CN105740822B - A kind of mechanical fault diagnosis optimization method and system - Google Patents
A kind of mechanical fault diagnosis optimization method and system Download PDFInfo
- Publication number
- CN105740822B CN105740822B CN201610069441.6A CN201610069441A CN105740822B CN 105740822 B CN105740822 B CN 105740822B CN 201610069441 A CN201610069441 A CN 201610069441A CN 105740822 B CN105740822 B CN 105740822B
- Authority
- CN
- China
- Prior art keywords
- diagnosis
- sample
- fault
- unknown
- state
- 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
- 238000003745 diagnosis Methods 0.000 title claims abstract description 501
- 238000000034 method Methods 0.000 title claims abstract description 28
- 238000005457 optimization Methods 0.000 title claims abstract description 14
- 238000012549 training Methods 0.000 claims abstract description 51
- 238000002360 preparation method Methods 0.000 claims abstract description 15
- 238000004458 analytical method Methods 0.000 claims abstract description 5
- 238000012545 processing Methods 0.000 claims description 33
- 238000004891 communication Methods 0.000 claims description 14
- 238000012795 verification Methods 0.000 claims description 12
- 230000035772 mutation Effects 0.000 claims description 9
- 238000013461 design Methods 0.000 claims description 8
- 238000003860 storage Methods 0.000 claims description 8
- 238000012986 modification Methods 0.000 claims description 5
- 230000004048 modification Effects 0.000 claims description 5
- 238000012790 confirmation Methods 0.000 claims description 4
- 238000010276 construction Methods 0.000 claims description 4
- 238000000354 decomposition reaction Methods 0.000 claims description 3
- 238000012423 maintenance Methods 0.000 claims description 3
- 238000005516 engineering process Methods 0.000 abstract description 7
- 238000004519 manufacturing process Methods 0.000 abstract description 7
- 230000007246 mechanism Effects 0.000 abstract description 2
- 238000003556 assay Methods 0.000 abstract 1
- 238000012512 characterization method Methods 0.000 abstract 1
- 230000003028 elevating effect Effects 0.000 abstract 1
- 238000010586 diagram Methods 0.000 description 10
- 238000013528 artificial neural network Methods 0.000 description 4
- 238000013473 artificial intelligence Methods 0.000 description 2
- 238000001514 detection method Methods 0.000 description 2
- 238000011161 development Methods 0.000 description 2
- 238000001228 spectrum Methods 0.000 description 2
- 230000004075 alteration Effects 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000006243 chemical reaction Methods 0.000 description 1
- 238000012217 deletion Methods 0.000 description 1
- 230000037430 deletion Effects 0.000 description 1
- 238000000605 extraction Methods 0.000 description 1
- 230000004927 fusion Effects 0.000 description 1
- 230000010365 information processing Effects 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 238000012706 support-vector machine Methods 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2218/00—Aspects of pattern recognition specially adapted for signal processing
- G06F2218/12—Classification; Matching
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2218/00—Aspects of pattern recognition specially adapted for signal processing
- G06F2218/08—Feature extraction
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Artificial Intelligence (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Signal Processing (AREA)
- Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Test And Diagnosis Of Digital Computers (AREA)
- Testing And Monitoring For Control Systems (AREA)
Abstract
The invention discloses a kind of mechanical fault diagnosis optimization method and systems, its method be include preparation process and diagnosis algorithm, and in preparation process, building is comprising the diagnostic sample library of whole known diagnosis samples at present and constructs several unknown diagnostic samples, further according to known diagnosis sample and unknown diagnostic sample, training simultaneously generates fault diagnosis algorithm;In diagnosis algorithm, with the diagnostic sample for the mechanical current operating conditions of characterization that the diagnosis of training fault diagnosis algorithm is obtained from scene, if diagnostic result is unknown operating status, Mechanical Running Condition is determined using technologies such as analysis expert or field assays, and diagnostic sample is updated to known diagnosis sample, and after deleting unknown diagnostic sample similar with the known diagnosis sample of recent renewal, re -training fault diagnosis algorithm.The present invention can be realized diagnostic sample library and the dynamic of fault diagnosis algorithm is perfect, improve the accuracy rate of fault diagnosis, so that the running quality of elevating mechanism equipment, improves production efficiency.
Description
Technical Field
The invention relates to the field of mechanical fault detection, in particular to a mechanical fault diagnosis optimization method and system.
Background
With the rapid development of modern industry and science and technology, the structure of modern equipment is more and more complex, the function is more and more perfect, the automation degree is higher and more, not only different parts of the same equipment are mutually related and tightly combined, but also different equipment have close relation, and form a whole in the production process. Therefore, one fault can cause chain reaction, so that the whole equipment and even the whole production process can not normally operate, and the diagnosis of mechanical faults is of great significance.
The fault diagnosis is an emerging comprehensive interdisciplinary which is inspired from medical examination and diagnosis and developed along with the development of modern scientific technologies such as system engineering, information theory, cybernetics, electronic technology, computer technology, information processing, artificial intelligence and the like, and the research content of the fault diagnosis relates to the technical fields of fault mechanisms, sensors and detection technology, signal analysis and data processing, automatic control, system identification, expert systems, computer software and hardware and the like.
At present, due to the technical limitation and the relative lack of fault samples, the fault diagnosis technology applied in the actual production process always has the problem of misdiagnosis, which is unacceptable in individual fields, so that a perfect fault sample library is established, and the design of a fault diagnosis algorithm with high diagnosis accuracy is an urgent problem to be solved.
