CN115186935B - Electromechanical device nonlinear fault prediction method and system - Google Patents
Electromechanical device nonlinear fault prediction method and system Download PDFInfo
- Publication number
- CN115186935B CN115186935B CN202211092188.8A CN202211092188A CN115186935B CN 115186935 B CN115186935 B CN 115186935B CN 202211092188 A CN202211092188 A CN 202211092188A CN 115186935 B CN115186935 B CN 115186935B
- Authority
- CN
- China
- Prior art keywords
- fault
- model
- electromechanical
- data
- acquiring
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Images
Classifications
-
- 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"
-
- 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
Abstract
The invention discloses a nonlinear fault prediction method and a system for electromechanical equipment, wherein the nonlinear fault prediction method comprises the following steps: 1: acquiring the equipment attribute of the electromechanical equipment, inputting the equipment attribute into a preset equipment model to establish an initial electromechanical model, and S2: acquiring the real-time working state of the electromechanical device, adjusting the initial model state of the initial electromechanical model to the real-time working state, establishing a prediction electromechanical model, and S3: analyzing and predicting the working trend of the electromechanical model in different real-time working states and after the electromechanical model continuously works for a preset time, and S4: and performing nonlinear fault prediction on the working trend based on a nonlinear prediction model to obtain a fault trend, and acquiring a fault device and a fault position corresponding to the fault trend so as to find a fault in time when the electromechanical equipment works and analyze a fault source, thereby reducing the overhaul work of light workers.
Description
Technical Field
The invention relates to the technical field of fault prediction, in particular to a nonlinear fault prediction method and a nonlinear fault prediction system for electromechanical equipment.
Background
The fault diagnosis technology of mechanical equipment in modern production is more and more emphasized, people invest a great deal of energy to research, the fault diagnosis technology of electromechanical equipment is greatly developed, a series of new theoretical methods and technologies are explored and applied to practice, the fault diagnosis efficiency of the electromechanical equipment is increased, a solid foundation for fault diagnosis analysis and repair of the electromechanical equipment is laid, obvious economic benefits and social benefits are generated, due to the fact that the general electromechanical equipment is large in size and high in difficulty in maintenance, particularly when the electromechanical equipment breaks down, whether the electromechanical equipment is linear fault or not is difficult to judge quickly, and a great amount of time and labor are easily wasted.
In view of this, the invention provides a nonlinear fault prediction method and system for electromechanical devices.
Disclosure of Invention
The invention aims to provide a nonlinear fault prediction method and a nonlinear fault prediction system for electromechanical equipment, which are used for establishing a prediction electromechanical model when the electromechanical equipment works, and predicting a fault trend by analyzing a working trend and an unlimited prediction model so as to reduce the overhaul work of light workers.
The invention provides a nonlinear fault prediction method for electromechanical equipment, which comprises the following steps:
s1: acquiring the equipment attribute of the electromechanical equipment, and inputting the equipment attribute into a preset equipment model to establish an initial electromechanical model;
s2: acquiring the real-time working state of the electromechanical equipment, adjusting the initial model state of the initial electromechanical model to the real-time working state, and establishing a prediction electromechanical model;
s3: analyzing and predicting the working trend of the electromechanical model in different real-time working states and after the electromechanical model continuously works for a preset time;
s4: and performing nonlinear fault prediction on the working trend based on a nonlinear prediction model to obtain a fault trend, and acquiring a fault device and a fault position corresponding to the fault trend.
Preferably, the obtaining of the device attribute of the electromechanical device and the inputting of the device attribute into the preset device model to establish the initial electromechanical model includes:
s11: acquiring the equipment attribute of the electromechanical equipment;
s12: analyzing the equipment attribute, and acquiring the equipment name and the structural data of the electromechanical equipment;
s13: and searching a corresponding preset equipment model in the model library according to the equipment name, inputting the structural data into the preset equipment model, and establishing an initial electromechanical model.
Preferably, the acquiring a real-time working state of the electromechanical device, adjusting an initial model state of the initial electromechanical model to the real-time working state, and establishing the predictive electromechanical model includes:
s21: collecting working data of the electromechanical equipment corresponding to different working devices at different time, preprocessing the working data, acquiring a processing result, and generating a real-time working state;
s22: and adjusting the initial model state of the initial electromechanical model according to the real-time working state, and establishing a prediction electromechanical model.
Preferably, analyzing and predicting the working trend of the electromechanical model in different real-time working states and working continuously for a preset time period includes:
s31: controlling the prediction electromechanical model to continuously work under different real-time working states according to corresponding preset duration;
s32: collecting a plurality of pieces of prediction data generated in the continuous working process of the prediction electromechanical model under different real-time working states;
s33: analyzing each predicted data, acquiring data stability corresponding to each predicted data, and extracting normal data in a corresponding preset stability range and abnormal data which are not in the corresponding preset stability range under the same real-time working state;
s34: according to the state attribute under the same real-time working state, and in combination with normal data and abnormal data, constructing a normal stable array and an abnormal stable array corresponding to the same real-time working state, and further constructing and obtaining a comprehensive stable array formed by different real-time working states;
s35: and performing pre-analysis on the comprehensive stable array and the independent stable array based on an array analysis database to obtain the working trend of the preset electromechanical model.
Preferably, based on the nonlinear prediction model, the nonlinear fault prediction is performed on the working trend to obtain a fault trend, and a fault device and a fault position corresponding to the fault trend are obtained, including:
s41: analyzing the working trend, extracting fault types contained in the working trend, and searching a corresponding target device in a preset fault type library;
s42: respectively judging whether each target device is on the electromechanical equipment;
acquiring a fault device contained in the electromechanical equipment according to the judgment result;
s43: and acquiring the fault position of the fault device on the electromechanical equipment, and transmitting the fault position to a specified terminal for displaying.
