CN116184988B - Multi-mode data-based fault prediction method, device, equipment and storage medium - Google Patents
Multi-mode data-based fault prediction method, device, equipment and storage medium Download PDFInfo
- Publication number
- CN116184988B CN116184988B CN202310487654.0A CN202310487654A CN116184988B CN 116184988 B CN116184988 B CN 116184988B CN 202310487654 A CN202310487654 A CN 202310487654A CN 116184988 B CN116184988 B CN 116184988B
- Authority
- CN
- China
- Prior art keywords
- data
- prediction
- fusion
- prediction model
- sub
- 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
Classifications
-
- G—PHYSICS
- G05—CONTROLLING; REGULATING
- G05B—CONTROL OR REGULATING SYSTEMS IN GENERAL; FUNCTIONAL ELEMENTS OF SUCH SYSTEMS; MONITORING OR TESTING ARRANGEMENTS FOR SUCH SYSTEMS OR ELEMENTS
- G05B23/00—Testing or monitoring of control systems or parts thereof
- G05B23/02—Electric testing or monitoring
- G05B23/0205—Electric testing or monitoring by means of a monitoring system capable of detecting and responding to faults
- G05B23/0218—Electric testing or monitoring by means of a monitoring system capable of detecting and responding to faults characterised by the fault detection method dealing with either existing or incipient faults
- G05B23/0243—Electric testing or monitoring by means of a monitoring system capable of detecting and responding to faults characterised by the fault detection method dealing with either existing or incipient faults model based detection method, e.g. first-principles knowledge model
-
- G—PHYSICS
- G05—CONTROLLING; REGULATING
- G05B—CONTROL OR REGULATING SYSTEMS IN GENERAL; FUNCTIONAL ELEMENTS OF SUCH SYSTEMS; MONITORING OR TESTING ARRANGEMENTS FOR SUCH SYSTEMS OR ELEMENTS
- G05B2219/00—Program-control systems
- G05B2219/20—Pc systems
- G05B2219/24—Pc safety
- G05B2219/24065—Real time diagnostics
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y04—INFORMATION OR COMMUNICATION TECHNOLOGIES HAVING AN IMPACT ON OTHER TECHNOLOGY AREAS
- Y04S—SYSTEMS INTEGRATING TECHNOLOGIES RELATED TO POWER NETWORK OPERATION, COMMUNICATION OR INFORMATION TECHNOLOGIES FOR IMPROVING THE ELECTRICAL POWER GENERATION, TRANSMISSION, DISTRIBUTION, MANAGEMENT OR USAGE, i.e. SMART GRIDS
- Y04S10/00—Systems supporting electrical power generation, transmission or distribution
- Y04S10/50—Systems or methods supporting the power network operation or management, involving a certain degree of interaction with the load-side end user applications
Landscapes
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Engineering & Computer Science (AREA)
- Automation & Control Theory (AREA)
- Testing And Monitoring For Control Systems (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention provides a fault prediction method, a device, equipment and a storage medium based on multi-mode data, which relate to the field of fault prediction, wherein the method comprises the following steps: acquiring a plurality of abnormal operation data of the target equipment, wherein each abnormal operation data corresponds to different data types respectively; carrying out data fusion on each abnormal operation data to obtain fusion data; and obtaining a prediction model, inputting the fusion data into the prediction model to obtain target fault prediction data, wherein the prediction model comprises a plurality of sub-prediction models, each sub-prediction model is respectively constructed based on different types of neural networks, and the weight value of each sub-prediction model is determined based on the data distribution characteristics of the fusion data. According to the invention, multi-mode data is adopted for prediction, and a plurality of different sub-prediction models are integrated in one prediction model, so that the fault prediction precision of the prediction model can be effectively improved.
Description
Technical Field
The present invention relates to the field of fault prediction technologies, and in particular, to a fault prediction method, device, equipment, and storage medium based on multi-mode data.
Background
The traditional numerical control machine tool fault prediction method generally adopts a manual diagnosis mode, namely, the running state and the fault condition of the machine tool are judged through experience of professionals, and the risk of misjudgment is possibly caused by the influence of experience of personnel and subjective factors. Therefore, an artificial intelligent diagnosis method is increasingly developed, wherein the artificial intelligent diagnosis is to input the collected operation data of the numerical control machine tool into a pre-trained fault prediction model, and predict whether the numerical control machine tool is in a normal state or a fault state currently through a deep learning model. However, the current fault prediction model adopts single-mode data to predict, and the prediction precision is not high.
Accordingly, there is a need for improvement and development in the art.
