CN109991314B - Mechanical sealing state judgment method and device based on machine learning - Google Patents

Mechanical sealing state judgment method and device based on machine learning Download PDF

Info

Publication number
CN109991314B
CN109991314B CN201910180532.0A CN201910180532A CN109991314B CN 109991314 B CN109991314 B CN 109991314B CN 201910180532 A CN201910180532 A CN 201910180532A CN 109991314 B CN109991314 B CN 109991314B
Authority
CN
China
Prior art keywords
sampling period
sample data
acoustic emission
emission signal
parameter values
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Active
Application number
CN201910180532.0A
Other languages
Chinese (zh)
Other versions
CN109991314A (en
Inventor
黄伟峰
刘向锋
尹源
王玉明
刘莹
李永健
王子羲
贾晓红
郭飞
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Tsinghua University
Original Assignee
Tsinghua University
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by Tsinghua University filed Critical Tsinghua University
Priority to CN201910180532.0A priority Critical patent/CN109991314B/en
Publication of CN109991314A publication Critical patent/CN109991314A/en
Application granted granted Critical
Publication of CN109991314B publication Critical patent/CN109991314B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Classifications

    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
    • G01N29/00—Investigating or analysing materials by the use of ultrasonic, sonic or infrasonic waves; Visualisation of the interior of objects by transmitting ultrasonic or sonic waves through the object
    • G01N29/14—Investigating or analysing materials by the use of ultrasonic, sonic or infrasonic waves; Visualisation of the interior of objects by transmitting ultrasonic or sonic waves through the object using acoustic emission techniques
    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
    • G01N29/00—Investigating or analysing materials by the use of ultrasonic, sonic or infrasonic waves; Visualisation of the interior of objects by transmitting ultrasonic or sonic waves through the object
    • G01N29/44—Processing the detected response signal, e.g. electronic circuits specially adapted therefor
    • G01N29/4409—Processing the detected response signal, e.g. electronic circuits specially adapted therefor by comparison
    • G01N29/4436—Processing the detected response signal, e.g. electronic circuits specially adapted therefor by comparison with a reference signal
    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
    • G01N29/00—Investigating or analysing materials by the use of ultrasonic, sonic or infrasonic waves; Visualisation of the interior of objects by transmitting ultrasonic or sonic waves through the object
    • G01N29/44—Processing the detected response signal, e.g. electronic circuits specially adapted therefor
    • G01N29/4454—Signal recognition, e.g. specific values or portions, signal events, signatures

Landscapes

  • Physics & Mathematics (AREA)
  • Biochemistry (AREA)
  • General Physics & Mathematics (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Chemical & Material Sciences (AREA)
  • Analytical Chemistry (AREA)
  • Pathology (AREA)
  • General Health & Medical Sciences (AREA)
  • Health & Medical Sciences (AREA)
  • Immunology (AREA)
  • Engineering & Computer Science (AREA)
  • Signal Processing (AREA)
  • Acoustics & Sound (AREA)
  • Examining Or Testing Airtightness (AREA)
  • Testing Of Devices, Machine Parts, Or Other Structures Thereof (AREA)

Abstract

The application relates to a mechanical seal state judgment method and device based on machine learning, computer equipment and a storage medium. The method comprises the following steps: acquiring a research target of a mechanical sealing process, and selecting a sample acquisition mode according to the research target of the mechanical sealing process; determining sample data according to the sample acquisition mode, wherein the sample data at least comprises a process of acquiring an acoustic emission signal according to the sample acquisition mode and extracting a characteristic quantity of the acoustic emission signal; and obtaining a target parameter value according to the sample data. By adopting the method, the efficiency of judging the sealing state can be improved, the manual participation is avoided, and the automatic judgment is realized.