Preferably, the preprocessing the working data includes:
respectively analyzing each working data to obtain effective information contained in each working data;
performing cluster analysis on all effective information to generate a plurality of cluster samples, and acquiring sample characteristics corresponding to each cluster sample;
extracting an isolated clustering sample only containing one effective message according to the information quantity of the effective message corresponding to each piece of working data contained in the first clustering sample, and obtaining a first sample characteristic of the isolated clustering sample;
acquiring the feature difference between the first sample feature and the residual sample features in the corresponding clustering samples, extracting a second sample feature with the highest similarity to the first sample based on the feature difference, and establishing a feature training pool by taking the second sample feature as an output target feature;
inputting first effective information corresponding to the first sample characteristic into a characteristic training pool for characteristic training, and regarding the trained first effective information as second effective information;
acquiring a redundant area contained in the second effective information, if the area length of the redundant area is greater than a preset length, determining the second effective information as invalid information, and rejecting invalid working data corresponding to the first effective information;
if the area length of the redundant area is smaller than or equal to the preset length, acquiring a target clustering sample corresponding to the second sample characteristic, and supplementing second effective information into the target clustering sample;
and generating a real-time working state according to the effective information related to each cluster sample.
Preferably, the constructing of the abnormal stable array corresponding to the same real-time working state includes:
acquiring a data source device corresponding to each abnormal data, searching a device parameter threshold corresponding to each data source device in the same real-time working state in a preset device library, and acquiring a first data stability corresponding to each abnormal data;
according to the first data stability and corresponding device parameter threshold values, analyzing first fault characteristics corresponding to each data source device;
inputting each first fault characteristic into a cyclic neural network model for cyclic training, and analyzing a cyclic training result to judge whether the first fault characteristic belongs to a linear fault;
if not, acquiring a second fault characteristic generated in the cyclic training process;
establishing a normal operation model for the corresponding data source device according to the device parameter threshold;
inputting abnormal data into a normal operation model, and establishing a first fault model;
operating the first fault model, and collecting target operation faults consistent with the second fault characteristics in the operation process;
analyzing a target operation fault, generating a fault parameter and inputting the fault parameter into a preset coordinate system to acquire a data trend corresponding to the fault parameter;
and acquiring a target device parameter threshold corresponding to the fault parameter, and constructing an abnormal stable array in the same real-time working state by combining the data trend.
Preferably, the analyzing each working data respectively to obtain the effective information included in each working data includes:
acquiring a data volume corresponding to each working data, and establishing an information domain consistent with a numerical relationship according to the numerical relationship between the data volume and the unit data volume;
dividing each working data into a plurality of data segments by taking the unit data volume as a division standard;
inputting each data segment into a corresponding information domain, and respectively obtaining the domain length of each information domain corresponding to the same working data;
sequencing all domain lengths corresponding to the same working data, extracting a first information domain, inputting the first information domain into a preset time domain model, and generating frequency spectrum information;
analyzing the frequency spectrum information, acquiring a plurality of frequency spectrum zero-crossing positions contained in the frequency spectrum information, and further acquiring a plurality of target zero-crossing positions on a first information domain;
respectively acquiring target information corresponding to each target zero-crossing position, and first information and second information adjacent to the target information in a first information domain;
after the target information is eliminated, analyzing the rejection degree between the first information and the second information;
if the rejection degree is greater than the preset degree, target information is reserved, otherwise, the target information is removed, and a first effective information domain is generated;
acquiring domain length difference information between each residual information domain corresponding to the same working data and the corresponding first effective information domain, and generating an extraction mode for extracting information of the corresponding residual information domain;
acquiring the extraction data amount corresponding to each extraction mode, and acquiring the target extraction mode with the maximum extraction data amount;
and acquiring target extraction data corresponding to the target extraction mode, and generating effective information contained in the corresponding working data by combining the first effective information contained in the first effective information domain.
Preferably, the obtaining of the feature difference between the first sample feature and the remaining sample features in the corresponding cluster sample and the extracting of the second sample feature with the highest similarity to the first sample based on the feature difference includes:
extracting all first feature points contained in the first sample feature, and respectively obtaining the first importance of each first feature point in the first sample feature;
extracting all second feature points contained in the corresponding residual sample features, and respectively obtaining a second importance degree of each second feature point in the corresponding residual sample features;
calculating a feature difference between the first sample feature and the remaining sample features;
wherein the content of the first and second substances,represents a characteristic difference between the first sample characteristic and the kth remaining sample characteristic, <' > is>Represents a first degree of importance of the ith first characteristic point in the first sample characteristic>Represents the greatest degree of importance in the first sample characteristic>Represents a minimum degree of importance in the first sample characteristic>Indicates the number of first feature points included in the first sample feature, device for selecting or keeping>A second significance, which represents a jth second feature point in the kth remaining sample feature>Represents the greatest degree of importance in the second sample characteristic>Representing a minimum importance in the second sample feature; wherein it is present>2 is greater than 1;
according to the calculation result, acquiring a feature difference between the first sample feature and the residual sample feature, and establishing a feature difference list;
and screening the sample characteristic corresponding to the minimum characteristic difference from the characteristic difference list to be used as a second sample characteristic.
The invention provides a nonlinear fault prediction system for electromechanical equipment, which comprises:
the acquisition processing module is used for acquiring the equipment attribute of the electromechanical equipment and inputting the equipment attribute into a preset equipment model to establish an initial electromechanical model;
the real-time acquisition module is used for acquiring the real-time working state of the electromechanical equipment, adjusting the initial model state of the initial electromechanical model to the real-time working state and establishing a prediction electromechanical model;
the fault analysis module is used for analyzing and predicting the working trend of the electromechanical model in different real-time working states and after the electromechanical model continuously works for corresponding preset duration;
and the positioning transmission module is used for carrying out nonlinear fault prediction on the working trend based on the nonlinear prediction model to obtain a fault trend and acquiring a fault device and a fault position corresponding to the fault trend.