Disclosure of Invention
The invention provides a fault prediction method, device, equipment and storage medium based on multi-mode data, which are used for solving the defect that in the prior art, a fault prediction model is predicted by adopting single-mode data, and the prediction precision is not high.
The invention provides a fault prediction method based on multi-mode data, which comprises the following steps:
acquiring a plurality of pieces of abnormal operation data of target equipment, wherein each piece of abnormal operation data corresponds to different data types respectively;
carrying out data fusion on each abnormal operation data to obtain fusion data;
obtaining a preset prediction model, inputting the fusion data into the prediction model to obtain target fault prediction data corresponding to the target equipment, wherein the prediction model comprises a plurality of sub-prediction models, each sub-prediction model is respectively constructed based on different types of neural networks, and the weight value of each sub-prediction model is determined based on the data distribution characteristics of the fusion data.
According to the fault prediction method based on multi-mode data provided by the invention, the method for acquiring the abnormal operation data corresponding to each data type comprises the following steps:
acquiring equipment operation data corresponding to the data type, and taking the equipment operation data as data to be filtered;
filtering the data to be filtered to obtain filtered data, and determining a data change value according to the data to be filtered and the filtered data;
judging whether the data change value is smaller than a preset change threshold value, if not, taking the filtered data as the data to be filtered, and continuing to execute the step of filtering the data to be filtered to obtain the filtered data until the data change value is smaller than the change threshold value, so as to obtain target filtered data corresponding to the equipment operation data;
and determining the abnormal operation data corresponding to the data type according to the equipment operation data and the target filtering data.
According to the fault prediction method based on multi-mode data provided by the invention, the determination method of the weight value of each sub-prediction model comprises the following steps:
acquiring the data distribution characteristics corresponding to the fusion data;
obtaining a confidence coefficient database corresponding to the sub-prediction model, wherein the confidence coefficient database is used for reflecting the prediction confidence coefficient of the sub-prediction model under different reference data distribution characteristics;
comparing the data distribution characteristics with the confidence coefficient database, and determining target prediction confidence coefficient corresponding to the data distribution characteristics according to the comparison result;
and determining the weight value corresponding to the sub-prediction model according to the target prediction confidence.
According to the fault prediction method based on multi-mode data provided by the invention, the determination method of the prediction confidence degree corresponding to each reference data distribution characteristic and each reference data distribution characteristic respectively comprises the following steps:
acquiring a plurality of historical fusion data and fault prediction labels corresponding to the historical fusion data respectively, acquiring similarity between the historical data distribution characteristics of the historical fusion data, and classifying the historical data distribution characteristics according to the similarity to obtain a plurality of characteristic sets;
according to the feature sets, determining fusion distribution features corresponding to the feature sets respectively, and taking the fusion distribution features as the reference data distribution features;
classifying each historical fusion data according to each feature set to obtain a plurality of test data sets, and determining the recognition accuracy of the prediction model under each reference data distribution feature according to each test data set and the corresponding fault prediction label;
and determining the prediction confidence corresponding to each reference data distribution characteristic according to each identification accuracy.
According to the fault prediction method based on the multi-mode data provided by the invention, the method for determining the target fault prediction data comprises the following steps:
obtaining fault prediction data corresponding to each sub prediction model respectively;
and carrying out weighting processing according to the weight value of each sub-prediction model and the fault prediction data to obtain the target fault prediction data.
According to the fault prediction method based on the multi-mode data, one of the sub-prediction models is an LSTM model, and a regularization term is included in a loss function of the LSTM model.
According to the fault prediction method based on the multi-mode data, each abnormal operation data comprises abnormal vibration data and abnormal sound wave data.
The invention also provides a fault prediction device based on the multi-mode data, which comprises:
the data acquisition module is used for acquiring a plurality of pieces of abnormal operation data of the target equipment, wherein each piece of abnormal operation data corresponds to different data types respectively;
the data fusion module is used for carrying out data fusion on each abnormal operation data to obtain fusion data;
the fault prediction module is used for acquiring a preset prediction model, inputting the fusion data into the prediction model to obtain target fault prediction data corresponding to the target equipment, wherein the prediction model comprises a plurality of sub-prediction models, each sub-prediction model is respectively constructed based on different types of neural networks, and the weight value of each sub-prediction model is determined based on the data distribution characteristics of the fusion data.
The invention also provides an electronic device comprising a memory, a processor and a computer program stored on the memory and executable on the processor, wherein the processor implements the fault prediction method based on multi-mode data according to any one of the above when executing the program.
