CN115221924A - Time-series multi-mode-based industrial equipment anomaly detection intelligent recognition algorithm framework - Google Patents
Time-series multi-mode-based industrial equipment anomaly detection intelligent recognition algorithm framework Download PDFInfo
- Publication number
- CN115221924A CN115221924A CN202210840517.6A CN202210840517A CN115221924A CN 115221924 A CN115221924 A CN 115221924A CN 202210840517 A CN202210840517 A CN 202210840517A CN 115221924 A CN115221924 A CN 115221924A
- Authority
- CN
- China
- Prior art keywords
- data
- equipment
- time
- industrial equipment
- collection module
- 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.)
- Pending
Links
Images
Landscapes
- Testing And Monitoring For Control Systems (AREA)
Abstract
The invention relates to the technical field of industrial equipment monitoring, and discloses an industrial equipment anomaly detection intelligent recognition algorithm framework based on time sequence multi-mode, which comprises the following steps: s1: the method comprises the following steps of (1) equipment installation, wherein detection equipment is installed on equipment to be detected, and the detection equipment comprises a camera, a sound pick-up, a temperature sensor, a humidity sensor, a voltage sensor, a current sensor and a timer; s2: data collection, namely collecting signal source data detected by the equipment to be detected through detection equipment, and collecting the collected data to a data platform through a signal port or a network for summarizing; s3: and (5) extracting features. The invention has wide universality, and can be trained by adopting the frame to collect data no matter depending on images collected by a camera, sound collected by a sound pick-up, temperature collected by a thermometer and the like, thereby improving the development efficiency and reducing the cost.
Description
Technical Field
The invention relates to the technical field of industrial equipment monitoring, in particular to an intelligent identification algorithm framework for industrial equipment anomaly detection based on time sequence multi-mode.
Background
The industrial production equipment refers to machine equipment which directly participates in the production process or directly serves for production in industrial enterprises, and mainly comprises machinery, power and conduction equipment and the like, and the industrial equipment is frequently damaged due to mechanical abrasion, service life, plant humidity, temperature and other reasons in the normal use process, so that the state of the industrial equipment needs to be monitored, and timely maintenance is facilitated.
The non-contact monitoring method for the abnormal state of the industrial production equipment mainly comprises image monitoring, sound monitoring or other single signals, the abnormal state of the equipment is judged by analyzing the signals at a certain moment, the abnormal state of the equipment can have various expression forms including an image layer, a sound layer, a temperature layer, a humidity layer and the like, in addition, the abnormal expression form can not be analyzed from a single moment, and the judgment can be made only by continuous sequence observed values.
Disclosure of Invention
Technical problem to be solved
Aiming at the defects of the prior art, the invention provides an intelligent identification algorithm framework for the abnormality detection of the industrial equipment based on the time series multi-mode, which solves the problems that the existing single-signal-source single-time analysis method has certain limitations, the judgment accuracy is reduced, a unified technical framework is lacked, different signal source analysis modes are different, and the development period is long and the cost is high.
(II) technical scheme
In order to achieve the purpose, the invention provides the following technical scheme:
an industrial equipment anomaly detection intelligent recognition algorithm framework based on time series multi-mode comprises the following steps:
s1: the method comprises the following steps of (1) equipment installation, wherein detection equipment is installed on the equipment to be detected, and the detection equipment comprises a camera, a sound pick-up, a temperature sensor, a humidity sensor, a voltage sensor, a current sensor and a timer;
s2: data collection, namely collecting signal source data detected by the equipment to be detected through detection equipment, and collecting the collected data to a data platform through a signal port or a network for summarizing;
s3: extracting characteristics, wherein the collected image data is subjected to characteristic extraction by adopting a convolutional neural network or a transform neural network, the sound data of the industrial equipment collected by a sound pick-up is subjected to characteristic extraction by adopting a Mel Frequency Cepstrum Coefficient (MFCC), and the numerical data of the industrial equipment collected by other equipment is not subjected to characteristic extraction except the image data and the sound data;
s4: extracting characteristics in a segmented manner, namely, acquiring signal source data of various different types in a certain time period, respectively extracting characteristics of different signal sources in a mode of S3, respectively inputting the extracted multi-modal signal source characteristics into a network based on Self-authorization after time alignment, respectively aligning time sequences of different signal sources based on time, and obtaining multi-modal signal source data characteristic representation containing time sequence information in the time period;
s5: calculating, namely integrating the features extracted in the S4, inputting the features into a full-connection discrimination network, and calculating the probability of each state of the industrial equipment according to softmax;
s6: and evaluating and detecting, namely judging whether the equipment is abnormal or not in the long-time state according to the calculated probability of the equipment state.