Compared with the prior art, the beneficial effects of this application are as follows:
when the electromechanical equipment works, a prediction electromechanical model is established, and the fault trend is predicted by analyzing the working trend and by using the non-limiting prediction model, so that the overhaul work of light workers is reduced.
Additional features and advantages of the invention will be set forth in the description which follows, and in part will be obvious from the description, or may be learned by practice of the invention. The objectives and other advantages of the invention will be realized and attained by the structure particularly pointed out in the written description and claims hereof as well as the appended drawings.
The technical solution of the present invention is further described in detail by the accompanying drawings and embodiments.
Drawings
Fig. 1 is a schematic workflow diagram of a nonlinear fault prediction method for an electromechanical device according to an embodiment of the present invention;
fig. 2 is a schematic composition diagram of a nonlinear fault prediction system for an electromechanical device according to an embodiment of the present invention.
Detailed Description
The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
Example 1
The invention provides a nonlinear fault prediction method for electromechanical equipment, which comprises the following steps of:
s1: acquiring the equipment attribute of the electromechanical equipment, and inputting the equipment attribute into a preset equipment model to establish an initial electromechanical model;
s2: acquiring a real-time working state of the electromechanical equipment, adjusting an initial model state of an initial electromechanical model to the real-time working state, and establishing a prediction electromechanical model;
s3: analyzing and predicting the working trend of the electromechanical model in different real-time working states and working continuously for a preset time;
s4: and performing nonlinear fault prediction on the working trend based on a nonlinear prediction model to obtain a fault trend, and acquiring a fault device and a fault position corresponding to the fault trend.
In this embodiment, the device attribute represents a standard for distinguishing the functions of different electromechanical devices, and may be a characteristic that can be clearly distinguished from other devices, such as a specification, a model, a technical parameter, and the like;
in this embodiment, the preset device model is preset, and the corresponding electromechanical model can be automatically established according to different device attributes.
In this embodiment, the initial electromechanical model represents a model consistent with the electromechanical device parameters;
in this embodiment, the real-time operating state represents the operation performed by the electromechanical device at the current time;
in this embodiment, the predictive electromechanical model represents a model that is consistent with the operating state of the electromechanical device;
in this embodiment, the operation trend represents a development trend after continuous operation by the predictive electromechanical model.
In this embodiment, the non-linear prediction model is trained in advance, and is mainly obtained by training samples of different working trends and linear and non-linear faults corresponding to the working trends.
In this embodiment, the nonlinear fault prediction is mainly analyzed for the conditions of vibration, stability and the like in different real-time working states.
The working principle and the beneficial effects of the technical scheme are as follows: when the electromechanical equipment works, a prediction electromechanical model is established, the working trend is analyzed, and the fault trend is predicted through an unlimited prediction model, so that the maintenance work of light workers is reduced.
Example 2
On the basis of embodiment 1, a nonlinear fault prediction method for electromechanical equipment, which obtains an equipment attribute of the electromechanical equipment, inputs the equipment attribute into a preset equipment model to establish an initial electromechanical model, includes:
s11: acquiring the equipment attribute of the electromechanical equipment;
s12: analyzing the equipment attribute, and acquiring the equipment name and the structural data of the electromechanical equipment;
s13: and searching a corresponding preset equipment model in the model library according to the equipment name, inputting the structural data into the preset equipment model, and establishing an initial electromechanical model.
The working principle and the beneficial effects of the technical scheme are as follows: in order to quickly establish an initial electromechanical model, a preset equipment model is searched according to the name of the electromechanical equipment, then the preset equipment model is adjusted according to the structural data of the electromechanical equipment, and finally the initial electromechanical model is generated at the highest speed, so that the modeling time is shortened, and the prediction efficiency is improved.
Example 3
On the basis of embodiment 1, a nonlinear fault prediction method for electromechanical equipment collects a real-time working state of the electromechanical equipment, adjusts an initial model state of an initial electromechanical model to the real-time working state, and establishes a predicted electromechanical model, including:
s21: collecting working data of the electromechanical equipment corresponding to different working devices at different time, preprocessing the working data, acquiring a processing result, and generating a real-time working state;
s22: and adjusting the initial model state of the initial electromechanical model according to the real-time working state, and establishing a prediction electromechanical model.
The working principle and the beneficial effects of the technical scheme are as follows: in order to generate an effective real-time working state, the working data of the electromechanical equipment is acquired and then preprocessed, then an initial electromechanical model is adjusted according to the real-time working state, and finally a prediction electromechanical model is established, so that the accuracy of the prediction electromechanical model is ensured, and the normal operation of the subsequent prediction work is ensured.
Example 4
On the basis of embodiment 1, a nonlinear fault prediction method for electromechanical equipment analyzes and predicts a working trend of an electromechanical model in different real-time working states and after the electromechanical model continuously works for a preset time, and includes:
s31: controlling the prediction electromechanical model to continuously work under different real-time working states according to corresponding preset duration;
s32: collecting and predicting a plurality of pieces of prediction data generated in the continuous working process of the electromechanical model in different real-time working states;
s33: analyzing each predicted data, acquiring data stability corresponding to each predicted data, and extracting normal data in a corresponding preset stability range and abnormal data which are not in the corresponding preset stability range under the same real-time working state;
s34: according to the state attribute under the same real-time working state, and in combination with normal data and abnormal data, a normal stable array and an abnormal stable array corresponding to the same real-time working state are constructed, and then a comprehensive stable array composed of different real-time working states is constructed;
s35: and performing pre-analysis on the comprehensive stable array and the independent stable array based on an array analysis database to obtain the working trend of the preset electromechanical model.
In this embodiment, the preset duration is preset, and the preset durations corresponding to different working states may be different, so as to be convenient for predicting data of the electromechanical model in different working states, and the predicted data is mainly related to the electromechanical operating condition, mainly for acquiring normal data and abnormal data.
In this embodiment, the status attribute refers to the operation type and the operation status of the operation status.