The invention also provides a non-transitory computer readable storage medium having stored thereon a computer program which, when executed by a processor, implements a multi-modal data-based fault prediction method as described in any of the above.
According to the fault prediction method, the device, the equipment and the storage medium based on the multi-mode data, provided by the invention, a plurality of abnormal operation data of the target equipment are obtained, wherein each abnormal operation data corresponds to different data types respectively; carrying out data fusion on each abnormal operation data to obtain fusion data; and obtaining a prediction model, inputting the fusion data into the prediction model to obtain target fault prediction data, wherein the prediction model comprises a plurality of sub-prediction models, each sub-prediction model is respectively constructed based on different types of neural networks, and the weight value of each sub-prediction model is determined based on the data distribution characteristics of the fusion data. According to the invention, multi-mode data is adopted for prediction, and a plurality of different sub-prediction models are integrated in one prediction model, so that the prediction precision of the fault prediction model can be effectively improved.
Drawings
In order to more clearly illustrate the invention or the technical solutions of the prior art, the following description will briefly explain the drawings used in the embodiments or the description of the prior art, and it is obvious that the drawings in the following description are some embodiments of the invention, and other drawings can be obtained according to the drawings without inventive effort for a person skilled in the art.
FIG. 1 is a flow chart of a multi-modal data-based fault prediction method provided by the invention;
FIG. 2 is a schematic structural diagram of a fault prediction device based on multi-modal data according to the present invention;
fig. 3 is a schematic structural diagram of an electronic device provided by the present invention.
Detailed Description
For the purpose of making the objects, technical solutions and advantages of the present invention more apparent, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings, and it is apparent that the described embodiments are some embodiments of the present invention, not all embodiments. All other embodiments, which can be made by those skilled in the art based on the embodiments of the invention without making any inventive effort, are intended to be within the scope of the invention.
The method, apparatus, device and storage medium for fault prediction based on multi-modal data according to the present invention are described below with reference to fig. 1 to 3.
As shown in fig. 1, the method comprises the steps of:
s110, acquiring a plurality of pieces of abnormal operation data of target equipment, wherein each piece of abnormal operation data corresponds to different data types;
s120, carrying out data fusion on each abnormal operation data to obtain fusion data;
s130, acquiring a preset prediction model, inputting the fusion data into the prediction model to obtain target fault prediction data corresponding to the target equipment, wherein the prediction model comprises a plurality of sub-prediction models, each sub-prediction model is respectively constructed based on different types of neural networks, and the weight value of each sub-prediction model is determined based on the data distribution characteristics of the fusion data.
Specifically, the target apparatus in the present embodiment may be a numerically controlled machine tool. In order to improve the prediction accuracy of the prediction model, it is first necessary to acquire abnormal operation data of a plurality of data types, such as abnormal vibration data and abnormal sound wave data of the target device. And in order to avoid instability of a single model, the prediction model in this embodiment integrates sub-prediction models constructed by a plurality of different neural networks. And (3) fusing a plurality of abnormal operation data and inputting the fused abnormal operation data into a prediction model, wherein the prediction model can analyze the data distribution characteristics of the fused data. Different neural networks have different sensitivity to different data distribution characteristics, some are good at processing data streams with higher sparsity (such as SP-CNN), some are good at processing data streams with higher density (such as LSTM), so that the distribution change of input data can affect the model precision of each sub-prediction model, and therefore, the embodiment can dynamically adjust the weight value of each sub-prediction model according to the data distribution characteristics of the fused data. And finally, synthesizing the fault prediction results output by the sub-prediction models to generate target fault prediction data output by the prediction models. In the embodiment, multi-mode data is adopted for prediction, and a plurality of different sub-prediction models are integrated in one prediction model, so that the prediction precision of the prediction model can be effectively improved.
In one implementation manner, the method for acquiring the abnormal operation data corresponding to each data type includes:
acquiring equipment operation data corresponding to the data type, and taking the equipment operation data as data to be filtered;
filtering the data to be filtered to obtain filtered data, and determining a data change value according to the data to be filtered and the filtered data;
judging whether the data change value is smaller than a preset change threshold value, if not, taking the filtered data as the data to be filtered, and continuing to execute the step of filtering the data to be filtered to obtain the filtered data until the data change value is smaller than the change threshold value, so as to obtain target filtered data corresponding to the equipment operation data;
and determining the abnormal operation data corresponding to the data type according to the equipment operation data and the target filtering data.