Disclosure of Invention
The invention aims to: aiming at the technical problems of low accuracy of mechanical fault identification and incomplete fault samples at present, a mechanical fault diagnosis optimization method and a mechanical fault diagnosis optimization system are provided.
In order to achieve the purpose of the invention, the invention adopts the technical scheme that: a mechanical fault diagnosis optimization method comprises a preparation step and a diagnosis step, wherein the preparation step comprises the following steps: firstly, a diagnosis sample library containing all known diagnosis samples at present is constructed, then a plurality of unknown diagnosis samples are constructed according to the known diagnosis samples, and the unknown diagnosis samples are stored in the diagnosis sample library; the operation states of the machine are divided into a normal state, a potential fault state and a fault state, the known diagnosis samples are respectively used for representing the known normal state, the known potential fault state and the known fault state of the machine, and the unknown diagnosis samples are used for representing the unknown operation state of the machine;
finally, training and generating a fault diagnosis algorithm according to all the known diagnosis samples and the unknown diagnosis samples in the diagnosis sample library, wherein the diagnosis results of the fault diagnosis algorithm respectively correspond to the known normal state, the known potential fault state, the known fault state and the unknown operation state;
the diagnosing step comprises:
the first step is as follows: diagnosing a diagnosis sample which is obtained from the site and represents the current operation state of the machine by using the fault diagnosis algorithm, and finally outputting a diagnosis result;
the second step is that: if the diagnosis result in the first step is the unknown operation state, determining the mechanical operation state represented by the diagnosis sample, and updating the diagnosis sample with the characterized mechanical operation state determined to be the known diagnosis sample;
the third step: after deleting the unknown diagnosis samples similar to the recently updated known diagnosis samples, retraining and generating a fault diagnosis algorithm according to all the known diagnosis samples and the unknown diagnosis samples in the diagnosis sample library;
the fourth step: and replacing the current fault diagnosis algorithm with the fault diagnosis algorithm generated by retraining in the third step.
According to a particular embodiment, in said first step, said known diagnostic sample, which characterizes said known normal state, is subjected to a mutation process to construct said unknown diagnostic sample; wherein,
the mutation treatment comprises the following steps: randomly superposing a wavelet packet obtained by wavelet analysis on the known diagnosis sample representing the known fault state onto the known diagnosis sample representing the known normal state; or the intrinsic mode function obtained by EMD decomposition of the known diagnosis sample representing the known fault state is superposed on the known diagnosis sample representing the known normal state.
According to a specific embodiment, in the second step, the diagnostic sample is analyzed by a fault diagnosis expert system or an expert person, or a field maintenance person overhauls a mechanical device diagnosed with an unknown operating state to determine a mechanical operating state characterized by the diagnostic sample, and the diagnostic sample of the determined characterized mechanical operating state is updated to the known diagnostic sample.
According to a specific embodiment, in the third step, the unknown diagnostic sample with the frequency band energy ratio of the fault feature exceeding the set range is deleted by extracting the fault features of the recently updated known diagnostic sample and the unknown diagnostic sample in the diagnostic sample library and comparing the frequency band energy ratios of the fault features of the unknown diagnostic sample and the recently updated known diagnostic sample.
According to a specific embodiment, training to generate the fault diagnosis algorithm includes:
extracting fault characteristics of the known diagnosis sample and the unknown diagnosis sample, taking the fault characteristics as input of the fault diagnosis algorithm, and training the diagnosis result of the fault diagnosis algorithm to respectively correspond to the known normal state or the known potential fault state or the known fault state represented by the known diagnosis sample or the unknown operation state represented by the unknown diagnosis sample;
and after the fault diagnosis algorithm is trained and generated, verifying the diagnosis accuracy of the fault diagnosis algorithm generated by training through a verification sample library, if the diagnosis accuracy is not less than a design value, successfully training, and if not, adjusting the fault diagnosis algorithm and re-training until the training is successful.
According to a specific implementation manner, in the diagnosing step, a diagnostic sample which is obtained from the site and is used for characterizing the current operation state of the machine is diagnosed by using the fault diagnosis algorithm, and after the diagnosis result is finally output, if the diagnosis result does not accord with the current operation state of the machine which is characterized by the diagnostic sample, the machine operation state which is characterized by the diagnostic sample is re-determined, and the diagnostic sample which is determined to characterize the machine operation state is updated to be the known diagnostic sample.
According to a particular embodiment, the system comprises an on-site processing unit and a server, wherein,
the field processing unit is used for acquiring a diagnosis sample representing the current operation state of the machine from the field, acquiring a fault diagnosis algorithm from the server, diagnosing the diagnosis sample by using the fault diagnosis algorithm, and if the diagnosis result of the diagnosis sample is an unknown operation state, sending the diagnosis sample of the current fault diagnosis to the server;
the server is used for constructing a diagnosis sample library containing all known diagnosis samples at present and constructing a plurality of unknown diagnosis samples in the preparation step, the known diagnosis samples are respectively used for representing the known normal state, the known potential fault state and the known fault state of the machine, and the unknown diagnosis samples are used for representing the unknown operation state of the machine; in the diagnosis step, when the diagnosis result obtained after the diagnosis of the diagnosis sample is the unknown operation state, the on-site processing unit determines that the operation state of the machine represented by the diagnosis sample is a normal state or a potential fault state or a fault state, and deletes the unknown diagnosis sample similar to the newly updated known diagnosis sample after updating the diagnosis sample into the known diagnosis sample; and training and generating a fault diagnosis algorithm according to all the known diagnosis samples and the unknown diagnosis samples in the diagnosis sample library in the preparation step and the diagnosis step respectively, wherein the diagnosis results of the fault diagnosis algorithm correspond to the known normal state, the known potential fault state, the known fault state and the unknown operation state respectively.
