CN117854245A - Abnormal equipment monitoring method and system based on equipment operation audio - Google Patents
Abnormal equipment monitoring method and system based on equipment operation audio Download PDFInfo
- Publication number
- CN117854245A CN117854245A CN202311795505.7A CN202311795505A CN117854245A CN 117854245 A CN117854245 A CN 117854245A CN 202311795505 A CN202311795505 A CN 202311795505A CN 117854245 A CN117854245 A CN 117854245A
- Authority
- CN
- China
- Prior art keywords
- audio
- equipment
- training
- target
- abnormal
- 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.)
- Granted
Links
- 230000002159 abnormal effect Effects 0.000 title claims abstract description 232
- 238000000034 method Methods 0.000 title claims abstract description 56
- 238000012544 monitoring process Methods 0.000 title claims abstract description 41
- 230000005856 abnormality Effects 0.000 claims abstract description 21
- 238000012512 characterization method Methods 0.000 claims abstract description 11
- 230000000977 initiatory effect Effects 0.000 claims abstract description 4
- 238000012549 training Methods 0.000 claims description 366
- 239000013598 vector Substances 0.000 claims description 140
- 230000007613 environmental effect Effects 0.000 claims description 59
- 238000012360 testing method Methods 0.000 claims description 29
- 230000005236 sound signal Effects 0.000 claims description 24
- 230000008569 process Effects 0.000 claims description 21
- 238000003491 array Methods 0.000 claims description 11
- 230000010354 integration Effects 0.000 claims description 10
- 238000006243 chemical reaction Methods 0.000 claims description 9
- 238000010586 diagram Methods 0.000 claims description 7
- 230000006835 compression Effects 0.000 claims description 5
- 238000007906 compression Methods 0.000 claims description 5
- 239000012634 fragment Substances 0.000 claims description 4
- 238000005520 cutting process Methods 0.000 claims description 3
- 238000013461 design Methods 0.000 abstract description 22
- 238000001514 detection method Methods 0.000 abstract description 5
- 238000012423 maintenance Methods 0.000 description 16
- 238000004519 manufacturing process Methods 0.000 description 9
- 238000003745 diagnosis Methods 0.000 description 7
- 238000013519 translation Methods 0.000 description 7
- 230000003993 interaction Effects 0.000 description 5
- 238000012545 processing Methods 0.000 description 5
- 238000004458 analytical method Methods 0.000 description 4
- 238000004422 calculation algorithm Methods 0.000 description 4
- 238000012986 modification Methods 0.000 description 4
- 230000004048 modification Effects 0.000 description 4
- 230000009286 beneficial effect Effects 0.000 description 3
- 238000004891 communication Methods 0.000 description 3
- 230000009471 action Effects 0.000 description 2
- 230000000052 comparative effect Effects 0.000 description 2
- 238000010801 machine learning Methods 0.000 description 2
- 230000003595 spectral effect Effects 0.000 description 2
- 206010000117 Abnormal behaviour Diseases 0.000 description 1
- 238000012935 Averaging Methods 0.000 description 1
- 230000005540 biological transmission Effects 0.000 description 1
- 238000004364 calculation method Methods 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 239000002131 composite material Substances 0.000 description 1
- 238000009826 distribution Methods 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 238000000605 extraction Methods 0.000 description 1
- 238000001914 filtration Methods 0.000 description 1
- 238000009776 industrial production Methods 0.000 description 1
- 238000007689 inspection Methods 0.000 description 1
- 238000011835 investigation Methods 0.000 description 1
- 238000007726 management method Methods 0.000 description 1
- 230000007246 mechanism Effects 0.000 description 1
- 238000005457 optimization Methods 0.000 description 1
- 238000013021 overheating Methods 0.000 description 1
- 238000003909 pattern recognition Methods 0.000 description 1
- 230000000737 periodic effect Effects 0.000 description 1
- 230000004044 response Effects 0.000 description 1
- 238000011524 similarity measure Methods 0.000 description 1
- 238000001228 spectrum Methods 0.000 description 1
- 238000010183 spectrum analysis Methods 0.000 description 1
- 238000003860 storage Methods 0.000 description 1
- 230000002123 temporal effect Effects 0.000 description 1
- 230000001960 triggered effect Effects 0.000 description 1
- 238000013024 troubleshooting Methods 0.000 description 1
Classifications
-
- G—PHYSICS
- G08—SIGNALLING
- G08B—SIGNALLING OR CALLING SYSTEMS; ORDER TELEGRAPHS; ALARM SYSTEMS
- G08B21/00—Alarms responsive to a single specified undesired or abnormal condition and not otherwise provided for
- G08B21/18—Status alarms
- G08B21/185—Electrical failure alarms
-
- G—PHYSICS
- G10—MUSICAL INSTRUMENTS; ACOUSTICS
- G10L—SPEECH ANALYSIS TECHNIQUES OR SPEECH SYNTHESIS; SPEECH RECOGNITION; SPEECH OR VOICE PROCESSING TECHNIQUES; SPEECH OR AUDIO CODING OR DECODING
- G10L25/00—Speech or voice analysis techniques not restricted to a single one of groups G10L15/00 - G10L21/00
- G10L25/48—Speech or voice analysis techniques not restricted to a single one of groups G10L15/00 - G10L21/00 specially adapted for particular use
- G10L25/51—Speech or voice analysis techniques not restricted to a single one of groups G10L15/00 - G10L21/00 specially adapted for particular use for comparison or discrimination
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Health & Medical Sciences (AREA)
- Signal Processing (AREA)
- Audiology, Speech & Language Pathology (AREA)
- Human Computer Interaction (AREA)
- Computational Linguistics (AREA)
- Acoustics & Sound (AREA)
- Multimedia (AREA)
- Business, Economics & Management (AREA)
- Emergency Management (AREA)
- General Physics & Mathematics (AREA)
- Measurement Of Mechanical Vibrations Or Ultrasonic Waves (AREA)
Abstract
The invention discloses an abnormal equipment monitoring method and system based on equipment operation audio, comprising the following steps: firstly, responding to an abnormal equipment monitoring instruction, and determining a target environment to be monitored; then collecting sound in the target environment, and extracting target equipment operation audio in the sound; determining abnormal conditions of the equipment according to the operation audios; and finally, according to the abnormal condition grade of the equipment abnormal condition characterization, initiating alarm information of the corresponding grade. By the design, the equipment state can be monitored in real time by automatically collecting and analyzing the equipment operation audio, the abnormal condition of the equipment can be found in time, and alarm information of corresponding grade is initiated according to the severity of the abnormality. This will greatly increase the efficiency and accuracy of the equipment monitoring, while also reducing reliance on manual detection.
Description
Technical Field
The invention relates to the technical field of equipment management, in particular to an abnormal equipment monitoring method and system based on equipment operation audio.
Background
In industrial production, home applications, etc., the proper functioning of the device is critical. However, the device may fail for various reasons, such as mechanical wear, circuit problems, software errors, etc. Generally, the operation state of the equipment is changed before the equipment fails, and the change can be represented by means of sound, vibration, temperature and the like. However, conventional equipment monitoring methods typically rely on periodic maintenance and manual inspection, which are inefficient and prone to error.
Disclosure of Invention
The invention aims to provide an abnormal equipment monitoring method and system based on equipment operation audio.
In a first aspect, an embodiment of the present invention provides a method for monitoring abnormal equipment based on equipment operation audio, where the method includes:
responding to an abnormal equipment monitoring instruction, and determining a target environment to be monitored;
collecting the target environment sound in the target environment, and extracting target equipment operation audio in the target environment sound;
determining equipment abnormality corresponding to the target equipment operation audio according to the target equipment operation audio;
and initiating alarm information of a corresponding grade according to the abnormal condition grade of the equipment abnormal condition characterization.
In a second aspect, an embodiment of the present invention provides a server system, including a server, where the server is configured to perform the method described in the first aspect.
Compared with the prior art, the invention has the beneficial effects that: by adopting the abnormal equipment monitoring method and system based on the equipment operation audio, the target environment to be monitored is determined by responding to the abnormal equipment monitoring instruction; then collecting sound in the target environment, and extracting target equipment operation audio in the sound; determining abnormal conditions of the equipment according to the operation audios; and finally, according to the abnormal condition grade of the equipment abnormal condition characterization, initiating alarm information of the corresponding grade. By the design, the equipment state can be monitored in real time by automatically collecting and analyzing the equipment operation audio, the abnormal condition of the equipment can be found in time, and alarm information of corresponding grade is initiated according to the severity of the abnormality. This will greatly increase the efficiency and accuracy of the equipment monitoring, while also reducing reliance on manual detection.
Drawings
In order to more clearly illustrate the technical solution of the embodiments of the present invention, the drawings that are required to be used in the embodiments will be briefly described. It is appreciated that the following drawings depict only certain embodiments of the invention and are therefore not to be considered limiting of its scope. Other relevant drawings may be made by those of ordinary skill in the art without undue burden from these drawings.
FIG. 1 is a schematic block diagram of a step flow of an abnormal equipment monitoring method based on equipment operation audio provided by an embodiment of the invention;
fig. 2 is a schematic block diagram of a computer device according to an embodiment of the present invention.
Detailed Description
In order to make the objects, technical solutions and advantages of the embodiments of the present invention more clear, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention. It will be apparent that the described embodiments are some, but not all, embodiments of the invention. The components of the embodiments of the present invention generally described and illustrated in the figures herein may be arranged and designed in a wide variety of different configurations.
The following describes specific embodiments of the present invention in detail with reference to the drawings.
In order to solve the foregoing technical problems in the background art, fig. 1 is a flowchart of an abnormal device monitoring method based on device operation audio according to an embodiment of the present disclosure, and the abnormal device monitoring method based on device operation audio is described in detail below.
Step S201, in response to an abnormal equipment monitoring instruction, determining a target environment to be monitored;
step S202, collecting the target environment sound in the target environment, and extracting target equipment operation audio in the target environment sound;
step S203, according to the operation audio of the target equipment, determining the equipment abnormality corresponding to the operation audio of the target equipment;
step S204, according to the abnormal condition grade of the equipment abnormal condition characterization, alarm information of a corresponding grade is initiated.