As a further scheme of the present invention, the S1 includes a data collection module, the data collection module includes a timing module and a data collection module, the data collection module is connected to the timing module, and the data collection module includes an image collection module, a sound collection module, a voltage collection module, a current collection module, a temperature collection module, and a humidity collection module.
Further, in S2, the summarized data are classified, and are associated with time, and are analyzed according to the corresponding time series.
On the basis of the scheme, the characteristics of the non-numerical signal sources such as the image and the voice are extracted in the S3, and the next step of calculation is facilitated.
Further, the S4 includes a feature extraction module in a time period, and calculates a time sequence feature formed by features of the multi-modal signal source in the time period.
On the basis of the scheme, the S5 comprises a calculation module, and the data calculated by the calculation module represents the probability of each state of the industrial equipment.
In a further aspect of the present invention, after the comparison in S6, the abnormal state of the industrial device is determined according to the probability in S5.
(III) advantageous effects
Compared with the prior art, the invention provides an intelligent identification algorithm framework for the anomaly detection of the industrial equipment based on the time series multimodal, which has the following beneficial effects:
1. according to the invention, by means of the multi-signal source data, different noises can be generated under certain scenes, such as equipment abnormality, and the noises are visually different from normal conditions, so that the accuracy of abnormality identification is improved by combining a plurality of signal sources.
2. In the invention, the summarized data are classified, correspond to time, and are analyzed according to the corresponding time sequence, so that the scene which is difficult to judge at a single step time can be solved by means of time sequence analysis, and the application universality and accuracy are further improved.
3. In the invention, the extracted multi-modal signal source characteristics are respectively input into a Self-authorization-based network after being aligned based on time, so that the signal source data characteristic representation in the time period is obtained, and the relation of time sequences can be captured, thereby solving the problem that the time sequence information is needed to judge.
4. The method has wide universality, and the frame can be used for acquiring data for training no matter depending on images acquired by a camera, sound acquired by a sound pick-up, temperature acquired by a thermometer and the like, so that the development efficiency is improved, and the cost is reduced.
Drawings
Fig. 1 is a schematic flow structure diagram of an industrial equipment anomaly detection intelligent recognition algorithm framework based on time series multi-modal.
Detailed Description
The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
Example 1
Referring to fig. 1, an intelligent industrial equipment anomaly detection and identification algorithm framework based on time series multi-mode comprises the following steps:
s1: the method comprises the following steps of installing equipment, namely installing detection equipment on the equipment to be detected, wherein the detection equipment comprises a camera, a sound pick-up, a temperature sensor, a humidity sensor, a voltage sensor, a current sensor and a timer, the S1 comprises a data acquisition module, the data acquisition module comprises a timing module and a data collection module, the data collection module is connected with the timing module, and the data collection module comprises an image acquisition module, a sound acquisition module, a voltage acquisition module, a current acquisition module, a temperature acquisition module and a humidity acquisition module;
s2: data collection, namely collecting signal source data detected by equipment to be detected through detection equipment, collecting the collected data to a data platform through a signal port or a network for summarization, sending different noises and generating vibration under certain scenes such as equipment abnormity by means of multi-signal source data, and improving the accuracy of abnormity identification by combining a plurality of signal sources, classifying the summarized data in S2, corresponding to time, analyzing according to a corresponding time sequence, and solving the scene difficult to judge at a single step moment by means of time sequence analysis, thereby further improving the application universality and accuracy;
s3: extracting characteristics, wherein the collected image data is subjected to characteristic extraction by adopting a convolutional neural network or a transform neural network, the sound data of the industrial equipment collected by a sound pick-up is subjected to characteristic extraction by adopting a Mel Frequency Cepstrum Coefficient (MFCC), and the numerical data of the industrial equipment collected by other equipment is not subjected to characteristic extraction except the image data and the sound data;
s4: extracting characteristics in a segmented manner, namely extracting various types of signal source data in a certain time period, wherein different signal sources respectively extract the characteristics in the S3 manner, and respectively inputting the extracted multi-modal signal source characteristics into a Self-Attention-based network after time alignment, so that the signal source data characteristic characterization in the time period is obtained, and the relation of time sequences can be captured to solve the problem that a scene needing time sequence information for judgment is solved;
s5: calculating, namely integrating the features extracted in the step S4 together, inputting the features into a full-connection discrimination network, and calculating the probability of each state of the industrial equipment according to softmax, wherein the method has wide universality, and can be used for training by adopting the frame to collect data no matter the method depends on images collected by a camera, sound collected by a sound pick-up, temperature collected by a thermometer and the like, so that the development efficiency is improved, and the cost is reduced;
s6: and evaluating and detecting, namely judging whether the equipment is abnormal or not in the long-time state according to the calculated probability of the equipment state.