According to a specific embodiment, the on-site processing unit comprises,
the acquisition subunit is used for acquiring the diagnosis sample representing the current operation state of the machine from the site;
the storage subunit is used for storing the diagnosis sample for the diagnosis and storing the fault diagnosis algorithm obtained from the server;
the operation subunit is used for calling the fault diagnosis algorithm in the storage subunit and the diagnosis sample for the current diagnosis, diagnosing the diagnosis sample by using the fault diagnosis algorithm and finally outputting the diagnosis result;
and the communication subunit is used for acquiring the fault diagnosis algorithm from the server and sending the diagnosis sample of the current diagnosis to the server when the diagnosis result of the current diagnosis is in an unknown running state.
According to a particular embodiment, the server comprises,
a network unit for establishing a communication connection with the communication subunit;
the database unit is used for constructing a diagnosis sample library containing all the known diagnosis samples at present, and storing the known diagnosis samples, the unknown diagnosis samples and the fault diagnosis algorithm;
the operation state confirmation unit is used for determining that the operation state of the machine represented by the diagnosis sample is a normal state or a potential fault state or a fault state, and updating the diagnosis sample to be a known diagnosis sample;
an unknown diagnosis sample construction unit, configured to perform mutation processing on the known diagnosis sample that represents the known normal state to construct the unknown diagnosis sample;
an unknown diagnosis sample deleting unit, configured to, after extracting fault features of the recently updated known diagnosis sample and the unknown diagnosis sample in the diagnosis sample library, compare a frequency band energy ratio of the fault features of the unknown diagnosis sample and the recently updated known diagnosis sample, and delete the unknown diagnosis sample in which the frequency band energy ratio of the fault features exceeds a set range;
a fault diagnosis algorithm training unit, configured to train and generate a fault diagnosis algorithm according to all the known diagnosis samples and the unknown diagnosis samples in the diagnosis sample library, where diagnosis results of the fault diagnosis algorithm correspond to the known normal state, the known latent fault state, the known fault state, and the unknown operating state, respectively;
the fault diagnosis algorithm verification unit is used for carrying out fault diagnosis on a verification sample base, verifying the diagnosis accuracy of the fault diagnosis algorithm generated by training, if the diagnosis accuracy is not less than a design value, successfully training, and storing the fault diagnosis algorithm in the database unit, otherwise, adjusting the fault diagnosis algorithm, and training again until the training is successful;
and the updating information sending unit is used for sending the information of updating the fault diagnosis algorithm to the field processing unit through the network unit after the fault diagnosis training unit successfully trains so that the field processing unit replaces the current fault diagnosis algorithm with the successfully trained fault diagnosis algorithm.
According to a specific embodiment, the field processing subunit further includes a sample modification subunit, configured to, when the current mechanical operating state represented by the diagnostic sample does not match the diagnostic result in the diagnosing step and after a field service man determines and modifies the mechanical operating state represented by the diagnostic sample to be a normal state, a potential fault state, or a fault state, send the diagnostic sample to the server through the communication subunit, and update the diagnostic sample to the known diagnostic sample.
Compared with the prior art, the invention has the beneficial effects that:
the invention can realize the dynamic perfection of the diagnosis sample library, simultaneously perfects the fault diagnosis algorithm according to the diagnosis samples of the perfected diagnosis sample library, updates the perfected fault diagnosis algorithm to the field processing unit, improves the operation quality of mechanical equipment, reduces the downtime and improves the production efficiency. Meanwhile, when the diagnostic sample library is completed once, the unknown diagnostic sample similar to the updated diagnostic sample in the diagnostic sample library is deleted, so that the fault diagnosis algorithm is simplified, and the accuracy of fault diagnosis is improved.
Drawings
FIG. 1 is a schematic flow diagram of the process of the present invention;
FIG. 2 is a flow chart of one embodiment of the diagnostic steps of the method of the present invention;
FIG. 3 is a schematic diagram of the system of the present invention;
FIG. 4 is a block diagram of one embodiment of the system of the present invention;
FIG. 5 is a block diagram of the field processing unit of the system of the present invention;
FIG. 6 is a block diagram of one embodiment of an on-site processing unit of the system of the present invention;
fig. 7 is a block diagram of a server of the system of the present invention.
Detailed Description
The present invention will be described in further detail with reference to specific embodiments. It should be understood that the scope of the above-described subject matter is not limited to the following examples, and any techniques implemented based on the disclosure of the present invention are within the scope of the present invention.
In connection with the schematic flow diagram of the process of the invention shown in FIG. 1; the mechanical fault diagnosis optimization method comprises a preparation step and a diagnosis step.
Wherein, the preparation step comprises: firstly, constructing a diagnosis sample library containing all known diagnosis samples at present, then constructing a plurality of unknown diagnosis samples according to the known diagnosis samples in the diagnosis sample library, and storing the unknown diagnosis samples in the diagnosis sample library; the operation states of the machine are divided into a normal state, a potential fault state and a fault state, known diagnosis samples are used for representing the known normal state, the known potential fault state and the known fault state of the machine respectively, and an unknown diagnosis sample is used for representing the unknown operation state of the machine.