In an embodiment of the present invention, multiple machines are running, illustratively, in a single manufacturing facility. When an operator issues an abnormal equipment monitoring command, the system selects a particular machine or whole production line to be monitored as a target environment according to the command. In a target environment, a plurality of microphones are installed to collect sounds in the environment. These microphones may capture various sounds, including sounds produced by the device while it is in operation. By analyzing the collected sound data, the system is able to extract therefrom audio information related to the operation of the target device. The system compares the extracted target device operating audio with known device normal operating audio and uses machine learning algorithms or other pattern recognition techniques to analyze the differences between them. If the extracted operation audio is inconsistent with the normal condition, the system judges that the target equipment has abnormal condition, and further determines specific abnormal type. Each anomaly may be assigned a particular anomaly level to indicate its severity. And according to the level of the abnormal condition of the equipment, the system generates corresponding alarm information. For example, for low-level anomalies, the system may generate only a slight alarm, while for high-level anomalies, the system may immediately notify the operator or take emergency action to ensure the safety of equipment and personnel.
By the design, the abnormal equipment monitoring method based on the equipment operation audio can monitor the operation state of the industrial equipment in real time, and timely send out corresponding alarm information when abnormal conditions are found, so that the reliability of the equipment and the operation safety are improved.
In one possible implementation, the aforementioned step S202 may be performed by the following steps.
(1) Collecting the target environmental sound in the target environment, wherein the target environmental sound comprises a plurality of operation audios of undetermined equipment;
(2) Collecting correlation information between at least one group of operation audios of the undetermined equipment, wherein the correlation information between any group of operation audios of the undetermined equipment is used for indicating the mutual influence degree between any group of operation audios of the undetermined equipment;
(3) Determining a correlation characterization result between the at least one set of pending device operational audio based on correlation information between the at least one set of pending device operational audio;
(4) And extracting target equipment operation audio in the target environment sound based on the correlation characterization result between the at least one group of undetermined equipment operation audio.
In an exemplary embodiment of the present invention, in a production plant of a large plant, there are multiple devices running simultaneously, such as machine a, machine B, and machine C. By arranging microphones at different locations, the system can collect the ambient sound of the whole shop at the same time. These ambient sounds include the operating audio from machine a, machine B, and machine C waiting for a particular device. The system analyzes the collected pending device operation audio and calculates correlation information between them. For example, in a plant, when machine a is operating, noise may be generated that affects the operation of machine B, while the operation of machine C has no significant correlation with them. The correlation information is used to indicate the degree of interaction between the operating audio of the pending device. Based on the analyzed correlation information, the system can determine correlation characterization results between running audios of different undetermined devices. For example, the correlation between machine a and machine B is higher, while the correlation between machine C and other devices is lower. By utilizing the previously determined correlation characterization results, the system is able to accurately extract the operating audio of the target device from the entire target ambient sound. For example, in a shop, based on the correlation results, the system may accurately extract the operating audio of machine A and machine B, while filtering out the audio of machine C that is not very relevant to them.
By the design, the system can effectively collect and extract the operation audio of the target equipment and accurately analyze according to the correlation information among the undetermined equipment, so that more accurate abnormal equipment monitoring and alarm processing are realized.
In one possible implementation, the foregoing step of collecting correlation information between at least one set of pending device operating audio may be performed by the following example.
(1) Collecting one or more of characteristics of the at least one set of pending equipment operation audio and azimuth characteristics among the at least one set of pending equipment operation audio, wherein characteristics of any pending equipment operation audio in the at least one set of pending equipment operation audio comprise one or more of environmental characteristics of a target environmental range in which the any pending equipment operation audio is located and audio characteristics of the any pending equipment operation audio, and the azimuth characteristics among any set of pending equipment operation audio are used for indicating azimuth relations among the target environmental ranges in which the any set of pending equipment operation audio is located;
(2) Correlation information between the at least one set of pending device operational audio is determined based on one or more of characteristics of the at least one set of pending device operational audio and orientation characteristics between the at least one set of pending device operational audio.
In an embodiment of the present invention, a plurality of sensors and microphones are installed for monitoring the operation state of the device, illustratively, in a plant room of one factory. These sensors may collect environmental characteristics of the pending device, such as temperature, humidity, vibration, etc. Meanwhile, the audio data collected by the microphone contains sound characteristics, such as frequency, amplitude, duration, etc., of the device when the device is in operation. The system can acquire the azimuth information of the operation audio of the undetermined equipment through the sensors and the microphones which are arranged at different positions. These orientation features are used to indicate the orientation relationship between the target environment ranges in which the audio is operated by different pending devices, e.g., device a is located on the east side of the plant, device B is located on the west side of the plant, etc. The system analyzes the collected characteristics and azimuth characteristics of the running audios of the undetermined equipment, and determines the correlation information between the audios through a machine learning algorithm or other related technologies. For example, by analyzing the audio characteristics of different devices and their positional relationship in the target environment, the system can determine the degree of interaction between certain devices and thus determine the correlation between them. By the design, the characteristics of the operation audio of the undetermined equipment and the azimuth characteristics of the undetermined equipment in the target environment can be comprehensively considered, so that the correlation information among the undetermined equipment can be accurately determined. This helps to improve the accuracy and reliability of the abnormal equipment monitoring system, making fault diagnosis and maintenance of the industrial equipment more efficient.
In a possible implementation manner, the characteristics of the running audio of the optional pending device include environmental characteristics of a target environmental range where the running audio of the optional pending device is located, and the step of collecting the characteristics of the running audio of the at least one set of pending devices may be implemented in the following manner.
(1) Collecting the environmental characteristics of the target environmental sound;
(2) And determining the environmental characteristics of the target environmental range of the optional equipment operation audio based on the environmental characteristics of the target environmental sound and the positioning information of the target environmental range of the optional equipment operation audio for the optional equipment operation audio in the at least one group of optional equipment operation audio.
In an exemplary embodiment of the present invention, in an industrial plant scenario, the system may collect sounds in a target environment and extract environmental features. For example, in a production plant, the system may collect machine running sounds, worker talking sounds, and other environmental noise, among others. By analyzing these environmental sounds, environmental characteristics such as noise level, spectral distribution, sound energy, etc. can be extracted. By combining the positioning information of the target environment range where the audio is located by the undetermined device and the environment characteristics of the target environment sound, the system can determine the environment characteristics of the target environment range where the audio is located by the undetermined device. For example, in a plant, if a pending device A is located near machine C and the target environmental sound analysis results indicate that machine C is producing a higher level of noise, then it may be determined that the environmental characteristics of the target environmental range in which device A is located may include a high noise level.
By the design, the environmental characteristics of the target environmental sound can be collected and extracted and combined with the positioning information of the operation audio of the equipment to be determined, so that the environmental characteristics of the target environmental range where the equipment to be determined is located are accurately determined. This helps to further analyze and understand the operational state of the device and its environmental impact factors, providing more comprehensive fault diagnosis and maintenance decision support.
In a possible implementation manner, the features of the running audio of the optional pending device include the audio features of the running audio of the optional pending device, and the step of collecting the features of the running audio of the at least one set of pending devices may be implemented in the following manner.
(1) For any undetermined equipment operation audio in the at least one set of undetermined equipment operation audio, collecting audio signal vectors of all audio signals in the any undetermined equipment operation audio;
(2) And executing integration operation on the audio signal vector of each audio signal in the running audio of the arbitrary equipment to obtain the audio characteristics of the running audio of the arbitrary equipment.
In an exemplary embodiment of the present invention, in an industrial equipment scenario, the system may collect audio signal vectors for individual audio signals in the audio of the pending equipment operation. For example, in a production line, multiple machines are running simultaneously, each of which emits an acoustic signal of a particular frequency and amplitude. The system may collect and analyze these audio signals and represent them as vectors of audio signals containing information about the frequency spectrum, amplitude and duration of the sound over a specific period of time. By integrating the audio signal vectors of the audio signals in the operation audio of any undetermined equipment, the system can obtain the audio characteristics of the operation audio of the undetermined equipment. For example, the audio may comprise a plurality of audio signal vectors representing sound characteristics over different time periods for a pending device operation of a machine. The system may perform averaging, weighting, or other integration operations on these audio signal vectors to obtain a composite audio feature of the operating audio of the pending device for further analysis and determination of the operating state of the device.
By the design, the audio signal vectors of all audio signals in the operation audio of the undetermined equipment can be collected and integrated, so that the audio characteristics representing the operation audio of the undetermined equipment can be obtained. This helps extract key information of the equipment operating audio, such as sound frequency variation, amplitude fluctuation, etc., for fault diagnosis, prediction and maintenance decision.
In a possible implementation manner, the characteristics of the running audio of the optional pending device include the environmental characteristics of the target environmental range where the running audio of the optional pending device is located and the audio characteristics of the running audio of the optional pending device, and the step of collecting the characteristics of the running audio of the at least one set of pending devices may be implemented in the following manner.
For any undetermined equipment operation audio in the at least one group of undetermined equipment operation audio, dividing the environmental characteristics of a target environmental range where the any undetermined equipment operation audio is positioned into a preset number of subdivision environmental characteristics, and dividing the audio characteristics of the any undetermined equipment operation audio into the preset number of subdivision audio characteristics;
for any subdivision environmental feature, executing integration operation on the random subdivision environmental feature and the corresponding subdivision audio feature to obtain subdivision integration feature;
And executing merging operation on each subdivision integration feature to obtain the feature of the running audio of the arbitrary undetermined equipment.
In an exemplary embodiment of the present invention, in an industrial device scenario, the system may divide the environmental features and audio features of the target environment range in which the audio is to be run by the pending device. For example, in a plant, the system may divide the environmental characteristics into several sub-divided environmental characteristics, such as temperature, humidity, vibration, etc.; at the same time, the audio features may also be divided into preset sub-divided audio features, such as frequency, amplitude, duration, etc. By integrating any subdivision environmental features and corresponding subdivision audio features, the system may obtain subdivision integration features. For example, in a shop, for the subdivided environmental features of temperature and the subdivided audio features of frequency, the system may integrate them into a subdivided integrated feature by using a weighted average method, which represents the integrated feature of the audio operated by the pending device in the subdivided environment. By combining the sub-division integrated features, the system can obtain the features of the operation audio of the undetermined equipment. For example, in a shop, the system may combine the different sub-divided ambient features and the sub-divided integrated features of the corresponding sub-divided audio features into one overall feature representing the integrated features of the operating audio of the pending device.
By the design, the environment characteristics and the audio characteristics of the target environment range of the operation audio of the equipment to be determined can be divided, integrated and combined, so that the characteristics representing the operation audio of the equipment to be determined can be obtained. The method is beneficial to extracting key information of the equipment operation audio, comprehensively considering influencing factors of the environment in which the equipment operation audio is located, and providing more comprehensive feature expression for fault diagnosis and maintenance decision.
In one possible implementation, the step of collecting the orientation features between the operating audios of the at least one group of pending devices may be implemented by the following example.