Example 2
Referring to fig. 1, an intelligent industrial equipment anomaly detection and identification algorithm framework based on time series multi-mode comprises the following steps:
s1: the method comprises the following steps of (1) mounting equipment, wherein the detection equipment is mounted on the equipment to be detected and comprises a camera, a pickup, a temperature sensor, a humidity sensor, a voltage sensor, a current sensor and a timer, the S1 comprises a data acquisition module, the data acquisition module comprises a timing module and a data collection module, the data collection module is connected with the timing module, and the data collection module comprises an image acquisition module, a sound acquisition module, a voltage acquisition module, a current acquisition module, a temperature acquisition module and a humidity acquisition module;
s2: the method comprises the steps of collecting signal source data detected by the device to be detected through the detection device, collecting the collected data to a data platform through a signal port or a network, summarizing the collected data, and generating different noises and vibration under certain scenes by means of multi-signal-source data if the device is abnormal, so that the accuracy of abnormal identification is improved by combining a plurality of signal sources;
s3: extracting characteristics, wherein the collected image data is subjected to characteristic extraction by adopting a convolutional neural network or a Transformer neural network, the sound data of the industrial equipment collected by a sound pick-up is subjected to characteristic extraction by adopting a Mel Frequency Cepstrum Coefficient (MFCC), and the numerical data of the industrial equipment collected by other equipment is not subjected to characteristic extraction except the image data and the sound data;
s4: extracting characteristics in a segmented manner, namely extracting various types of signal source data in a certain time period, wherein different signal sources respectively extract the characteristics in the S3 manner, and respectively inputting the extracted multi-modal signal source characteristics into a Self-Attention-based network after time alignment, so that the signal source data characteristic characterization in the time period is obtained, and the relation of time sequences can be captured to solve the problem that a scene needing time sequence information for judgment is solved;
s5: calculating, namely integrating the features extracted in the step S4 together, inputting the features into a full-connection discrimination network, and calculating the probability of each state of the industrial equipment according to softmax, wherein the method has wide universality, and can be used for training by adopting the frame to collect data no matter the method depends on images collected by a camera, sound collected by a sound pick-up, temperature collected by a thermometer and the like, so that the development efficiency is improved, and the cost is reduced;
s6: and evaluating and detecting, namely judging whether the equipment is abnormal or not in the long-time state according to the calculated probability of the equipment state.
In the description herein, it is noted that relational terms such as first and second, and the like, are used solely to distinguish one entity or action from another entity or action without necessarily requiring or implying any actual such relationship or order between such entities or actions. Also, the terms "comprises," "comprising," or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements does not include only those elements but may include other elements not expressly listed or inherent to such process, method, article, or apparatus.
Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and alterations can be made in these embodiments without departing from the principles and spirit of the invention, the scope of which is defined in the appended claims and their equivalents.
Claims (7)
1. An industrial equipment anomaly detection intelligent recognition algorithm framework based on a time series multi-mode is characterized by comprising the following steps:
s1: the method comprises the following steps of (1) equipment installation, wherein detection equipment is installed on equipment to be detected, and the detection equipment comprises a camera, a sound pick-up, a temperature sensor, a humidity sensor, a voltage sensor, a current sensor and a timer;
s2: data collection, namely collecting signal source data detected by the equipment to be detected through detection equipment, and collecting the collected data to a data platform through a signal port or a network for summarizing;
s3: extracting characteristics, wherein the collected image data is subjected to characteristic extraction by adopting a convolutional neural network or a transform neural network, the sound data of the industrial equipment collected by a sound pick-up is subjected to characteristic extraction by adopting a Mel Frequency Cepstrum Coefficient (MFCC), and the numerical data of the industrial equipment collected by other equipment is not subjected to characteristic extraction except the image data and the sound data;
s4: extracting characteristics in a segmented manner, namely taking various types of signal source data in a certain time period, respectively extracting the characteristics of different signal sources in a mode of S3, respectively inputting the extracted multi-modal signal source characteristics into a Self-Attention-based network after time alignment, respectively aligning the time sequences of different signal sources based on time, and obtaining multi-modal signal source data characteristic representation containing time sequence information in the time period;
s5: calculating, namely integrating the features extracted in the step S4, inputting the features into a full-connection discrimination network, and calculating the probability of each state of the industrial equipment according to softmax;
s6: and evaluating and detecting, namely judging whether the equipment is abnormal or not in the long-time state according to the calculated probability of the equipment state.