And finally, training and generating a fault diagnosis algorithm according to all known diagnosis samples and unknown diagnosis samples in the diagnosis sample library, wherein the diagnosis results of the fault diagnosis algorithm respectively correspond to a known normal state, a known potential fault state, a known fault state and an unknown operation state.
In an embodiment, the preparation step constructs the unknown diagnostic sample in a manner that: and performing mutation treatment on the known diagnosis sample which is characterized by the known normal state to construct an unknown diagnosis sample.
Specifically, the mutation treatment is as follows: randomly superposing a wavelet packet obtained by wavelet analysis on a known diagnosis sample representing a known normal state; or superposing an intrinsic mode function obtained by EMD decomposition on a known diagnosis sample representing a known fault state on a known diagnosis sample representing a known normal state.
The diagnostic steps of the present invention include:
the first step is as follows: and diagnosing a diagnosis sample which is obtained from the site and represents the current operation state of the machine by using a fault diagnosis algorithm, and finally outputting a diagnosis result.
In the embodiment, usually, a diagnostic sample representing the current operation state of the machine is obtained from the site at a certain frequency, a fault diagnosis algorithm is used to diagnose the diagnostic sample, if the diagnostic result is a known normal state, a known potential fault state or a known fault state, the diagnostic sample representing the current operation state of the machine is the normal state, the potential fault state or the known fault state, and meanwhile, when waiting for obtaining the diagnostic sample next time, the fault diagnosis algorithm is used to diagnose. And if the diagnosis result is the unknown operation state, entering the second step.
The second step is that: and if the diagnosis result in the first step is the unknown operation state, determining the mechanical operation state characterized by the diagnosis sample, and updating the diagnosis sample with the characterized mechanical operation state determined to be the known diagnosis sample.
In an embodiment, in the second step, the diagnostic sample is analyzed by a fault diagnosis expert system or an expert person, or a field maintenance person overhauls the mechanical equipment diagnosed with unknown operating state to determine that the mechanical operating state represented by the diagnostic sample is one of a normal state, a potential fault state or a fault state, and after the determination, the diagnostic sample is updated to the known diagnostic sample, so that the mechanical operating state represented by the diagnostic sample is determined to be one of a known normal state, a known potential fault state or a known fault state. And updating the diagnostic sample to the known diagnostic sample after the machine operating state characterized by the diagnostic sample is determined.
The third step: and after the unknown diagnosis sample similar to the newly updated known diagnosis sample is deleted, retraining and generating a fault diagnosis algorithm according to all the known diagnosis samples and the unknown diagnosis samples in the diagnosis sample library.
In practice, the way to delete an unknown diagnostic sample that is similar to the most recently updated known diagnostic sample is: and deleting the unknown diagnosis sample with the frequency band energy ratio of the fault feature exceeding the set range by extracting the fault features of the recently updated known diagnosis sample and the unknown diagnosis sample in the diagnosis sample library and comparing the frequency band energy ratios of the fault features of the unknown diagnosis sample and the recently updated known diagnosis sample. And when the diagnosis sample library is updated and improved once, the unknown diagnosis sample similar to the updated diagnosis sample in the diagnosis sample library is deleted, so that the fault diagnosis algorithm is simplified, and the fault diagnosis accuracy is improved.
In practice, training the generated fault diagnosis algorithm includes: and extracting the fault characteristics of the known diagnosis sample and the unknown diagnosis sample, taking the fault characteristics as the input of a fault diagnosis algorithm, and respectively corresponding the diagnosis result of the training fault diagnosis algorithm to the known normal state or the known potential fault state or the known fault state represented by the known diagnosis sample or the unknown operation state represented by the unknown diagnosis sample.
After the fault diagnosis algorithm is trained and generated, the diagnosis accuracy of the fault diagnosis algorithm generated by training is verified through a verification sample library, if the diagnosis accuracy is not smaller than a design value, the training is successful, otherwise, the fault diagnosis algorithm is adjusted and the training is repeated until the training is successful.
Taking the three-layer BP neural network fault diagnosis algorithm as an example to train a diagnosis sample library, generating the three-layer BP neural network fault diagnosis algorithm according to all known diagnosis samples and unknown diagnosis samples in the diagnosis sample library, then verifying through a verification sample library, and continuously modifying the number of hidden layer nodes of the BP neural network until the diagnosis accuracy reaches the design requirement, thereby ensuring the diagnosis accuracy of the BP neural network algorithm generated by training. Of course, the present invention may also adopt artificial intelligence algorithms such as support vector machine algorithm in the prior art.
In specific implementation, the fault feature extraction method can select a short-time fourier transform method, a wavelet transform method, an IHT method and an EMD method to extract a vibration signal spectrum, and then construct a fault feature vector by using methods of fusion energy spectrum, wavelet packet band energy, eigenmode function energy and the like. In addition, in the process of training and generating the fault diagnosis algorithm, the adopted verification sample library can select a plurality of diagnosis samples from the diagnosis sample library, the diagnosis samples are diagnosed by using the fault diagnosis algorithm generated by training, the diagnosis accuracy is calculated, and the accuracy of fault diagnosis of the fault diagnosis algorithm generated by training is objectively evaluated.
The fourth step: and replacing the current fault diagnosis algorithm with the fault diagnosis algorithm generated by retraining in the third step. Therefore, the fault diagnosis algorithm is perfected, the operation quality of mechanical equipment is improved, the downtime is reduced, and the production efficiency is improved.