For any group of operation audios of the undetermined equipment, collecting positioning information of a target environment range where the operation audios of the undetermined equipment are positioned;
and determining azimuth characteristics among the operation audios of any group of undetermined equipment based on the positioning information of the target environment range where the operation audios of any group of undetermined equipment are positioned and the audio energy of the target environment sound.
In an exemplary embodiment of the present invention, in an industrial equipment scenario, the system may collect positioning information of a target environment range where the audio is operated by the pending equipment. For example, in a large factory floor, there are multiple machines running at different locations simultaneously. The system may acquire and record specific location coordinates or area identifications for each machine using sensors, locating devices, or other locating techniques. By combining the location information of the target environment range in which the audio is located by the pending device and the audio energy of the target environment sound, the system can determine the azimuth characteristics between the pending device and the audio. For example, in a shop, if the pending device a is located to the left of the machine C and the audio energy display of the target ambient sound shows a high energy peak to the left, then the relationship between the pending device a in azimuth and the machine C may be determined. This orientation feature may be expressed as a relative position, such as left, right, front-to-back, etc.
By the design, the azimuth characteristics between operation audios of the undetermined equipment can be collected. By combining the location information of the target environment range and the audio energy of the target environment sound, the system can determine the relative positional relationship between the devices. This helps to further analyze interactions, synergies, and possible fault propagation paths between devices, providing more comprehensive device state assessment and maintenance decision support.
In a possible implementation, the step of determining the correlation information between the at least one set of pending device operational audio based on the characteristics of the at least one set of pending device operational audio and the orientation characteristics between the at least one set of pending device operational audio may be implemented by the following example.
(1) Establishing an audio relationship graph based on the characteristics of the at least one set of pending equipment operation audio and the azimuth characteristics between the at least one set of pending equipment operation audio, wherein the audio relationship graph comprises at least two elements and at least one connecting line, the elements are used for indicating the characteristics of the pending equipment operation audio, and the connecting line is used for indicating the azimuth characteristics between the set of pending equipment operation audio;
(2) And determining correlation information between operation audios of the at least one group of undetermined devices based on the audio relation diagram.
In an exemplary embodiment of the present invention, in an industrial equipment scenario, the system builds an audio relationship graph based on characteristics of the audio of the operation of the device to be determined and orientation characteristics between the audio of the operation of the device to be determined. For example, in a production line, multiple machines are running simultaneously, each machine having a different operating state and position. The system may represent the characteristics of the pending device operating audio for each machine as an element and use the wiring to indicate the positional relationship between the machines. By analyzing the audio relationship graph, the system can determine correlation information between the operational audio of the pending device. For example, by observing elements and wiring in an audio relationship graph, the system may identify that similar operational features or synergies exist between certain machines. Such correlation information may be used to determine the degree of inter-association between devices, the path of fault propagation, and possibly common faults or maintenance requirements. In one detailed embodiment, three machines are assumed: machine a, machine B, and machine C. The system presents the undetermined device operation audio characteristics of each machine, such as frequency, amplitude, duration, etc., as elements of an audio relationship graph by collecting and analyzing them. In addition, the system collects the orientation characteristics of these machines, such as their specific location coordinates or area identification within the plant. Wires are used to connect elements, indicating the azimuthal relationship between machines. Assuming that machine a is placed on the left side of the shop, machine B is located in the center position and machine C is located on the right side of the shop. By building an audio relationship graph, the system may represent the pending device operation audio features of machine A, machine B, and machine C as corresponding elements and represent the azimuthal relationship between them with a wire. For example, a wire may be directed from machine A to machine B, indicating that machine A is on the left side of machine B, while another wire may be directed from machine C to machine B, indicating that machine C is on the right side of machine B. By observing the elements and wires in the audio relationship graph, the system can further analyze the correlation information between machines. For example, if the connection between machine A and machine B is relatively strong, this may indicate that there is a close synergy or similar operating characteristics between them. In contrast, the wiring between machine C and other machines may be weaker, indicating a lower correlation between them. Such correlation information may be used to identify the degree of inter-association between devices, fault propagation paths, and possibly common faults or maintenance requirements. For example, if the correlation between machine a and machine B is high and machine a fails, the system may infer that machine B may also be affected. Therefore, in making maintenance decisions, it is necessary to give priority to checking the association part between machine a and machine B.
By the design, a detailed audio relation diagram can be established based on the characteristics and the azimuth characteristics of the operation audio of the undetermined equipment, and the correlation information between the equipment can be determined by analyzing the diagram. This helps reveal interactions between devices, fault propagation mechanisms, and possibly common faults or maintenance requirements, providing a more accurate and comprehensive basis for maintenance decisions, predictive analysis, and fault diagnosis.
In a possible implementation manner, before the step of collecting correlation information between operation audios of at least one group of pending devices, the embodiment of the present invention further provides the following steps.
(1) Collecting the audio types of the operation audio of each undetermined device and the correlation information between the operation audio of each two undetermined devices;
the foregoing step of collecting correlation information between at least one set of pending device operating tones may be implemented by the following example execution.
(1) And determining the correlation information between the at least one group of the operation audios of the undetermined equipment from the correlation information between every two operation audios of the undetermined equipment based on the audio types of the operation audios of the undetermined equipment.
In an exemplary embodiment of the present invention, in an industrial equipment scenario, the system first collects the operating audio of the various pending devices and determines their audio categories. The audio categories may represent different types of sound signals, such as machine noise, vibration frequencies, or other specific sound patterns. Meanwhile, the system also collects and calculates the correlation information between the operation audios of each two undetermined devices so as to quantify the similarity or the degree of correlation between the operation audios. Assume that there are three machines: machine a, machine B, and machine C. The system collected 0.8 for the correlation information between machine a and machine B, 0.6 for the correlation information between machine a and machine C, and 0.4 for the correlation information between machine B and machine C. In addition, the system determines the audio class of each machine, such as machine A for vibration frequency, machine B for noise level, and machine C for sound mode. By analyzing the audio class of each pending device operational audio and the correlation information obtained from step 1, the system may determine correlation information between at least one set of pending device operational audio. This step is to correlate the correlation information with specific audio categories in order to more accurately understand the relationship between the devices. Assuming that the system determines that the correlation information related to the vibration frequency is 0.8 (between machine a and machine B) according to the vibration frequency audio type of machine a. Thus, it can be concluded that the frequency dependence of vibration between machine a and machine B is high. However, depending on the noise level and the audio type of the sound pattern, and the corresponding correlation information, the system may draw different conclusions.
By the design, the audio types of the operation audios of the various undetermined devices and the correlation information between the operation audios of every two undetermined devices can be collected, and specific correlation information between the undetermined devices can be determined from the correlation information based on the audio types. This helps identify devices with similar audio characteristics or with specific relevance, providing more accurate guidance and analytical basis in maintenance decision-making and fault diagnosis processes.
In a possible implementation manner, the step of determining the correlation information between the at least one group of pending device operation audio from the correlation information between every two pending device operation audio based on the audio types of the respective pending device operation audio may be implemented by the following example.
(1) Determining candidate equipment operation audio with the audio type as a target audio type from the plurality of the equipment operation audio based on the audio types of the equipment operation audio;
(2) And determining the correlation information between every two candidate device operation audios from the correlation information between every two undetermined device operation audios to obtain the correlation information between at least one group of undetermined device operation audios.
In the embodiment of the invention, the system firstly analyzes the collected correlation information between every two operation audios of the undetermined devices according to different audio types of the operation audios of the undetermined devices in an industrial device scene by way of example. In this way specific correlation information between the pending devices can be determined. Assume that there are three machines: machine a, machine B, and machine C. The system has analyzed the correlation information between them and determined that the correlation between machine a and machine B is 0.8, the correlation between machine a and machine C is 0.6, and the correlation between machine B and machine C is 0.4. In addition, the system determines the audio class of each machine, such as machine A for vibration frequency, machine B for noise level, and machine C for sound mode. In this step, the system further analyzes the audio category and correlation information based on the previous determination. Firstly, the system screens candidate device operation audio with the audio type as a target audio type from a plurality of operation audio of the undetermined devices based on the audio types of the operation audio of the undetermined devices. For example, all the operation audio of machine B having a noise level of the target audio class is selected. Then, the system analyzes the selected candidate device operation audio according to the correlation information between every two candidate device operation audios. For example, in the candidate device operating audio that has been screened, the system calculates the correlation information between machine B and machine C to be 0.7. Through this analysis, the system obtains correlation information between machine B and machine C and operates the same as the correlation information between at least one set of pending device operating audio.
By the design, the candidate device operation audio can be screened from the correlation information based on the audio types of the operation audio of each undetermined device, and the correlation information among the candidate devices can be further determined. This helps to more accurately identify devices with targeted audio categories and specific correlations, providing a more accurate and comprehensive basis for maintenance decisions and fault diagnostics.
In one possible implementation, the aforementioned step S203 may be implemented by the following example execution.
(1) Acquiring the running audio of the target equipment in the target environment;
(2) Performing feature selection operation on the target equipment operation audio through an abnormal equipment identification model corresponding to the target environment to obtain an equipment operation audio vector to be detected corresponding to the target equipment operation audio; the abnormal equipment identification model is obtained by performing iterative training according to training equipment operation audios corresponding to a plurality of test environments and then performing iterative training according to training equipment operation audios corresponding to the target environments;
(3) And determining equipment abnormality corresponding to the target equipment operation audio according to the equipment operation audio vector to be detected.
In the embodiment of the invention, for example, in an industrial equipment scene, the system first needs to acquire the operation audio of the target equipment in the target environment, namely, the operation audio is determined through the scheme. The system performs feature selection operations on the operating audio of the target device using the model identified for the abnormal device. The model is obtained by iterative training based on training equipment operation audio of a plurality of test environments. Thus, the model already has the ability to accurately classify and identify abnormal devices. The system inputs the operation audio of the target device into the abnormal device identification model and performs a feature selection operation to extract specific features related to the abnormal device identification. These features may be spectral analysis, temporal features, or other audio features related to device anomalies. According to the operation, the system generates a device operation audio vector to be detected corresponding to the target device operation audio, wherein the vector contains characteristic information related to device abnormal condition identification. The system uses the device operation audio vector to be detected to determine the device abnormality corresponding to the target device operation audio. By comparing the difference between the operational audio vector of the target device and the known device normal and abnormal conditions in the training data, the system is able to determine whether the target device is abnormal. For example, if the operational audio vector of the target device has a high similarity to the existing normal device operational audio vector, it may be determined that the target device is operating normally. Conversely, if the operational audio vector of the target device is closer to the existing abnormal device operational audio vector, then the target device may be considered to have an abnormal condition.
