CN115855463A - Rotary mechanical equipment fault detection method and device - Google Patents
Rotary mechanical equipment fault detection method and device Download PDFInfo
- Publication number
- CN115855463A CN115855463A CN202211490528.2A CN202211490528A CN115855463A CN 115855463 A CN115855463 A CN 115855463A CN 202211490528 A CN202211490528 A CN 202211490528A CN 115855463 A CN115855463 A CN 115855463A
- Authority
- CN
- China
- Prior art keywords
- fault
- mode component
- mode
- magnitude
- information
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Images
Abstract
The application provides a method and a device for detecting faults of rotary mechanical equipment, which are applied to the field of machinery, and are used for carrying out decomposition analysis on collected vibration information and diagnosing the fault conditions of the mechanical equipment more accurately by using a mode component with finer granularity. The method comprises the following steps: acquiring vibration information, which may include acquired information of vibration generated by at least one piece of mechanical equipment, and specifically may be acquired by a sensor, such as a vibration sensor or an optical fiber distributed acoustic sensor; demodulating the vibration information to obtain a plurality of mode components, wherein the plurality of mode components can be used for representing information of at least one working mode of at least one piece of mechanical equipment, and the working mode of the mechanical equipment can comprise a normal working mode or a fault mode and the like; fault information for the at least one mechanical device is then determined based on the plurality of mode components, the fault information being usable to indicate whether the at least one mechanical device is malfunctioning.
Description
Technical Field
The application relates to the field of machinery, in particular to a fault detection method and device for rotary mechanical equipment.
Background
For some transport devices, such as belt conveyors or conveyors, idlers are important mechanical components. For example, a belt conveyor is a main machine for conveying raw coal in a coal mine, and a carrier roller is used as a core component of the belt conveyor, and the normal operation of the carrier roller is directly related to the material conveying function of the belt conveyor. The failure of the carrier roller easily causes local temperature rise, is an important factor threatening the coal mining safety of a coal mine, and greatly improves the operation and maintenance strength and the cost of the belt conveyor due to frequent failure of the carrier roller.
In some scenes, for fault detection of the carrier roller, vibration information can be picked up through the optical fiber, phase demodulation can be carried out on the optical fiber signal, a waterfall graph of the phase signal after the optical fiber demodulation is drawn, and a carrier roller fault point can be located according to the phase waterfall graph. However, when the vibration information is picked up, the vibration of the belt conveyor itself interferes strongly with the vibration of the failure point, resulting in a decrease in accuracy of failure recognition.
Disclosure of Invention
The application provides a fault detection method and device for rotary mechanical equipment, which are used for carrying out decomposition analysis on collected vibration information and diagnosing the fault condition of the mechanical equipment more accurately by using a mode component with finer granularity.
In view of the above, a first aspect of the present application provides a method for detecting a fault of a rotary machine, including: firstly, acquiring vibration information, which may include information of acquired vibration generated by at least one piece of mechanical equipment, specifically, acquiring the vibration information by using a vibration sensor, such as an inductive vibration sensor or a fiber Distributed Acoustic Sensor (DAS); then, demodulating the vibration information to obtain a plurality of mode components, where the plurality of mode components may be used to represent information of at least one operation mode of at least one piece of mechanical equipment, and the operation mode of the mechanical equipment may include a normal operation mode or a failure mode, etc.; fault information for the at least one mechanical device is then determined based on the plurality of mode components, the fault information being usable to indicate whether the at least one mechanical device is malfunctioning.
In the embodiment of the application, after the vibration information is collected, the vibration information can be demodulated, so that a plurality of mode components can be obtained. The multiple mode components can be analyzed respectively, and when the working mode represented by one or more mode components is determined to be a fault mode, the roller fault can be determined. Therefore, the collected vibration information can be decomposed to obtain a plurality of components which can be used for representing the working mode of the carrier roller, whether mechanical equipment fails or not is identified through the more detailed components of the representation information, namely whether the carrier roller fails or not can be analyzed from the dimension of the mode components, and whether the carrier roller fails or not can be accurately judged.
In a possible implementation, the aforementioned determining the fault information of the at least one mechanical device according to the plurality of mode components may include: acquiring a plurality of index values of each mode component in the plurality of mode components, wherein the plurality of indexes comprise evaluation values obtained by evaluating each component from a plurality of dimensions, and the plurality of dimensions can comprise at least one of signal impact, periodic impact or signal-to-noise ratio of the plurality of mode components; then, whether the working mode of the corresponding mechanical equipment is a failure mode is identified according to the index values of each mode component; and if the plurality of mode components comprise at least one fault mode component, generating fault information indicating whether at least one mechanical device generates faults, wherein at least one fault mode component indicates a mode component indicating that the operation of the mechanical device is a fault mode.
Therefore, in the embodiment of the application, the collected vibration information can be decomposed to obtain multiple components, each component is measured from dimensions such as signal impact, periodic impact or signal-to-noise ratio, whether the indexes of each mode are in the range of the normal working mode is judged, if one or more indexes of the modes are not in the range of the normal working mode, the mechanical equipment is indicated to have a fault, and whether the mechanical equipment has the fault is accurately identified.
In a possible implementation, the aforementioned identifying whether the operation mode of the corresponding mechanical device is a failure mode according to the index values of each mode component may include: if the index values corresponding to the first mode component are not within the preset boundary, determining that the index values of the first mode component identify that the working mode of the corresponding mechanical equipment is the fault mode, wherein the preset boundary is the distribution range of the index values corresponding to the mechanical equipment when the working mode of the mechanical equipment is in the non-fault mode, and the first mode component is any one of the mode components. In the embodiment of the present application, when the indexes of the respective mode components are clustered, that is, when abnormal noise or abnormal vibration occurs in the mechanical equipment, it is recognized that a fault occurs in the mechanical equipment.
In one possible embodiment, the plurality of indicators for each mode component includes at least one of: kurtosis, relative kurtosis, including harmonic-to-noise ratio, negative entropy, gini exponent, variance, or standard deviation. Therefore, in the embodiment of the application, whether the working mode represented by each component is abnormal or not can be measured through kurtosis, related kurtosis, indexes including a harmonic-to-noise ratio, a negative entropy, a Gini index, a variance or a standard deviation, and the like, so that the fault mode component can be accurately identified.
In one possible implementation, the method provided by the present application may further include: and determining the fault type corresponding to each fault mode component in the at least one fault mode component according to a plurality of index values corresponding to the at least one fault mode component. Generally, different fault types may correspond to different index ranges, and therefore in the embodiment of the present application, the fault type of the mechanical device may be further identified by the index of the component, so that a user may accurately perform operations such as maintenance or replacement of the mechanical device according to the fault type.
In a possible implementation manner, the determining a fault type corresponding to each fault mode component in the at least one fault mode component according to a plurality of index values corresponding to the at least one fault mode component may include: taking a plurality of index values corresponding to at least one fault mode component as input of a fault type identification model, and outputting a fault type corresponding to each fault mode component in the at least one fault mode component, wherein the fault type identification model is obtained by training by using a first training set, the first training set comprises historical data of at least one piece of mechanical equipment, and the historical data comprises data of the at least one piece of mechanical equipment, which are generated when the at least one piece of mechanical equipment works in a non-fault mode, and a fault type obtained when the at least one piece of mechanical equipment works in a fault mode; or comparing a plurality of index values corresponding to at least one fault mode component with corresponding threshold values, and determining the fault type corresponding to each fault mode component in the at least one fault mode component.
Generally, when the historical data is sufficient, the model can be trained by using the historical data to obtain a trained model, so that the fault type can be efficiently and accurately identified. When the historical data is insufficient, the threshold value can be calculated through the historical data, so that the fault type can be identified. The method can be suitable for more scenes, and can accurately identify the fault type under various scenes.
In one possible implementation, the method provided by the present application may further include: training a fault type recognition model by using a first training set to obtain a trained fault type recognition model, wherein in the process of training the fault type recognition model, samples in the first training set are used as input of the fault type recognition model, and the fault type recognition model is used for carrying out fault classification on samples corresponding to multiple types of faults and calculating a classification center of each fault classification;
the aforementioned step of taking a plurality of index values corresponding to at least one fault mode component as an input of the fault type identification model, and outputting a fault type corresponding to each fault mode component in the at least one fault mode component may specifically include: extracting features from a plurality of index values corresponding to at least one fault mode component through a fault type identification model; obtaining the distance between the characteristic corresponding to each fault mode component and the classification center of each fault classification through a fault type identification model; and determining the fault type corresponding to each fault mode component according to the distance between the characteristic corresponding to each fault mode component and the classification center of each fault classification.
In the embodiment of the application, the fault category identification model can be trained through historical data, and the fault type closer to the fault mode component can be identified according to the distance between each fault mode component and various types of faults in the historical data, so that the fault type corresponding to the fault mode component can be identified more accurately based on the historical data.
In one possible implementation, the method provided by the present application may further include: and determining a fault magnitude corresponding to each fault mode component in the at least one fault mode component according to a plurality of index values corresponding to the at least one fault mode component, wherein the fault magnitude represents the fault degree of mechanical equipment corresponding to each fault mode component in the at least one fault mode component.
Therefore, in the embodiment of the application, the magnitude of the fault mode component can be identified, so that a user can accurately know the fault degree of the mechanical equipment, and the user can conveniently perform subsequent operations such as maintenance or replacement on the mechanical equipment.
In a possible implementation manner, the determining, according to a plurality of index values corresponding to at least one fault mode component, a fault magnitude corresponding to each fault mode component in the at least one fault mode component may specifically include: taking a plurality of index values corresponding to at least one fault mode component as input of a fault magnitude recognition model, and outputting a fault magnitude corresponding to each fault mode component in the at least one fault mode component, wherein the fault magnitude recognition model is obtained by using a second training set for training, the second training set comprises historical data of at least one piece of mechanical equipment, and the historical data comprises data of the at least one piece of mechanical equipment generated when the at least one piece of mechanical equipment works in a non-fault mode and the fault magnitude obtained when the at least one piece of mechanical equipment works in a fault mode;
or comparing a plurality of index values corresponding to at least one fault mode component with ranges corresponding to a plurality of fault magnitude levels, and determining the fault magnitude level corresponding to each fault mode component in the at least one fault mode component.
Therefore, in the embodiment of the application, the fault type recognition model can be trained by using the historical data, so that the model is trained by combining the historical empirical data, and the fault type can be accurately and efficiently recognized when the fault type is inferred. And establishing a mapping relation between indexes of all dimensions of the mode components and fault types based on historical data. Therefore, when the fault type is identified, the fault type corresponding to the fault mode component can be identified based on the mapping relation. And under the scene that the negative sample is less, can discern the fault type according to the empirical value to be convenient for maintain to mechanical equipment trouble.
In one possible implementation, the method provided by the present application may further include: training the fault magnitude recognition model by using a second training set to obtain a trained fault magnitude recognition model, wherein in the process of training the fault magnitude recognition model, samples in the second training set are used as the input of the fault type recognition model, and the fault type recognition model is used for classifying magnitudes corresponding to multiple types of faults and calculating the classification center of each fault magnitude;
the aforementioned step of taking a plurality of index values corresponding to at least one fault mode component as an input of the fault magnitude recognition model and outputting the fault magnitude corresponding to each fault mode component in the at least one fault mode component may include: extracting features from a plurality of index values corresponding to at least one fault mode component through a fault magnitude recognition model; obtaining the distance between the characteristic corresponding to each fault mode component and the classification center of each fault magnitude through a fault magnitude recognition model; and determining the fault magnitude corresponding to each fault mode component according to the distance between the characteristic corresponding to each fault mode component and the classification center of each fault magnitude.
According to the fault magnitude recognition method and device, the fault magnitude recognition model can be trained by using historical data, so that the model is trained by combining historical experience data, and the fault magnitude can be accurately and efficiently recognized when the fault magnitude is inferred. And establishing a mapping relation between indexes of all dimensions of the mode components and fault magnitude based on historical data. Therefore, when the fault magnitude is identified, the fault magnitude corresponding to the fault mode component can be identified based on the mapping relation. And under the scene that the negative sample is less, can discern the trouble magnitude according to the empirical value to be convenient for maintain to mechanical equipment trouble.