Specifically, a specific acquisition process of abnormal operation data corresponding to a data type is described by taking the data type as an example. First, equipment operation data corresponding to the data type is obtained. And then the equipment operation data is used as data to be filtered and is input into a preset filter for filtering treatment, and the filtering treatment can filter out some data points with larger deviation, wherein the filtered data points are abnormal data points. And comparing the data before and after the filtering treatment, and judging the change degree between the data before and after the filtering treatment to obtain a data change value. When the data change value is larger than or equal to a preset change threshold value, the filtering processing is continued until the data change value is smaller than the change threshold value, the current obtained filtering data is used as final target filtering data after the abnormal data point is completely filtered. The filtered abnormal data points can be obtained by comparing the original acquired equipment operation data with the target filtering data, and the abnormal data points can form abnormal operation data corresponding to the data type. Compared with the traditional abnormality detection method, the embodiment can simply, conveniently and rapidly screen out abnormal operation data in the equipment operation data by using the filter.
In one implementation, the filter is an adaptive filter, such as an LMS (least mean square) algorithm, an NLMS (normalized least mean square) algorithm, an RLS (recursive least square) algorithm, or the like. The selection of different adaptive filter algorithms may produce different filtering effects and thus may be compared experimentally. The self-adaptive filter can automatically adjust the filter coefficient, and more accurately process the data, so that the problem of poor filtering effect caused by the difference between preset filter parameters and actual signals is avoided.
In one implementation, the method for determining the weight value of each sub-prediction model includes:
acquiring the data distribution characteristics corresponding to the fusion data;
obtaining a confidence coefficient database corresponding to the sub-prediction model, wherein the confidence coefficient database is used for reflecting the prediction confidence coefficient of the sub-prediction model under different reference data distribution characteristics;
comparing the data distribution characteristics with the confidence coefficient database, and determining target prediction confidence coefficient corresponding to the data distribution characteristics according to the comparison result;
and determining the weight value corresponding to the sub-prediction model according to the target prediction confidence.
Specifically, a specific determination process of the weight value of a sub-prediction model is described by taking the sub-prediction model as an example. Firstly, a confidence coefficient database of the sub-prediction model is pre-built, the confidence coefficient database comprises a plurality of reference data distribution characteristics, each reference data distribution characteristic has a prediction confidence coefficient which is stored in an associated mode, and the prediction confidence coefficient is used for reflecting the recognition accuracy of the sub-prediction model under the corresponding reference data distribution characteristics. Comparing the data distribution characteristics of the fusion data obtained currently with each reference data distribution characteristic in the confidence coefficient database, and determining the target prediction confidence coefficient corresponding to the data distribution characteristics according to the reference data distribution characteristic with the highest similarity, wherein the larger the target prediction confidence coefficient is, the larger the weight value of the sub prediction model is. According to the method and the device, the weight value of the sub-prediction model is dynamically adjusted through the data distribution characteristics of the fusion data, the effect of the high-precision sub-prediction model on the final prediction result can be pertinently improved, the interference of the low-precision sub-prediction model on the final prediction result is reduced, and the prediction precision of the prediction model is further effectively improved.
In one implementation, the method for determining the prediction confidence of each of the reference data distribution features and each of the reference data distribution features includes:
acquiring a plurality of historical fusion data and fault prediction labels corresponding to the historical fusion data respectively, acquiring similarity between the historical data distribution characteristics of the historical fusion data, and classifying the historical data distribution characteristics according to the similarity to obtain a plurality of characteristic sets;
according to the feature sets, determining fusion distribution features corresponding to the feature sets respectively, and taking the fusion distribution features as the reference data distribution features;
classifying each historical fusion data according to each feature set to obtain a plurality of test data sets, and determining the recognition accuracy of the prediction model under each reference data distribution feature according to each test data set and the corresponding fault prediction label;
and determining the prediction confidence corresponding to each reference data distribution characteristic according to each identification accuracy.
Specifically, a large amount of historical fusion data of the target device is firstly obtained, and each historical fusion data is marked, namely a fault prediction label is generated. And classifying the historical data distribution characteristics of the historical fusion data, and classifying the historical data distribution characteristics with high characteristic similarity into one characteristic set (for example, the similarity between every two historical data distribution characteristics in each characteristic set is larger than a first similarity threshold value). And fusing all the historical data distribution characteristics in the characteristic set aiming at each characteristic set to obtain the fused distribution characteristics of the characteristic set, and taking the fused distribution characteristics as a reference data distribution characteristic. And then generating a test data set according to all the historical fusion data corresponding to the feature set, wherein the historical fusion data have fault prediction labels, so that the identification accuracy of the sub-prediction model under the test data set can be calculated, and the prediction confidence of the corresponding reference data distribution feature can be determined according to the identification accuracy. According to the embodiment, the prediction confidence of the sub-prediction model under different data distribution characteristics can be accurately judged by adopting a data classification analysis mode.