By the design, feature selection can be performed by utilizing the abnormal equipment identification model according to the operation audio of the target equipment, and finally the abnormal condition of the target equipment is determined. The method is beneficial to monitoring the state of the equipment in real time, finding and processing potential problems in time, and improving the reliability and safety of industrial equipment.
In a possible implementation manner, the step of performing the feature selection operation on the target device operation audio through the abnormal device identification model corresponding to the target environment to obtain the device operation audio vector to be detected corresponding to the target device operation audio may be implemented through the following example implementation.
(1) Performing vector projection operation on the target equipment operation audio through an input layer of the abnormal equipment identification model to obtain a primary equipment operation audio vector;
(2) Converting the primary equipment operation audio vector through a conversion layer of the abnormal equipment identification model to obtain a secondary equipment operation audio vector, wherein the conversion layer comprises an abnormal equipment identification sub-model obtained by performing iterative training according to training equipment operation audio corresponding to a plurality of test environments;
(3) And executing feature selection operation on the secondary equipment operation audio vector through the performance layer of the abnormal equipment identification model to obtain the equipment operation audio vector to be detected, wherein the dimension of the equipment operation audio vector to be detected does not exceed the dimension of the primary equipment operation audio vector.
In an exemplary embodiment of the present invention, in an industrial device scenario, an abnormal device identification model is assumed, and an input layer of the abnormal device identification model is used to perform a vector projection operation on running audio of a target device. This means that the running audio of the target device is converted into a numerical vector representation. Consider, for example, an industrial machine, such as an engine. The system collects the sound signal of the engine while it is running and inputs it to the input layer of the abnormal equipment identification model. The input layer will process these sound signals and convert them into initial device-running audio vectors. The system uses the translation layer of the abnormal device identification model to further translate the primary device operational audio vector. The conversion layer comprises an abnormal equipment identification sub-model obtained by performing iterative training according to training equipment operation audios corresponding to a plurality of test environments. For example, the system uses the abnormal device identification submodel that has been iteratively trained to transform the primary device operational audio vector. The sub-model is obtained by training according to the running audio of the training equipment corresponding to a plurality of test environments in a similar scene. The purpose of the translation layer is to further extract useful features in order to better distinguish between normal and abnormal conditions of the device. Taking the previous engine example as an example, the system would input the initial device running audio vector into the translation layer of the abnormal device identification model. The conversion layer applies the abnormal equipment identification submodel obtained through training to convert the initial equipment operation audio vector to generate a secondary equipment operation audio vector. The system uses the presentation layer of the abnormal device identification model to perform feature selection operations on the secondary device operational audio vectors to obtain device operational audio vectors to be detected. The feature selection operation is based on knowledge and experience learned by the model and aims to extract important features related to equipment abnormality recognition. Taking the foregoing engine example as an example, the system may input the secondary device operational audio vector to the performance layer of the abnormal device identification model for feature selection operations. Through the step, the system can select important characteristics related to equipment abnormal condition identification from the secondary equipment operation audio vectors to obtain the equipment operation audio vectors to be detected. Notably, the dimension of the device to be detected running audio vector does not exceed the dimension of the primary device running audio vector, in order to reduce computational complexity and improve algorithm efficiency.
In a more detailed embodiment, consider, for example, a production line in a factory where multiple machines are operating. These devices would normally emit specific operating audio. The system needs to monitor the operating state of these devices and detect anomalies in time.
Step A: the system collects the operating audio of the target device (e.g., a piece of machinery) and inputs it into the input layer of the abnormal device identification model. The input layer performs a vector projection operation on the running audio, converting it into an initial device running audio vector.
For example, if it is desired to monitor the operating state of a motor in a factory, the system will collect the sound signal of the motor as it is operating and convert it to an initial device operating audio vector.
And (B) step (B): the system uses a translation layer of the abnormal equipment identification model to carry out conversion processing on the initial equipment operation audio vector so as to obtain a secondary equipment operation audio vector. The translation layer performs translation operations using an abnormal device identification submodel trained based on a plurality of test environments.
In the case of a motor, the system would input the initial device running audio vector into the translation layer of the abnormal device identification model. The conversion layer applies an abnormal equipment identification submodel trained based on a plurality of test environments to convert the initial equipment operation audio vector to obtain a secondary equipment operation audio vector.
Step C: the system uses the presentation layer of the abnormal device identification model to perform feature selection operations on the secondary device operational audio vectors to obtain device operational audio vectors to be detected. This vector is an important feature extracted from the secondary device operational audio vector that is relevant to device anomaly identification.
In the example of a motor, the system would input the secondary device operational audio vector into the presentation layer of the abnormal device identification model. Through the feature selection operation of the presentation layer, the system can select important features related to motor abnormal condition identification from the secondary device operation audio vectors, and generate final device operation audio vectors to be detected.
By the design, the operation of vector projection, conversion and feature selection on the operation audio of the target equipment is realized in the scene of the industrial equipment, and finally the operation audio vector of the equipment to be detected is obtained. This vector contains important information related to the identification of the device anomalies and can be used for subsequent anomaly detection and diagnostic tasks.
In a possible implementation manner, the step of performing the vector projection operation on the running audio of the target device to obtain the running audio vector of the primary device may be implemented by the following example.
(1) Executing cutting operation on the audio operated by the target equipment to obtain a set of undetermined audio fragments;
(2) And performing vector projection operation on each undetermined audio segment in the undetermined audio segment set to obtain the primary equipment operation audio vector.
In an embodiment of the present invention, the system first performs a cutting operation on the running audio of a target device (e.g., a piece of machinery) to obtain a set of pending audio clips. This means that the running audio of the target device is divided into a plurality of segments. For example, in a production line, it is desirable to monitor the operation of a sander. The system collects the sound signal of the sander as it operates and cuts it into a plurality of pending audio segments, each segment representing the operating audio for a short period of time. The system performs a vector projection operation on each of the set of pending audio segments to obtain a primary device operational audio vector. This means that each pending audio segment is converted into a numerical vector representation. Continuing with the grinder example, the system may perform a vector projection operation on each segment in the set of pending audio segments that was just cut. By inputting these audio clips into the input layer of the abnormal device identification model, they can be converted into primary device operational audio vectors.
So designed, the operating audio of the target device is cut into a plurality of pending audio segments; and then, vector projection operation is carried out on each undetermined audio segment, and corresponding primary equipment operation audio vectors are obtained. These vectors can more accurately characterize the operating state of the device and provide more information for subsequent anomaly detection and diagnostic tasks.
In a possible implementation manner, the step of performing the feature selection operation on the secondary device operation audio vector to obtain the device operation audio vector to be detected may be implemented by the following example.
Performing a downsampling operation on the secondary device run audio vector to obtain a reduced audio vector;
and performing dimension compression operation on the reduced audio vector to obtain the operation audio vector of the equipment to be detected.
In an embodiment of the present invention, the system performs a downsampling operation from the running audio vector of the secondary device (e.g., assembly robot) to obtain a reduced audio vector. Downsampling is to reduce computational complexity, reduce memory requirements, or adapt to the needs of a particular algorithm. For example, on an automated assembly line, it is desirable to monitor the motion state of a robot. The system captures the robot's operating audio and downsamples it to reduce the amount of data. This saves computational resources and increases processing efficiency. The system performs a dimension compression operation on the reduced audio vector to obtain a running audio vector for the device to be detected. Dimensional compression is to reduce the number of features, removing redundant or unimportant information. Continuing with the example of the robot above, the system performs a dimension compression operation on the previously obtained reduced audio vector. By this operation, the most important features related to the robot motion state recognition can be chosen to remain, while those features that do not have significant impact in this task are removed.
By the design, the calculation and storage requirements can be reduced, and key characteristics related to the abnormal equipment identification task can be extracted, so that the abnormal detection and fault diagnosis can be more effectively carried out.
In a possible implementation manner, before the step of performing the feature selection operation on the running audio of the target device through the abnormal device identification model corresponding to the target environment, the following steps are further provided in the embodiment of the present invention.
(1) Acquiring a real training data set and a false training data set corresponding to abnormal conditions of each comparison device in the target environment; the real training data set comprises a plurality of positive training device operation audios conforming to the abnormal condition of the comparison device, and the false training data set comprises a plurality of negative training device operation audios not conforming to the abnormal condition of the comparison device;
(2) Setting a training sample set according to a real training data set and a false training data set corresponding to abnormal conditions of each comparison device, and setting a device abnormal condition judgment model according to an abnormal device identification submodel obtained by performing iterative training by training device operation audios corresponding to a plurality of test environments;
(3) And executing a training process on the equipment abnormal condition judgment model based on the training sample set to obtain the abnormal equipment identification model.
In an embodiment of the present invention, the system collects real training data sets and false training data sets corresponding to abnormal conditions of each comparison device in the environment of the target device. The real training data set includes the operating audio of a plurality of devices being trained, the anomalies of which are consistent with the comparison device. The set of false training data includes a plurality of device operating tones of false training data whose anomalies are inconsistent with the comparison device. For example, while monitoring an engine for anomalies, the system may collect and prepare operating audio for a plurality of normally operating engines as a real training data set. These engines are compatible with the anomalies of the comparison equipment. At the same time, the system prepares false training data for a plurality of malfunctioning engines, which do not correspond to the anomalies of the comparison equipment. In the step, the system sets a training sample set according to a real training data set and a false training data set corresponding to the abnormal condition of each comparison device, and sets a device abnormal condition judging model based on an abnormal device identification sub-model obtained by performing iterative training by training device operation audios corresponding to a plurality of test environments. For example, when monitoring for an abnormal condition of the engine, the system may set up a training sample set using the previously prepared real training data set and the false training data set. For each engine anomaly, the system creates a training sample set that includes a certain amount of real training data and false training data. Then, by performing iterative training using training device operation audio corresponding to a plurality of test environments, the system obtains a sub-model for device anomaly determination. This sub-model may be used to determine whether the target device is in an abnormal state. And the system executes a training process on the equipment abnormal condition judgment model according to the training sample set, so that a final abnormal equipment identification model is obtained. In an engine monitoring scenario, the system will apply the prepared training sample set to the equipment anomaly decision model for training. The training process may include multiple iterative loops where the system adjusts and optimizes model parameters based on feedback from the training data. Through the iterative training processes, the system gradually improves the equipment abnormal condition judgment model and obtains an abnormal equipment identification model with higher accuracy. During training, the system uses the device audio data for both normal and abnormal states in the training sample set, as well as the sub-models obtained from the foregoing as references. The system can continuously adjust and learn the model to improve the judging capability of the model on the abnormal state of the target equipment. Finally, the system will obtain an abnormal device identification model trained based on the training sample set. The model can be applied to monitoring the running state of target equipment on a production line in real time and judging whether the equipment is in abnormal conditions according to the audio signals. When the model detects an abnormality in the target device, an alarm may be triggered or appropriate action may be taken, such as notifying a worker to perform maintenance or shutdown, etc.