The working principle and the beneficial effects of the technical scheme are as follows: in order to timely stop the failure before the failure occurs, a model is established, then a comprehensive stable array in the same real-time working state is generated, and further analysis is carried out, so that the purpose of prediction is achieved, the failure trend of the electromechanical equipment can be predicted, and the prediction effect is improved.
Example 5
On the basis of embodiment 1, a nonlinear fault prediction method for an electromechanical device, which performs nonlinear fault prediction on a working trend based on a nonlinear prediction model to obtain a fault trend, and obtains a fault device and a fault position corresponding to the fault trend, includes:
s41: analyzing the working trend, extracting fault types contained in the working trend, and searching a corresponding target device in a preset fault type library;
s42: respectively judging whether each target device is on the electromechanical equipment;
acquiring a fault device contained in the electromechanical equipment according to the judgment result;
s43: and acquiring the fault position of the fault device on the electromechanical equipment, and transmitting the fault position to a specified terminal for displaying.
In the embodiment, the preset fault type base represents a statistical base of all faults of the electromechanical equipment;
in this embodiment, the target device represents a failed device corresponding to the failure trend.
The working principle and the beneficial effects of the technical scheme are as follows: in order to quickly locate the position with the fault, the device with the fault is determined according to the fault trend, and then the position of the device is searched on the electromechanical equipment to realize the locating work.
Example 6
On the basis of embodiment 3, a nonlinear fault prediction method for electromechanical equipment, which is a process for preprocessing working data, includes:
respectively analyzing each working data to obtain effective information contained in each working data;
performing cluster analysis on all effective information to generate a plurality of cluster samples, and acquiring sample characteristics corresponding to each cluster sample;
extracting an isolated clustering sample only containing one effective message according to the information quantity of the effective message corresponding to each piece of working data contained in the first clustering sample, and obtaining a first sample characteristic of the isolated clustering sample;
acquiring the feature difference between the first sample feature and the residual sample features in the corresponding clustering samples, extracting a second sample feature with the highest similarity to the first sample based on the feature difference, and establishing a feature training pool by taking the second sample feature as an output target feature;
inputting first effective information corresponding to the first sample characteristic into a characteristic training pool for characteristic training, and regarding the trained first effective information as second effective information;
acquiring a redundant area contained in the second effective information, if the area length of the redundant area is greater than a preset length, determining the second effective information as invalid information, and rejecting invalid working data corresponding to the first effective information;
if the area length of the redundant area is smaller than or equal to the preset length, acquiring a target clustering sample corresponding to the second sample characteristic, and supplementing second effective information into the target clustering sample;
and generating a real-time working state according to the effective information related to each cluster sample.
In this embodiment, the cluster analysis indicates a process of dividing the effective information into a class according to the source of the effective information;
in this embodiment, the cluster samples represent samples composed of source-consistent valid information;
in this embodiment, an isolated sample represents a clustered sample containing only one valid information;
in this embodiment, the redundant area indicates a blank area, a repeated area, and a scrambling code area included in the second valid information.
The working principle and the beneficial effects of the technical scheme are as follows: in order to quickly generate a real-time working state, eliminate redundant data in the working data and perform data cleaning on the working data, in order to avoid mistakenly deleting useful information in the cleaning process, the useful information is subjected to cluster analysis, then an isolated sample is analyzed, the relation between the isolated sample and the residual sample is obtained, and the isolated sample is classified or deleted, so that the cleaning purpose is realized, the real-time working state is quickly generated, the interference of the redundant data is eliminated, and the generated real-time working state is more suitable for the self state of the electromechanical equipment.
Example 7
On the basis of embodiment 6, a method for predicting a nonlinear fault of an electromechanical device, which obtains a feature difference between a first sample feature and a residual sample feature in a corresponding cluster sample, and extracts a second sample feature having the highest similarity with the first sample based on the feature difference, includes:
extracting all first feature points contained in the first sample feature, and respectively obtaining the first importance of each first feature point in the first sample feature;
extracting all second feature points contained in the corresponding residual sample features, and respectively obtaining a second importance degree of each second feature point in the corresponding residual sample features;
calculating a feature difference between the first sample feature and the remaining sample features;
wherein the content of the first and second substances,represents a feature difference between a first sample feature and a kth remaining sample feature>Represents a first degree of importance of the ith first characteristic point in the first sample characteristic>Represents the greatest degree of importance in the first sample feature>Indicates the minimum degree of importance in the first sample characteristic, <' > is>Represents the number of first characteristic points comprised in a first sample characteristic, and>a second significance, which represents a jth second feature point in the kth remaining sample feature>Represents the greatest degree of importance in the second sample characteristic>Representing a minimum importance in the second sample feature; wherein it is present>2 is greater than 1;
according to the calculation result, acquiring a feature difference between the first sample feature and the residual sample feature, and establishing a feature difference list;
and screening the sample characteristic corresponding to the minimum characteristic difference from the characteristic difference list to be used as a second sample characteristic.
In this example, the value of the importance is in the range of [0,1].
In this example, the importance degree represents the specific gravity of the feature corresponding to the feature point in the sample feature.
The working principle and the beneficial effects of the technical scheme are as follows: in order to classify the isolated effective data in time and avoid data loss, a second sample characteristic with the lowest rejection degree with the first sample characteristic is obtained through a formula, and a basis is made for subsequently classifying the data.