In one implementation, the method for determining the target fault prediction data includes:
obtaining fault prediction data corresponding to each sub prediction model respectively;
and carrying out weighting processing according to the weight value of each sub-prediction model and the fault prediction data to obtain the target fault prediction data.
Specifically, in the embodiment, the weight value of the sub-prediction model is dynamically adjusted through the data distribution characteristics of the fusion data, and the fault prediction data of each sub-prediction model is summarized in a weighted fusion mode to obtain final target fault prediction data, so that the effect of the high-precision sub-prediction model on a final prediction result can be pertinently improved, the interference of the low-precision sub-prediction model on the final prediction result is reduced, and the prediction precision of the prediction model is further effectively improved.
In one implementation, one of the sub-prediction models is an LSTM model, and a regularization term is included in a loss function of the LSTM model.
Specifically, the present embodiment uses an LSTM network added with regularization terms as a sub-prediction model. Regularization is a method for controlling model complexity and avoiding overfitting, which can be achieved by adding regularization terms to the loss function. The complexity of the LSTM network is controlled by adding regularization items, so that the occurrence of the overfitting phenomenon can be avoided, and the generalization capability and the classification accuracy of the model are improved.
In one implementation, each of the abnormal operation data includes abnormal vibration data and abnormal sound wave data.
In particular, the vibration signal may reflect the motion state of the target device, while the acoustic wave signal may reflect sound during processing of the target device. The abnormal data corresponding to the two signals are fused, so that the actual running state of the target equipment can be reflected more comprehensively, and the accuracy of fault prediction is improved.
In one implementation, each of the abnormal operation data needs to be subjected to data preprocessing after acquisition, including noise removal, down-sampling of the signal, and the like.
The fault prediction device based on the multi-mode data provided by the invention is described below, and the fault prediction device based on the multi-mode data described below and the fault prediction method based on the multi-mode data described above can be correspondingly referred to each other.
As shown in fig. 2, the apparatus includes:
the data acquisition module 210 is configured to acquire a plurality of abnormal operation data of the target device, where each abnormal operation data corresponds to a different data type;
the data fusion module 220 is configured to perform data fusion on each abnormal operation data to obtain fusion data;
the fault prediction module 230 is configured to obtain a preset prediction model, input the fused data into the prediction model, and obtain target fault prediction data corresponding to the target device, where the prediction model includes a plurality of sub-prediction models, each of the sub-prediction models is respectively constructed based on different types of neural networks, and a weight value of each of the sub-prediction models is determined based on a data distribution feature of the fused data.
Fig. 3 illustrates a physical schematic diagram of an electronic device, as shown in fig. 3, where the electronic device may include: processor 310, communication interface (Communications Interface) 320, memory 330 and communication bus 340, wherein processor 310, communication interface 320, memory 330 accomplish communication with each other through communication bus 340. The processor 310 may invoke logic instructions in the memory 330 to perform a fault prediction method based on multi-modal data, the method comprising:
acquiring a plurality of pieces of abnormal operation data of target equipment, wherein each piece of abnormal operation data corresponds to different data types respectively;
carrying out data fusion on each abnormal operation data to obtain fusion data;
obtaining a preset prediction model, inputting the fusion data into the prediction model to obtain target fault prediction data corresponding to the target equipment, wherein the prediction model comprises a plurality of sub-prediction models, each sub-prediction model is respectively constructed based on different types of neural networks, and the weight value of each sub-prediction model is determined based on the data distribution characteristics of the fusion data.
Further, the logic instructions in the memory 330 described above may be implemented in the form of software functional units and may be stored in a computer-readable storage medium when sold or used as a stand-alone product. Based on this understanding, the technical solution of the present invention may be embodied essentially or in a part contributing to the prior art or in a part of the technical solution, in the form of a software product stored in a storage medium, comprising several instructions for causing a computer device (which may be a personal computer, a server, a network device, etc.) to perform all or part of the steps of the method according to the embodiments of the present invention. And the aforementioned storage medium includes: a U-disk, a removable hard disk, a Read-Only Memory (ROM), a random access Memory (RAM, random Access Memory), a magnetic disk, or an optical disk, or other various media capable of storing program codes.