So designed, the system will use the training sample set to train the device anomaly determination model. Through iterative training and optimization, the system can obtain an abnormal equipment identification model with higher accuracy. The model can be used for judging whether the target equipment is in an abnormal state or not and providing corresponding alarm basis.
In one possible implementation, the step of setting the training sample set according to the real training data set and the false training data set corresponding to the abnormal condition of each comparison device may be implemented by the following example.
(1) Executing association operation on the running audio of the positive training equipment in the real training data set corresponding to the abnormal condition of the comparison equipment to generate a plurality of real training data pairs;
(2) Performing association operation on positive training equipment operation audio in the real training data set corresponding to the abnormal condition of the comparison equipment and negative training equipment operation audio in the false training data set corresponding to the abnormal condition of the comparison equipment to form a plurality of false training data pairs;
(3) Generating a training sample set corresponding to the abnormal condition of the comparison equipment according to the plurality of real training data pairs and the plurality of false training data pairs;
(4) And setting the training sample set based on the training sample set corresponding to the abnormal condition of each comparison device.
In an exemplary embodiment of the present invention, in this step, the system collects real training data sets and dummy training data sets corresponding to respective comparison device anomalies corresponding to the environment in which the target device is located. The real training data set includes the operating audio of a plurality of devices being trained, the anomalies of which are consistent with the comparison device. The false training data set includes operational audio of a plurality of negative training devices, and anomalies of the data are inconsistent with the comparison device. For example, while monitoring an engine for anomalies, the system may collect and prepare operating audio for a plurality of normally operating engines as a real training data set. These engines are compatible with the anomalies of the comparison equipment. At the same time, the system prepares multiple false training data sets for the malfunctioning engine that do not correspond to the anomalies of the contrast device. The system will perform an association operation that generates a plurality of real training data pairs and false training data pairs by combining the real training data set and the false training data set. For example, in an engine monitoring scenario, the system may select the operating audio of a properly functioning engine from a real training data set as a real training data pair. This real training data pair contains normal engine audio. Meanwhile, the system also selects the operation audio of an abnormal engine from the false training data set and performs related operation on the abnormal engine and the previously selected normal engine audio to form a false training data pair. This pair of false training data contains engine audio for an abnormal condition. The system generates a training sample set corresponding to the abnormal condition of the comparison equipment based on the plurality of real training data pairs and the false training data pairs. The system combines the previously generated pairs of real training data and pairs of false training data to form a set of training sample sets. The training sample set comprises training samples aiming at different abnormal conditions of the engine. For example, the system may generate multiple sets of training samples, each set being directed to a different engine anomaly, such as overheating, pressure anomalies, and the like. Each set contains a plurality of real training data pairs and false training data pairs for training the recognition capability of the model to the corresponding abnormal situation. The system sets a required training sample set based on training sample sets corresponding to abnormal conditions of each comparison device. Taking an engine monitoring scene as an example, three abnormal conditions of contrast equipment are assumed: overheat, pressure abnormality, and vibration abnormality. The system will set the corresponding set of training samples based on the set of training samples generated for each anomaly. Likewise, for pressure anomalies and vibration anomalies, the system will also generate corresponding training sample sets, respectively.
By the design, the system can more accurately identify the abnormal state of the target equipment and improve the accuracy and the robustness of the training model by collecting the real training data and the false training data and carrying out the association operation on the real training data and the false training data.
In one possible implementation, the set of training samples includes a plurality of training sample arrays including a first instance, a second instance, and configuration tags for the first instance and the second instance; the foregoing provides the following in performing a training procedure on the device abnormal situation determination model based on the training sample set.
(1) Respectively executing feature selection operation on a first instance and a second instance in the training sample array to obtain training data features corresponding to the first instance and training data features corresponding to the second instance;
(2) And determining a cost parameter according to the difference value between the training data characteristics corresponding to the first instance and the training data characteristics corresponding to the second instance and the configuration labels of the first instance and the second instance.
In the embodiment of the invention, the system performs feature selection operation on a first instance and a second instance in each training sample array in the training sample set to obtain corresponding training data features. For example, in monitoring an abnormal condition of a press, the first instance may be a section of a press vibration signal under normal operation, and the second instance may be a section of dummy training data associated therewith, including a press vibration signal inconsistent with the normal operation. The system will perform feature selection operations on these two instances, extracting frequency domain features, time domain features, etc. of the vibration signal. And determining the cost parameter according to the difference value between the training data characteristics corresponding to the first instance and the training data characteristics corresponding to the second instance and the configuration labels of the first instance and the second instance. For example, in a press monitoring scenario, the system calculates the characteristic differences between the vibration signal and the spurious training data under normal operating conditions. This difference value may reflect the degree of difference between the normal state and the abnormal state. At the same time, each training sample array also has a configuration tag for identifying which kind of comparison equipment abnormal condition the sample array belongs to, such as high voltage, low voltage and the like. By calculating the feature variance value and using the configuration tag, the system can determine the cost parameter of each training sample array, i.e., its importance in the training process. The feature selection operation and the determination of cost parameters can help the system to more effectively utilize the information of the training sample set, and improve the accuracy and generalization capability of the training model.
In a possible implementation manner, the foregoing step before performing the feature selection operation on the running audio of the target device through the abnormal device identification model corresponding to the target environment is further provided in the embodiment of the present invention.
(1) Distributing the training equipment operation audio corresponding to the plurality of test environments into a plurality of training data pools;
(2) And executing a training process on a preset model according to the training data pools to obtain an abnormal equipment identification sub-model, wherein the abnormal equipment identification sub-model is used for setting the abnormal equipment identification model.
In an embodiment of the present invention, the system allocates training device operation audio corresponding to a plurality of test environments to different training data pools. Each training data pool contains a set of training samples from a particular test environment, which may contain normal state device audio and abnormal state device audio associated therewith. For example, in a vibration signal monitoring scenario, there may be multiple test environments, such as different models of presses, rotors, etc. For each test environment, vibration signal audios in a plurality of sections of normal working states and abnormal states from the environment can be collected and respectively stored in a corresponding training data pool. The system performs a training process on the pre-set model using a set of training samples in a plurality of training data pools to obtain an abnormal device identification sub-model for each test environment. These abnormal device identification sub-models may be used to further set the abnormal device identification model corresponding to the target environment. For example, in a vibration signal monitoring scenario, the system may train a preset model using a set of training samples from different training data pools, resulting in a plurality of abnormal device identification sub-models. Each abnormal device identification sub-model is directed to a particular test environment for identifying device abnormal conditions in that environment.
By the design, the training data are distributed to different training data pools, and the training process is executed for each pool, the system can generate corresponding abnormal equipment identification submodels aiming at different test environments, so that the accuracy and the effect of equipment abnormal states in the target environment are improved.
In a possible implementation manner, the step of distributing the training device operation audio corresponding to the plurality of test environments into a plurality of training data pools may be implemented through the following example implementation.
(1) Performing class division operation on the training equipment operation audios corresponding to the multiple test environments to obtain training equipment operation audio sets corresponding to equipment abnormal conditions under the environments;
(2) Determining a target equipment abnormal condition and a target number of reference equipment abnormal conditions from the equipment abnormal conditions in each environment;
(3) Extracting two pieces of training equipment operation audio from the training equipment operation audio set corresponding to the abnormal condition of the target equipment to form a training audio array corresponding to the abnormal condition of the target equipment, and extracting two pieces of training equipment operation audio from the training equipment operation audio set corresponding to the abnormal condition of each reference equipment to form a training audio array corresponding to the abnormal condition of each reference equipment;
(4) And generating the training data pool according to the training audio array corresponding to the abnormal condition of the target equipment and the training audio arrays corresponding to the abnormal condition of each reference equipment.
In an embodiment of the present invention, it is assumed, by way of example, that abnormal equipment identification is performed using vibration signals in an industrial equipment scenario. There are two test environments: environment a represents the press and environment B represents the rotor. The training device operation audio corresponding to the test environments is used for multi-environment abnormal device identification, and an abnormal device identification model applicable to the target environment is generated. The system performs class division operation according to the training equipment operation audio corresponding to the test environment, and obtains training equipment operation audio sets corresponding to equipment abnormal conditions under each environment. For example, in the press environment a, vibration signal audio data in a normal operation state and a high-pressure abnormal state are collected; in the rotor environment B, vibration signal audio data in a normal operation state and a low rotation speed abnormal state are collected. And respectively placing the audio data into a training data pool corresponding to the press environment A and a training data pool corresponding to the rotor environment B through classification operation. It is necessary to determine a target device anomaly and a target number of reference device anomalies. It is assumed that it is desirable to identify a high-pressure abnormality in the press environment a, and therefore a high-pressure abnormality is selected as the target equipment abnormality. Meanwhile, in order to provide a comparison reference, a low rotation speed abnormality in the rotor environment B is selected as a reference device abnormality. Extracting two pieces of training equipment operation audio from the training equipment operation audio set corresponding to the abnormal condition of the target equipment to form a training audio array corresponding to the abnormal condition of the target equipment. Meanwhile, two pieces of training equipment operation audio are extracted from the training equipment operation audio set corresponding to the abnormal conditions of each reference equipment, so that a training audio array corresponding to the abnormal conditions of each reference equipment is formed. For example, in the training device operation audio set corresponding to the high-voltage abnormal condition, two sections of typical vibration signal audio under the high-voltage abnormal condition are selected, such as samples with obviously increased vibration frequency. The two pieces of audio form a training audio array corresponding to the abnormal condition of the target equipment. Meanwhile, in the running audio set of the training equipment corresponding to the abnormal condition of the low rotating speed, two sections of typical vibration signal audio under the abnormal condition of the low rotating speed are selected, such as samples with reduced vibration amplitude. The two pieces of audio form a training audio array corresponding to the abnormal condition of the reference equipment. And generating a training data pool for training the abnormal equipment identification model according to the training audio array corresponding to the abnormal condition of the target equipment and the training audio arrays corresponding to the abnormal conditions of the reference equipment. For example, in a scenario, a high voltage anomaly sample is used as a training audio array for a target device anomaly and a low speed anomaly sample is used as a training audio array for a reference device anomaly. These training audio arrays are combined to form a training data pool. The training data pool comprises vibration signal audios of different abnormal conditions under different environments and is used for training an abnormal equipment identification model.