Example 8
On the basis of the embodiment 4, the method for constructing the abnormal stable array corresponding to the same real-time working state comprises the following steps:
acquiring a data source device corresponding to each abnormal data, searching a device parameter threshold corresponding to each data source device in the same real-time working state in a preset device library, and acquiring a first data stability corresponding to each abnormal data;
analyzing a first fault characteristic corresponding to each data source device according to the first data stability and the corresponding device parameter threshold;
inputting each first fault feature into a cyclic neural network model for cyclic training, and analyzing a cyclic training result to judge whether the first fault feature belongs to a linear fault;
if not, acquiring a second fault characteristic generated in the cyclic training process;
establishing a normal operation model for the corresponding data source device according to the device parameter threshold;
inputting abnormal data into a normal operation model, and establishing a first fault model;
operating the first fault model, and collecting target operation faults consistent with the second fault characteristics in the operation process;
analyzing a target operation fault, generating a fault parameter and inputting the fault parameter into a preset coordinate system to acquire a data trend corresponding to the fault parameter;
and acquiring a target device parameter threshold corresponding to the fault parameter, and constructing an abnormal stable array in the same real-time working state by combining the data trend.
In this embodiment, the data source device represents a source of the abnormal data on the electromechanical device;
in this embodiment, the device parameter threshold represents the maximum and minimum values of each parameter of the data sourcing device;
in this embodiment, the data flatness indicates the fluctuation intensity of the data;
in this embodiment, the first fault signature is indicative of a fault predicted to occur at the data source device;
in this embodiment, the second failure characteristic represents a new failure of the predictive data source device due to the first failure.
In this embodiment, an abnormally smooth array may refer to an array of abnormally fluctuating intensities over different normal ranges.
The working principle and the beneficial effects of the technical scheme are as follows: in order to quickly predict the fault of the electromechanical equipment, a data source device corresponding to the electromechanical equipment is quickly obtained according to abnormal data, a normal operation model is established according to a parameter threshold value of the device, then fault interference is carried out on the normal operation model by using the abnormal data, generated fault parameters are analyzed, and finally a fault trend of the electromechanical equipment is generated by combining the device parameter threshold value for reference of working personnel.
Example 9
On the basis of embodiment 6, a method for predicting nonlinear faults of electromechanical devices, which respectively analyzes each working datum and obtains effective information included in each working datum, includes:
acquiring a data volume corresponding to each working data, and establishing an information domain consistent with a numerical relationship according to the numerical relationship between the data volume and the unit data volume;
dividing each working data into a plurality of data segments by taking the unit data amount as a division standard;
inputting each data segment into a corresponding information domain, and respectively obtaining the domain length of each information domain corresponding to the same working data;
sequencing all domain lengths corresponding to the same working data, extracting a first information domain, inputting the first information domain into a preset time domain model, and generating frequency spectrum information;
analyzing the frequency spectrum information, acquiring a plurality of frequency spectrum zero-crossing positions contained in the frequency spectrum information, and further acquiring a plurality of target zero-crossing positions on a first information domain;
respectively acquiring target information corresponding to each target zero-crossing position, and first information and second information adjacent to the target information in a first information domain;
after the target information is eliminated, analyzing the rejection degree between the first information and the second information;
if the rejection degree is greater than the preset degree, target information is reserved, otherwise, the target information is removed, and a first effective information domain is generated;
acquiring domain length difference information between each residual information domain corresponding to the same working data and the corresponding first effective information domain, and generating an extraction mode for extracting information of the corresponding residual information domain;
acquiring the extraction data amount corresponding to each extraction mode, and acquiring the target extraction mode with the maximum extraction data amount;
and acquiring target extraction data corresponding to the target extraction mode, and generating effective information contained in the corresponding working data by combining the first effective information contained in the first effective information domain.
In this embodiment, the information field indicates an area for storing the data segment;
in this embodiment, the preset time domain model represents a model for converting an information domain into spectral information;
in this embodiment, the spectrum zero-crossing position represents a point of 0 in the spectrum information.
In this embodiment, the unit data amount is preset, for example, the unit data amount is 10, the data amount of the corresponding working data is 15, at this time, the numerical relationship is 15-10, and an information field consistent with 15-10 is established.
In this embodiment, for example, the data amount 15 is divided into two segments, one segment is 10 and one segment is 5.
In this embodiment, for example, the field length of an information field with a field length of 10 is 1, and the field length of an information field with a field length of 5 is 2, at this time, the field length of the information field corresponding to the same working data is 1.5, and the corresponding first information field may be the information field corresponding to the largest field length.
In this embodiment, the preset time domain model is preset.
In this embodiment, the frequency information refers to a distribution curve of the first information domain.
In this embodiment, the target information, the first information, and the second information all refer to related work related information and the like.
In this embodiment, the degree of rejection refers to the correlation between the first information and the second information, and the less correlated the degree of rejection is.
In this embodiment, the preset degree is preset.
In this embodiment, the remaining information field refers to the remaining information field excluding the first valid information field in the information field corresponding to the same working data.
In this embodiment, according to the domain length difference information, the corresponding extraction manner is obtained from the length difference-extraction database, and the length difference-extraction database includes different domain length differences and extraction manners corresponding to the different differences.
In this embodiment, the valid information of the same working data includes: the analysis of the most effective data can be ensured, including the target extraction data extracted by the target extraction mode and the first effective information.
The working principle and the beneficial effects of the technical scheme are as follows: the effective information is the core of the working data, so the aim of not missing and comprehensively acquiring is achieved when the effective information is acquired, in order to achieve the aim, the working data is divided into a plurality of data sections with data volume as unit data volume, then the data sections are input into an information domain, a plurality of useful information points are extracted by processing the information domain, and finally the useful information is formed to be used as the basis for subsequent prediction work, so that accurate prediction is achieved.
Example 10
The invention provides a nonlinear fault prediction system of electromechanical equipment, as shown in fig. 2, comprising:
the acquisition processing module is used for acquiring the equipment attribute of the electromechanical equipment and inputting the equipment attribute into a preset equipment model to establish an initial electromechanical model;
the real-time acquisition module is used for acquiring the real-time working state of the electromechanical equipment, adjusting the real-time model state of the initial electromechanical model to the real-time working state and establishing a prediction electromechanical model;
the fault analysis module is used for analyzing and predicting the fault trend of the electromechanical model in a real-time working state after the electromechanical model continuously works for a preset time;
and the positioning transmission module is used for analyzing the fault device corresponding to the fault trend, acquiring the fault position of the fault device on the electromechanical equipment and transmitting the fault position to the specified terminal for displaying.