In a possible implementation manner, the plurality of indicators of each mode component include a root-mean-square RMS, and before identifying whether the operation mode of the corresponding mechanical device is the failure mode according to the plurality of indicator values of each mode component, the method provided by the present application may further include: screening out at least one mode component with a change value of RMS within a preset time period smaller than a preset value from the multiple mode components; and identifying whether the working mode of the corresponding mechanical equipment is a failure mode or not according to the index values of the at least one mode component.
Therefore, in the embodiment of the present application, the filtering may be performed according to the RMS of the signal, and the mode component of some abnormal non-fault modes may be filtered, such as the state change when the mechanical device is started or stopped, so as to avoid the influence of the state when the mechanical device is started or stopped on the fault diagnosis.
In a possible implementation, the demodulating the vibration information to obtain the plurality of mode components may include: demodulating the vibration information to obtain a demodulation signal; decomposing the demodulated signal to obtain a plurality of decomposed signals; calculating an envelope signal and an envelope spectrum signal of each of the plurality of decomposed signals; and identifying whether each decomposed signal is a mode component or not according to the envelope signal and the envelope spectrum signal of each decomposed signal and the objective function so as to obtain a plurality of mode components.
Therefore, in the embodiment of the application, the acquired vibration information can be decomposed, so that the dimensions of the envelope signal and the envelope spectrum signal are decomposed, a mode component with a finer granularity is obtained, and whether mechanical equipment fails or not can be identified more accurately in the follow-up process.
In a possible implementation, the aforementioned determining whether each decomposed signal is a mode component according to the envelope signal and the envelope spectrum signal of each decomposed signal and the objective function may include: acquiring a first frequency according to a period in the envelope signal; acquiring a second frequency according to the harmonic wave of the envelope spectrum signal; constructing an objective function according to the difference between the first frequency and the second frequency; whether each decomposed signal is a mode component is determined according to the value of the objective function.
Therefore, in the embodiment of the present application, the frequency variation of the mode component can be obtained according to the period of the envelope signal and the harmonic of the envelope spectrum signal, so as to identify whether the mode represented by the mode component is a failure mode or not according to the frequency variation of the mode component.
In one possible embodiment, the vibration information comprises information generated by sound when the at least one mechanical device is operating or information generated by vibration when the at least one mechanical device is operating. Therefore, the fault diagnosis can be performed based on vibration information generated by vibration of the mechanical device or sound generated by the mechanical device, so that the fault can be accurately identified for a scene such as vibration or abnormal sound.
In a possible implementation, the acquiring of the vibration information may include: information generated by sound of at least one mechanical device or information generated by vibration of at least one mechanical device is collected by a fiber optic Distributed Acoustic Sensor (DAS). Wherein, the sensing optical fiber for collecting data can be disposed on the side surface of at least one mechanical device, such as the side surface or the bottom of the rack. When the vibration information comprises information generated by the sound of at least one mechanical device in operation, the outside of the optical fiber can be wrapped with a sound sensitizing structure, the sound sensitizing structure is used for increasing the induction intensity of the optical fiber to the sound, and when the vibration information comprises the information generated by the vibration of at least one mechanical device in operation, the at least one mechanical device is connected with the optical fiber through a connecting piece.
Therefore, in the embodiment of the application, the sound sensitizing material can be arranged, or the optical fiber is directly connected with the mechanical equipment, so that the accuracy of data acquisition is improved.
In one possible embodiment, the at least one mechanical device comprises at least one idler. Therefore, the diagnosis of the roller fault can be realized.
In a second aspect, the present application provides a rotary machine fault detection apparatus, comprising:
the acquisition module is used for acquiring vibration information, and the vibration information comprises acquired information of vibration generated by at least one piece of mechanical equipment;
the demodulation module demodulates the vibration information to obtain a plurality of mode components, and the plurality of mode components are used for representing information of at least one working mode of at least one mechanical device;
and the fault identification module is used for determining fault information of the at least one mechanical device according to the multiple mode components, and the fault information is used for indicating whether the at least one mechanical device generates faults or not.
The effect achieved by the second aspect or any optional implementation manner of the second aspect may refer to the related description of the first aspect, and is not described herein again.
In a possible implementation, the fault identification module may be specifically configured to: acquiring a plurality of index values of each mode component in the plurality of mode components, wherein the plurality of index values comprise evaluation values for evaluating each mode component from a plurality of dimensions, and the plurality of dimensions can comprise at least one of signal impact, periodic impact or signal-to-noise ratio of the plurality of mode components; identifying whether the working mode of the corresponding mechanical equipment is a failure mode or not according to the index values of each mode component; and if the plurality of mode components comprise at least one fault mode component, generating fault information indicating whether at least one mechanical device has faults or not, wherein at least one fault mode component indicates a mode component indicating that the operation of the mechanical device is a fault mode.
In a possible implementation, the fault identification module may be specifically configured to: if the index values corresponding to the first mode component are not within the preset boundary, determining that the index values of the first mode component identify that the working mode of the corresponding mechanical equipment is the fault mode, wherein the preset boundary is the distribution range of the index values corresponding to the mechanical equipment when the working mode of the mechanical equipment is in the non-fault mode, and the first mode component is any one of the mode components.
In one possible embodiment, the plurality of indicators for each mode component includes at least one of: kurtosis, relative kurtosis, including harmonic-to-noise ratio, negative entropy, gini exponent, variance, or standard deviation.
In one possible embodiment, the apparatus provided herein may further include:
and the fault type identification module is used for determining the fault type corresponding to each fault mode component in the at least one fault mode component according to a plurality of index values corresponding to the at least one fault mode component.
In a possible implementation manner, the fault type identifying module may be specifically configured to: taking a plurality of index values corresponding to at least one fault mode component as input of a fault type identification model, and outputting a fault type corresponding to each fault mode component in the at least one fault mode component, wherein the fault type identification model is obtained by training by using a first training set, the first training set comprises historical data of at least one piece of mechanical equipment, and the historical data comprises data of the at least one piece of mechanical equipment, which are generated when the at least one piece of mechanical equipment works in a non-fault mode, and a fault type obtained when the at least one piece of mechanical equipment works in a fault mode; or comparing a plurality of index values corresponding to at least one fault mode component with corresponding threshold values, and determining the fault type corresponding to each fault mode component in the at least one fault mode component.
In one possible embodiment, the apparatus provided herein may further include: the first training module is used for training the fault type recognition model by using a first training set to obtain the trained fault type recognition model, wherein in the process of training the fault type recognition model, samples in the first training set are used as the input of the fault type recognition model, and the fault type recognition model is used for carrying out fault classification on samples corresponding to multiple types of faults and calculating the classification center of each fault classification;
the fault type identification module may be specifically configured to: extracting features from a plurality of index values corresponding to at least one fault mode component through a fault type identification model; obtaining the distance between the characteristic corresponding to each fault mode component and the classification center of each fault classification through a fault type identification model; and determining the fault type corresponding to each fault mode component according to the distance between the characteristic corresponding to each fault mode component and the classification center of each fault classification.
In one possible embodiment, the apparatus provided herein may further include:
and the fault magnitude identification module is used for determining the fault magnitude corresponding to each fault mode component in the at least one fault mode component according to a plurality of index values corresponding to the at least one fault mode component, and the fault magnitude represents the fault degree of the mechanical equipment corresponding to each fault mode component in the at least one fault mode component.
In a possible implementation manner, the failure magnitude identification module is specifically configured to: taking a plurality of index values corresponding to at least one fault mode component as input of a fault magnitude recognition model, and outputting a fault magnitude corresponding to each fault mode component in the at least one fault mode component, wherein the fault magnitude recognition model is obtained by using a second training set for training, the second training set comprises historical data of at least one piece of mechanical equipment, and the historical data comprises data of the at least one piece of mechanical equipment generated when the at least one piece of mechanical equipment works in a non-fault mode and the fault magnitude obtained when the at least one piece of mechanical equipment works in a fault mode; or comparing a plurality of index values corresponding to at least one fault mode component with ranges corresponding to a plurality of fault magnitude levels, and determining the fault magnitude level corresponding to each fault mode component in the at least one fault mode component.
In one possible embodiment, the apparatus may further comprise: the second training module is used for training the fault magnitude recognition model by using a second training set to obtain a trained fault magnitude recognition model, wherein in the process of training the fault magnitude recognition model, samples in the second training set are used as the input of the fault type recognition model, and the fault type recognition model is used for classifying magnitudes corresponding to multiple types of faults and calculating the classification center of each fault magnitude;
the failure magnitude identification module may be specifically configured to: extracting features from a plurality of index values corresponding to at least one fault mode component through a fault magnitude recognition model; obtaining the distance between the characteristic corresponding to each fault mode component and the classification center of each fault magnitude through a fault magnitude recognition model; and determining the fault magnitude corresponding to each fault mode component according to the distance between the characteristic corresponding to each fault mode component and the classification center of each fault magnitude.
In a possible implementation manner, the plurality of indicators of each mode component includes a root mean square RMS, and the apparatus provided in this application may further include a screening module, which may be configured to: before identifying whether the working mode of the corresponding mechanical equipment is a fault mode according to the index values of each mode component, screening out at least one mode component with a change value of RMS within a preset time period smaller than a preset value from the multiple mode components; and identifying whether the working mode of the corresponding mechanical equipment is a failure mode or not according to the index values of the at least one mode component.
In a possible implementation, the demodulation module may be specifically configured to: demodulating the vibration information to obtain a demodulation signal; decomposing the demodulated signal to obtain a plurality of decomposed signals; calculating an envelope signal and an envelope spectrum signal of each of the plurality of decomposed signals; and identifying whether each decomposed signal is a mode component or not according to the envelope signal and the envelope spectrum signal of each decomposed signal and the objective function so as to obtain a plurality of mode components.
In a possible implementation, the demodulation module may be specifically configured to: acquiring a first frequency according to a period in the envelope signal; acquiring a second frequency according to the harmonic wave of the envelope spectrum signal; constructing an objective function according to the difference between the first frequency and the second frequency; whether each decomposed signal is a mode component is determined according to the value of the objective function.
In one possible embodiment, the vibration information includes information generated by sound when the at least one mechanical device is operating or information generated by vibration when the at least one mechanical device is operating.
In a possible embodiment, the acquisition module is specifically configured to acquire information generated by sound of at least one piece of mechanical equipment or information generated by vibration of at least one piece of mechanical equipment through the optical fiber distributed acoustic sensor DAS, where an optical fiber for acquiring data is disposed outside the at least one piece of mechanical equipment, and when the vibration information includes the information generated by sound of at least one piece of mechanical equipment during operation, the optical fiber externally wraps a sound-sensitizing structure, the sound-sensitizing structure is configured to increase an intensity of sensing of the optical fiber for sound, and when the vibration information includes the information generated by vibration of at least one piece of mechanical equipment during operation, the at least one piece of mechanical equipment and the optical fiber are connected by a connecting member.
In one possible embodiment, the at least one mechanical device comprises at least one idler.
In a third aspect, the present application provides a fault detection device for a rotary machine, including: a processor, a memory, an input-output device, and a bus; the memory having stored therein computer instructions; when the processor executes the computer instructions in the memory, the memory stores the computer instructions; the processor, when executing the computer instructions in the memory, is adapted to implement any of the implementations of the first aspect.
In a possible embodiment, the fault detection device for the rotary mechanical equipment provided by the present application may further include a vibration sensor for collecting information of vibration generated when the mechanical equipment is in operation.
In a fourth aspect, an embodiment of the present application provides a chip system, where the chip system includes a processor and an input/output port, where the processor is configured to implement a processing function related to the method for detecting a fault of a rotating mechanical device according to the first aspect, and the input/output port is configured to implement a transceiver function related to the method for detecting a fault of a rotating mechanical device according to the first aspect.
In a possible design, the chip system further includes a memory for storing program instructions and data for implementing the functions of the method according to the first aspect or any of the embodiments of the first aspect.
The chip system may be constituted by a chip, or may include a chip and other discrete devices.