By means of the design, model training can be conducted through the training data pool, so that abnormal conditions of industrial equipment can be accurately identified later.
In one possible implementation, the training data pool includes a plurality of training audio arrays; the foregoing performing a training process on a preset model according to a plurality of training data pools may be implemented by the following example execution.
(1) Selecting one training audio array from the training data pool, taking one training equipment operation audio in the selected training audio array as a to-be-determined training equipment operation audio, taking the other training equipment operation audio as a real training equipment operation audio which is consistent with the to-be-determined training equipment operation audio, and taking the training equipment operation audio except the selected training audio array in the training data pool as a false training equipment operation audio which is not consistent with the to-be-determined training equipment operation audio;
(2) And executing a training process on a preset model according to the operation audio of the undetermined training equipment, the operation audio of the real training equipment and the operation audio of the false training equipment.
In an embodiment of the present invention, exemplary, based on the foregoing flow, the running audio of the training device corresponding to the multiple test environments has been divided into different training data pools. Each training data pool contains a set of training audio arrays representing equipment anomalies under different circumstances. For example, in a vibration signal monitoring scenario, there are three test environments: environment a represents the press, environment B represents the rotor, and environment C represents the bearing. The training device operating audio for environments A, B and C, respectively, has been partitioned into three different training data pools. For example, the training data pool a includes a training audio array of a normal state and a high-pressure abnormal state in a press machine environment, the training data pool B includes a training audio array of a normal state and a low-rotation speed abnormal state in a rotor environment, and the training data pool C includes a training audio array of a normal state and a vibration abnormal state in a bearing environment. And executing a training process on the preset model according to the plurality of training data pools. First, a training audio array is selected from a training data pool. Suppose that training data pool A is selected, which contains training audio arrays for normal and high pressure anomalies in the press environment. Then, randomly selecting one training device operation audio from the selected training audio array as the undetermined training device operation audio, and taking the undetermined training device operation audio as a target sample of model input. And simultaneously, selecting another training equipment operation audio from the same training audio array as a real training equipment operation audio which is consistent with the operation audio of the undetermined training equipment, so that the model learns the characteristics of the abnormal situation. Next, one or more false training device operation audio frequencies which do not coincide with the operation audio frequency of the training device to be determined are randomly selected from the operation audio frequencies of other training devices except the selected training audio frequency array in the training data pool a, so that the distinguishing capability of the model on other abnormal conditions is improved. And finally, executing a training process on the preset model according to the operation audio of the undetermined training equipment, the operation audio of the real training equipment and the operation audio of the false training equipment, so that the training process can learn the ability of correctly identifying the abnormal equipment.
By the design, the model can learn the characteristics from data of different environments and different abnormal conditions and has the capability of identifying abnormal equipment.
In one possible implementation, the foregoing training process performed on the preset model according to the pending training device running audio, the real training device running audio, and the dummy training device running audio may be implemented by the following example.
(1) Performing feature selection operations on the undetermined training equipment operation audio, the real training equipment operation audio and the false training equipment operation audio respectively to obtain undetermined training data features, real training data features and false training data features;
(2) Determining a first observed difference value between the pending training data feature and the real training data feature, and determining a second observed difference value between the pending training data feature and a target training data feature, the target training data feature comprising the real training data feature and the false training data feature;
(3) And determining a cost parameter according to the first observed difference value and the second observed difference value.
In the embodiment of the invention, the feature selection operation is performed on the audio of the training equipment to be determined, the audio of the real training equipment to be trained and the audio of the false training equipment to be trained respectively, so as to acquire the respective features of the audio. For example, in a vibration signal monitoring scenario, signal processing techniques may be used to extract a set of features from the vibration signal, such as frequency domain features (e.g., power spectral density, band energy, etc.) and time domain features (e.g., mean, variance, etc.). And respectively carrying out feature extraction operation on the operation audio of the undetermined training equipment, the operation audio of the real training equipment and the operation audio of the false training equipment to obtain respective feature vectors of the undetermined training equipment, the operation audio of the real training equipment and the operation audio of the false training equipment. And determining a first observed difference value between the undetermined training data characteristic and the real training data characteristic and a second observed difference value between the undetermined training data characteristic and the target training data characteristic according to the extracted characteristic vector. The target training data features include real training data features and dummy training data features. For the first observed difference value, it may be determined by calculating a distance or degree of difference between the pending training data feature and the real training data feature; for the second observed difference value, a distance or degree of difference between the pending training data feature and the target training data feature may be calculated for determination. These observed differences reflect the similarity or variability between the pending training data features and other data features. And finally, determining a cost parameter according to the first observation difference value and the second observation difference value, and evaluating the loss degree or error of the model in the training process. The cost parameter may be determined by different methods, for example, based on least squares or cross entropy.
The design is that the feature selection operation is executed in the training process, and the training process of the model is adjusted according to the observed difference value and the cost parameter. Therefore, the recognition accuracy of the model to the abnormal equipment can be improved, and different abnormal conditions can be well adapted.
In a possible implementation manner, the step of determining the device abnormal situation corresponding to the target device operation audio according to the device operation audio vector to be detected may be implemented through the following example.
(1) Determining matching values between the equipment operation audio vector to be detected and a plurality of comparison equipment operation audio vectors, wherein the comparison equipment operation audio vectors are obtained by performing feature selection operation on equipment operation audio of comparison equipment abnormal conditions through the abnormal equipment identification model;
(2) And determining the equipment abnormality corresponding to the target equipment operation audio according to the matching value between the equipment operation audio vector to be detected and each comparison equipment operation audio vector.
In an embodiment of the present invention, the matching value between the device operation audio vector to be detected and the plurality of comparison device operation audio vectors is determined by comparing them, for example. The comparison equipment operation audio vector is obtained by executing feature selection operation on equipment operation audio of comparison equipment abnormal conditions through an abnormal equipment identification model. For example, in a vibration signal monitoring scenario, there are multiple contrasting devices, such as presses, rotors, bearings, and the like. And aiming at each comparison device, performing feature selection operation by using an abnormal device identification model to obtain a corresponding comparison device operation audio vector. Then, a matching value between the device operation audio vector to be detected and each of the comparison device operation audio vectors is calculated. This matching value may use different similarity measures, such as cosine similarity or euclidean distance. The matching value reflects the degree of similarity between the device operational audio vector to be detected and the comparative device operational audio vector. And determining the equipment abnormality corresponding to the target equipment operation audio according to the matching value between the equipment operation audio vector to be detected and each contrast equipment operation audio vector. For example, assume that the matching values of the device operation audio vector to be detected and the three comparison device operation audio vectors of the press, rotor and bearing are 0.8, 0.6 and 0.4, respectively. From these matching values, it can be determined that the device to be detected is more likely to be an abnormal state of the press because the matching value of the running audio vector of the device against the press is highest.
By the design, the equipment abnormality corresponding to the target equipment operation audio can be determined according to the matching value between the equipment operation audio vector to be detected and the plurality of comparison equipment operation audio vectors. In this way, abnormal behavior in the industrial equipment can be timely discovered and identified for further maintenance or troubleshooting.
In one possible implementation manner, the step of determining the device abnormality corresponding to the target device operation audio according to the matching value between the device operation audio vector to be detected and each comparison device operation audio vector may be implemented by the following example.
(1) Determining a maximum matching value from matching values between the equipment operation audio vector to be detected and each contrast equipment operation audio vector;
(2) And when the maximum matching value exceeds a preset matching value threshold, taking the abnormal condition of the contrast equipment corresponding to the contrast equipment operation audio vector corresponding to the maximum matching value as the abnormal condition of the equipment corresponding to the target equipment operation audio.
In an embodiment of the present invention, the maximum matching value is illustratively determined from matching values between the device operation audio vector to be detected and the respective comparison device operation audio vectors. The maximum matching value indicates the matching degree of the running audio vector of the device to be detected and which contrast device runs the audio vector. For example, in the vibration signal monitoring scenario, matching values between the device operation audio vector to be detected and three comparative device operation audio vectors of the press, rotor and bearing have been calculated to be 0.8, 0.6 and 0.4, respectively. Here, the maximum matching value is 0.8, which means that the matching degree of the running audio vector of the equipment to be detected and the running audio vector of the comparison equipment of the press is the highest. And when the maximum matching value exceeds a preset matching value threshold, taking the abnormal condition of the contrast equipment corresponding to the contrast equipment operation audio vector corresponding to the maximum matching value as the abnormal condition of the equipment corresponding to the target equipment operation audio. For example, assume that a preset match value threshold is set to 0.7. According to the maximum match value in the previous example of 0.8, the preset match value threshold is exceeded. Therefore, the abnormal condition of the comparison equipment of the press machine can be used as the abnormal condition of the equipment corresponding to the operation audio of the target equipment.
By the design, the maximum matching value can be determined according to the matching value between the to-be-detected equipment operation audio vector and each comparison equipment operation audio vector, and when the maximum matching value exceeds a preset matching value threshold, the corresponding comparison equipment abnormal condition is used as the equipment abnormal condition corresponding to the target equipment operation audio. Therefore, the abnormal condition of the target equipment can be judged according to the comparison equipment with high matching degree, and corresponding maintenance or fault investigation can be carried out.
The embodiment of the invention provides a determiner 100, wherein the determiner 100 comprises a processor and a nonvolatile memory storing determiner instructions, and when the determiner instructions are executed by the processor, the determiner 100 executes the abnormal equipment monitoring method based on equipment operation audio. As shown in fig. 2, the block diagram of the structure of the determiner device 100 according to the embodiment of the present invention is shown. The determiner device 100 comprises a memory 111, a processor 112 and a communication unit 113. For data transmission or interaction, the memory 111, the processor 112 and the communication unit 113 are electrically connected to each other directly or indirectly. For example, the elements may be electrically connected to each other via one or more communication buses or signal lines.
The foregoing description, for purpose of explanation, has been presented with reference to particular embodiments. However, the illustrative discussions above are not intended to be exhaustive or to limit the disclosure to the precise forms disclosed. Many modifications and variations are possible in light of the above teaching. The embodiments were chosen and described in order to best explain the principles of the disclosure and its practical application, to thereby enable others skilled in the art to best utilize the disclosure and various embodiments with various modifications as are suited to the particular use contemplated. The foregoing description, for purpose of explanation, has been presented with reference to particular embodiments. However, the illustrative discussions above are not intended to be exhaustive or to limit the disclosure to the precise forms disclosed. Many modifications and variations are possible in light of the above teaching. The embodiments were chosen and described in order to best explain the principles of the disclosure and its practical application, to thereby enable others skilled in the art to best utilize the disclosure and various embodiments with various modifications as are suited to the particular use contemplated.