In the invention, the operation state of the machine is divided into a normal state, a potential fault state and a fault state, which means that all normal states of the machine are summarized as the normal state, all potential fault states of the machine are summarized as the potential fault state, and all fault states of the machine are summarized as the fault state. A potentially faulty situation is one in which the machine is able to operate in such a situation despite a deviation from the normal situation and which has the risk of a faulty situation.
In connection with FIG. 2, a flow chart of one embodiment of the diagnostic steps of the method of the present invention is shown; in the diagnosis step, a diagnosis sample which is obtained from the site and represents the current operation state of the machine is diagnosed by using a fault diagnosis algorithm, after a diagnosis result is finally output, if the diagnosis result does not conform to the current operation state of the machine represented by the diagnosis sample, the operation state of the machine represented by the diagnosis sample is determined to be a normal state or a potential fault state or a fault state again, and after the determination, the diagnosis sample is updated to be a known diagnosis sample. Therefore, the sources of the diagnosis samples of the invention are widened, the wrong diagnosis samples are modified, the diagnosis accuracy of the fault diagnosis algorithm is improved, and the practicability of the invention is improved.
Based on the same inventive concept as the mechanical fault diagnosis optimization method, the invention also provides a mechanical fault diagnosis system. The structure of the system of the invention shown in fig. 3 is schematically illustrated; the mechanical fault diagnosis system comprises a field processing unit and a server.
The field processing unit is used for obtaining a diagnosis sample representing the current operation state of the machine from the field and a fault diagnosis algorithm from the server, diagnosing the diagnosis sample by using the fault diagnosis algorithm, and sending the diagnosis sample of the current fault diagnosis to the server if the diagnosis result of the diagnosis sample is the unknown operation state.
The server is used for constructing a diagnosis sample library containing all known diagnosis samples at present and constructing a plurality of unknown diagnosis samples in the preparation step, wherein the known diagnosis samples are respectively used for representing the known normal state, the known potential fault state and the known fault state of the machine, and the unknown diagnosis samples are used for representing the unknown operation state of the machine.
And the server is also used for determining that the operation state of the machine represented by the diagnosis sample is a normal state or a potential fault state or a fault state when a diagnosis result obtained after the diagnosis sample is diagnosed by the field processing unit in the diagnosis step is an unknown operation state, and deleting the unknown diagnosis sample similar to the newly updated known diagnosis sample after the diagnosis sample is updated to the known diagnosis sample.
And the server is used for training and generating a fault diagnosis algorithm according to all known diagnosis samples and unknown diagnosis samples in the diagnosis sample library in the preparation step and the diagnosis step respectively, and the diagnosis results of the fault diagnosis algorithm correspond to the known normal state, the known potential fault state, the known fault state and the unknown operation state respectively.
FIG. 4 is a block diagram of one embodiment of the system of the present invention; the server cluster is formed by a plurality of servers, mechanical equipment and the field processing unit are numbered, and all the mechanical equipment are equipment of the same model or a certain specific part of the same equipment. Due to the fact that a large number of mechanical devices are included in the system, the probability of the unknown operation condition is increased, and therefore the completeness of the diagnosis sample library is accelerated.
The block diagram of the field processing unit of the system of the present invention shown in connection with FIG. 5; the field processing unit comprises an acquisition subunit, a storage subunit, an operation subunit and a communication subunit.
The acquisition subunit is used for acquiring a diagnostic sample representing the current operation state of the machine from the field and storing the acquired diagnostic sample in the storage subunit. The storage subunit is used for storing the diagnosis sample for the diagnosis and storing the fault diagnosis algorithm obtained from the server. The operation subunit is used for calling the fault diagnosis algorithm in the storage subunit and the diagnosis sample for the diagnosis, diagnosing the diagnosis sample by using the fault diagnosis algorithm, and finally outputting a diagnosis result. The communication subunit is used for acquiring the fault diagnosis algorithm from the server and sending the diagnosis sample of the diagnosis to the server when the diagnosis result of the diagnosis is in an unknown operation state.
With reference to FIG. 6, there is shown a block diagram of one embodiment of a system on-site processing unit of the present invention; the field processing subunit further comprises a sample modification subunit, and the sample modification subunit is used for sending the diagnostic sample to the server through the communication subunit and updating the diagnostic sample into a known diagnostic sample after the field maintainer determines and modifies the mechanical operating state represented by the diagnostic sample when the current mechanical operating state represented by the diagnostic sample does not accord with the diagnostic result in the diagnostic step.
The architecture of the server of the system of the present invention shown in connection with FIG. 7; the server comprises a network unit, a database unit, an operation state confirmation unit, an unknown diagnosis sample construction unit, a fault diagnosis algorithm training unit, a fault diagnosis algorithm verification unit, an unknown diagnosis sample deletion unit and an update information sending unit.
The network unit is used for establishing communication connection with the communication subunits, and the network unit can simultaneously realize connection with a plurality of communication subunits. The database unit is used for constructing a diagnosis sample library containing all the known diagnosis samples at present, and storing the known diagnosis samples, the unknown diagnosis samples and the fault diagnosis algorithm. The operation state confirmation unit is used for determining the operation state of the machine represented by the diagnosis sample and updating the diagnosis sample into a known diagnosis sample. The unknown diagnosis sample construction unit is used for carrying out mutation processing on the known diagnosis sample representing the known normal state so as to construct the unknown diagnosis sample.
