space without being connected through a weight, when the central point of the RBF is determined, the mapping relation is also determined, the mapping from the hidden layer space to an output space is linear, namely the output of a network is the linear weighted sum of the output of the hidden unit, and the weight is a network adjustable parameter; the activation function of the radial basis function neural network can be expressed as:
wherein xpFor the p-th input sample, ciIs the ith center point, h is the node number of the hidden layer, and n is the number of samples or classifications output. The structure of the radial basis function neural network yields the output of the network as:
of course, a least squares penalty function representation is used:
the parameters for the solution are 3: the center and variance of the basis function and the weight from the hidden layer to the output layer;
adopting a self-organizing selection center learning method:
the first step is as follows: in the unsupervised learning process, the center and the variance of the underlying layer basis function are solved;
the second step is that: a supervised learning process is adopted, and weights from the hidden layer to the output layer are solved;
firstly, h centers are selected to be subjected to k-means clustering, and for the radial basis of the Gaussian kernel function, the variance is solved by a formula:
cmaxthe maximum distance between the selected central points is taken;
the connection weight of the neuron between the hidden layer and the output layer can be directly calculated by a least square method, namely, the partial derivative of the loss function with respect to w is solved to be equal to 0, and the calculation formula can be simplified as follows:
selecting parameters with important influences on the running reliability and the service life of the equipment as monitoring parameters, considering the mutual relation, analyzing and optimizing the parameters by an expert, giving parameter weights to the parameters, inputting an extreme working condition, a critical working condition and a normal working condition for multiple times respectively in a training mode, correcting an algorithm and the parameters according to an output result, supplementing a rich rule base (initially a static rule) at the same time, and simulating online running to verify the accuracy and timeliness of the output result according to the real working condition equipment state;
the object historical database in the step 3 collects the historical data values of the objects and the influence factors thereof, particularly the data in the time periods before and after the occurrence of the abnormality, and records the data into the object historical database according to the manufacturer, the model and the type;
the artificial neural network in the step 4 learns an algorithm by utilizing the adaptivity of the ANN, such as back propagation, learns any function by training the connection weight in a multilayer network, is very suitable for approximating a function and fitting a curve based on a Radial Basis Function (RBF) network of the ANN, is suitable for clustering and the like, can establish various models, trains and screens a model algorithm with the highest accuracy;
the basic principle of the BP neural network in the step 5 is as follows: the input signal Xi acts on an output node through an intermediate node (hidden layer point), and an output signal Yk is generated through nonlinear transformation; each sample of the network training comprises an input vector X and an expected output quantity t, and the deviation between a network output value Y and the expected output value t is adjusted by adjusting the weight W between an input node and a hidden nodeijAnd the weight T between the hidden node and the output nodejkAnd a threshold value, which makes the output of the network and the output of the sample as close as possible, and determines the network parameters (weight value and threshold value) corresponding to the minimum error through repeated learning training. The trained BP neural network can automatically process and output the information which is subjected to nonlinear conversion and has the minimum error to the input information of similar samples;
an organization expert gives each evaluation characteristic value as a sample initial value by using a rated parameter, an operation condition, historical operation time and a historical fault weight parameter, and introduces a fault database and a knowledge base obtained after RBF network training as input Xi in the later period, wherein in the training process of the neural network, a transfer activation function of the neural network requires that the data Xi input into the network is between 0 and 1, so that the sample initial value can be processed in such a way, XI is the sample initial value/MAX (sample initial value), and the larger the value is, the more reliable the representation is, and the smaller the maintenance possibility is;
the initial design of the weight can calculate a weight vector by utilizing a square root method from the aspects of the rated performance, the running time, the fault maintenance frequency, the reliability of maintenance spare parts and the like of each object;
the neural transfer function may be generally Sigmoid function f (x) 1/(1+ e)-x)；
In the training process, the weight can be adjusted by utilizing a square error function,
(network output value Yi-expected output value ti)2；
BP neural network evaluation process: after the initial sample, the reliability standard sample and the expected output value sample are input, the error between the correction and the expected output value sample is minimum (after the weight W is adjusted) through the learning training of the neural network continuouslyij) And obtaining a preliminary fault evaluation library.