2. The time-series multi-modal based industrial equipment anomaly detection intelligent recognition algorithm framework as claimed in claim 1, wherein S1 comprises a data collection module, the data collection module comprises a timing module and a data collection module, the data collection module is connected with the timing module, and the data collection module comprises an image collection module, a sound collection module, a voltage collection module, a current collection module, a temperature collection module and a humidity collection module.
3. The time-series multimodal industrial equipment anomaly detection intelligent recognition algorithm framework as claimed in claim 2, wherein the summarized data are classified in S2, are corresponding to time, and are analyzed according to the corresponding time series.
4. The time-series multi-modal-based industrial equipment anomaly detection intelligent recognition algorithm framework is characterized in that in the step S3, non-numerical data such as images and voices are subjected to feature extraction, so that the next calculation is facilitated.
5. The framework of claim 4, wherein the S4 comprises a feature extraction module in a time period, and calculates a time series feature formed by features of multi-modal signal sources in the time period.
6. The framework of claim 3, wherein the S5 comprises a calculation module, and the data calculated by the calculation module represents the probability of each state of the industrial equipment.
7. The time-series multi-modal-based industrial equipment anomaly detection intelligent recognition algorithm framework as claimed in claim 1, wherein in S6, after probability calculation, whether the industrial equipment state is abnormal or not is judged.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210840517.6A CN115221924A (en) | 2022-07-18 | 2022-07-18 | Time-series multi-mode-based industrial equipment anomaly detection intelligent recognition algorithm framework |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210840517.6A CN115221924A (en) | 2022-07-18 | 2022-07-18 | Time-series multi-mode-based industrial equipment anomaly detection intelligent recognition algorithm framework |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN115221924A true CN115221924A (en) | 2022-10-21 |
Family
ID=83611037
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202210840517.6A Pending CN115221924A (en) | 2022-07-18 | 2022-07-18 | Time-series multi-mode-based industrial equipment anomaly detection intelligent recognition algorithm framework |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN115221924A (en) |
-
2022
- 2022-07-18 CN CN202210840517.6A patent/CN115221924A/en active Pending
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110366031B (en) | Vision-based abnormal state monitoring and fault diagnosis method for MES (manufacturing execution system) of digital workshop | |
| CN112660745B (en) | Intelligent diagnosis method and system for carrier roller fault and readable storage medium | |
| KR102249113B1 (en) | Automatic equipment failure prediction system in smart factory | |
| CN112925279A (en) | Fault comprehensive analysis system based on MES system | |
| CN112173636B (en) | Method for detecting faults of belt conveyor carrier roller by inspection robot | |
| KR102575629B1 (en) | Equipment failure prediction system and method | |
| US20210149387A1 (en) | Facility failure prediction system and method for using acoustic signal of ultrasonic band | |
| CN116383633A (en) | Method and system for detecting faults of machine-made sand vibrating screen through multi-factor comprehensive analysis | |
| CN114371353A (en) | Power equipment abnormity monitoring method and system based on voiceprint recognition | |
| CN110287827A (en) | A kind of bridge strain data outliers recognition methods based on data correlation | |
| CN113866455A (en) | Bridge acceleration monitoring data anomaly detection method, system and device based on deep learning | |
| CN113962308A (en) | Aviation equipment fault prediction method | |
| CN116522156A (en) | Equipment state data analysis system and method based on energy management platform | |
| CN115220396A (en) | Intelligent monitoring method and system for numerical control machine tool | |
| CN109598309B (en) | Detection system and monitoring method of metal packaging punching machine | |
| CN115221924A (en) | Time-series multi-mode-based industrial equipment anomaly detection intelligent recognition algorithm framework | |
| CN117277566A (en) | Power grid data analysis power dispatching system and method based on big data | |
| KR20210077496A (en) | Equipment failure prediction system having multi-channel sensors for receiving acoustic signals in the ultrasonic band | |
| KR102325439B1 (en) | Failure Prediction System of Injection Molding Equipment | |
| CN113672859A (en) | Switch point machine fault acoustic diagnosis system | |
| JP2004125641A (en) | Abnormal sound detecting apparatus | |
| KR102500140B1 (en) | Equipment failure prediction system and method using sound signal of ultrasonic band | |
| CN111157812A (en) | Power equipment fault monitoring method based on electromagnetic induction | |
| CN117672255B (en) | Abnormal equipment identification method and system based on artificial intelligence and equipment operation sound | |
| CN116663875B (en) | Intelligent installation engineering monitoring system and method based on big data |
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 |