In a fifth aspect, embodiments of the present application provide a computer-readable storage medium. The computer readable storage medium having stored therein computer instructions; the computer instructions, when executed on a computer, cause the computer to perform a method as set forth in the first aspect or any one of the possible implementations of any one of the embodiments of the first aspect.
In a sixth aspect, embodiments of the present application provide a computer program product. The computer program product comprises a computer program or instructions which, when run on a computer, causes the computer to perform the method according to the first aspect or any one of the possible implementations of the first aspect.
Drawings
FIG. 1 is a schematic structural diagram of a mechanical apparatus provided herein;
FIG. 2 is a schematic diagram of another embodiment of the present disclosure;
FIG. 3 is a schematic flow chart illustrating a method for detecting faults of a rotating machine according to the present disclosure;
FIG. 4 is a schematic flow chart of another method for fault detection of a rotary machine provided herein;
FIG. 5 is a schematic flow chart of another method for fault detection of a rotary machine provided herein;
FIG. 6 is a schematic structural diagram of a fault detection apparatus for a rotary machine according to the present disclosure;
FIG. 7 is a schematic diagram of another fault detection apparatus for a rotary machine provided herein;
fig. 8 is a schematic structural diagram of a chip provided in the present application.
Detailed Description
The technical solutions in the embodiments of the present application will be described below with reference to the drawings in the embodiments of the present application, and it is obvious that the described embodiments are only a part of the embodiments of the present application, and not all of the embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present application.
With the development of the technology, mechanical equipment is used in all fields to replace all or part of manpower, and the working efficiency is improved. Failure then occurs because the mechanical equipment will likely be affected by wear, impact, load bearing, etc. However, as the industry develops, the scale of the mechanical equipment is also larger and larger, if manual detection is still performed, a large amount of labor cost is increased, and the condition of missing detection may occur in the manual detection, so that the application provides a method for detecting the fault of the rotary mechanical equipment, which can be used for detecting whether the mechanical equipment has a fault, such as a belt conveyor (or simply referred to as a belt conveyor), a conveyor with a crawler, and the like. Specifically, whether the mechanical equipment is in failure or not can be identified according to the picked vibration information by picking up vibration information of the mechanical equipment, including information generated by sound vibration or direct vibration.
The present application exemplarily describes the fault detection of the idler roller, and the idler roller mentioned below may also be replaced by other mechanical devices, for example, by a mechanical device having periodicity in some operations, which is not limited in the present application.
Taking a belt conveyor (or simply a belt conveyor) as an example, the belt conveyor plays an important role in some transportation scenarios. For example, in the scene of conveying raw coal in a coal mine, a belt conveyor is a main conveying machine. The carrier roller is an important part in the belt conveyor, the fault of the carrier roller easily causes local temperature rise, and is an important factor threatening the coal mining safety of a coal mine, and the operation and maintenance strength and the cost of the belt conveyor are greatly improved due to frequent fault of the carrier roller. Especially, the mining belt conveyor has complex working condition, severe service environment and high safety requirement, and provides higher requirement for roller fault monitoring. Although the vibration testing technology has been developed greatly, the belt conveyor is often over ten kilometers and the quantity of the carrier roller is large, so that the working efficiency of carrying out signal acquisition and carrier roller fault diagnosis by using the vibration sensor in a point-to-point manner is low, the cost is high, and the engineering deployment is difficult to complete. Therefore, the existing carrier roller monitoring scheme still mainly uses manual inspection as a main part and is limited by severe roadway environment, and the manual inspection has high strength and is easy to miss diagnosis.
The optical fiber sensor is simple to lay and convenient to transmit signals, is widely tried to be used for belt conveyor carrier roller fault diagnosis, can acquire abnormal vibration information of the carrier roller to carry out fault diagnosis by laying optical fibers, but needs to customize a specific carrier roller frame to realize a complex optical fiber laying method, needs to coil a large number of optical fibers to improve the signal-to-noise ratio of signals, and is high in monitoring cost and large in engineering realization difficulty.
After the carrier roller breaks down, abnormal vibration and accompanying abnormal sound can be generated in the operation process, so that the carrier roller abnormal information can be acquired by using the optical fiber sound pickup sensor, the signal transmission is convenient, and the engineering deployment of a large number of carrier roller fault diagnosis is conveniently carried out. However, the problem of insufficient signal-to-noise ratio still exists when the fault diagnosis is carried out only by means of optical fiber picking up abnormal vibration of a fault carrier roller or sound information, and fault diagnosis omission is easily caused.
For example, in some scenarios, vibration information of the belt conveyor can be picked up, and the sensing optical fiber is arranged on the bottom frame or the side frame of each groove-shaped carrier roller. And picking up abnormal vibration information of the carrier roller by using the optical fiber, drawing a phase waterfall diagram based on the demodulated signal, determining a fault point of the carrier roller according to the phase waterfall diagram, and carrying out fault diagnosis and positioning. However, the signal to noise ratio of the picked-up vibration information is limited, the vibration interference of the belt conveyor frame is strong, in addition, the factors causing the difference of the signal amplitude energy change are more, the relevance with the fault is not strong, and the party based on the signal waterfall diagram or the signal power is very easy to cause misdiagnosis or missed diagnosis.
For another example, in some scenes, a multifunctional distributed optical fiber vibration monitoring signal enhanced clamp is provided, and a set of carrier roller state monitoring device is developed based on the clamp, aiming at reducing the interference of belt conveyor frame vibration on optical fibers and further improving the signal-to-noise ratio of vibration perception through the clamp, but the decoupling capacity of a mechanical clamp on signals is limited, the signals still contain responsible interference components, which easily causes failure missed diagnosis and false alarm, in addition, the vibration of the belt conveyor frame easily causes fatigue failure of the clamp, which leads to premature failure of a testing instrument, and the testing reliability needs to be further improved.
For example, in some scenes, sound information generated by abnormal operation of the carrier roller is directly picked up by using an array of sound sensors to evaluate damage of the carrier roller, each sound sensor is provided with a data acquisition unit, collection of sound information of operation of the carrier roller can be realized, and the state of the carrier roller is judged by comparing sound signals of corresponding carrier roller frames with standard noise signals. However, the signal-to-noise ratio of the sound signals picked up by the sound sensor array is high, and each sensor needs to be provided with modules for power supply, signal acquisition, transmission and the like, so that the engineering deployment difficulty is high.
The idler is taken as a typical rotating part, the movement process of the idler has typical periodicity, when the idler breaks down, impact components, strong harmonic components and the like can be generated, and meanwhile, the installation structures of the idler frame and the like can also generate complex interference components in signals, so that the fault characteristics are difficult to identify. Extracting fault characteristics from strong background noise, and successively providing indexes of kurtosis, harmonic-to-noise ratio, correlation kurtosis and the like for measuring signal impact, periodicity and periodic impact. When a fault carrier roller signal shows an impact characteristic, impact fault characteristics can be effectively extracted by utilizing a kurtosis index, and when the fault carrier roller signal shows a strong-period harmonic characteristic, although the impact is weak, the fault characteristics can be identified by utilizing a harmonic-to-noise ratio index; when the signal contains a large number of periodic modulation components and discrete impact components, the target periodic impact components can be effectively extracted by utilizing the correlation kurtosis index, so that the purpose of carrying out fault diagnosis is achieved.
However, an abnormal state is identified by monitoring a certain data expression characteristic of a fault idler, but in operation in a scene, a fault idler signal does not express a stable single characteristic; particularly, when the sound information is picked up by using the optical fiber, the transmission, coupling and analysis processes of the sound information can cause serious attenuation of fault information, the carrier roller fault information is difficult to identify, fault characteristics are difficult to accurately extract by using the indexes, and the fault diagnosis accuracy is difficult to meet the requirement of carrier roller diagnosis.
Therefore, the method for detecting the faults of the rotary mechanical equipment can be used for decomposing the collected vibration or sound signals into multiple mode components after the collected vibration or sound signals are demodulated, and whether the carrier roller has the faults or not can be accurately judged through the multiple mode components.
The method provided by the application can be applied to various belt conveyors, such as various forms of a groove type belt conveyor, a flat type belt conveyor, a climbing belt conveyor, a turning belt conveyor, a telescopic belt conveyor and the like, and can be particularly applied to various scenes, such as mine transportation scenes, material transmission in factories, logistics center transmission or security inspection article transmission and the like. In the following embodiment of the present application, for example, by taking the application of the method provided by the present application to a belt conveyor as an example, the belt conveyor mentioned below may be specifically replaced by a groove-type belt conveyor, a flat belt conveyor, a climbing belt conveyor, a turning belt conveyor, a telescopic belt conveyor, and the like, and details are not described below.
For example, the structure of the belt conveyor applied by the method provided by the application can be shown in fig. 1. Wherein, the conveyer belt for bearing the weight of the material sets up on the bearing roller, and the bearing roller sets up on the frame, and in addition, this belt feeder can also include assemblies such as motor, drive drum or turnabout drum, still can add annex such as promotion baffle, shirt rim on the conveyer belt, and here is no longer repeated one by one.
Generally, when the belt conveyor is in operation, the conveyor belt encircles the drive rollers and the direction-changing rollers. The upper and lower branches between the two rollers are supported by carrier rollers. The materials are placed on the upper branch, and the conveying belt and the materials are dragged to run by utilizing the friction force between the driving roller and the belt. The conveying device is suitable for conveying bulk materials and finished articles in horizontal and inclined directions, and can also be used in a production line for certain process operations. The device has the advantages of simple structure, stable and reliable work, strong adaptability to materials, larger conveying capacity, low power consumption and wide application.
For example, the structure of the idler provided by the present application can refer to fig. 2, and the idler can specifically include a roller, a vertical roller, a ratchet sleeve, a base, an upper frame body, a lower frame body, a ratchet sleeve bearing, a base bearing, or a connecting rod.
When the belt conveyor works, the carrier roller also needs to move along with the roller so as to drag the belt to move. The carrier roller can also be out of order, for example, the screw becomes flexible, the roller is damaged, section of thick bamboo skin wearing and tearing, the carrier roller support is unusual, ratchet cover or base are not flexible etc. when the carrier roller trouble, will produce abnormal sound or unusual vibration usually, this application can discern the carrier roller through gathering abnormal sound or unusual vibration and whether produce the trouble, or can also discern trouble type or trouble magnitude etc. when producing the trouble. In the bearing roller that this application provided, as shown in fig. 2, can set up sensing optical fiber in the side of frame, when needs gather sound, can be outside including the sensitization material at sensing optical fiber, if the parcel has empty plastic film or parcel bubble to roll up etc. when the vibration information of bearing roller needs to be gathered, can be with sensing optical fiber and bearing roller lug connection, or set up the connecting piece between sensing optical fiber and bearing roller, this connecting piece can include buckle, connecting wire etc. to the vibration that makes the bearing roller produce can be transmitted to sensing optical fiber.
The following describes the flow of the method for detecting faults of rotary mechanical equipment provided by the application in combination with the belt conveyor and the carrier roller.
First, the method provided by the application can be deployed in a cloud platform, a local server or a terminal and other devices, and can be used for diagnosing the fault of a mechanical device. For example, when the method provided by the application is deployed on a cloud platform, information of mechanical equipment during operation can be collected through local equipment and uploaded to the cloud platform, and the cloud platform performs fault diagnosis on the mechanical equipment through the method provided by the application to determine whether the mechanical equipment generates faults. For another example, if the method provided by the present application is deployed in a local server, the collected data may be imported into the local server, and the local server may perform fault diagnosis on the mechanical device by using the method provided by the present application, and determine whether the mechanical device generates a fault. For another example, if the method provided by the present application is deployed in a terminal, information generated by a mechanical device may be collected or received by the terminal, and fault diagnosis may be performed on the mechanical device by the method provided by the present application, and when a fault occurs, fault information, such as indication of the occurrence of the fault, a fault location, a fault type, or a fault magnitude, may be displayed at the terminal.
Referring to fig. 3, a flow chart of a method for detecting a fault of a rotary machine according to the present application is shown as follows.