The invention has the beneficial effects that: the invention relates to an equipment fault diagnosis method, which comprises the following steps: determining the category, rated parameters and initial life cycle of research equipment; establishing an expert mathematical model aiming at the object category characteristics; establishing an object historical database; introducing an artificial neural network to model a high-dimensional nonlinear problem; realizing fault early warning by using a BP neural network; the invention relates to an equipment fault diagnosis method, which introduces technologies such as big data and artificial intelligence algorithm to digitize, model and standardize equipment rated parameters, equipment operation working conditions, equipment operation and maintenance history and experience, gives intelligent learning alternation ability to a diagnosis system, carries out feedback correction on debugging parameters according to result output, and finally displays and reflects the full life cycle state of equipment through a visual platform to provide data support for fault early warning, diagnosis analysis and fault maintenance decision, thereby reducing equipment fault rate, improving diagnosis accuracy and effectiveness and fully exerting economic benefits of equipment operation.
Drawings
Fig. 1 is an operation diagram of an expert system model rule base of an equipment fault diagnosis method according to the present invention.
Fig. 2 is a diagram of a fault diagnosis structure based on an artificial neural network according to an apparatus fault diagnosis method of the present invention.
Fig. 3 is a diagram of a fault handling structure model of an equipment fault diagnosis method according to the present invention.
Fig. 4 is a linear flow chart of an artificial neural network of the equipment fault diagnosis method of the present invention.
Fig. 5 is a BP network topology structure diagram of an apparatus fault diagnosis method of the present invention.
Detailed Description
A method of diagnosing equipment faults, comprising the steps of:
step 1: determining the category, rated parameters and initial life cycle of research equipment;
step 2: establishing an expert system model rule base aiming at the object category characteristics;
and step 3: establishing an object historical database;
and 4, step 4: introducing an artificial neural network to model a high-dimensional nonlinear problem;
and 5: realizing fault diagnosis by using a BP neural network;
the equipment category in the step 1 comprises machinery, motors, valves or other equipment, and the rated parameters of the equipment comprise power, voltage, frequency, rotating speed, temperature and other rated parameters influencing the use of the equipment;
the expert system model rule base in the step 2 comprises a fault database, a rule base and a knowledge base; as a beneficial supplement of an artificial neural network, the RBF neural network is a special case of the artificial neural network, the RBF neural network adopts a deep learning structure model to perform self-adaption extraction of hidden characteristics of monitoring sensing signals from an online data monitoring database, time domain and frequency domain fault information acquisition is expanded, the defect that time domain indexes have poor fault characteristic expression capacity is overcome, and the accuracy of the model is improved; the RBF network is a three-layer neural network which comprises an input layer, a hidden layer and an output layer, wherein the transformation from an input space to the hidden layer space is nonlinear, and the transformation from the hidden layer space to the output layer space is linear;
as shown in fig. 4 and 5, the basic principle of the RBF network is: an RBF is used as a base of a hidden unit to form a hidden layer space, so that an input vector can be directly mapped to the hidden space without being connected through a weight, when a central point of an RBF network is determined, the mapping relation is also determined, the mapping from the hidden layer space to an output space is linear, namely the output of the network is a linear weighted sum of the output of the hidden unit, and the weight is a network adjustable parameter; the activation function of the radial basis function neural network can be expressed as:
wherein xpFor the p-th input sample, ciIs the ith center point, h is the node number of the hidden layer, and n is the number of samples or classifications output. The structure of the radial basis function neural network yields the output of the network as:
of course, a least squares penalty function representation is used:
the parameters for the solution are 3: the center and variance of the basis function and the weight from the hidden layer to the output layer;
adopting a self-organizing selection center learning method:
the first step is as follows: in the unsupervised learning process, the center and the variance of the underlying layer basis function are solved;
the second step is that: a supervised learning process is adopted, and weights from the hidden layer to the output layer are solved;
firstly, h centers are selected to be subjected to k-means clustering, and for the radial basis of the Gaussian kernel function, the variance is solved by a formula:
cmaxthe maximum distance between the selected central points is taken;
the connection weight of the neuron between the hidden layer and the output layer can be directly calculated by a least square method, namely, the partial derivative of the loss function with respect to w is solved to be equal to 0, and the calculation formula can be simplified as follows:
the selection of the parameters can be based on the operation attributes of the equipment such as pressure, rotating speed, water level, voltage, power, temperature and the like, the parameters with important influences on the operation reliability and the service life cycle of the equipment are mainly considered and selected as monitoring parameters, the expert analyzes and optimizes the parameters and gives the parameter weight after the mutual relation is considered, the extreme working condition, the critical working condition and the normal working condition are respectively input for many times under a training mode, an algorithm and the parameters are corrected according to the output result, meanwhile, a rich rule base (initially, a static rule) is supplemented, and the accuracy and the timeliness of the output result are verified according to the real working condition equipment state in the simulation of online operation;
the object historical database in the step 3 collects the historical data values of the objects and the influence factors thereof, particularly the data in the time periods before and after the occurrence of the abnormality, and records the data into the object historical database according to the manufacturer, the model and the type;
the artificial neural network described in step 4 utilizes the adaptivity of the ANN to learn algorithms, such as back propagation, any function is learned by training the connection weights in the multilayer network, and since the BP neural network is suitable for diagnosing a plurality of faults, but the misdiagnosis rate is high, the abnormal fault recognition rate is not high without learning training, the Radial Basis Function (RBF) network based on the ANN is very suitable for approximating functions and fitting curves, the Adaptive Resonance Theory (ART) model is suitable for clustering and the like, can establish various models, train and screen out a model algorithm with the highest accuracy, the RBF network adopts a Gaussian kernel function to diagnose single fault and is superior to a BP network, therefore, according to different points of emphasis on scheduled maintenance early warning and fault analysis of the power plant equipment, a BP neural network and an RBF neural network are mixed to respectively process the scheduled maintenance early warning auxiliary decision and fault judgment, and intelligent fault early warning and diagnosis and analysis of the equipment are comprehensively completed;
the basic principle of the BP neural network in the step 5 is as follows: as shown in fig. 5, an input signal Xi acts on an output node through an intermediate node (hidden layer point), and an output signal Yk is generated through nonlinear transformation, which is simply obtained through an inner product of an input vector and a weight vector and then through a nonlinear transfer functionTo a scalar result; each sample of the network training comprises an input vector X and an expected output quantity t, and the deviation between a network output value Y and the expected output value t is adjusted by adjusting the weight W between an input node and a hidden nodeijAnd the weight T between the hidden node and the output nodejkAnd a threshold value, which makes the output of the network and the output of the sample as close as possible, and determines the network parameters (weight value and threshold value) corresponding to the minimum error through repeated learning training. The trained BP neural network can automatically process and output the information which is subjected to nonlinear conversion and has the minimum error to the input information of similar samples;
an organization expert gives each evaluation characteristic value as a sample initial value for rated parameters, operation conditions (real-time data can be obtained if an equipment online monitoring and acquisition system exists), historical operation time and historical fault weight parameters, and introduces a fault database and a knowledge base obtained after RBF network training as input Xi in the later period;
the initial design of the weight can calculate a weight vector by utilizing a square root method from the aspects of the rated performance, the running time, the fault maintenance frequency, the reliability of maintenance spare parts and the like of each object;
the neural transfer function may be generally Sigmoid function f (x) 1/(1+ e)-x)；
In the training process, the weight can be adjusted by utilizing a square error function,
(network output value Yi-expected output value ti)2；
BP neural network evaluation process: after the initial sample, the reliability standard sample and the expected output value sample are input, the error between the correction and the expected output value sample is minimum (after the weight W is adjusted) through the learning training of the neural network continuouslyij) And obtaining a preliminary equipment fault evaluation library.
The invention has the beneficial effects that: the invention relates to an equipment fault diagnosis method, which comprises the following steps: determining the category, rated parameters and initial life cycle of research equipment; establishing an expert system model rule base aiming at the object category characteristics; establishing an object historical database; introducing an artificial neural network to model a high-dimensional nonlinear problem; realizing fault diagnosis by using a BP neural network; according to the equipment fault diagnosis method, technologies such as big data and artificial intelligence algorithms are introduced to carry out datamation, modeling and standardization on equipment rated parameters, equipment operation conditions, equipment operation and maintenance history and experience, intelligent learning and alternation capabilities are given to a diagnosis system, the debugging parameters are fed back and corrected according to result output, and finally the full life cycle state of the equipment is displayed and reflected through a visual platform, so that data support is provided for fault early warning, diagnosis analysis and fault maintenance decision, the equipment fault rate is reduced, the accuracy and effectiveness of diagnosis are improved, and the equipment operation can fully exert economic benefits.