And the unknown diagnosis sample deleting unit is used for extracting the fault characteristics of the recently updated known diagnosis sample and the unknown diagnosis sample in the diagnosis sample library, then comparing the frequency band energy ratio of the fault characteristics of the unknown diagnosis sample and the recently updated known diagnosis sample, and deleting the unknown diagnosis sample with the frequency band energy ratio of the fault characteristics exceeding the set range.
And the fault diagnosis algorithm training unit is used for training and generating a fault diagnosis algorithm according to all known diagnosis samples and unknown diagnosis samples in the diagnosis sample library, wherein the diagnosis results of the fault diagnosis algorithm correspond to a known normal state, a known potential fault state, a known fault state and an unknown operation state respectively.
And the fault diagnosis algorithm verification unit is used for verifying the diagnosis accuracy of the fault diagnosis algorithm generated by the sample library verification training, if the diagnosis accuracy is not less than the design value, the training is successful, the fault diagnosis algorithm is stored in the database unit, and otherwise, the fault diagnosis algorithm is adjusted and the training is repeated until the training is successful.
And the updating information sending unit is used for sending the information of updating the fault diagnosis algorithm to the field processing unit through the network unit after the fault diagnosis training unit successfully trains so that the field processing unit replaces the current fault diagnosis algorithm with the successfully trained fault diagnosis algorithm.
The invention can realize the dynamic perfection of the diagnosis sample library, simultaneously perfects the fault diagnosis algorithm according to the diagnosis samples of the perfected diagnosis sample library, updates the perfected fault diagnosis algorithm to the field processing unit, and improves the accuracy of fault diagnosis, thereby improving the operation quality of mechanical equipment, reducing the downtime and improving the production efficiency.
While the present invention has been described in detail with reference to the embodiments shown in the drawings, the present invention is not limited to the above embodiments, and various modifications or alterations can be made by those skilled in the art without departing from the spirit and scope of the claims of the present application.
Claims (10)
1. A method for optimizing the diagnosis of mechanical faults, characterized in that it comprises a preparation step and a diagnosis step, in which,
the preparing step includes: firstly, a diagnosis sample library containing all known diagnosis samples at present is constructed, then a plurality of unknown diagnosis samples are constructed according to the known diagnosis samples, and the unknown diagnosis samples are stored in the diagnosis sample library; the operation states of the machine are divided into a normal state, a potential fault state and a fault state, the known diagnosis samples are respectively used for representing the known normal state, the known potential fault state and the known fault state of the machine, and the unknown diagnosis samples are used for representing the unknown operation state of the machine;
finally, training and generating a fault diagnosis algorithm according to all the known diagnosis samples and the unknown diagnosis samples in the diagnosis sample library, wherein the diagnosis results of the fault diagnosis algorithm respectively correspond to the known normal state, the known potential fault state, the known fault state and the unknown operation state;
the diagnosing step comprises:
the first step is as follows: diagnosing a diagnosis sample which is obtained from the site and represents the current operation state of the machine by using the fault diagnosis algorithm, and finally outputting a diagnosis result;
the second step is that: if the diagnosis result in the first step is the unknown operation state, determining the mechanical operation state represented by the diagnosis sample, and updating the diagnosis sample with the characterized mechanical operation state determined to be the known diagnosis sample;
the third step: after deleting the unknown diagnosis samples similar to the recently updated known diagnosis samples, retraining and generating a fault diagnosis algorithm according to all the known diagnosis samples and the unknown diagnosis samples in the diagnosis sample library;
the fourth step: and replacing the current fault diagnosis algorithm with the fault diagnosis algorithm generated by retraining in the third step.
2. The mechanical failure diagnosis optimization method according to claim 1, wherein in the preparation step, the known diagnosis samples that characterize the known normal state are subjected to mutation processing to construct the unknown diagnosis samples; wherein,
the mutation treatment comprises the following steps: randomly superposing a wavelet packet obtained by wavelet analysis on the known diagnosis sample representing the known fault state onto the known diagnosis sample representing the known normal state; or the intrinsic mode function obtained by EMD decomposition of the known diagnosis sample representing the known fault state is superposed on the known diagnosis sample representing the known normal state.
3. The mechanical failure diagnosis optimization method according to claim 2, wherein in the second step, the diagnosis sample is analyzed by a failure diagnosis expert system or an expert, or a mechanical device diagnosed with an unknown operating state is overhauled by a field maintenance person to determine a mechanical operating state characterized by the diagnosis sample, and the diagnosis sample in which the characterized mechanical operating state is determined is updated to the known diagnosis sample.
4. The mechanical failure diagnosis optimization method according to claim 3, wherein in the third step, the unknown diagnostic sample whose band energy ratio of the failure feature is out of a set range is deleted by extracting the failure feature of the known diagnostic sample that is updated most recently and the unknown diagnostic sample in the diagnostic sample library, and comparing the band energy ratio of the failure feature of the unknown diagnostic sample with the band energy ratio of the known diagnostic sample that is updated most recently.
5. The mechanical fault diagnosis optimization method of claim 1, wherein training to generate the fault diagnosis algorithm comprises:
extracting fault characteristics of the known diagnosis sample and the unknown diagnosis sample, taking the fault characteristics as input of the fault diagnosis algorithm, and training diagnosis results of the fault diagnosis algorithm to respectively correspond to the known normal state, the known potential fault state, the known fault state and the unknown operation state;
and after the fault diagnosis algorithm is trained and generated, verifying the diagnosis accuracy of the fault diagnosis algorithm generated by training through a verification sample library, if the diagnosis accuracy is not less than a design value, successfully training, and if not, adjusting the fault diagnosis algorithm and re-training until the training is successful.