301. Vibration information is acquired.
For convenience of understanding, in the following embodiments, the mechanical device is taken as at least one idler roller for example, and the idler rollers mentioned below may also be replaced by other rotary mechanical devices and the like having periodicity or vibrating during operation, such as a belt conveyor, a machine tool and other mechanical devices, and will not be described below again.
In particular, the vibration information may include information about the vibration of the support roller during operation or information about the sound generated during vibration. For example, sensing optical fibers can be arranged on a frame of the belt conveyor, and sound or vibration generated by the carrier roller is sensed through the DAS to generate vibration information, or the vibration information is referred to as an initial vibration signal.
Optionally, when the sound-sensitizing structure is arranged in the sensing optical fiber, the sensing optical fiber can be used for collecting the sound signal generated by the carrier roller, so that the sensitivity of sound signal collection is improved through the sound-sensitizing structure, and the sound signal with a higher signal-to-noise ratio is obtained.
Optionally, when a connecting piece (such as a buckle) is arranged between the sensing optical fiber and the carrier roller, the vibration of the carrier roller drives the sensing optical fiber to vibrate, so that a signal generated by the vibration of the carrier roller can be acquired through the sensing optical fiber, the connecting piece can sense the vibration of the carrier roller and generate a vibration signal, and whether the carrier roller fails or not can be judged based on the vibration condition of the carrier roller subsequently.
302. And demodulating the vibration information to obtain a plurality of mode components.
After the optical fiber senses the vibration information, the optical fiber signal can be subjected to phase demodulation, the demodulated signal is subjected to signal decomposition, and a plurality of mode components are obtained, wherein the plurality of mode components can be used for representing information of at least one working mode of the carrier roller.
In general, different operating modes of the idler will likely produce vibrations of different frequencies or different periods. The working modes of the carrier roller can be divided into a normal working mode or a fault mode, and the like, and different types of faults can be divided into different fault modes. Such as idlers, may not be identical in frequency or period of vibration generated in a normal operating mode or in different types of failure modes. Therefore, after demodulating the vibration information, a plurality of mode components can be obtained, and each mode component can be used for representing the working mode of the carrier roller.
Specifically, in the demodulation process, the vibration information may be demodulated to obtain a demodulated signal. And decomposing the demodulated signal to obtain a plurality of decomposed signals. And calculating an envelope signal and an envelope spectrum signal of each decomposed signal, and constructing an objective function according to the envelope signal and the envelope spectrum signal of each decomposed signal, thereby identifying whether each decomposed signal can be used as a mode component for representing an operation mode. That is, whether each of the resolved signals can be used to represent one mode of operation of the idler, including a normal mode of operation or a failure mode, etc., is determined, thereby resulting in a plurality of mode components that are used to represent the mode of operation of the idler. Therefore, in the embodiment of the application, the demodulation signal can be decomposed, so that multiple working modes of the carrier roller included in the vibration information are disassembled, and a mode component with a finer granularity is obtained, so that whether the carrier roller fails or not can be accurately judged according to various mode components subsequently.
More specifically, the process of constructing the objective function may include: calculating a first frequency according to the envelope signal and the corresponding period, calculating a second frequency according to the envelope spectrum signal and the corresponding harmonic, constructing an objective function based on a difference value of the first frequency and the second frequency, and judging whether the decomposition signal is a mode component according to the constructed objective function. Therefore, in the process of decomposing to obtain a plurality of mode components, the objective function can be constructed based on the frequency, and different working modes generally correspond to different frequencies, so that the mode components representing the working modes of the carrier roller from the frequency dimension are obtained.
303. Determining whether the mechanical device is malfunctioning based on the plurality of mode components.
After obtaining the multiple mode components, it can be determined whether the working modes represented by the various mode components include a failure mode. When the working modes represented by the multiple mode components include a fault mode, the fault of the carrier roller can be determined, and fault information is output, wherein the fault information represents that the carrier roller is in fault.
In the embodiment of the application, after the vibration information is collected, the vibration information can be demodulated, so that various mode components can be obtained. The multiple mode components can be analyzed respectively, and when the working mode represented by one or more mode components is determined to be a fault mode, the roller fault can be determined. Therefore, whether the carrier roller is in fault or not can be analyzed from the dimension of the mode component, and whether the carrier roller is in fault or not can be accurately judged.
Specifically, each mode component may be evaluated from a plurality of dimensions, and a plurality of index values for each mode component may be calculated, where the plurality of dimensions may include at least one of signal impulse, periodic impulse, or signal-to-noise ratio of the mode component; then, according to the index values of each mode component, whether the working mode of the corresponding mechanical equipment is a fault mode is identified; and if the plurality of mode components comprise at least one fault mode component, generating fault information indicating whether at least one mechanical device has faults, wherein the at least one fault mode component indicates the mode component of the operation of the mechanical device, which is indicated as a fault mode. Therefore, whether the carrier roller fails or not can be comprehensively judged from dimensions such as signal impact, periodic impact or signal to noise ratio, and whether the carrier roller fails or not can be accurately identified.
Further, the manner of determining whether the working mode represented by the mode component is the failure mode may specifically include: taking any mode component as an example, for convenience of understanding, it is referred to as a first mode component, and if the index values corresponding to the first mode component are not within the preset boundary range, it may be determined that the index values of the first mode component identify that the operation mode of the corresponding mechanical device is a failure mode, where the preset boundary is a distribution range of the index values corresponding to the operation mode of the mechanical device when the operation mode is in a non-failure mode, that is, when the operation mode is in a normal operation mode, and specifically may be calculated from historical operation data of the mechanical device, and the first mode component may be any one of multiple mode components. Therefore, whether the work representation represented by each mode component is a fault mode or not can be comprehensively identified by combining the indexes of the multiple dimensions of each mode component, and whether the carrier roller is in fault or not can be identified more accurately.
The aforementioned multiple indexes of multiple dimensions may specifically include one or more of the following: kurtosis, relative kurtosis, including harmonic-to-noise ratio, negative entropy, gini exponent, variance, or standard deviation, etc. Therefore, the characteristics of signal impact, periodic impact or signal-to-noise ratio and other dimensions can be represented by various indexes, so that whether the carrier roller fails or not can be judged more accurately.
The method provided by the present application is introduced in the foregoing, and for convenience of understanding, the method provided by the present application is described in more detail below with reference to specific application scenarios.
In addition, in the method provided by the present application, in addition to identifying whether the idler roller is faulty or not, the type of fault or the magnitude of the fault can be identified, and the like, referring to fig. 4, a flow diagram of another method for detecting faults of the rotary mechanical equipment is provided by the present application.
Specifically, the method for detecting faults of rotary mechanical equipment provided by the application can be divided into a plurality of parts, as shown in fig. 4: signal sensitization, fault mode identification, fault diagnosis, fault magnitude diagnosis and the like.
In the signal sensitization part, a sound sensitization material is wrapped outside the sensing optical fiber, so that the sensing strength of the sensing optical fiber to sound is increased, and vibration information with higher signal-to-noise ratio is obtained. Or a connecting piece is added between the sensing optical fiber and the carrier roller, so that the vibration of the carrier roller is transmitted to the sensing optical fiber through the connecting piece, and accurate vibration information is obtained. And demodulating the vibration information to obtain a demodulated signal.
In the failure mode identifying section, the demodulated signal may be decomposed into a plurality of mode components, and the mode component corresponding to the failure mode may be identified. Specifically, the demodulated signal may be decomposed to obtain a plurality of mode components, an objective function is constructed for each mode component, and whether each mode component is a mode component corresponding to the failure mode is identified according to the constructed objective function. Therefore, fault identification of the carrier roller is realized.
Besides the fact that whether the carrier roller has faults or not is recognized, fault type diagnosis and fault magnitude diagnosis can be further conducted. In the fault type diagnosis stage, fault types can be identified according to indexes of multiple dimensions of each fault mode component.
In the fault magnitude diagnosis stage, data joint distribution corresponding to each magnitude can be determined by combining full-band statistical information of historical data based on each index of the fault mode component, and therefore the fault magnitude corresponding to the fault of the carrier roller is calculated and obtained based on the multi-dimensional characteristics of each mode component.
The flow of the method for detecting faults of a rotary mechanical device provided by the present application is described in more detail with reference to fig. 4. Referring to fig. 5, a flow chart of another method for detecting a fault of a rotary machine according to the present disclosure is shown as follows.
501. Vibration information is picked up.
Generally, when the idler is degraded, abnormal impact, squeaking and other noises are generated or vibration and the like are generated during operation due to factors such as bearing failure, drum skin breakage, eccentricity, collision and abrasion and the like, and the state is generally called as an abnormal state or a failure state of the idler.
When the picked-up information comprises vibration information, a buckle or other connecting piece can be arranged between the sensing optical fiber and each carrier roller, so that the vibration information of the carrier roller can be collected through the sensing optical fiber. Generally, the vibration of the carrier roller can directly reflect the working mode of the carrier roller, the vibration information of the carrier roller can be picked up more accurately by the sensing optical fiber through the connecting piece by picking up the vibration information of the carrier roller, and the vibration information with higher signal-to-noise ratio is obtained.
When the pickup information comprises the sound of the carrier, the sensing optical fiber can be wrapped with a sound sensitizing material to pick up the abnormal sound signal of the carrier roller. Such as a bubble wrap or plastic film, which is sensitive to sound or vibration, on the outside of the sensing fiber. When the sensitizing material is used to pick up the sound emitted by the idler in an abnormal state, the optical fiber needs to be arranged at a position as close as possible to the idler. For example, the carrier roller is not more than 1 meter perpendicular to the fiber, and if the carrier roller is too far away, the signal will be severely attenuated. The positioning requirement of carrier roller monitoring is considered at the intervals of the sound sensitizing materials, the intervals are set to be 5 meters, the specific intervals can be adjusted according to the positioning precision requirement and the data transmission pressure, if the abnormal positioning precision of the carrier roller is reduced to a range of 10 meters, the intervals between the sensitizing materials can be set to be 10 meters, the corresponding data quantity collection is reduced by one time, the data are stored, the network transmission pressure is correspondingly reduced, on the contrary, if the positioning precision is to be improved, the intervals between the sensitizing materials are required to be reduced, the data transmission pressure is increased, and if the intervals between the sensitizing materials are 3 meters, the data transmission pressure is increased. Therefore, in the embodiment, the optical fiber can be coated outside the sound-sensitizing material, the abnormal sound signal of the carrier roller is enhanced through the sound-sensitizing material, the signal-to-noise ratio of the abnormal sound signal of the carrier roller picked up by the optical fiber is improved, the sound information of the operation of the full-channel carrier roller is obtained, and a foundation is laid for extracting fault characteristics based on signal processing and developing fault diagnosis based on signal statistical characteristics.
502. And (5) signal demodulation.
After vibration information is picked up, signals collected through the sensing optical fiber can be demodulated through the DAS, for example, coherent demodulation or non-coherent demodulation can be adopted, for example, hilbert phase demodulation can be performed, and demodulated optical fiber phase demodulation signals xx (t) are obtained.
503. And storing necessary information.
Generally, the characteristic of the demodulated signal has strong correlation with the roller fault process, and when the demodulated signal is normalized, such correlation information may be lost, so that before signal normalization, data screening can be performed, and information which needs to be retained is screened out, so as to avoid data loss after signal normalization.
Specifically, information can be stored through several dimensions of vibration intensity, impact, amplitude energy and amplitude trend, and the corresponding measurement indexes can include indexes such as vibration intensity, negative entropy, gini index, kurtosis, RMS value, variance, standard value or mean value, and the indexes are stored for subsequent calculation.
The vibration of the idler may also be abnormal, as is often the case when conditions such as starting the idler, changing loads, or shutting the idler change significantly. The mean value of necessary reserved information can be used for identifying a variable rotating speed process, for example, RMS is used for identifying load change, data with the mean value and the RMS not changing greatly can be selected as approximate working condition data for analysis, the fluctuation characteristic of kurtosis is used for measuring the random impact characteristic of the service environment of the carrier roller, if the rotating speed or the load changes, the necessary reserved information is not in fault factor outlier, the sample weighting factor can be set to be zero, namely, when whether the carrier roller is in fault or not is considered, the data can not be used for fault analysis.