The accompanying drawings, which are included to provide a further understanding of the invention and are incorporated in and constitute a part of this application, illustrate embodiment(s) of the invention and together with the description serve to explain the invention without limiting the invention,
fig. 1 is an expert system model rule base operation diagram, which is characterized in that an initial fault phenomenon is firstly input, a fault database is set to be in an initial state, then a proper rule is selected from the rule base, a knowledge base is searched to see whether data matched with the initial fault phenomenon exist or not, if the data can be matched, a conclusion is output through diagnosis and reasoning, if the data cannot be matched, an unprocessed rule exists, and the rule base records the novel rule so as to achieve the purpose of continuous learning.
FIG. 2 is a diagram of a fault diagnosis architecture based on an artificial neural network, which is constructed by first establishing a recurrent neural network, generating a test set and a training set through signal acquisition, inputting a training set signal into the training network, inputting a test set signal into a verification network, outputting a diagnosis result through the training network and the verification network, and seeing whether an expected target is achieved, if not, continuing to exercise through the training network, if so, completing the recurrent neural network,
FIG. 3 is a model diagram of a fault early warning processing structure, wherein a rated parameter of a device is input, historical data is read by artificial intelligence, the operation condition of the device is predicted by continuous working time of the device, whether hidden dangers are eliminated and an adjustment mode are adjusted, the state of the device is recorded to the historical data at the same time, the current state of the device is obtained by the artificial intelligence through historical data accumulation experience, a system platform analyzes the state data of the device based on the current working condition, the life cycle stage of the device is evaluated, maintenance time, maintenance strategies and maintenance contents are guided, a maintenance plan and a standby material plan are formed, and the closed-loop management of the data of the device is formed according to a maintenance condition feedback system after the maintenance is completed. The system can greatly improve the accuracy, timeliness and planning of equipment maintenance, and achieve the purposes of equipment fault diagnosis and analysis and auxiliary equipment decision maintenance strategy.
And in the later stage, simulation is carried out by combining with an evaluation object index description database for further training until an accurate training result is obtained, and formal evaluation is carried out by applying a network so as to determine the equipment state and trigger the generation of a production overhaul plan and a material plan.
The method has the advantages that the method continuously learns under real working conditions, and intelligently alternates parameter adjustment through expert adjustment of weight factors, so that functions of early warning of equipment faults, fault diagnosis, reason analysis reference and the like are completed, meanwhile, a fault database, a rule base and a knowledge base are continuously perfected, and a foundation is laid for later-stage deep data mining of fault rules.
Claims (6)
1. An equipment fault diagnosis method is characterized by comprising the following steps:
step 1: determining the category, rated parameters and initial life cycle of research equipment;
step 2: establishing an expert system model rule base aiming at the object category characteristics;
and step 3: establishing an object historical database;
and 4, step 4: introducing an artificial neural network to model a high-dimensional nonlinear problem;
and 5: and (4) utilizing the BP neural network to realize fault diagnosis.
2. The apparatus fault diagnosis method according to claim 1, characterized in that: the equipment types in the step 1 comprise machinery, a motor and a valve, and the rated parameters of the equipment comprise power, voltage, frequency, rotating speed and temperature.