The working principle and the beneficial effects of the technical scheme are as follows: the fault prediction is realized by mutual cooperation of a plurality of modules, an implementation basis is established for the method, an operation platform of the method is guaranteed, the multiple modules cooperate, the prediction speed is accelerated by the aid of work division cooperation, and the prediction precision is improved.
The above description is only for the preferred embodiment of the present invention, but the scope of the present invention is not limited thereto, and any person skilled in the art should be considered as the technical solutions and the inventive concepts of the present invention within the technical scope of the present invention.
Claims (7)
1. A nonlinear fault prediction method for electromechanical equipment is characterized by comprising the following steps:
s1: acquiring the equipment attribute of the electromechanical equipment, and inputting the equipment attribute into a preset equipment model to establish an initial electromechanical model;
s2: acquiring a real-time working state of the electromechanical equipment, adjusting an initial model state of an initial electromechanical model to the real-time working state, and establishing a prediction electromechanical model;
s3: analyzing and predicting the working trend of the electromechanical model in different real-time working states and after the electromechanical model continuously works for a preset time;
s4: performing nonlinear fault prediction on the working trend based on a nonlinear prediction model to obtain a fault trend, and acquiring a fault device and a fault position corresponding to the fault trend;
the method comprises the following steps of obtaining the equipment attribute of the electromechanical equipment, inputting the equipment attribute into a preset equipment model to establish an initial electromechanical model, and comprises the following steps:
s11: acquiring the equipment attribute of the electromechanical equipment;
s12: analyzing the equipment attribute, and acquiring the equipment name and the structural data of the electromechanical equipment;
s13: searching a corresponding preset equipment model in a model library according to the equipment name, inputting structural data into the preset equipment model, and establishing an initial electromechanical model;
the analyzing and predicting of the working trend of the electromechanical model in different real-time working states and after the electromechanical model continuously works for a preset time length comprises the following steps:
s31: controlling and predicting the electromechanical model to continuously work in different real-time working states according to corresponding preset time length;
s32: collecting a plurality of pieces of prediction data generated in the continuous working process of the prediction electromechanical model under different real-time working states;
s33: analyzing each predicted data, acquiring data stability corresponding to each predicted data, and extracting normal data in a corresponding preset stability range and abnormal data which are not in the corresponding preset stability range under the same real-time working state;
s34: according to the state attribute under the same real-time working state, and in combination with normal data and abnormal data, constructing a normal stable array and an abnormal stable array corresponding to the same real-time working state, and further constructing and obtaining a comprehensive stable array formed by different real-time working states;
s35: performing pre-analysis on the comprehensive stable array and the independent stable array based on an array analysis database to obtain the working trend of a preset electromechanical model;
the method comprises the following steps of performing nonlinear fault prediction on a working trend based on a nonlinear prediction model to obtain a fault trend, and acquiring a fault device and a fault position corresponding to the fault trend, wherein the method comprises the following steps:
s41: analyzing the working trend, extracting fault types contained in the working trend, and searching a corresponding target device in a preset fault type library;
s42: respectively judging whether each target device is on the electromechanical equipment;
acquiring a fault device contained in the electromechanical equipment according to a judgment result;
s43: and acquiring the fault position of the fault device on the electromechanical equipment, and transmitting the fault position to a specified terminal for displaying.
2. The method of claim 1, wherein the collecting the real-time operating state of the electromechanical device and adjusting the initial model state of the initial electromechanical model to the real-time operating state to create the predictive electromechanical model comprises:
s21: collecting working data of the electromechanical equipment corresponding to different working devices at different time, preprocessing the working data, acquiring a processing result, and generating a real-time working state;
s22: and adjusting the initial model state of the initial electromechanical model according to the real-time working state, and establishing a prediction electromechanical model.
3. The electromechanical device nonlinear fault prediction method of claim 2, wherein preprocessing the operational data comprises:
respectively analyzing each working data to obtain effective information contained in each working data;
performing cluster analysis on all effective information to generate a plurality of cluster samples, and acquiring sample characteristics corresponding to each cluster sample;
extracting an isolated clustering sample only containing one effective message according to the information quantity of the effective message corresponding to each piece of working data contained in the first clustering sample, and obtaining a first sample characteristic of the isolated clustering sample;
acquiring the feature difference between the first sample feature and the residual sample features in the corresponding clustering samples, extracting a second sample feature with the highest similarity with the first sample based on the feature difference, and establishing a feature training pool by taking the second sample feature as an output target feature;
inputting first effective information corresponding to the first sample characteristic into a characteristic training pool for characteristic training, and regarding the trained first effective information as second effective information;
acquiring a redundant area contained in the second effective information, if the area length of the redundant area is greater than a preset length, determining the second effective information as invalid information, and rejecting invalid working data corresponding to the first effective information;
if the region length of the redundant region is smaller than or equal to the preset length, acquiring a target clustering sample corresponding to the second sample characteristic, and supplementing second effective information into the target clustering sample;
and generating a real-time working state according to the effective information related to each cluster sample.
4. The method of nonlinear fault prediction of electromechanical devices of claim 1 wherein constructing an abnormally smooth array corresponding to the same real-time operating condition comprises:
acquiring a data source device corresponding to each abnormal data, searching a device parameter threshold corresponding to each data source device in the same real-time working state in a preset device library, and acquiring a first data stability corresponding to each abnormal data;
analyzing a first fault characteristic corresponding to each data source device according to the first data stability and the corresponding device parameter threshold;
inputting each first fault characteristic into a cyclic neural network model for cyclic training, and analyzing a cyclic training result to judge whether the first fault characteristic belongs to a linear fault;
if not, acquiring a second fault characteristic generated in the cyclic training process;
establishing a normal operation model for the corresponding data source device according to the device parameter threshold;
inputting abnormal data into a normal operation model, and establishing a first fault model;
operating the first fault model, and collecting target operation faults consistent with the second fault characteristics in the operation process;
analyzing a target operation fault, generating a fault parameter and inputting the fault parameter into a preset coordinate system to acquire a data trend corresponding to the fault parameter;
and acquiring a target device parameter threshold corresponding to the fault parameter, and constructing an abnormal stable array in the same real-time working state by combining the data trend.