In this application embodiment, can discern the information that picks up whether the data of gathering under the condition that the operating mode of bearing roller changes obviously takes place. Usually, the difference exists between the data acquired under the condition that the working condition of the carrier roller is obviously changed and the data in the working mode of the carrier roller, and if the difference is used for fault diagnosis, misdiagnosis can be caused. Therefore, in the embodiment of the application, when the condition of the carrier roller is obviously changed, the weight of the acquired data is set to be 0, that is, whether the carrier roller is in fault is not diagnosed through the data, so that the condition of misdiagnosis is avoided, and the fault diagnosis accuracy is improved.
504. And (4) signal normalization.
The signals can be standardized subsequently, so that differences of the individual working environments of the carrier rollers are eliminated, and a standardized signal x (t) is obtained.
Specifically, the demodulated signals can be normalized through a plurality of normalization algorithms, for example, the demodulated signals are normalized through a normalization algorithm, a centering algorithm or a Z-score normalization algorithm, so as to obtain normalized signals x (t), so that the influence caused by the difference of the working environments of the carrier rollers is eliminated, and only the characteristics generated by the working of the carrier rollers can be possibly reserved.
505. And (5) signal decomposition.
The normalized signal x (t) may then be decomposed into a number of decomposed signals x i (t) of (d). E.g. by signal decomposition algorithms such as Singular Value Decomposition (SVD), trigonometric decomposition, orthogonal triangle decompositionMethods, and the like. And then screening out a mode component which can be used for representing the fault mode of the carrier roller from the decomposed signal so as to facilitate fault identification based on the mode component.
The present application exemplarily describes the decomposition algorithm as SVD, and the SVD mentioned below may be replaced by other decomposition algorithms.
Illustratively, the normalized signal may be decomposed into several decomposed signals x by SVD i (t) and singular value σ for each mode component i . Each decomposed signal may then be envelope demodulated to obtain an envelope signal Xenv (t) and its envelope spectrum signal Xenv (f) for each decomposed signal.
Subsequently, for a single decomposed signal, a mode component which can be used for representing the operation mode of the carrier roller can be screened by constructing an objective function, and the specific process of constructing the objective function can include:
xenv is obtained by performing envelope correlation analysis on the ith decomposition signal i (t) and calculating xenv using correlation analysis i (t) frequency f corresponding to the strongest periodic component i0 ，f i0 The calculation process is as follows:
wherein, xenv i (t) represents the time-domain envelope signal of the ith decomposition signal, [ t ] 0 ,t 1 ]Representing the range of time delay, t, of tau when calculating the autocorrelation of the envelope signal 0 The time corresponding to the zero crossing point of the autocorrelation spectrum of the envelope signal, t, is usually selected 1 The duration of the analysis signal, argmax, is usually chosen τ (-) represents the maximum likelihood estimation operator.
Performing spectrum correlation analysis through the envelope spectrum signal of the ith decomposition signal, and calculating the frequency corresponding to the component with the strongest sparse harmonic of the envelope spectrumf i1 ，f i1 The calculation process is as follows:
f i1 ＝τ i1
wherein, xenv i (f) Envelope spectrum signal [0, f ] representing the i-th mode component s /2]Representing the range of frequency shift, f, of tau when calculating the autocorrelation of the envelope spectrum s The sampling frequency is usually taken.
Constructing whether the ith decomposition signal is an objective function of a mode component may be:
|f i0 -f i1 |<ε
wherein epsilon is a specified period estimation deviation range, and can be calculated according to historical data, such as 0.3 or other values. The decomposed signal satisfying the objective function is defined as a mode component y j (N) wherein j =1,2, \8230;, N j ，N j Is the number of mode components. Therefore, in the embodiment of the application, whether each decomposed signal can be used for representing the operation mode of the carrier roller or not can be identified from the frequency represented by the decomposed signal, such as the operation frequency or the harmonic frequency, so as to screen the component which can be used for representing the operation mode of the carrier roller, and to facilitate accurate fault diagnosis in the following process.
506. And calculating the multidimensional index.
Screening out a plurality of mode components y from the decomposed signal by constructing an objective function j And (n), calculating a multi-dimensional characteristic indicator of the mode component, wherein the multi-dimensional characteristic indicator is used for measuring whether the mode component is a key input of the fault mode component. These statistics can only be correlated with the failure signature to avoid misidentification or missing identification of the failure mode. Therefore, the method and the device can identify the dimensionalities such as signal impact, periodic impact, signal-to-noise ratio and the like, can ensure the identification precision of the fault mode component, can select corresponding indexes such as kurtosis K, correlation kurtosis CK, envelope harmonic-to-noise ratio EHNR and the like, and can be replaced by similar indexes such as kurtosis K, correlation kurtosis CK, envelope harmonic-to-noise ratio EHNR and the likeThe method and the device are used for representing other indexes of dimensionalities such as signal impact, periodic impact, signal to noise ratio and the like, and are not described in detail in the application.
Specifically, the process of calculating the multi-dimensional index may include the following.
Kurtosis K of jth information mode component j The calculation process is as follows:
wherein: n denotes the signal length.
Correlation kurtosis CK of jth information mode component j The calculation process of (2) is as follows:
wherein, T represents a period, and the correlation kurtosis is often calculated by a conventional method through a specified period, so that it is often difficult for the indicator to effectively reflect fault information due to insufficient accuracy of the specified period. According to the method, because the frequency of the information mode component is calculated in the fault mode component identification link, and the frequency is associated with the period, the period for calculating the correlation kurtosis can be set as
Envelope harmonic-to-noise ratio EHNR of the jth information mode component j The calculation process is as follows:
r j (τ)＝∫yenv j (t)yenv j (t+τ)dt
wherein: yenv j (t) an envelope signal representing the jth information pattern component, r j (τ) represents yenv j (t) autocorrelation function, τ max Is expressed such that the autocorrelation function r j (τ) taking the maximum extremumAnd (4) time delay.
507. And identifying whether the mode component has a fault, if not, executing step 508, and if not, executing step 501.
After the multidimensional index corresponding to each mode component is calculated, whether each mode component is a failure mode component can be judged based on the multidimensional index. When the plurality of mode components include a fault mode component, a fault indication indicating a fault of the carrier roller can be directly output, or the fault types are further classified, that is, step 508 is executed; if no fault mode component is included in the plurality of mode components, that is, no fault is identified in the idler, the vibration information of the idler may be continuously picked up, that is, the step 501 is continuously executed, or the step 501 is executed after a certain period.
For example, the manner of determining whether the mode component is the failure mode component may specifically include:
and analyzing the three-dimensional combined distribution of K, CK and EHNR through historical data to obtain the distribution range and the distribution boundary of the normal carrier roller. And when the three-dimensional coordinates of the mode component to be diagnosed exceed the distribution boundary of the normal carrier roller, judging the index outlier of the mode component, and recognizing the mode component as a fault mode component if the mode is recognizable. The number of failure mode components N can also be calculated f And the energy of the fault mode component is compared with FER, and the information of the fault mode component can be saved for subsequent fault diagnosis and magnitude. When the mode component is within the normal idler distribution boundary, the idler is determined to be a normal idler, i.e. the idler is indicated as being in a normal operating mode.
Wherein, the FER is calculated as follows:
energy contribution to FER of the kth failure mode component i The index is defined as:
the energy to energy ratio FER of the identified failure mode component for each failure sample is defined as:
wherein N is 0 The number of mode components of the SVD decomposition is shown, and σ shows a singular value of the SVD decomposition.
Therefore, in the embodiment of the present application, the mode component is defined, and the failure mode component is identified by the multi-dimensional feature joint distribution outlier of the mode component. The accuracy of the abnormal identification of the fault carrier roller can be improved, and the abnormal state of the carrier roller can be identified under different fault types.
508. And (4) fault classification.
In the embodiment of the application, the fault types can be identified by utilizing the joint distribution of the multi-dimensional statistical information of the fault mode components, and the multi-dimensional information can improve the identification precision of the carrier rollers at different positions and different fault types.
The fault classification method according to the embodiment of the present application may include various methods, such as classification by a model trained in advance, classification by an empirical threshold, and the like, which are described below.
1. Model training
And taking a plurality of index values corresponding to at least one fault mode component as the input of the fault type identification model, and outputting the fault type corresponding to each fault mode component in the at least one fault mode component.
The used fault type identification model is obtained by training through a first training set, the first training set can comprise historical data of at least one piece of mechanical equipment, and the historical data comprises data of the at least one piece of mechanical equipment generated when the at least one piece of mechanical equipment works in a non-fault mode and a fault type obtained when the at least one piece of mechanical equipment works in a fault mode.
The specific training process may include: and training the fault type recognition model by using a first training set to obtain the trained fault type recognition model, wherein in the process of training the fault type recognition model, samples in the first training set are used as the input of the fault type recognition model, and the fault type recognition model is used for carrying out fault classification on samples corresponding to various types of faults and calculating the classification center of each fault classification.
Accordingly, the process of identifying a fault type using the fault type identification model may include: extracting characteristics from a plurality of index values corresponding to at least one fault mode component through a fault type identification model; obtaining the distance between the characteristic corresponding to each fault mode component and the classification center of each fault classification through a fault type identification model; and determining the fault type corresponding to each fault mode component according to the distance between the characteristic corresponding to each fault mode component and the classification center of each fault classification, for example, taking the fault classification which is closest to the fault mode component or has the distance within a preset range as the fault type corresponding to the fault mode component.
For example, a training set is generated from historical failure data based on statistical information of the failure mode components. The collected historical data can include data of different states such as normal carrier roller data, carrier roller bearing faults, cylinder skin damage, carrier roller bracket abnormity, carrier roller collision and abrasion and the like. Each type of fault data is multiple, a certain amount of negative sample size is usually needed, and data of more fault types are collected as much as possible, so that the purpose of model training is achieved. If each type of fault comprises ten groups of fault data and 80 groups of normal data, a total of 120 groups of negative samples can be selected, each training sample x (t) comprises a plurality of inputs, namely four indexes K, CK, EHNR and sigma of the mode component with the highest correlation kurtosis and energy ratios FER and N of all fault mode components f . Subsequently, a feature matrix TR is generated using the training set (120×6) And using Principal Component Analysis (PCA) mode to reduce dimension of the feature matrix, calculating and arranging feature values delta in descending order, and selecting the feature values delta of the first m i And corresponding feature vector PC i Where i = 1-m, the m eigenvectors corresponding to the classification accuracy satisfying 90% or more (although other ranges may be substituted) may be selected to generate the eigenvector matrix PC (6×m) And calculating the classification center O corresponding to the training set of different fault types i Calculating an input vector X for the test data (1×6) By X (1×6) ·PC (6×m) Calculating the projection of the test point P on each feature vector, and projecting the point P to various centers O according to the test set i Distance | PO of i And select min (| PO) i |) the corresponding fault type is the fault type of the fault mode.
Therefore, in the embodiment of the application, the fault type identification model can be trained by using the historical data, so that the model is trained by combining the historical empirical data, and the fault type can be accurately and efficiently identified when the fault type is inferred. And establishing a mapping relation between indexes of all dimensions of the mode components and fault types based on historical data. Therefore, when the fault type is identified, the fault type corresponding to the fault mode component can be identified based on the mapping relation.
2. Comparing empirical thresholds
The plurality of index values corresponding to at least one fault mode component may be compared with corresponding threshold values to determine a fault type corresponding to each fault mode component in the at least one fault mode component. Wherein the threshold value can be calculated according to historical data or set manually.
For example, to more effectively perform idler roller abnormal state identification and failure diagnosis with a relatively small number of negative samples, the information pattern component y is identified j (n), after the information mode component is selected by using the objective function, the embodiment of the present application may perform failure mode identification by using multidimensional characteristic statistical index outlier, which is performed by using the associated kurtosis CK j ＞CK 0 Or kurtosis K j ＞K 0 (the condition information pattern component is identified as the failure pattern component, and the condition information pattern component and the failure pattern component can be set to be satisfied at the same time so as to improve the application capability of the method under the condition of few samples, wherein CK 0 And K 0 The outlier thresholds representing the relative kurtosis and kurtosis, respectively, where the relative kurtosis is still computed over a specified period
Thus obtaining the product. The multidimensional characteristic outlier determination requires a large number of negative samples, and the embodiment can utilize a small number of samples for the determination。
Therefore, in a scene with fewer negative samples, the fault type can be identified according to the empirical value, so that the roller fault can be conveniently repaired.