6. The method according to claim 1, wherein in the diagnosing step, a diagnostic sample representing a current operating state of the machine obtained from a field is diagnosed by using the fault diagnosis algorithm, and after the diagnosis result is finally output, if the diagnosis result does not match the current operating state of the machine represented by the diagnostic sample, the operating state of the machine represented by the diagnostic sample is determined again.
7. A mechanical failure diagnosis system using the mechanical failure diagnosis optimization method according to claim 1, comprising an on-site processing unit and a server, wherein,
the field processing unit is used for acquiring a diagnosis sample representing the current operation state of the machine from the field, acquiring a fault diagnosis algorithm from the server, diagnosing the diagnosis sample by using the fault diagnosis algorithm, and if the diagnosis result of the diagnosis sample is an unknown operation state, sending the diagnosis sample of the current fault diagnosis to the server;
the server is used for constructing a diagnosis sample library containing all known diagnosis samples at present and constructing a plurality of unknown diagnosis samples in the preparation step, the known diagnosis samples are respectively used for representing the known normal state, the known potential fault state and the known fault state of the machine, and the unknown diagnosis samples are used for representing the unknown operation state of the machine; in the diagnosis step, when the diagnosis result obtained after the diagnosis of the diagnosis sample is the unknown operation state, the field processing unit determines the operation state of the machine characterized by the diagnosis sample, and after the diagnosis sample is updated to a known diagnosis sample, deletes the unknown diagnosis sample similar to the newly updated known diagnosis sample; and training and generating a fault diagnosis algorithm according to all the known diagnosis samples and the unknown diagnosis samples in the diagnosis sample library in the preparation step and the diagnosis step respectively, wherein the diagnosis results of the fault diagnosis algorithm correspond to the known normal state, the known potential fault state, the known fault state and the unknown operation state respectively.
8. The mechanical fault diagnostic system of claim 7, wherein the field processing unit comprises,
the acquisition subunit is used for acquiring the diagnosis sample representing the current operation state of the machine from the site;
the storage subunit is used for storing the diagnosis sample for the diagnosis and storing the fault diagnosis algorithm obtained from the server;
the operation subunit is used for calling the fault diagnosis algorithm in the storage subunit and the diagnosis sample for the current diagnosis, diagnosing the diagnosis sample by using the fault diagnosis algorithm and finally outputting the diagnosis result;
and the communication subunit is used for acquiring the fault diagnosis algorithm from the server and sending the diagnosis sample of the current diagnosis to the server when the diagnosis result of the current diagnosis is in an unknown running state.
9. The mechanical fault diagnostic system of claim 8, wherein the server comprises,
a network unit for establishing a communication connection with the communication subunit;
the database unit is used for constructing a diagnosis sample library containing all the known diagnosis samples at present, and storing the known diagnosis samples, the unknown diagnosis samples and the fault diagnosis algorithm;
the operation state confirmation unit is used for determining the operation state of the machine represented by the diagnosis sample and updating the diagnosis sample into a known diagnosis sample;
an unknown diagnosis sample construction unit, configured to perform mutation processing on the known diagnosis sample that represents the known normal state to construct the unknown diagnosis sample;
an unknown diagnosis sample deleting unit, configured to, after extracting fault features of the recently updated known diagnosis sample and the unknown diagnosis sample in the diagnosis sample library, compare a frequency band energy ratio of the fault features of the unknown diagnosis sample and the recently updated known diagnosis sample, and delete the unknown diagnosis sample in which the frequency band energy ratio of the fault features exceeds a set range;
a fault diagnosis algorithm training unit, configured to train and generate a fault diagnosis algorithm according to all the known diagnosis samples and the unknown diagnosis samples in the diagnosis sample library, where diagnosis results of the fault diagnosis algorithm correspond to the known normal state, the known latent fault state, the known fault state, and the unknown operating state, respectively;
the fault diagnosis algorithm verification unit is used for carrying out fault diagnosis on a verification sample base, verifying the diagnosis accuracy of the fault diagnosis algorithm generated by training, if the diagnosis accuracy is not less than a design value, successfully training, and storing the fault diagnosis algorithm in the database unit, otherwise, adjusting the fault diagnosis algorithm, and training again until the training is successful;
and the updating information sending unit is used for sending the information of updating the fault diagnosis algorithm to the field processing unit through the network unit after the fault diagnosis training unit successfully trains so that the field processing unit replaces the current fault diagnosis algorithm with the successfully trained fault diagnosis algorithm.