Claims (10)
1. An abnormal equipment monitoring method based on equipment operation audio, which is characterized by comprising the following steps:
Responding to an abnormal equipment monitoring instruction, and determining a target environment to be monitored;
collecting the target environment sound in the target environment, and extracting target equipment operation audio in the target environment sound;
determining equipment abnormality corresponding to the target equipment operation audio according to the target equipment operation audio;
and initiating alarm information of a corresponding grade according to the abnormal condition grade of the equipment abnormal condition characterization.
2. The method of claim 1, wherein the collecting the target environmental sounds in the target environment and extracting target device operating audio in the target environmental sounds comprises:
collecting the target environmental sound in the target environment, wherein the target environmental sound comprises a plurality of operation audios of undetermined equipment;
collecting correlation information between at least one group of operation audios of the undetermined equipment, wherein the correlation information between any group of operation audios of the undetermined equipment is used for indicating the mutual influence degree between any group of operation audios of the undetermined equipment;
determining a correlation characterization result between the at least one set of pending device operational audio based on correlation information between the at least one set of pending device operational audio;
And extracting target equipment operation audio in the target environment sound based on the correlation characterization result between the at least one group of undetermined equipment operation audio.
3. The method of claim 2, wherein the collecting correlation information between the operational audio of the at least one set of pending devices comprises:
collecting one or more of characteristics of the at least one set of pending equipment operation audio and azimuth characteristics among the at least one set of pending equipment operation audio, wherein characteristics of any pending equipment operation audio in the at least one set of pending equipment operation audio comprise one or more of environmental characteristics of a target environmental range in which the any pending equipment operation audio is located and audio characteristics of the any pending equipment operation audio, and the azimuth characteristics among any set of pending equipment operation audio are used for indicating azimuth relations among the target environmental ranges in which the any set of pending equipment operation audio is located;
establishing an audio relationship graph based on the characteristics of the at least one set of pending equipment operation audio and the azimuth characteristics between the at least one set of pending equipment operation audio, wherein the audio relationship graph comprises at least two elements and at least one connecting line, the elements are used for indicating the characteristics of the pending equipment operation audio, and the connecting line is used for indicating the azimuth characteristics between the set of pending equipment operation audio;
And determining correlation information between operation audios of the at least one group of undetermined devices based on the audio relation diagram.
4. The method of claim 3, wherein the characteristics of the arbitrary pending device operational audio include environmental characteristics of a target environmental range in which the arbitrary pending device operational audio is located, and the collecting the characteristics of the at least one set of pending device operational audio includes:
collecting the environmental characteristics of the target environmental sound;
for any undetermined equipment operation audio in the at least one group of undetermined equipment operation audio, determining the environmental characteristics of the target environment range in which the any undetermined equipment operation audio is positioned based on the environmental characteristics of the target environment sound and the positioning information of the target environment range in which the any undetermined equipment operation audio is positioned;
alternatively, the feature of the optional device operation audio includes an audio feature of the optional device operation audio, and the collecting the feature of the at least one set of the optional device operation audio includes:
for any undetermined equipment operation audio in the at least one set of undetermined equipment operation audio, collecting audio signal vectors of all audio signals in the any undetermined equipment operation audio;
Performing an integration operation on the audio signal vector of each audio signal in the operation audio of the arbitrary equipment to obtain the audio characteristics of the operation audio of the arbitrary equipment;
or the characteristics of the running audio of the arbitrary pending equipment comprise the environmental characteristics of the target environmental range where the running audio of the arbitrary pending equipment is located and the audio characteristics of the running audio of the arbitrary pending equipment, and the collecting the characteristics of the running audio of at least one group of pending equipment comprises the following steps:
for any undetermined equipment operation audio in the at least one group of undetermined equipment operation audio, dividing the environmental characteristics of a target environmental range where the any undetermined equipment operation audio is positioned into a preset number of subdivision environmental characteristics, and dividing the audio characteristics of the any undetermined equipment operation audio into the preset number of subdivision audio characteristics;
for any subdivision environmental feature, executing integration operation on the random subdivision environmental feature and the corresponding subdivision audio feature to obtain subdivision integration feature;
and executing merging operation on each subdivision integration feature to obtain the feature of the running audio of the arbitrary undetermined equipment.
5. The method of claim 3, wherein collecting the orientation features between the at least one set of pending device operational audio comprises:
For any group of operation audios of the undetermined equipment, collecting positioning information of a target environment range where the operation audios of the undetermined equipment are positioned;
and determining azimuth characteristics among the operation audios of any group of undetermined equipment based on the positioning information of the target environment range where the operation audios of any group of undetermined equipment are positioned and the audio energy of the target environment sound.
6. The method of claim 2, wherein before collecting correlation information between the operational audio of the at least one set of pending devices, further comprising:
collecting the audio types of the operation audio of each undetermined device and the correlation information between the operation audio of each two undetermined devices;
the collecting correlation information between at least one set of pending device operation audio comprises:
determining candidate equipment operation audio with the audio type as a target audio type from the plurality of the equipment operation audio based on the audio types of the equipment operation audio;
and determining the correlation information between every two candidate device operation audios from the correlation information between every two undetermined device operation audios to obtain the correlation information between at least one group of undetermined device operation audios.
7. The method of claim 1, wherein the determining, according to the target device operation audio, a device abnormality corresponding to the target device operation audio includes:
acquiring the running audio of the target equipment in the target environment;
executing cutting operation on the running audio of the target equipment through an input layer of the abnormal equipment identification model to obtain a set of undetermined audio fragments;
vector projection operation is carried out on each undetermined audio fragment in the undetermined audio fragment set, and an operation audio vector of the primary equipment is obtained;
converting the primary equipment operation audio vector through a conversion layer of the abnormal equipment identification model to obtain a secondary equipment operation audio vector, wherein the conversion layer comprises an abnormal equipment identification sub-model obtained by performing iterative training according to training equipment operation audio corresponding to a plurality of test environments;
performing downsampling operation on the secondary device operation audio vector through the performance layer of the abnormal device identification model to obtain a reduced audio vector;
performing dimension compression operation on the reduced audio vector to obtain an equipment operation audio vector to be detected, wherein the dimension of the equipment operation audio vector to be detected does not exceed the dimension of the primary equipment operation audio vector; the abnormal equipment identification model is obtained by performing iterative training according to training equipment operation audios corresponding to a plurality of test environments and then performing iterative training according to training equipment operation audios corresponding to the target environments;
Determining matching values between the equipment operation audio vector to be detected and a plurality of comparison equipment operation audio vectors, wherein the comparison equipment operation audio vectors are obtained by performing feature selection operation on equipment operation audio of comparison equipment abnormal conditions through the abnormal equipment identification model;
determining a maximum matching value from matching values between the equipment operation audio vector to be detected and each contrast equipment operation audio vector;
and when the maximum matching value exceeds a preset matching value threshold, taking the abnormal condition of the contrast equipment corresponding to the contrast equipment operation audio vector corresponding to the maximum matching value as the abnormal condition of the equipment corresponding to the target equipment operation audio.
8. The method of claim 7, wherein prior to performing feature selection operations on the target device run audio by the abnormal device identification model corresponding to the target environment, the method further comprises:
acquiring a real training data set and a false training data set corresponding to abnormal conditions of each comparison device in the target environment; the real training data set comprises a plurality of positive training device operation audios conforming to the abnormal condition of the comparison device, and the false training data set comprises a plurality of negative training device operation audios not conforming to the abnormal condition of the comparison device;
Executing association operation on the running audio of the positive training equipment in the real training data set corresponding to the abnormal condition of the comparison equipment to generate a plurality of real training data pairs;
performing association operation on positive training equipment operation audio in the real training data set corresponding to the abnormal condition of the comparison equipment and negative training equipment operation audio in the false training data set corresponding to the abnormal condition of the comparison equipment to form a plurality of false training data pairs;
generating a training sample set corresponding to the abnormal condition of the comparison equipment according to the plurality of real training data pairs and the plurality of false training data pairs;
setting a training sample set based on training sample sets corresponding to abnormal conditions of each comparison device, and setting a device abnormal condition judgment model according to an abnormal device identification submodel obtained by performing iterative training on training device operation audios corresponding to a plurality of test environments; the training sample set comprises a plurality of training sample arrays, wherein the training sample arrays comprise a first instance, a second instance and configuration labels of the first instance and the second instance;
respectively executing feature selection operation on a first instance and a second instance in the training sample array to obtain training data features corresponding to the first instance and training data features corresponding to the second instance;
Determining a cost parameter according to the difference value between the training data characteristics corresponding to the first instance and the training data characteristics corresponding to the second instance and the configuration labels of the first instance and the second instance;
and executing a training process on the equipment abnormal condition judgment model according to the cost parameter to obtain the abnormal equipment identification model.
9. The method of claim 7, wherein prior to performing feature selection operations on the target device run audio by the abnormal device identification model corresponding to the target environment, the method further comprises:
performing class division operation on the training equipment operation audios corresponding to the multiple test environments to obtain training equipment operation audio sets corresponding to equipment abnormal conditions under the environments;
determining a target equipment abnormal condition and a target number of reference equipment abnormal conditions from the equipment abnormal conditions in each environment;
extracting two pieces of training equipment operation audio from the training equipment operation audio set corresponding to the abnormal condition of the target equipment to form a training audio array corresponding to the abnormal condition of the target equipment, and extracting two pieces of training equipment operation audio from the training equipment operation audio set corresponding to the abnormal condition of each reference equipment to form a training audio array corresponding to the abnormal condition of each reference equipment;
Generating a training data pool according to the training audio array corresponding to the abnormal condition of the target equipment and the training audio arrays corresponding to the abnormal conditions of the reference equipment; the training data pool comprises a plurality of training audio arrays;
selecting one training audio array from the training data pool, taking one training equipment operation audio in the selected training audio array as a to-be-determined training equipment operation audio, taking the other training equipment operation audio as a real training equipment operation audio which is consistent with the to-be-determined training equipment operation audio, and taking the training equipment operation audio except the selected training audio array in the training data pool as a false training equipment operation audio which is not consistent with the to-be-determined training equipment operation audio;
performing feature selection operations on the undetermined training equipment operation audio, the real training equipment operation audio and the false training equipment operation audio respectively to obtain undetermined training data features, real training data features and false training data features;
determining a first observed difference value between the pending training data feature and the real training data feature, and determining a second observed difference value between the pending training data feature and a target training data feature, the target training data feature comprising the real training data feature and the false training data feature;
And determining a cost parameter according to the first observation difference value and the second observation difference value, and executing a training process on a preset model to obtain an abnormal equipment identification sub-model, wherein the abnormal equipment identification sub-model is used for setting the abnormal equipment identification model.