5. The method according to claim 3, wherein the analyzing each working datum to obtain valid information included in each working datum comprises:
acquiring a data volume corresponding to each working data, and establishing an information domain consistent with a numerical relationship according to the numerical relationship between the data volume and the unit data volume;
dividing each working data into a plurality of data segments by taking the unit data amount as a division standard;
inputting each data segment into a corresponding information domain, and respectively obtaining the domain length of each information domain corresponding to the same working data;
sequencing all domain lengths corresponding to the same working data, extracting a first information domain, inputting the first information domain into a preset time domain model, and generating frequency spectrum information;
analyzing the frequency spectrum information, acquiring a plurality of frequency spectrum zero-crossing positions contained in the frequency spectrum information, and further acquiring a plurality of target zero-crossing positions on a first information domain;
respectively acquiring target information corresponding to each target zero-crossing position, and first information and second information adjacent to the target information in a first information domain;
after the target information is eliminated, analyzing the rejection degree between the first information and the second information;
if the rejection degree is greater than the preset degree, target information is reserved, otherwise, the target information is removed, and a first effective information domain is generated;
acquiring domain length difference information between each residual information domain corresponding to the same working data and the corresponding first effective information domain, and generating an extraction mode for extracting information of the corresponding residual information domain;
acquiring the extraction data amount corresponding to each extraction mode, and acquiring the target extraction mode with the maximum extraction data amount;
and acquiring target extraction data corresponding to the target extraction mode, and generating effective information contained in the corresponding working data by combining the first effective information contained in the first effective information domain.
6. The nonlinear fault prediction method for electromechanical equipment as recited in claim 3, wherein obtaining feature differences between the first sample feature and the remaining sample features in the corresponding cluster samples, and extracting a second sample feature with the highest similarity to the first sample based on the feature differences comprises:
extracting all first feature points contained in the first sample feature, and respectively obtaining the first importance of each first feature point in the first sample feature;
extracting all second feature points contained in the corresponding residual sample features, and respectively obtaining a second importance degree of each second feature point in the corresponding residual sample features;
calculating a feature difference between the first sample feature and the remaining sample features;
wherein the content of the first and second substances,represents a characteristic difference between the first sample characteristic and the kth remaining sample characteristic, <' > is>Represents a first degree of importance of the ith first characteristic point in the first sample characteristic>Represents the greatest degree of importance in the first sample characteristic>Indicates the minimum degree of importance in the first sample characteristic, <' > is>Indicates the number of first feature points included in the first sample feature, device for selecting or keeping>A second significance, which represents a jth second feature point in the kth remaining sample feature>Representing the greatest importance in the second sample feature,representing a minimum importance in the second sample feature; wherein max (n, m) -2 is greater than 1;
according to the calculation result, obtaining a feature difference between the first sample feature and the residual sample features, and establishing a feature difference list;
and screening the sample characteristic corresponding to the minimum characteristic difference from the characteristic difference list as a second sample characteristic.
7. An electromechanical device nonlinear fault prediction system, comprising:
the acquisition processing module is used for acquiring the equipment attribute of the electromechanical equipment and inputting the equipment attribute into a preset equipment model to establish an initial electromechanical model;
the real-time acquisition module is used for acquiring the real-time working state of the electromechanical equipment, adjusting the initial model state of the initial electromechanical model to the real-time working state and establishing a prediction electromechanical model;
the fault analysis module is used for analyzing and predicting the working trend of the electromechanical model in different real-time working states after the electromechanical model continuously works for a preset time;
the positioning transmission module is used for carrying out nonlinear fault prediction on the working trend based on a nonlinear prediction model to obtain a fault trend and acquiring a fault device and a fault position corresponding to the fault trend;
the method comprises the following steps of obtaining the equipment attribute of the electromechanical equipment, inputting the equipment attribute into a preset equipment model to establish an initial electromechanical model, and comprises the following steps:
s11: acquiring the equipment attribute of the electromechanical equipment;
s12: analyzing the equipment attribute, and acquiring the equipment name and the structural data of the electromechanical equipment;
s13: searching a corresponding preset equipment model in a model library according to the equipment name, inputting structural data into the preset equipment model, and establishing an initial electromechanical model;
the analyzing and predicting of the working trend of the electromechanical model in different real-time working states and after the electromechanical model continuously works for a preset time length comprises the following steps:
s31: controlling and predicting the electromechanical model to continuously work in different real-time working states according to corresponding preset time length;
s32: collecting and predicting a plurality of pieces of prediction data generated in the continuous working process of the electromechanical model in different real-time working states;
s33: analyzing each predicted data, acquiring data stability corresponding to each predicted data, and extracting normal data in a corresponding preset stability range and abnormal data which are not in the corresponding preset stability range under the same real-time working state;
s34: according to the state attribute under the same real-time working state, and in combination with normal data and abnormal data, constructing a normal stable array and an abnormal stable array corresponding to the same real-time working state, and further constructing and obtaining a comprehensive stable array formed by different real-time working states;