509. And calculating the fault magnitude.
After the fault of the carrier roller is determined, the fault magnitude can be further calculated, so that a user can know the fault degree of the fault of the carrier roller, and the carrier roller can be maintained according to the fault degree.
Specifically, the failure magnitude calculation method may also be divided into multiple methods, for example, the failure magnitude may be identified by inputting a failure magnitude identification model, or by comparing an index of a failure mode component with an empirical value, which are described below separately.
1. Training model
The plurality of index values corresponding to the at least one fault mode component may be used as an input of the fault magnitude recognition model, and the fault magnitude corresponding to each fault mode component in the at least one fault mode component may be output.
The fault magnitude recognition model is obtained by training through a second training set, the second training set comprises historical data of at least one piece of mechanical equipment, and the historical data comprises data of the at least one piece of mechanical equipment generated when the at least one piece of mechanical equipment works in a non-fault mode and fault magnitude obtained when the at least one piece of mechanical equipment works in a fault mode.
The specific training process may include: and training the fault magnitude recognition model by using a second training set to obtain a trained fault magnitude recognition model, wherein in the process of training the fault magnitude recognition model, samples in the second training set are used as the input of the fault type recognition model, and the fault type recognition model is used for classifying magnitudes corresponding to multiple types of faults and calculating the classification center of each fault magnitude.
Accordingly, the specific process of calculating the magnitude of the fault may include: extracting features from a plurality of index values corresponding to at least one fault mode component through a fault magnitude recognition model; obtaining the distance between the characteristic corresponding to each fault mode component and the classification center of each fault magnitude through a fault magnitude recognition model; and determining the fault type corresponding to each fault mode component according to the distance between the characteristic corresponding to each fault mode component and the classification center of each fault magnitude, for example, taking the fault magnitude corresponding to the classification with the closest distance or the distance within a preset range as the fault magnitude corresponding to the fault mode component.
Specifically, for example, after the fault type is diagnosed, the health state of the carrier roller may be classified, 30 groups of test samples labeled with different fault classifications are selected, a 120 × 10 dimensional matrix is generated by using the multidimensional information calculated in step 507 and the information stored in step 503, and the first four components δ are selected by using PCA to reduce dimensions i I = 1-4 as the principal component of the fault magnitude and calculates the classification center L under different grades i For the test data with unknown fault magnitude, calculating a projection point M on the selected principal component, and calculating the projection point to the class center L i Distance | ML of i If the ML distance is the minimum, the fault classification corresponding to the class is the fault classification of the fault mode.
According to the fault magnitude recognition method and device, the fault magnitude recognition model can be built by using historical data, so that the model is trained by combining historical experience data, and the fault magnitude can be accurately and efficiently recognized when the fault magnitude is inferred. And establishing a mapping relation between indexes of all dimensions of the mode components and fault magnitude based on historical data. Therefore, when the fault magnitude is identified, the fault type corresponding to the fault mode component can be identified based on the mapping relation. Therefore, in the embodiment of the application, the fault magnitude recognition model can be constructed based on the historical data, so that the fault magnitude distribution rule can be learned according to the historical data, and the fault magnitude of the carrier roller can be accurately recognized.
2. Comparing empirical values
The plurality of index values corresponding to the at least one fault mode component may be compared with the preset range corresponding to the plurality of fault magnitude levels, and the fault magnitude level corresponding to each fault mode component in the at least one fault mode component may be determined. The preset range can be calculated according to historical data or can be obtained by manual setting.
For example, in order to deal with the problem of insufficient fault classification negative samples, the health state of the carrier roller can be divided into four stages of normal, early fault, medium fault and serious fault according to the correlation kurtosis and the envelope harmonic-to-noise ratio according to the numerical values; when CK j ≤CK 0 The carrier roller is in a normal state and is in a state of being CK when being satisfied j ＞CK 0 Idler in faulted state when CK 0 ＜CK j ＜CK 1 And EHNR j ＜EHNR 0 When the carrier roller is considered to be in an early failure state, when CK 0 ＜CK j ＜CK 1 And EHNR j ＞EHNR 0 When the carrier roller is in the middle-stage fault state, when CK j ＞CK 1 And EHNR j ＞EHNR 0 In the process, the carrier roller is considered to be in a serious fault state and should be replaced or maintained as soon as possible.
Therefore, in the embodiment of the application, even if the negative sample is insufficient, namely, the fault data in the historical data is less, the fault magnitude of the carrier roller can be accurately identified by the experience value set in advance.
It should be noted that, the execution order of step 508 and step 509 is not limited in the present application, step 508 may be executed first, step 509 may be executed first, step 508 and step 509 may also be executed simultaneously, and the present application does not limit this.
510. And outputting fault information.
After carrying out fault detection to the bearing roller, can be based on fault detection result output fault information.
If a fault component is detected as being present after step 507 is performed, the idler failure may be determined, and an indication of idler failure may be carried in the failure information.
Alternatively, if step 508 is performed, the type of failure of the idler may be carried in the failure information, so that the user may perform further maintenance or replacement of the idler based on the failure information, and the like.
Optionally, if step 509 is executed, the fault magnitude of the idler may be carried in the individual fault information, so that the user may know the fault degree of the idler through the fault magnitude, thereby deciding the operation for the idler.
In addition, warning information can be generated according to the fault type or the fault magnitude of the carrier roller, and the warning information can be carried in the fault information. The warning information may include a warning for the idler failure, for example, a warning identifier may be generated and displayed, so as to warn a user to process the idler failure; or generating an alarm prompt tone and playing the alarm prompt tone so as to remind a user of processing the roller fault through sound; or generating and displaying an alarm text, reminding a user to process the fault carrier roller through the text, and the like, wherein the alarm text can be specifically used according to actual application scenarios.
Therefore, in the embodiment of the application, the component which can be used for representing the working mode of each carrier roller can be obtained by decomposing the demodulated signal, so that the mode component can be comprehensively analyzed by combining the signal impact, the periodic impact, the signal-to-noise ratio and other dimensions of the component, whether the fault mode component exists can be accurately identified, and whether the carrier roller has a fault can be accurately identified. Further, the fault type or the fault magnitude can be comprehensively analyzed according to indexes of multiple dimensions, so that a user can more accurately know the type or the severity of the fault of the carrier roller, and the user can conveniently process the fault carrier roller.
The foregoing describes the process of the method provided in the present application, and the following describes the structure of the apparatus provided in the present application with reference to the foregoing process.
Referring to fig. 6, a schematic structural diagram of a fault detection apparatus for a rotary mechanical device provided in the present application may include:
the acquisition module 601 is configured to acquire vibration information, where the vibration information includes information of vibration generated by at least one mechanical device;
a demodulation module 602, configured to demodulate the vibration information to obtain multiple mode components, where the multiple mode components are used to represent information of at least one operation mode of at least one mechanical device;
and a fault identifying module 603, configured to determine fault information of the at least one mechanical device according to the plurality of mode components, where the fault information is used to indicate whether the at least one mechanical device generates a fault.
In a possible implementation, the fault identifying module 603 may be specifically configured to: evaluating each mode component from a plurality of dimensions, and calculating a plurality of index values of each mode component in the plurality of mode components, wherein the plurality of dimensions comprise at least one of signal impact, periodic impact or signal-to-noise ratio of the plurality of mode components; identifying whether the working mode of the corresponding mechanical equipment is a failure mode or not according to the index values of each mode component; and if the plurality of mode components comprise at least one fault mode component, generating fault information indicating whether at least one mechanical device generates faults, wherein at least one fault mode component indicates a mode component indicating that the operation of the mechanical device is a fault mode.
In a possible implementation, the fault identifying module 603 may be specifically configured to: if the index values corresponding to the first mode component are not within the preset boundary, determining that the index values of the first mode component identify that the working mode of the corresponding mechanical equipment is the fault mode, wherein the preset boundary is the distribution range of the index values corresponding to the mechanical equipment when the working mode of the mechanical equipment is in the non-fault mode, and the first mode component is any one of the mode components.
In one possible embodiment, the plurality of indicators for each mode component includes at least one of: kurtosis, relative kurtosis, including harmonic-to-noise ratio, negative entropy, gini exponent, variance, or standard deviation.
In one possible embodiment, the apparatus provided herein may further include:
the fault type identifying module 604 is configured to determine a fault type corresponding to each fault mode component in the at least one fault mode component according to a plurality of index values corresponding to the at least one fault mode component.
In a possible implementation, the fault type identifying module 604 may be specifically configured to: taking a plurality of index values corresponding to at least one fault mode component as input of a fault type identification model, and outputting a fault type corresponding to each fault mode component in the at least one fault mode component, wherein the fault type identification model is obtained by training by using a first training set, the first training set comprises historical data of at least one piece of mechanical equipment, and the historical data comprises data of the at least one piece of mechanical equipment, which is generated when the at least one piece of mechanical equipment works in a non-fault mode, and a fault type obtained when the at least one piece of mechanical equipment works in a fault mode; or comparing a plurality of index values corresponding to at least one fault mode component with corresponding threshold values, and determining the fault type corresponding to each fault mode component in the at least one fault mode component.
In one possible embodiment, the apparatus provided herein may further include: the first training module 605 is configured to train the fault type identification model by using a first training set to obtain a trained fault type identification model, where in the process of training the fault type identification model, samples in the first training set are used as inputs of the fault type identification model, and the fault type identification model is configured to perform fault classification on samples corresponding to multiple types of faults and calculate a classification center of each fault classification;
the fault type identification module 604 may be specifically configured to: extracting features from a plurality of index values corresponding to at least one fault mode component through a fault type identification model; obtaining the distance between the characteristic corresponding to each fault mode component and the classification center of each fault classification through a fault type identification model; and determining the fault type corresponding to each fault mode component according to the distance between the characteristic corresponding to each fault mode component and the classification center of each fault classification.
In one possible embodiment, the apparatus provided herein may further include:
the failure magnitude identification module 606 is configured to determine a failure magnitude corresponding to each failure mode component in the at least one failure mode component according to a plurality of index values corresponding to the at least one failure mode component, where the failure magnitude represents a failure degree of a mechanical device corresponding to each failure mode component in the at least one failure mode component.
In a possible implementation, the failure magnitude identification module 606 is specifically configured to: taking a plurality of index values corresponding to at least one fault mode component as input of a fault magnitude recognition model, and outputting a fault magnitude corresponding to each fault mode component in the at least one fault mode component, wherein the fault magnitude recognition model is obtained by using a second training set for training, the second training set comprises historical data of at least one piece of mechanical equipment, and the historical data comprises data of the at least one piece of mechanical equipment generated when the at least one piece of mechanical equipment works in a non-fault mode and the fault magnitude obtained when the at least one piece of mechanical equipment works in a fault mode; or comparing a plurality of index values corresponding to at least one fault mode component with ranges corresponding to a plurality of fault magnitude levels, and determining the fault magnitude level corresponding to each fault mode component in the at least one fault mode component.
In one possible embodiment, the apparatus may further comprise: a second training module 607, configured to train the fault magnitude recognition model using a second training set to obtain a trained fault magnitude recognition model, where in the process of training the fault magnitude recognition model, samples in the second training set are used as inputs of the fault type recognition model, and the fault type recognition model is used to classify magnitudes corresponding to multiple types of faults and calculate a classification center of each fault magnitude;
the failure magnitude identification module 606 may be specifically configured to: extracting features from a plurality of index values corresponding to at least one fault mode component through a fault magnitude recognition model; obtaining the distance between the characteristic corresponding to each fault mode component and the classification center of each fault magnitude through a fault magnitude recognition model; and determining the fault magnitude corresponding to each fault mode component according to the distance between the characteristic corresponding to each fault mode component and the classification center of each fault magnitude.