10. The mechanical fault diagnosis system according to claim 8 or 9, wherein the field processing subunit further includes a sample modification subunit, configured to, when the current machine operating state represented by the diagnosis sample does not match the diagnosis result in the diagnosis step, and after a field service man determines and modifies the machine operating state represented by the diagnosis sample to be a normal state or a potential fault state or a fault state, send the diagnosis sample to the server through the communication subunit, and update the diagnosis sample to the known diagnosis sample.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201610069441.6A CN105740822B (en) | 2016-02-01 | 2016-02-01 | A kind of mechanical fault diagnosis optimization method and system |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201610069441.6A CN105740822B (en) | 2016-02-01 | 2016-02-01 | A kind of mechanical fault diagnosis optimization method and system |
Publications (2)
Publication Number | Publication Date |
---|---|
CN105740822A CN105740822A (en) | 2016-07-06 |
CN105740822B true CN105740822B (en) | 2019-02-19 |
Family
ID=56242124
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201610069441.6A Active CN105740822B (en) | 2016-02-01 | 2016-02-01 | A kind of mechanical fault diagnosis optimization method and system |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN105740822B (en) |
Families Citing this family (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN107092247B (en) * | 2017-06-16 | 2019-11-22 | 温州大学 | A kind of packaging production line method for diagnosing faults based on status data |
CN110207967B (en) * | 2019-06-13 | 2020-12-01 | 大连海事大学 | State identification method and system based on wavelet packet energy characteristics and cross correlation |
Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN102243140A (en) * | 2011-04-18 | 2011-11-16 | 杨彦利 | Mechanical equipment state monitoring method based on sub-band signal analysis |
CN102721537A (en) * | 2012-06-01 | 2012-10-10 | 西安交通大学 | Mechanical impact type fault diagnosis method based on variable space-scale frame |
CN103743554A (en) * | 2013-06-28 | 2014-04-23 | 国家电网公司 | High-voltage circuit breaker mechanical failure diagnosis method based on vibration signal analysis |
CN103914735A (en) * | 2014-04-17 | 2014-07-09 | 北京泰乐德信息技术有限公司 | Failure recognition method and system based on neural network self-learning |
Family Cites Families (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JP2005033559A (en) * | 2003-07-14 | 2005-02-03 | Fuji Xerox Co Ltd | Failure diagnostic device |
-
2016
- 2016-02-01 CN CN201610069441.6A patent/CN105740822B/en active Active
Patent Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN102243140A (en) * | 2011-04-18 | 2011-11-16 | 杨彦利 | Mechanical equipment state monitoring method based on sub-band signal analysis |
CN102721537A (en) * | 2012-06-01 | 2012-10-10 | 西安交通大学 | Mechanical impact type fault diagnosis method based on variable space-scale frame |
CN103743554A (en) * | 2013-06-28 | 2014-04-23 | 国家电网公司 | High-voltage circuit breaker mechanical failure diagnosis method based on vibration signal analysis |
CN103914735A (en) * | 2014-04-17 | 2014-07-09 | 北京泰乐德信息技术有限公司 | Failure recognition method and system based on neural network self-learning |
Non-Patent Citations (1)
Title |
---|
基于智能理论的高压断路器机械故障诊断;田涛等;《江苏电机工程》;20141130;第33卷(第6期);第12-15页 |
Also Published As
Publication number | Publication date |
---|---|
CN105740822A (en) | 2016-07-06 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
US11379723B2 (en) | Method and apparatus for compressing neural network | |
US20210089900A1 (en) | Transformer dga data prediction method based on multi-dimensional time sequence frame convolution lstm | |
US10795753B2 (en) | Log-based computer failure diagnosis | |
CN111046027B (en) | Missing value filling method and device for time series data | |
US11341415B2 (en) | Method and apparatus for compressing neural network | |
CN110969194B (en) | Cable early fault positioning method based on improved convolutional neural network | |
CN108809948A (en) | A kind of abnormal network connecting detection method based on deep learning | |
CN110019519A (en) | Data processing method, device, storage medium and electronic device | |
CN115514620B (en) | Anomaly detection method and cloud network platform | |
CN112418175A (en) | Rolling bearing fault diagnosis method and system based on domain migration and storage medium | |
CN105740822B (en) | A kind of mechanical fault diagnosis optimization method and system | |
CN117076955A (en) | Fault detection method and system for high-voltage frequency converter | |
WO2021047184A1 (en) | System failure analysis and processing method and apparatus, storage medium, and electronic device | |
CN113314146A (en) | Method, software and apparatus for training an alert system to audio classification of an event | |
CN105606353B (en) | A kind of mechanical failure diagnostic method and system | |
CN117667495B (en) | Association rule and deep learning integrated application system fault prediction method | |
CN111522705A (en) | Intelligent operation and maintenance solution method for industrial big data | |
CN114413409A (en) | Detection method and device for air conditioner fault probability and intelligent air conditioner | |
CN117493797A (en) | Fault prediction method and device of Internet of things equipment, electronic equipment and storage medium | |
CN108427744B (en) | Intelligent data association relation determining method based on information operation and maintenance | |
CN114650211A (en) | Fault repairing method, device, electronic equipment and computer readable storage medium | |
CN113360306B (en) | Real-time fault detection method for airborne information system | |
CN115080445B (en) | Game test management method and system | |
US20240273412A1 (en) | Unlearnable tasks in machine learning | |
US11861509B2 (en) | Automated positive train control event data extraction and analysis engine for performing root cause analysis of unstructured data |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
C06 | Publication | ||
PB01 | Publication | ||
C10 | Entry into substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
GR01 | Patent grant | ||
GR01 | Patent grant | ||
PE01 | Entry into force of the registration of the contract for pledge of patent right | ||
PE01 | Entry into force of the registration of the contract for pledge of patent right |
Denomination of invention: An optimization method and system for mechanical fault diagnosis Effective date of registration: 20220217 Granted publication date: 20190219 Pledgee: Zhejiang Xiaoshan rural commercial bank Limited by Share Ltd. Jingjiang branch Pledgor: HANGZHOU JIE DRIVE TECHNOLOGY Co.,Ltd. Registration number: Y2022330000233 |