In another aspect, the present invention also provides a computer program product, where the computer program product includes a computer program, where the computer program can be stored on a non-transitory computer readable storage medium, and when the computer program is executed by a processor, the computer can execute the method for predicting a fault based on multi-mode data provided by the above methods, and the method includes:
acquiring a plurality of pieces of abnormal operation data of target equipment, wherein each piece of abnormal operation data corresponds to different data types respectively;
carrying out data fusion on each abnormal operation data to obtain fusion data;
obtaining a preset prediction model, inputting the fusion data into the prediction model to obtain target fault prediction data corresponding to the target equipment, wherein the prediction model comprises a plurality of sub-prediction models, each sub-prediction model is respectively constructed based on different types of neural networks, and the weight value of each sub-prediction model is determined based on the data distribution characteristics of the fusion data.
In yet another aspect, the present invention also provides a non-transitory computer readable storage medium having stored thereon a computer program which, when executed by a processor, is implemented to perform the method of multimodal data based fault prediction provided by the methods above, the method comprising:
acquiring a plurality of pieces of abnormal operation data of target equipment, wherein each piece of abnormal operation data corresponds to different data types respectively;
carrying out data fusion on each abnormal operation data to obtain fusion data;
obtaining a preset prediction model, inputting the fusion data into the prediction model to obtain target fault prediction data corresponding to the target equipment, wherein the prediction model comprises a plurality of sub-prediction models, each sub-prediction model is respectively constructed based on different types of neural networks, and the weight value of each sub-prediction model is determined based on the data distribution characteristics of the fusion data.
The apparatus embodiments described above are merely illustrative, wherein the elements illustrated as separate elements may or may not be physically separate, and the elements shown as elements may or may not be physical elements, may be located in one place, or may be distributed over a plurality of network elements. Some or all of the modules may be selected according to actual needs to achieve the purpose of the solution of this embodiment. Those of ordinary skill in the art will understand and implement the present invention without undue burden.
From the above description of the embodiments, it will be apparent to those skilled in the art that the embodiments may be implemented by means of software plus necessary general hardware platforms, or of course may be implemented by means of hardware. Based on this understanding, the foregoing technical solution may be embodied essentially or in a part contributing to the prior art in the form of a software product, which may be stored in a computer readable storage medium, such as ROM/RAM, a magnetic disk, an optical disk, etc., including several instructions for causing a computer device (which may be a personal computer, a server, or a network device, etc.) to execute the method described in the respective embodiments or some parts of the embodiments.
Finally, it should be noted that: the above embodiments are only for illustrating the technical solution of the present invention, and are not limiting; although the invention has been described in detail with reference to the foregoing embodiments, it will be understood by those of ordinary skill in the art that: the technical scheme described in the foregoing embodiments can be modified or some technical features thereof can be replaced by equivalents; such modifications and substitutions do not depart from the spirit and scope of the technical solutions of the embodiments of the present invention.
Claims (8)
1. A method for fault prediction based on multimodal data, the method comprising:
acquiring a plurality of pieces of abnormal operation data of target equipment, wherein each piece of abnormal operation data corresponds to different data types respectively;
carrying out data fusion on each abnormal operation data to obtain fusion data;
acquiring a preset prediction model, inputting the fusion data into the prediction model to obtain target fault prediction data corresponding to the target equipment, wherein the prediction model comprises a plurality of sub-prediction models, each sub-prediction model is respectively constructed based on different types of neural networks, and the weight value of each sub-prediction model is determined based on the data distribution characteristics of the fusion data;
the method for determining the weight value of each sub-prediction model comprises the following steps:
acquiring the data distribution characteristics corresponding to the fusion data;
obtaining a confidence coefficient database corresponding to the sub-prediction model, wherein the confidence coefficient database is used for reflecting the prediction confidence coefficient of the sub-prediction model under different reference data distribution characteristics;
comparing the data distribution characteristics with the confidence coefficient database, and determining target prediction confidence coefficient corresponding to the data distribution characteristics according to the comparison result;
determining the weight value corresponding to the sub-prediction model according to the target prediction confidence;
the method for determining the prediction confidence of each reference data distribution feature and each reference data distribution feature comprises the following steps:
acquiring a plurality of historical fusion data and fault prediction labels corresponding to the historical fusion data respectively, acquiring similarity between the historical data distribution characteristics of the historical fusion data, and classifying the historical data distribution characteristics according to the similarity to obtain a plurality of characteristic sets;
according to the feature sets, determining fusion distribution features corresponding to the feature sets respectively, and taking the fusion distribution features as the reference data distribution features;
classifying each historical fusion data according to each feature set to obtain a plurality of test data sets, and determining the recognition accuracy of the prediction model under each reference data distribution feature according to each test data set and the corresponding fault prediction label;
and determining the prediction confidence corresponding to each reference data distribution characteristic according to each identification accuracy.