In a possible implementation, the plurality of indicators for each mode component includes a root mean square RMS, and the apparatus provided herein may further include a filtering module 608 configured to: before identifying whether the working mode of the corresponding mechanical equipment is a fault mode according to the index values of each mode component, screening out at least one mode component with a change value of RMS within a preset time period smaller than a preset value from the multiple mode components; and identifying whether the working mode of the corresponding mechanical equipment is a failure mode or not according to the index values of the at least one mode component.
In a possible implementation, the demodulation module 602 may specifically be configured to: demodulating the vibration information to obtain a demodulation signal; decomposing the demodulated signal to obtain a plurality of decomposed signals; calculating an envelope signal and an envelope spectrum signal of each of the plurality of decomposed signals; and identifying whether each decomposed signal is a mode component or not according to the envelope signal and the envelope spectrum signal of each decomposed signal and the objective function so as to obtain a plurality of mode components.
In a possible implementation, the demodulation module 602 may specifically be configured to: acquiring a first frequency according to a period in the envelope signal; acquiring a second frequency according to the harmonic wave of the envelope spectrum signal; constructing an objective function according to the difference between the first frequency and the second frequency; whether each decomposed signal is a mode component is determined according to the value of the objective function.
In one possible embodiment, the vibration information includes information generated by sound when the at least one mechanical device is operating or information generated by vibration when the at least one mechanical device is operating.
In a possible embodiment, the acquisition module 601 is specifically configured to acquire information generated by sound of at least one piece of mechanical equipment or information generated by vibration of at least one piece of mechanical equipment through the optical fiber distributed acoustic sensor DAS, where an optical fiber for acquiring data is deployed outside the at least one piece of mechanical equipment, and when the vibration information includes the information generated by sound of at least one piece of mechanical equipment in operation, a sound-sensitizing structure is wrapped outside the optical fiber, and the sound-sensitizing structure is configured to increase the sensing strength of the optical fiber to sound, and when the vibration information includes the information generated by vibration of at least one piece of mechanical equipment in operation, the at least one piece of mechanical equipment and the optical fiber are connected by a connecting member.
In a possible embodiment, the aforementioned at least one mechanical device may comprise at least one idler.
Referring to fig. 7, a schematic structural diagram of another fault detection apparatus for a rotary mechanical device provided in the present application is as follows.
The rotary machine equipment fault detection apparatus may include a processor 701 and a memory 702. The processor 701 and the memory 702 are interconnected by wires. The memory 702 has stored therein program instructions and data.
The memory 702 stores program instructions and data corresponding to the steps of fig. 3-5 described above.
Optionally, the rotating machine fault detection apparatus may further include a transceiver 703 for receiving or transmitting data.
In one possible embodiment, the device for detecting a fault in a rotating machine may further include a vibration sensor (not shown in fig. 7) for collecting information about the vibration generated by the machine. The vibration sensor may specifically comprise a DAS or an inductive vibration sensor or the like.
Also provided in embodiments of the present application is a computer-readable storage medium, which stores a program that, when executed on a computer, causes the computer to perform the steps in the method described in the foregoing embodiments shown in fig. 3-5.
Alternatively, the fault detection device of the rotary mechanical equipment shown in the aforementioned fig. 7 is a chip
The embodiment of the present application further provides a fault detection apparatus for a rotary machine, where the fault detection apparatus for a rotary machine may also be referred to as a digital processing chip or a chip, and the chip includes a processing unit and a communication interface, where the processing unit obtains a program instruction through the communication interface, and the program instruction is executed by the processing unit, and the processing unit is configured to execute the method steps in fig. 3 to fig. 5.
The embodiment of the application also provides a digital processing chip. Integrated with circuitry and one or more interfaces to implement the processor 701, or functions of the processor 701, as described above. When integrated with memory, the digital processing chip may perform the method steps of any one or more of the foregoing embodiments. When the digital processing chip is not integrated with the memory, the digital processing chip can be connected with the external memory through the communication interface. The digital processing chip implements the actions performed by the rotary machine fault detection device or the rotary machine fault detection device in the above embodiments according to the program codes stored in the external memory.
Also provided in embodiments of the present application is a computer program product, which when executed on a computer, causes the computer to perform the steps of the method as described in the embodiments of fig. 3-5 above.
The rotary mechanical equipment fault detection device or the rotary mechanical equipment fault detection device provided by the embodiment of the application can be a chip, and the chip comprises: a processing unit, which may be, for example, a processor, and a communication unit, which may be, for example, an input/output interface, a pin or a circuit, etc. The processing unit may execute computer-executable instructions stored by the storage unit to cause a chip within the server to perform the method steps described in the embodiments of fig. 3-5 above. Optionally, the storage unit is a storage unit in the chip, such as a register, a cache, and the like, and the storage unit may also be a storage unit located outside the chip in the wireless access device, such as a read-only memory (ROM) or another type of static storage device that can store static information and instructions, a Random Access Memory (RAM), and the like.
Specifically, the aforementioned processing unit or processor may be a Central Processing Unit (CPU), a Network Processor (NPU), a Graphics Processing Unit (GPU), a Digital Signal Processor (DSP), an Application Specific Integrated Circuit (ASIC), a Field Programmable Gate Array (FPGA), or other programmable logic devices (programmable gate array), discrete gate or transistor logic devices (discrete hardware components), or the like. The general purpose processor may be a microprocessor or may be any conventional processor or the like.
Referring to fig. 8, fig. 8 is a schematic structural diagram of a chip according to an embodiment of the present disclosure, where the chip may be represented as a neural network processor NPU 80, and the NPU 80 is mounted on a main CPU (Host CPU) as a coprocessor, and the Host CPU allocates tasks. The core portion of the NPU is an arithmetic circuit 803, and the controller 804 controls the arithmetic circuit 803 to extract matrix data in the memory and perform multiplication.
In some implementations, the arithmetic circuit 803 includes a plurality of processing units (PEs) therein. In some implementations, the operational circuitry 803 is a two-dimensional systolic array. The arithmetic circuit 803 may also be a one-dimensional systolic array or other electronic circuit capable of performing mathematical operations such as multiplication and addition. In some implementations, the arithmetic circuitry 803 is a general-purpose matrix processor.
For example, assume that there is an input matrix A, a weight matrix B, and an output matrix C. The arithmetic circuit fetches the data corresponding to the matrix B from the weight memory 802 and buffers the data on each PE in the arithmetic circuit. The arithmetic circuit takes the matrix a data from the input memory 801 and performs matrix operation with the matrix B, and partial or final results of the obtained matrix are stored in an accumulator (accumulator) 808.
The unified memory 806 is used to store input data as well as output data. The weight data is directly passed through a Direct Memory Access Controller (DMAC) 805 and the DMAC is carried into the weight memory 802. The input data is also carried through the DMAC into the unified memory 806.
A Bus Interface Unit (BIU) 810 for interaction of the AXI bus with the DMAC and an Instruction Fetch Buffer (IFB) 809.
The bus interface unit 810 (BIU) is configured to obtain an instruction from the external memory by the instruction fetch memory 809, and obtain the original data of the input matrix a or the weight matrix B from the external memory by the storage unit access controller 805.
The DMAC is mainly used to carry input data in the external memory DDR to the unified memory 806, or carry weight data into the weight memory 802, or carry input data into the input memory 801.
The vector calculation unit 807 includes a plurality of operation processing units, and further processes the output of the operation circuit, such as vector multiplication, vector addition, exponential operation, logarithmic operation, magnitude comparison, and the like, if necessary. The method is mainly used for non-convolution/full-connection layer network calculation in the neural network, such as batch normalization (batch normalization), pixel-level summation, up-sampling of a feature plane and the like.
In some implementations, the vector calculation unit 807 can store the vector of processed outputs to the unified memory 806. For example, the vector calculation unit 807 may apply a linear function and/or a nonlinear function to the output of the operation circuit 803, such as linear interpolation of the feature planes extracted by the convolution layer, and further such as a vector of accumulated values to generate the activation values. In some implementations, the vector calculation unit 807 generates normalized values, pixel-level summed values, or both. In some implementations, the vector of processed outputs can be used as activation inputs to the arithmetic circuitry 803, e.g., for use in subsequent layers in a neural network.
An instruction fetch buffer 809 connected to the controller 804, configured to store instructions used by the controller 804;
the unified memory 806, the input memory 801, the weight memory 802, and the instruction fetch memory 809 are all On-Chip memories. The external memory is private to the NPU hardware architecture.
The operation of each layer in the recurrent neural network can be performed by the operation circuit 803 or the vector calculation unit 807.
Where any of the aforementioned processors may be a general purpose central processing unit, microprocessor, ASIC, or one or more integrated circuits configured to control the execution of the programs of the methods of fig. 3-5.
It should be noted that the above-described embodiments of the apparatus are merely illustrative, where the units described as separate parts may or may not be physically separate, and parts displayed as units may or may not be physical units, may be located in one place, or may be distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the solution of the present embodiment. In addition, in the drawings of the embodiments of the apparatus provided in the present application, the connection relationship between the modules indicates that there is a communication connection therebetween, and may be implemented as one or more communication buses or signal lines.
Through the above description of the embodiments, those skilled in the art will clearly understand that the present application can be implemented by software plus necessary general-purpose hardware, and certainly can also be implemented by special-purpose hardware including special-purpose integrated circuits, special-purpose CPUs, special-purpose memories, special-purpose components and the like. Generally, functions performed by computer programs can be easily implemented by corresponding hardware, and specific hardware structures for implementing the same functions may be various, such as analog circuits, digital circuits, or dedicated circuits. However, for the present application, the implementation of a software program is more preferable. Based on such understanding, the technical solutions of the present application may be embodied in the form of a software product, which is stored in a readable storage medium, such as a floppy disk, a usb disk, a removable hard disk, a Read Only Memory (ROM), a Random Access Memory (RAM), a magnetic disk or an optical disk of a computer, and includes instructions for enabling a computer device (which may be a personal computer, a server, or a network device) to execute the methods described in the embodiments of the present application.
In the above embodiments, the implementation may be wholly or partially realized by software, hardware, firmware, or any combination thereof. When implemented in software, may be implemented in whole or in part in the form of a computer program product.
The computer program product includes one or more computer instructions. When loaded and executed on a computer, cause the processes or functions described in accordance with the embodiments of the application to occur, in whole or in part. The computer may be a general purpose computer, a special purpose computer, a network of computers, or other programmable device. The computer instructions may be stored in a computer readable storage medium or transmitted from one computer readable storage medium to another, for example, from one website site, computer, server, or data center to another website site, computer, server, or data center via wired (e.g., coaxial cable, fiber optic, digital Subscriber Line (DSL)) or wireless (e.g., infrared, wireless, microwave, etc.). The computer-readable storage medium can be any available medium that a computer can store or a data storage device, such as a server, a data center, etc., that is integrated with one or more available media. The usable medium may be a magnetic medium (e.g., floppy disk, hard disk, magnetic tape), an optical medium (e.g., DVD), or a semiconductor medium (e.g., solid State Disk (SSD)), among others.
The terms "first," "second," "third," "fourth," and the like in the description and in the claims of the present application and in the drawings described above, if any, are used for distinguishing between similar elements and not necessarily for describing a particular sequential or chronological order. It will be appreciated that the data so used may be interchanged under appropriate circumstances such that the embodiments described herein may be practiced otherwise than as specifically illustrated or described herein. Furthermore, the terms "comprises," "comprising," and "having," and any variations thereof, are intended to cover a non-exclusive inclusion, such that a process, method, system, article, or apparatus that comprises a list of steps or elements is not necessarily limited to those steps or elements expressly listed, but may include other steps or elements not expressly listed or inherent to such process, method, article, or apparatus.
Claims (19)
1. A method for detecting a fault of a rotary machine, comprising:
acquiring vibration information, wherein the vibration information comprises information of collected vibration generated by at least one piece of mechanical equipment;
demodulating the vibration information to obtain a plurality of mode components, wherein the plurality of mode components are used for representing information of at least one working mode of the at least one mechanical device;
and determining fault information of the at least one mechanical device according to the plurality of mode components, wherein the fault information is used for indicating whether the at least one mechanical device generates faults or not.