3. The apparatus fault diagnosis method according to claim 1, characterized in that: the expert system model rule base in the step 2 comprises a fault database, a rule base and a knowledge base; the RBF network is a three-layer neural network which comprises an input layer, a hidden layer and an output layer, wherein the transformation from an input space to the hidden layer space is nonlinear, and the transformation from the hidden layer space to the output layer space is linear;
the basic principle of the RBF network is: an RBF is used as a base of a hidden unit to form a hidden layer space, so that an input vector can be directly mapped to the hidden space without being connected through a weight, when the central point of the RBF is determined, the mapping relation is also determined, the mapping from the hidden layer space to an output space is linear, namely the output of a network is the linear weighted sum of the output of the hidden unit, and the weight is a network adjustable parameter; the activation function of the radial basis function neural network can be expressed as:
wherein xpFor the p-th input sample, ciIs the ith center point, h is the node number of the hidden layer, and n is the number of samples or classifications output. The structure of the radial basis function neural network yields the output of the network as:
of course, a least squares penalty function representation is used:
the parameters for the solution are 3: the center and variance of the basis function and the weight from the hidden layer to the output layer;
adopting a self-organizing selection center learning method:
the first step is as follows: in the unsupervised learning process, the center and the variance of the underlying layer basis function are solved;
the second step is that: a supervised learning process is adopted, and weights from the hidden layer to the output layer are solved;
firstly, h centers are selected to be subjected to k-means clustering, and for the radial basis of the Gaussian kernel function, the variance is solved by a formula:
cmaxthe maximum distance between the selected central points is taken;
the connection weight of the neuron between the hidden layer and the output layer can be directly calculated by a least square method, namely, the partial derivative of the loss function with respect to w is solved to be equal to 0, and the calculation formula can be simplified as follows:
the method mainly considers that parameters with important influences on the running reliability and the service life of the equipment are selected as monitoring parameters, and after mutual relations are considered, experts analyze and optimize the parameters and give parameter weights to the parameters, extreme working conditions, critical working conditions and normal working conditions are respectively input for many times in a training mode, an algorithm and the parameters are corrected according to output results, meanwhile, a rich rule base (initially, static rules) is supplemented, and the accuracy and timeliness of the output results are verified according to the real working condition equipment states in a simulation on-line running mode.
4. The apparatus fault diagnosis method according to claim 1, characterized in that: and 3, collecting historical data values of the objects and the influence factors thereof, particularly data in time periods before and after the occurrence of the abnormality, and inputting the data into the historical object database according to manufacturers, models and types.
5. The apparatus fault diagnosis method according to claim 1, characterized in that: the artificial neural network in the step 4 learns an algorithm by utilizing the adaptivity of the ANN, such as back propagation, learns any function by training the connection weight in the multilayer network, is very suitable for approximating functions and curve fitting based on the Radial Basis Function (RBF) network of the ANN, is suitable for clustering and the like, can establish various models, trains and screens the model algorithm with the highest accuracy.
6. The apparatus fault diagnosis method according to claim 1, characterized in that: the basic principle of the BP neural network in the step 5 is as follows: the input signal Xi acts on an output node through an intermediate node (hidden layer point), and an output signal Yk is generated through nonlinear transformation; each sample of the network training comprises an input vector X and an expected output quantity t, and the deviation between a network output value Y and the expected output value t is adjusted by adjusting the weight W between an input node and a hidden nodeijAnd the weight T between the hidden node and the output nodejkAnd a threshold value, which makes the output of the network and the output of the sample as close as possible, and determines the network parameters (weight value and threshold value) corresponding to the minimum error through repeated learning training. The trained BP neural network can automatically process and output the information which is subjected to nonlinear conversion and has the minimum error to the input information of similar samples;