2. The multi-modal data-based fault prediction method as set forth in claim 1, wherein the method for obtaining the abnormal operation data corresponding to each data type includes:
acquiring equipment operation data corresponding to the data type, and taking the equipment operation data as data to be filtered;
filtering the data to be filtered to obtain filtered data, and determining a data change value according to the data to be filtered and the filtered data;
judging whether the data change value is smaller than a preset change threshold value, if not, taking the filtered data as the data to be filtered, and continuing to execute the step of filtering the data to be filtered to obtain the filtered data until the data change value is smaller than the change threshold value, so as to obtain target filtered data corresponding to the equipment operation data;
and determining the abnormal operation data corresponding to the data type according to the equipment operation data and the target filtering data.
3. The multi-modal data-based fault prediction method of claim 1, wherein the target fault prediction data determination method comprises:
obtaining fault prediction data corresponding to each sub prediction model respectively;
and carrying out weighting processing according to the weight value of each sub-prediction model and the fault prediction data to obtain the target fault prediction data.
4. The multi-modal data-based fault prediction method of claim 1, wherein one of the sub-prediction models is an LSTM model, and a regularization term is included in a loss function of the LSTM model.
5. The method of claim 1, wherein each of the abnormal operation data includes abnormal vibration data and abnormal sound wave data.
6. A multi-modal data-based fault prediction apparatus, the apparatus comprising:
the data acquisition module is used for acquiring a plurality of pieces of abnormal operation data of the target equipment, wherein each piece of abnormal operation data corresponds to different data types respectively;
the data fusion module is used for carrying out data fusion on each abnormal operation data to obtain fusion data;
the fault prediction module is used for acquiring a preset prediction model, inputting the fusion data into the prediction model to obtain target fault prediction data corresponding to the target equipment, wherein the prediction model comprises a plurality of sub-prediction models, each sub-prediction model is respectively constructed based on different types of neural networks, and the weight value of each sub-prediction model is determined based on the data distribution characteristics of the fusion data;
the method for determining the weight value of each sub-prediction model comprises the following steps:
acquiring the data distribution characteristics corresponding to the fusion data;
obtaining a confidence coefficient database corresponding to the sub-prediction model, wherein the confidence coefficient database is used for reflecting the prediction confidence coefficient of the sub-prediction model under different reference data distribution characteristics;
comparing the data distribution characteristics with the confidence coefficient database, and determining target prediction confidence coefficient corresponding to the data distribution characteristics according to the comparison result;
determining the weight value corresponding to the sub-prediction model according to the target prediction confidence;
the method for determining the prediction confidence of each reference data distribution feature and each reference data distribution feature comprises the following steps:
acquiring a plurality of historical fusion data and fault prediction labels corresponding to the historical fusion data respectively, acquiring similarity between the historical data distribution characteristics of the historical fusion data, and classifying the historical data distribution characteristics according to the similarity to obtain a plurality of characteristic sets;
according to the feature sets, determining fusion distribution features corresponding to the feature sets respectively, and taking the fusion distribution features as the reference data distribution features;
classifying each historical fusion data according to each feature set to obtain a plurality of test data sets, and determining the recognition accuracy of the prediction model under each reference data distribution feature according to each test data set and the corresponding fault prediction label;
and determining the prediction confidence corresponding to each reference data distribution characteristic according to each identification accuracy.
7. An electronic device comprising a memory, a processor and a computer program stored on the memory and executable on the processor, wherein the processor implements the multimodal data based fault prediction method of any of claims 1 to 5 when the program is executed.