2. The method of claim 1, wherein determining fault information for the at least one mechanical device based on the plurality of mode components comprises:
acquiring a plurality of index values of each mode component in the plurality of mode components, wherein the plurality of index values comprise evaluation values obtained by evaluating each mode component from a plurality of dimensions, and the plurality of dimensions comprise at least one of signal impact, periodic impact or signal-to-noise ratio of the plurality of mode components;
identifying whether the working mode of the corresponding mechanical equipment is a failure mode or not according to the index values of each mode component;
and if the plurality of mode components comprise at least one fault mode component, generating fault information which indicates whether the at least one mechanical device has faults or not, wherein the at least one fault mode component indicates a mode component of the mechanical device, wherein the operation of the mechanical device is indicated as a fault mode.
3. The method of claim 2, wherein said identifying whether the operating mode of the corresponding machine is a failure mode based on the plurality of indicator values for each mode component comprises:
if the index values corresponding to the first mode component are not within a preset boundary, determining that the index values of the first mode component identify that the working mode of the corresponding mechanical equipment is a fault mode, wherein the preset boundary is a distribution range of the index values corresponding to the working mode of the mechanical equipment when the working mode of the mechanical equipment is in a non-fault mode, and the first mode component is any one of the mode components.
4. The method of claim 2 or 3, wherein the plurality of indicators for each mode component comprises at least one of: kurtosis, relative kurtosis, including harmonic-to-noise ratio, negative entropy, gini exponent, variance, or standard deviation.
5. The method according to any one of claims 2-4, further comprising:
and determining the fault type corresponding to each fault mode component in the at least one fault mode component according to a plurality of index values corresponding to the at least one fault mode component.
6. The method according to claim 5, wherein the determining the fault type corresponding to each fault mode component in the at least one fault mode component according to a plurality of index values corresponding to the at least one fault mode component comprises:
taking a plurality of index values corresponding to the at least one fault mode component as input of a fault type identification model, and outputting a fault type corresponding to each fault mode component in the at least one fault mode component, wherein the fault type identification model is obtained by training by using a first training set, the first training set comprises historical data of the at least one mechanical device, and the historical data comprises data of the at least one mechanical device generated when the at least one mechanical device works in a non-fault mode and the fault type obtained when the at least one mechanical device works in a fault mode;
or comparing a plurality of index values corresponding to the at least one fault mode component with corresponding threshold values, and determining the fault type corresponding to each fault mode component in the at least one fault mode component.
7. The method of claim 6, further comprising:
training the fault type recognition model by using the first training set to obtain the trained fault type recognition model, wherein in the process of training the fault type recognition model, samples in the first training set are used as the input of the fault type recognition model, and the fault type recognition model is used for carrying out fault classification on samples corresponding to multiple types of faults and calculating the classification center of each fault classification;
the outputting the fault type corresponding to each fault mode component in the at least one fault mode component by taking the index values corresponding to the at least one fault mode component as the input of the fault type identification model comprises:
extracting features from a plurality of index values corresponding to the at least one fault mode component through the fault type identification model;
obtaining the distance between the feature corresponding to each fault mode component and the classification center of each fault classification through the fault type identification model;
and determining the fault type corresponding to each fault mode component according to the distance between the characteristic corresponding to each fault mode component and the classification center of each fault classification.
8. The method according to any one of claims 2-7, further comprising:
and determining a fault magnitude corresponding to each fault mode component in the at least one fault mode component according to a plurality of index values corresponding to the at least one fault mode component, wherein the fault magnitude represents the fault degree of mechanical equipment corresponding to each fault mode component in the at least one fault mode component.
9. The method according to claim 8, wherein the determining the fault magnitude corresponding to each fault mode component in the at least one fault mode component according to a plurality of index values corresponding to the at least one fault mode component comprises:
taking a plurality of index values corresponding to the at least one fault mode component as input of a fault magnitude recognition model, and outputting a fault magnitude corresponding to each fault mode component in the at least one fault mode component, where the fault magnitude recognition model is obtained by using a second training set for training, the second training set includes historical data of the at least one mechanical device, and the historical data includes data of the at least one mechanical device generated when the at least one mechanical device works in a non-fault mode and the fault magnitude obtained when the at least one mechanical device works in a fault mode;
or comparing a plurality of index values corresponding to the at least one fault mode component with ranges corresponding to a plurality of fault magnitude levels, and determining the fault magnitude level corresponding to each fault mode component in the at least one fault mode component.
10. The method of claim 9, further comprising:
training the fault magnitude recognition model by using the second training set to obtain the trained fault magnitude recognition model, wherein in the process of training the fault magnitude recognition model, samples in the second training set are used as the input of the fault type recognition model, and the fault type recognition model is used for classifying magnitudes corresponding to multiple types of faults and calculating the classification center of each fault magnitude;
the outputting the fault magnitude corresponding to each fault mode component in the at least one fault mode component by taking the index values corresponding to the at least one fault mode component as the input of the fault magnitude recognition model comprises:
extracting features from a plurality of index values corresponding to the at least one fault mode component through the fault magnitude recognition model;
obtaining the distance between the feature corresponding to each fault mode component and the classification center of each fault magnitude through the fault magnitude recognition model;
and determining the fault magnitude corresponding to each fault mode component according to the distance between the characteristic corresponding to each fault mode component and the classification center of each fault magnitude.
11. The method according to any one of claims 2-10, wherein the plurality of indicators for each mode component comprise a root mean square RMS, and before the identifying whether the operating mode of the corresponding mechanical device is a failure mode according to the plurality of indicator values for each mode component, the method further comprises:
screening out at least one mode component with a change value of the RMS within a preset time period smaller than a preset value from the multiple mode components;
and identifying whether the working mode of the corresponding mechanical equipment is a failure mode or not according to the index values of the at least one mode component.
12. The method according to any one of claims 1-11, wherein said demodulating said vibration information resulting in a plurality of mode components comprises:
demodulating the vibration information to obtain a demodulation signal;
decomposing the demodulated signal to obtain the plurality of decomposed signals;
calculating an envelope signal and an envelope spectrum signal for each of the plurality of decomposed signals;
and identifying whether each decomposed signal is a mode component or not according to the envelope signal and the envelope spectrum signal of each decomposed signal and an objective function so as to obtain the multiple mode components.
13. The method of claim 12, wherein said determining whether each of said decomposed signals is a mode component based on said envelope signal and said envelope spectrum signal of said each of said decomposed signals and an objective function comprises:
acquiring a first frequency according to a period in the envelope signal;
acquiring a second frequency according to the harmonic wave of the envelope spectrum signal;
constructing the objective function from a difference between the first frequency and the second frequency;
determining whether each of the decomposed signals is a mode component according to the value of the objective function.
14. The method according to any one of claims 1-13, wherein the obtaining vibration information comprises:
the method comprises the steps that information generated by sound of at least one piece of mechanical equipment or information generated by vibration of at least one piece of mechanical equipment is collected through a fiber optic distributed acoustic sensor DAS, wherein an optical fiber used for collecting data is deployed outside the at least one piece of mechanical equipment, when the vibration information comprises the information generated by the sound of the at least one piece of mechanical equipment during working, a sound sensitizing structure is wrapped outside the optical fiber and used for increasing the sensing strength of the optical fiber on the sound, and when the vibration information comprises the information generated by the vibration of the at least one piece of mechanical equipment during working, the at least one piece of mechanical equipment is connected with the optical fiber through a connecting piece.
15. The method of any one of claims 1-14, wherein the at least one mechanical device comprises at least one idler.
16. A rotary machine equipment failure detection apparatus, characterized in that the communication processing means includes: a processor coupled with a memory;
the memory for storing a computer program;
the processor configured to execute the computer program stored in the memory to cause the rotating machine equipment failure detection apparatus to perform the method of any one of claims 1 to 15.
17. The apparatus of claim 16, wherein the rotating machinery fault detection apparatus further comprises a vibration sensor for collecting information generated by vibration of the machinery.
18. A computer program product comprising instructions which, when run on a computer, cause the computer to perform the method of any one of claims 1 to 15.
19. A computer-readable storage medium comprising instructions that, when executed on a computer, cause the computer to perform the method of any of claims 1 to 15.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211490528.2A CN115855463A (en) | 2022-11-25 | 2022-11-25 | Rotary mechanical equipment fault detection method and device |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211490528.2A CN115855463A (en) | 2022-11-25 | 2022-11-25 | Rotary mechanical equipment fault detection method and device |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN115855463A true CN115855463A (en) | 2023-03-28 |
Family
ID=85666511
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202211490528.2A Pending CN115855463A (en) | 2022-11-25 | 2022-11-25 | Rotary mechanical equipment fault detection method and device |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN115855463A (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116608908A (en) * | 2023-07-21 | 2023-08-18 | 四川省华盾防务科技股份有限公司 | Fast-jump frequency source abnormity monitoring system and method |
| CN116793653A (en) * | 2023-06-21 | 2023-09-22 | 北京谛声科技有限责任公司 | Acoustic signal-based full life cycle monitoring method and system for rotating equipment |
-
2022
- 2022-11-25 CN CN202211490528.2A patent/CN115855463A/en active Pending
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116793653A (en) * | 2023-06-21 | 2023-09-22 | 北京谛声科技有限责任公司 | Acoustic signal-based full life cycle monitoring method and system for rotating equipment |
| CN116793653B (en) * | 2023-06-21 | 2024-02-27 | 北京谛声科技有限责任公司 | Acoustic signal-based full life cycle monitoring method and system for rotating equipment |
| CN116608908A (en) * | 2023-07-21 | 2023-08-18 | 四川省华盾防务科技股份有限公司 | Fast-jump frequency source abnormity monitoring system and method |
| CN116608908B (en) * | 2023-07-21 | 2023-10-27 | 四川省华盾防务科技股份有限公司 | Fast-jump frequency source abnormity monitoring system and method |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN115855463A (en) | Rotary mechanical equipment fault detection method and device | |
| Cui et al. | Quantitative trend fault diagnosis of a rolling bearing based on Sparsogram and Lempel-Ziv | |
| US8380447B2 (en) | Quantification of condition indicators in the presence of synchronous noise | |
| CN109397703B (en) | Fault detection method and device | |
| CN108363952A (en) | Diagnostic device | |
| Neupane et al. | Bearing fault detection using scalogram and switchable normalization-based CNN (SN-CNN) | |
| US20170052060A1 (en) | Method and system for automatically detecting faults in a rotating shaft | |
| JP2017227624A (en) | Method and apparatus for acoustic emissions testing | |
| US20040199368A1 (en) | Poor data quality identification | |
| US11521435B2 (en) | Method and device for diagnosing problematic noise source based on big data information | |
| CN106537111A (en) | Machine component diagnosis system and server thereof | |
| US10677765B2 (en) | Structural health monitoring of cyclically loaded structures | |
| CN110056640B (en) | Speed reducer wireless fault diagnosis method based on acceleration signal and edge calculation | |
| CN111665066B (en) | Equipment fault self-adaptive upper and lower early warning boundary generation method based on convolutional neural network | |
| CN110398384A (en) | The machine health monitoring of rotating machinery | |
| GB2277151A (en) | Machine monitoring using neural network | |
| CN110223489A (en) | A kind of monitoring method and device of engineering object | |
| CN111678698A (en) | Rolling bearing fault detection method based on sound and vibration signal fusion | |
| Harris | A Kohonen SOM based, machine health monitoring system which enables diagnosis of faults not seen in the training set | |
| Zhao et al. | Rolling bearing composite fault diagnosis method based on EEMD fusion feature | |
| Kim et al. | Failure diagnosis system using a new nonlinear mapping augmentation approach for deep learning algorithm | |
| CN110319957B (en) | Fault diagnosis method for irregular abnormal value of sensor of ship structure stress monitoring system | |
| CN109990803A (en) | The method, apparatus of method, apparatus and the sensor processing of detection system exception | |
| TW202045898A (en) | Analysis apparatus, analysis method, and non-transitory computer readable storage medium | |
| Tastimur et al. | Defect diagnosis of rolling element bearing using deep learning |
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 |