an organization expert gives each evaluation characteristic value as a sample initial value by using a rated parameter, an operation condition, historical operation time and a historical fault weight parameter, and introduces a fault database and a knowledge base obtained after RBF network training as input Xi in the later period, wherein in the training process of the neural network, a transfer activation function of the neural network requires that the data Xi input into the network is between 0 and 1, so that the sample initial value can be processed in such a way, XI is the sample initial value/MAX (sample initial value), and the larger the value is, the more reliable the representation is, and the smaller the maintenance possibility is;
the initial design of the weight can calculate a weight vector by utilizing a square root method from the aspects of the rated performance, the running time, the fault maintenance frequency, the reliability of maintenance spare parts and the like of each object;
the neural transfer function may be generally Sigmoid function f (x) 1/(1+ e)-x)；
In the training process, the weight can be adjusted by utilizing a square error function,
(network output value Yi-expected output value ti)2；
BP neural network evaluation process: after the initial sample, the reliability standard sample and the expected output value sample are input, the error between the correction and the expected output value sample is minimum (after the weight W is adjusted) through the learning training of the neural network continuouslyij) And obtaining a preliminary equipment fault evaluation library.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111491009.3A CN114386312A (en) | 2021-12-08 | 2021-12-08 | Equipment fault diagnosis method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111491009.3A CN114386312A (en) | 2021-12-08 | 2021-12-08 | Equipment fault diagnosis method |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN114386312A true CN114386312A (en) | 2022-04-22 |
Family
ID=81195048
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202111491009.3A Withdrawn CN114386312A (en) | 2021-12-08 | 2021-12-08 | Equipment fault diagnosis method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114386312A (en) |
Cited By (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN115056824A (en) * | 2022-05-06 | 2022-09-16 | 北京和利时系统集成有限公司 | Method and device for determining vehicle control parameters, computer storage medium and terminal |
| CN115374881A (en) * | 2022-10-17 | 2022-11-22 | 图林科技(深圳)有限公司 | Fault diagnosis method for running state of refrigeration equipment |
| CN115409255A (en) * | 2022-08-24 | 2022-11-29 | 广东电网有限责任公司广州供电局 | Method for managing life cycle of electric power material |
| CN115828452A (en) * | 2022-11-16 | 2023-03-21 | 华能国际电力股份有限公司上海石洞口第二电厂 | Belt sampling equipment diagnosis and optimization method and system based on Internet of things and 5G |
| CN116189868A (en) * | 2023-03-02 | 2023-05-30 | 杭州电子科技大学信息工程学院 | Medical equipment management method and system based on big data |
| CN116381480A (en) * | 2023-03-30 | 2023-07-04 | 湖南雪墨电气科技有限公司 | Monitoring method, system and medium of intelligent desulfurization, denitrification and dedusting equipment |
| CN116821774A (en) * | 2023-08-28 | 2023-09-29 | 山东理工昊明新能源有限公司 | Power generation fault diagnosis method based on artificial intelligence |
| CN117074839A (en) * | 2023-10-16 | 2023-11-17 | 南通西塔自动化科技有限公司 | Electromagnetic valve fault diagnosis method and system |
| CN117390468A (en) * | 2023-10-19 | 2024-01-12 | 应急管理部大数据中心 | Dust removal equipment state monitoring and early warning method and system based on data mining |
| CN117830309A (en) * | 2024-03-05 | 2024-04-05 | 深圳华强电子网集团股份有限公司 | Electronic component fault diagnosis system based on artificial intelligence |
-
2021
- 2021-12-08 CN CN202111491009.3A patent/CN114386312A/en not_active Withdrawn
Cited By (17)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN115056824B (en) * | 2022-05-06 | 2023-11-28 | 北京和利时系统集成有限公司 | Method and device for determining vehicle control parameters, computer storage medium and terminal |
| CN115056824A (en) * | 2022-05-06 | 2022-09-16 | 北京和利时系统集成有限公司 | Method and device for determining vehicle control parameters, computer storage medium and terminal |
| CN115409255A (en) * | 2022-08-24 | 2022-11-29 | 广东电网有限责任公司广州供电局 | Method for managing life cycle of electric power material |
| CN115409255B (en) * | 2022-08-24 | 2023-09-08 | 广东电网有限责任公司广州供电局 | Electric power material life cycle management method |
| CN115374881A (en) * | 2022-10-17 | 2022-11-22 | 图林科技(深圳)有限公司 | Fault diagnosis method for running state of refrigeration equipment |