8. A non-transitory computer readable storage medium having stored thereon a computer program, wherein the computer program when executed by a processor implements the multi-modal data-based fault prediction method of any of claims 1 to 5.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310487654.0A CN116184988B (en) | 2023-05-04 | 2023-05-04 | Multi-mode data-based fault prediction method, device, equipment and storage medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310487654.0A CN116184988B (en) | 2023-05-04 | 2023-05-04 | Multi-mode data-based fault prediction method, device, equipment and storage medium |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN116184988A CN116184988A (en) | 2023-05-30 |
| CN116184988B true CN116184988B (en) | 2023-07-21 |
Family
ID=86447489
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202310487654.0A Active CN116184988B (en) | 2023-05-04 | 2023-05-04 | Multi-mode data-based fault prediction method, device, equipment and storage medium |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN116184988B (en) |
Families Citing this family (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117102950B (en) * | 2023-10-17 | 2023-12-22 | 上海诺倬力机电科技有限公司 | Fault analysis method, device, electronic equipment and computer readable storage medium |
| CN118013444B (en) * | 2024-04-08 | 2024-06-11 | 徐州硕博电子科技有限公司 | Abnormality monitoring method and device for intelligent mine multi-autonomous cluster operation equipment |
| CN118152901B (en) * | 2024-05-13 | 2024-07-26 | 西北工业大学 | Equipment fault prediction method and system based on data driving |
| CN118503884B (en) * | 2024-07-17 | 2024-09-17 | 浪潮通用软件有限公司 | Equipment state identification method, equipment and medium |
Family Cites Families (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112651534B (en) * | 2019-10-10 | 2024-07-02 | 顺丰科技有限公司 | Method, device and storage medium for predicting resource supply chain demand |
| CN112578213A (en) * | 2020-12-23 | 2021-03-30 | 交控科技股份有限公司 | Fault prediction method and device for rail power supply screen |
| CN113128793A (en) * | 2021-05-19 | 2021-07-16 | 中国南方电网有限责任公司 | Photovoltaic power combination prediction method and system based on multi-source data fusion |
| CN113723716B (en) * | 2021-11-02 | 2022-03-18 | 深圳市城市交通规划设计研究中心股份有限公司 | Passenger flow classification early warning abnormity warning method, device and storage medium |
| CN114239885A (en) * | 2022-01-11 | 2022-03-25 | 中国科学院深圳先进技术研究院 | Operation fault prediction method and device |
| CN115238831B (en) * | 2022-09-21 | 2023-04-14 | 中国南方电网有限责任公司超高压输电公司广州局 | Fault prediction method, device, computer equipment and storage medium |
| CN115758121A (en) * | 2022-11-22 | 2023-03-07 | 中国电信股份有限公司 | Bearing fault classification method and device, electronic equipment and readable storage medium |
| CN115758225B (en) * | 2023-01-06 | 2023-08-29 | 中建科技集团有限公司 | Fault prediction method and device based on multi-mode data fusion and storage medium |
-
2023
- 2023-05-04 CN CN202310487654.0A patent/CN116184988B/en active Active
Also Published As
| Publication number | Publication date |
|---|---|
| CN116184988A (en) | 2023-05-30 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN116184988B (en) | Multi-mode data-based fault prediction method, device, equipment and storage medium | |
| CN113899577A (en) | Abnormal sound detection device, abnormal sound generation device, and abnormal sound generation method | |
| CN111310814A (en) | Method and device for training business prediction model by utilizing unbalanced positive and negative samples | |
| US20230161842A1 (en) | Parameter setting method, parameter setting device, and electronical device | |
| CN110008082B (en) | Abnormal task intelligent monitoring method, device, equipment and storage medium | |
| KR20190124846A (en) | The design of GRU-based cell structure robust to missing value and noise of time-series data in recurrent neural network | |
| CN112527604A (en) | Deep learning-based operation and maintenance detection method and system, electronic equipment and medium | |
| CN112232370A (en) | Fault analysis and prediction method for engine | |
| CN113989519B (en) | Long-tail target detection method and system | |
| CN109656818B (en) | Fault prediction method for software intensive system | |
| CN115758225A (en) | Fault prediction method and device based on multi-mode data fusion and storage medium | |
| CN114962390A (en) | Hydraulic system fault diagnosis method and system and working machine | |
| CN116595465A (en) | High-dimensional sparse data outlier detection method and system based on self-encoder and data enhancement | |
| CN114328821A (en) | Multi-round conversation control method and device based on control slot position and service data slot position | |
| CN107979606B (en) | Self-adaptive distributed intelligent decision-making method | |
| CN117097541A (en) | API service attack detection method, device, equipment and storage medium | |
| CN116361695A (en) | Abnormal data detection method and device | |
| CN116502177A (en) | Fault prediction method, device, equipment and medium for passive optical network optical module | |
| CN113807541B (en) | Fairness repair method, system, equipment and storage medium for decision system | |
| CN115758086A (en) | Method, device and equipment for detecting faults of cigarette cut-tobacco drier and readable storage medium | |
| JP2022088341A (en) | Apparatus learning device and method | |
| CN114202110A (en) | Service fault prediction method and device based on RF-XGBOOST | |
| CN111027678A (en) | Data migration method and device | |
| CN113094504A (en) | Self-adaptive text classification method and device based on automatic machine learning | |
| CN112598020A (en) | Target identification 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 |