10. A server system comprising a server for performing the method of any of claims 1-9.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311795505.7A CN117854245B (en) | 2023-12-25 | 2023-12-25 | Abnormal equipment monitoring method and system based on equipment operation audio |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311795505.7A CN117854245B (en) | 2023-12-25 | 2023-12-25 | Abnormal equipment monitoring method and system based on equipment operation audio |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN117854245A true CN117854245A (en) | 2024-04-09 |
| CN117854245B CN117854245B (en) | 2024-06-14 |
Family
ID=90528040
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202311795505.7A Active CN117854245B (en) | 2023-12-25 | 2023-12-25 | Abnormal equipment monitoring method and system based on equipment operation audio |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN117854245B (en) |
Citations (21)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103953490A (en) * | 2014-04-23 | 2014-07-30 | 浙江工业大学 | Implementation method for monitoring status of hydraulic turbine set based on HLSNE |
| CN104272183A (en) * | 2012-03-04 | 2015-01-07 | 约翰·比蒂 | System and method for mapping and displaying audio source locations |
| CN104700686A (en) * | 2013-12-09 | 2015-06-10 | 上海仪电信息网络有限公司 | Real-object simulated maintenance training system for domestic electronic products |
| CN111048114A (en) * | 2019-12-30 | 2020-04-21 | 深圳江行联加智能科技有限公司 | Equipment and method for detecting abnormal sound of equipment |
| US20200233397A1 (en) * | 2019-01-23 | 2020-07-23 | New York University | System, method and computer-accessible medium for machine condition monitoring |
| US20210164786A1 (en) * | 2019-12-02 | 2021-06-03 | Here Global B.V. | Method, apparatus, and computer program product for road noise mapping |
| CN113376457A (en) * | 2021-05-10 | 2021-09-10 | 首钢京唐钢铁联合有限责任公司 | Equipment operation state detection method, system, device, equipment and medium |
| CN113421588A (en) * | 2021-06-18 | 2021-09-21 | 青岛海尔工业智能研究院有限公司 | Method and device for detecting abnormal sound of household appliance, electronic equipment and storage medium |
| CN113470695A (en) * | 2021-06-30 | 2021-10-01 | 平安科技（深圳）有限公司 | Sound abnormality detection method, sound abnormality detection device, computer device, and storage medium |
| CN114360581A (en) * | 2021-12-30 | 2022-04-15 | 北京百度网讯科技有限公司 | Method and device for identifying equipment fault and electronic equipment |
| CN114462475A (en) * | 2021-12-21 | 2022-05-10 | 南京邮电大学 | Unsupervised machine abnormal sound detection method and unsupervised machine abnormal sound detection device based on single classification algorithm |
| CN114495975A (en) * | 2022-01-27 | 2022-05-13 | 贵州盘江煤层气开发利用有限责任公司 | Abnormal sound source positioning method and device based on sound pressure amplitude ratio |
| CN115291056A (en) * | 2022-07-29 | 2022-11-04 | 国网安徽省电力有限公司超高压分公司 | Transformer working state identification system and method based on voiceprint identification model |
| CN115547356A (en) * | 2022-11-25 | 2022-12-30 | 杭州兆华电子股份有限公司 | Wind noise processing method and system based on abnormal sound detection of unmanned aerial vehicle |
| JPWO2023276495A1 (en) * | 2021-07-01 | 2023-01-05 | ||
| CN116206625A (en) * | 2022-12-26 | 2023-06-02 | 郑州信大先进技术研究院 | Self-supervision abnormal sound detection method based on combination of frequency spectrum and time information |
| CN116754855A (en) * | 2023-04-11 | 2023-09-15 | 华能灌云清洁能源发电有限责任公司 | Transformer fault detection method based on sound frequency spectrum |
| CN116881832A (en) * | 2023-09-07 | 2023-10-13 | 山东能源数智云科技有限公司 | Construction method and device of fault diagnosis model of rotary mechanical equipment |
| CN116880309A (en) * | 2023-07-31 | 2023-10-13 | 重庆环科源博达环保科技有限公司 | Factory environment monitoring management system and method |
| CN117116293A (en) * | 2023-08-23 | 2023-11-24 | 深圳前海慧联科技发展有限公司 | Machine equipment fault diagnosis system in complex sound field environment |
| CN117129154A (en) * | 2023-07-31 | 2023-11-28 | 华南理工大学 | Shunt disc aluminum liquid leakage monitoring method and system based on voice recognition |
-
2023
- 2023-12-25 CN CN202311795505.7A patent/CN117854245B/en active Active
Patent Citations (21)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104272183A (en) * | 2012-03-04 | 2015-01-07 | 约翰·比蒂 | System and method for mapping and displaying audio source locations |
| CN104700686A (en) * | 2013-12-09 | 2015-06-10 | 上海仪电信息网络有限公司 | Real-object simulated maintenance training system for domestic electronic products |
| CN103953490A (en) * | 2014-04-23 | 2014-07-30 | 浙江工业大学 | Implementation method for monitoring status of hydraulic turbine set based on HLSNE |
| US20200233397A1 (en) * | 2019-01-23 | 2020-07-23 | New York University | System, method and computer-accessible medium for machine condition monitoring |
| US20210164786A1 (en) * | 2019-12-02 | 2021-06-03 | Here Global B.V. | Method, apparatus, and computer program product for road noise mapping |
| CN111048114A (en) * | 2019-12-30 | 2020-04-21 | 深圳江行联加智能科技有限公司 | Equipment and method for detecting abnormal sound of equipment |
| CN113376457A (en) * | 2021-05-10 | 2021-09-10 | 首钢京唐钢铁联合有限责任公司 | Equipment operation state detection method, system, device, equipment and medium |
| CN113421588A (en) * | 2021-06-18 | 2021-09-21 | 青岛海尔工业智能研究院有限公司 | Method and device for detecting abnormal sound of household appliance, electronic equipment and storage medium |
| CN113470695A (en) * | 2021-06-30 | 2021-10-01 | 平安科技（深圳）有限公司 | Sound abnormality detection method, sound abnormality detection device, computer device, and storage medium |
| JPWO2023276495A1 (en) * | 2021-07-01 | 2023-01-05 | ||
| CN114462475A (en) * | 2021-12-21 | 2022-05-10 | 南京邮电大学 | Unsupervised machine abnormal sound detection method and unsupervised machine abnormal sound detection device based on single classification algorithm |
| CN114360581A (en) * | 2021-12-30 | 2022-04-15 | 北京百度网讯科技有限公司 | Method and device for identifying equipment fault and electronic equipment |
| CN114495975A (en) * | 2022-01-27 | 2022-05-13 | 贵州盘江煤层气开发利用有限责任公司 | Abnormal sound source positioning method and device based on sound pressure amplitude ratio |
| CN115291056A (en) * | 2022-07-29 | 2022-11-04 | 国网安徽省电力有限公司超高压分公司 | Transformer working state identification system and method based on voiceprint identification model |
| CN115547356A (en) * | 2022-11-25 | 2022-12-30 | 杭州兆华电子股份有限公司 | Wind noise processing method and system based on abnormal sound detection of unmanned aerial vehicle |
| CN116206625A (en) * | 2022-12-26 | 2023-06-02 | 郑州信大先进技术研究院 | Self-supervision abnormal sound detection method based on combination of frequency spectrum and time information |
| CN116754855A (en) * | 2023-04-11 | 2023-09-15 | 华能灌云清洁能源发电有限责任公司 | Transformer fault detection method based on sound frequency spectrum |
| CN116880309A (en) * | 2023-07-31 | 2023-10-13 | 重庆环科源博达环保科技有限公司 | Factory environment monitoring management system and method |
| CN117129154A (en) * | 2023-07-31 | 2023-11-28 | 华南理工大学 | Shunt disc aluminum liquid leakage monitoring method and system based on voice recognition |
| CN117116293A (en) * | 2023-08-23 | 2023-11-24 | 深圳前海慧联科技发展有限公司 | Machine equipment fault diagnosis system in complex sound field environment |
| CN116881832A (en) * | 2023-09-07 | 2023-10-13 | 山东能源数智云科技有限公司 | Construction method and device of fault diagnosis model of rotary mechanical equipment |
Also Published As
| Publication number | Publication date |
|---|---|
| CN117854245B (en) | 2024-06-14 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Langarica et al. | An industrial internet application for real-time fault diagnosis in industrial motors | |
| EP3098681B1 (en) | Artificial intelligence based health management of host system | |
| CN115034248A (en) | Automatic diagnostic method, system and storage medium for equipment | |
| US9443201B2 (en) | Systems and methods for learning of normal sensor signatures, condition monitoring and diagnosis | |
| CN116842423A (en) | Aeroengine fault diagnosis method and system based on multi-mode deep learning | |
| KR102301201B1 (en) | Apparatus and method for precise state diagnosis of rotating machinery based on IoT sensor | |
| JP7068246B2 (en) | Abnormality judgment device and abnormality judgment method | |
| CN110133500B (en) | Motor online monitoring and fault precursor diagnosis system and method based on multi-layer architecture | |
| CN115437358A (en) | Intelligent state monitoring and fault diagnosis system and fault diagnosis method for industrial robot | |
| CN117672255B (en) | Abnormal equipment identification method and system based on artificial intelligence and equipment operation sound | |
| CN118051822A (en) | Equipment running state abnormality detection method based on voiceprint recognition technology | |
| CN112711850A (en) | Unit online monitoring method based on big data | |
| Ribeiro et al. | Rotating machinery fault diagnosis using similarity-based models | |
| KR20210006832A (en) | Method and apparatus for machine fault diagnosis | |
| Alghassi et al. | Machine performance monitoring and fault classification using vibration frequency analysis | |
| CN117423345A (en) | Voiceprint recognition monitoring system for power equipment | |
| CN112664410B (en) | Big data-based modeling method for unit online monitoring system | |
| CN117435908A (en) | Multi-fault feature extraction method for rotary machine | |
| CN117854245B (en) | Abnormal equipment monitoring method and system based on equipment operation audio | |
| CN116699400A (en) | Generator rotor short-circuit fault monitoring system, method and readable storage medium | |
| CN116453437A (en) | Display screen module testing method, device, equipment and storage medium | |
| CN113557414A (en) | Device for monitoring equipment | |
| US20230394301A1 (en) | Device and method for tracking basis of abnormal state determination by using neural network model | |
| Boukra et al. | Identifying new prognostic features for remaining useful life prediction | |
| Hu et al. | Degradation assessment of bearing fault using SOM network |
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 |