| CN115374881B (en) * | 2022-10-17 | 2023-01-17 | 图林科技(深圳)有限公司 | Fault diagnosis method for running state of refrigeration equipment |
| CN115828452A (en) * | 2022-11-16 | 2023-03-21 | 华能国际电力股份有限公司上海石洞口第二电厂 | Belt sampling equipment diagnosis and optimization method and system based on Internet of things and 5G |
| CN116189868A (en) * | 2023-03-02 | 2023-05-30 | 杭州电子科技大学信息工程学院 | Medical equipment management method and system based on big data |
| CN116381480A (en) * | 2023-03-30 | 2023-07-04 | 湖南雪墨电气科技有限公司 | Monitoring method, system and medium of intelligent desulfurization, denitrification and dedusting equipment |
| CN116821774A (en) * | 2023-08-28 | 2023-09-29 | 山东理工昊明新能源有限公司 | Power generation fault diagnosis method based on artificial intelligence |
| CN116821774B (en) * | 2023-08-28 | 2023-11-28 | 山东理工职业学院 | Power generation fault diagnosis method based on artificial intelligence |
| CN117074839A (en) * | 2023-10-16 | 2023-11-17 | 南通西塔自动化科技有限公司 | Electromagnetic valve fault diagnosis method and system |
| CN117074839B (en) * | 2023-10-16 | 2023-12-26 | 南通西塔自动化科技有限公司 | Electromagnetic valve fault diagnosis method and system |
| CN117390468A (en) * | 2023-10-19 | 2024-01-12 | 应急管理部大数据中心 | Dust removal equipment state monitoring and early warning method and system based on data mining |
| CN117390468B (en) * | 2023-10-19 | 2024-06-04 | 应急管理部大数据中心 | Dust removal equipment state monitoring and early warning method and system based on data mining |
| CN117830309A (en) * | 2024-03-05 | 2024-04-05 | 深圳华强电子网集团股份有限公司 | Electronic component fault diagnosis system based on artificial intelligence |
| CN117830309B (en) * | 2024-03-05 | 2024-05-07 | 深圳华强电子网集团股份有限公司 | Electronic component fault diagnosis system based on artificial intelligence |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN114386312A (en) | Equipment fault diagnosis method | |
| CN111596604B (en) | Intelligent fault diagnosis and self-healing control system and method for engineering equipment based on digital twinning | |
| US11840998B2 (en) | Hydraulic turbine cavitation acoustic signal identification method based on big data machine learning | |
| CN111222549B (en) | Unmanned aerial vehicle fault prediction method based on deep neural network | |
| CN110455537A (en) | A kind of Method for Bearing Fault Diagnosis and system | |
| CN105302848A (en) | Evaluation value calibration method of equipment intelligent early warning system | |
| CN109492790A (en) | Wind turbines health control method based on neural network and data mining | |
| CN111680875A (en) | Unmanned aerial vehicle state risk fuzzy comprehensive evaluation method based on probability baseline model | |
| CN115859077A (en) | Multi-feature fusion motor small sample fault diagnosis method under variable working conditions | |
| CN118157132B (en) | Data mining method and device for voltage monitoring system based on neural network | |
| Hu et al. | A hybrid model for predicting the degradation trend of hydropower units based on deep learning | |
| CN117151684A (en) | Wind power fan data analysis early warning method, system, device and readable storage medium | |
| CN117556347A (en) | Power equipment fault prediction and health management method based on industrial big data | |
| CN114548494B (en) | Visual cost data prediction intelligent analysis system | |
| CN118154174A (en) | Intelligent operation and maintenance cloud platform for industrial equipment | |
| CN116595319A (en) | Prediction method and system applied to rail transit motor health state evaluation | |
| Yao et al. | Multi-step-ahead tool state monitoring using clustering feature-based recurrent fuzzy neural networks | |
| Pérez-Pérez et al. | Fault detection and isolation in wind turbines based on neuro-fuzzy qLPV zonotopic observers | |
| CN113126489A (en) | CNN-GRU-BINN-based intelligent BIT design method for heavy-duty gas turbine control system | |
| CN114548701B (en) | Full-measurement-point-oriented coupling structure analysis and estimation process early warning method and system | |
| CN116662925A (en) | Industrial process soft measurement method based on weighted sparse neural network | |
| CN116401545A (en) | Multimode fusion type turbine runout analysis method | |
| CN102830341A (en) | Online intelligent fault prediction method for power electronic circuit based on RS-CMAC (rough sets and cerebellar model articulation controller) | |
| Gao et al. | DRN-GAN: an integrated deep learning-based health degradation assessment model for naval propulsion system | |
| CN116738689B (en) | Hydroelectric generating set state health assessment system and method based on Elman model |
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 | ||
| WW01 | Invention patent application withdrawn after publication | ||
| WW01 | Invention patent application withdrawn after publication |
Application publication date: 20220422 |