s35: performing pre-analysis on the comprehensive stable array and the independent stable array based on an array analysis database to obtain the working trend of a preset electromechanical model;
the method comprises the following steps of performing nonlinear fault prediction on a working trend based on a nonlinear prediction model to obtain a fault trend, and acquiring a fault device and a fault position corresponding to the fault trend, wherein the method comprises the following steps:
s41: analyzing the working trend, extracting fault types contained in the working trend, and searching a corresponding target device in a preset fault type library;
s42: respectively judging whether each target device is on the electromechanical equipment;
acquiring a fault device contained in the electromechanical equipment according to the judgment result;
s43: and acquiring the fault position of the fault device on the electromechanical equipment, and transmitting the fault position to a specified terminal for displaying.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202211092188.8A CN115186935B (en) | 2022-09-08 | 2022-09-08 | Electromechanical device nonlinear fault prediction method and system |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202211092188.8A CN115186935B (en) | 2022-09-08 | 2022-09-08 | Electromechanical device nonlinear fault prediction method and system |
Publications (2)
Publication Number | Publication Date |
---|---|
CN115186935A CN115186935A (en) | 2022-10-14 |
CN115186935B true CN115186935B (en) | 2023-04-07 |
Family
ID=83523894
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202211092188.8A Active CN115186935B (en) | 2022-09-08 | 2022-09-08 | Electromechanical device nonlinear fault prediction method and system |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN115186935B (en) |
Families Citing this family (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN117077040B (en) * | 2023-09-04 | 2024-02-23 | 武汉蓝海科创技术有限公司 | Large-scale complex equipment fault diagnosis and prediction system based on machine learning |
Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110514967A (en) * | 2019-10-12 | 2019-11-29 | 国网江苏省电力有限公司徐州供电分公司 | A kind of Fault Locating Method based on NARNN model prediction wavefront arrival time |
CN111815056A (en) * | 2020-07-10 | 2020-10-23 | 中国人民解放军空军工程大学 | Aircraft external field aircraft fuel system fault prediction method based on flight parameter data |
CN113077172A (en) * | 2021-04-19 | 2021-07-06 | 瑞湖智科数据(苏州)有限公司 | Equipment state trend analysis and fault diagnosis method |
CN113704075A (en) * | 2021-09-23 | 2021-11-26 | 中国人民解放军国防科技大学 | Fault log-based high-performance computing system fault prediction method |
Family Cites Families (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US6682669B2 (en) * | 2001-09-29 | 2004-01-27 | Van Dorn Demag Corporation | Model predictive control apparatus and methods for motion and/or pressure control of injection molding machines |
CN101799367B (en) * | 2010-01-27 | 2011-08-10 | 北京信息科技大学 | Electromechanical device neural network failure trend prediction method |
CN101799368B (en) * | 2010-01-27 | 2011-05-25 | 北京信息科技大学 | Electromechanical device nonlinear failure prediction method |
CN108508863A (en) * | 2017-10-12 | 2018-09-07 | 上海智容睿盛智能科技有限公司 | A kind of electromechanical equipment method for diagnosing faults based on gray model |
CN112810504B (en) * | 2021-01-20 | 2023-03-31 | 同济大学 | Fuel cell automobile energy management method based on nonlinear prediction model control |
-
2022
- 2022-09-08 CN CN202211092188.8A patent/CN115186935B/en active Active
Patent Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110514967A (en) * | 2019-10-12 | 2019-11-29 | 国网江苏省电力有限公司徐州供电分公司 | A kind of Fault Locating Method based on NARNN model prediction wavefront arrival time |
CN111815056A (en) * | 2020-07-10 | 2020-10-23 | 中国人民解放军空军工程大学 | Aircraft external field aircraft fuel system fault prediction method based on flight parameter data |
CN113077172A (en) * | 2021-04-19 | 2021-07-06 | 瑞湖智科数据(苏州)有限公司 | Equipment state trend analysis and fault diagnosis method |
CN113704075A (en) * | 2021-09-23 | 2021-11-26 | 中国人民解放军国防科技大学 | Fault log-based high-performance computing system fault prediction method |
Also Published As
Publication number | Publication date |
---|---|
CN115186935A (en) | 2022-10-14 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN105677791B (en) | For analyzing the method and system of the operation data of wind power generating set | |
CN111259947A (en) | Power system fault early warning method and system based on multi-mode learning | |
CN113870260A (en) | Welding defect real-time detection method and system based on high-frequency time sequence data | |
CN115186935B (en) | Electromechanical device nonlinear fault prediction method and system | |
CN110968703B (en) | Method and system for constructing abnormal metering point knowledge base based on LSTM end-to-end extraction algorithm | |
CN105930255A (en) | Method and apparatus for predicting health degree of system | |
CN107742162B (en) | Multidimensional feature association analysis method based on allocation monitoring information | |
CN113689079A (en) | Transformer area line loss prediction method and system based on multivariate linear regression and cluster analysis | |
CN112508278A (en) | Multi-connected system load prediction method based on evidence regression multi-model | |
CN116561519A (en) | Electric carbon emission process monitoring method and system based on big data of power grid | |
CN113902222B (en) | New forms of energy wisdom fortune dimension system | |
CN115392663A (en) | Data acquisition and processing method based on big data | |
CN112257937B (en) | Power distribution network fault prediction system and method based on big data technology | |
CN110298585B (en) | Hierarchical automatic auditing method for monitoring information of substation equipment | |
CN113888231A (en) | Marketing data screening analysis system based on artificial intelligence | |
CN109767062B (en) | Dynamic generation method of power grid task disposal scheme | |
CN113515560A (en) | Vehicle fault analysis method and device, electronic equipment and storage medium | |
CN111709594A (en) | Economic management data analysis system | |
CN111210147A (en) | Sintering process operation performance evaluation method and system based on time sequence feature extraction | |
CN110569277A (en) | Method and system for automatically identifying and classifying configuration data information | |
CN113159131B (en) | Hierarchical prediction method and hierarchical prediction system for running conditions of bioreactor | |
CN110261693A (en) | A kind of online thermal test method and system based on stable state sampling technique | |
CN117277566B (en) | Power grid data analysis power dispatching system and method based on big data | |
CN112149969A (en) | Method and system for evaluating running state of extra-high voltage direct current control protection merging unit | |
CN117455406A (en) | Automatic man-hour collecting and analyzing method and system |
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 |