Description

Mechanical sealing state judgment method and device based on machine learning
Technical Field
The present disclosure relates to the field of mechanical sealing technologies, and in particular, to a method and an apparatus for determining a mechanical sealing state based on machine learning, a computer device, and a storage medium.
Background
A mechanical seal is a dynamic seal for shaft ends that needs to reduce or eliminate the frictional wear of the friction pair (formed by the two end faces that move relative to each other and the fluid medium) while maintaining low or no leakage for extended life.
With the development of mechanical seal research technology, the acoustic emission technology is used for collecting acoustic emission signals generated by a mechanical seal friction pair, and the technology has wide application prospect due to the fact that high-density information can be obtained and engineering convenience is achieved.
However, interpreting more information about the sealing state from the acoustic emission signals remains a challenge: on the one hand, the current technology can roughly judge the severity of contact friction, and no good solution exists as to how to further judge the cause of contact friction abnormality and how to obtain information about the sealing state when the contact is released. On the other hand, the analysis of the acoustic emission signals of the sealing friction pairs at present depends on human involvement and cannot be automated.
Disclosure of Invention
In view of the above, it is necessary to provide a method and an apparatus for determining a mechanical seal state based on machine learning, a computer device, and a storage medium.
A machine learning-based mechanical seal state determination method, the method comprising:
acquiring a research target of a mechanical sealing process, and selecting a sample acquisition mode according to the research target of the mechanical sealing process;
determining sample data according to the sample acquisition mode, wherein the sample data at least comprises a process of acquiring an acoustic emission signal according to the sample acquisition mode and extracting a characteristic quantity of the acoustic emission signal;
and obtaining a target parameter value according to the sample data.
In one embodiment, the determining sample data according to the sample obtaining manner includes:
acquiring at least one test parameter value within preset time in a test process, processing the at least one test parameter value, and determining the test parameter value of at least one sampling period within the preset time, wherein the test parameter value at least comprises an acoustic emission signal and an auxiliary parameter value;
processing the acoustic emission signal of at least one sampling period in the preset time to obtain a characteristic quantity of the acoustic emission signal;
and processing the auxiliary parameter values of at least one sampling period in the preset time, and determining the average value of the at least one auxiliary parameter value in the at least one sampling period.
In one embodiment, the processing the acoustic emission signal of at least one sampling period within the preset time to obtain the characteristic quantity of the acoustic emission signal includes:
acquiring an acoustic emission signal of at least one sampling period in the preset time, and performing periodic secondary division on the acoustic emission signal of at least one sampling period to obtain the acoustic emission signal of at least one sampling period after periodic secondary division;
determining a signal energy value of at least one sampling period after the periodic secondary division according to the acoustic emission signal of at least one sampling period after the periodic secondary division;
and sequencing the signal energy values according to time, performing feature extraction on the sequenced signal energy values, and determining at least one characteristic quantity of at least one sampling period after the period is divided twice.
In one embodiment, the processing the auxiliary parameter values for at least one sampling period within the preset time, and determining an average value of the at least one auxiliary parameter value for the at least one sampling period includes:
acquiring an output parameter value of at least one sampling period within a preset time in the test process;
and summarizing the average value of the auxiliary parameter values of the at least one sampling period, the output parameter value of the at least one sampling period and the characteristic quantity of the transmission signal of the at least one sampling period, and taking the average value of the auxiliary parameter values of the at least one sampling period, the output parameter value of the at least one sampling period and the characteristic quantity of the transmission signal of the at least one sampling period as sample data.
In one embodiment, the determining sample data according to the sample obtaining manner further includes:
obtaining a test parameter value of at least one physical process in a test process, and matching the test parameter value of at least one physical process in the test process with a standard parameter value of at least one physical process in the test process;
and selecting the test parameter value of the physical process with the maximum correlation between the test parameter value of the at least one physical process in the test process and the standard parameter value of the at least one physical process in the test process.
In one embodiment, the selecting the test parameter value of the physical process having the largest correlation between the test parameter value of the at least one physical process in the test process and the standard parameter value of the at least one physical process in the test process includes:
processing the test parameter value of the physical process with the maximum correlation degree to obtain an acoustic emission signal of at least one sampling period;
determining a characteristic quantity of the acoustic emission signal of the at least one sampling period according to the acoustic emission signal of the at least one sampling period;
and acquiring the average value of the auxiliary parameter values of at least one sampling period and the output parameter value of at least one sampling period, and taking the average value of the auxiliary parameter values of at least one sampling period, the output parameter value of at least one sampling period and the characteristic quantity of the transmission signal of at least one sampling period as sample data.
In one embodiment, the determining the characteristic quantity of the acoustic emission signal of the at least one sampling period according to the acoustic emission signal of the at least one sampling period includes:
extracting signal energy values of the acoustic emission signals in the vicinity of at least one frequency band, and sorting the signal energy values by time;
and carrying out cycle division on the sequenced signal energy values for at least one time to obtain the signal energy value of at least one cycle, and taking the signal energy value of at least one cycle as a characterization quantity.
In one embodiment, the obtaining the target parameter according to the sample data includes:
acquiring the sample data, and determining a trained model according to the sample data and the initial model;
and acquiring the average value of the auxiliary parameter values of at least one sampling period and the characteristic quantity of the transmitting signal of at least one sampling period in a preset time in the using process, inputting the average value of the auxiliary parameter values of at least one sampling period and the characteristic quantity of the transmitting signal of at least one sampling period into the trained model, and determining a target parameter value.
In one embodiment, the inputting the average value of the auxiliary parameter values of the at least one sampling period and the characteristic quantity of the transmission signal of the at least one sampling period into the trained model, and the determining the target parameter value includes:
inputting the average value of the auxiliary parameter values of the at least one sampling period and the characteristic quantity of the transmitting signal of the at least one sampling period into the trained model to obtain target parameter values corresponding to different sample data, and calculating the standard deviation of the target parameter values corresponding to the different sample data;
the method comprises the steps of obtaining an average value of target parameter values of at least one sampling period within preset time in a test process, and determining a standard deviation corresponding to the average value of the target parameter values of the at least one sampling period according to the average value of the target parameter values of the at least one sampling period.
In one embodiment, the obtaining an average value of the target parameter values of at least one sampling period within a preset time in the test process, and determining a standard deviation corresponding to the average value of the target parameter values of the at least one sampling period according to the average value of the target parameter values of the at least one sampling period includes:
and if the standard deviations of the target parameter values corresponding to the different sample data are all smaller than the standard deviation corresponding to the average value of the target parameter values of at least one sampling period, and the standard deviations of the target parameter values corresponding to the different sample data have no statistical correlation, calculating the average value of the target parameter values corresponding to the different sample data, and taking the average value of the target parameter values corresponding to the different sample data as the target parameter value.
A machine learning-based mechanical seal state judgment device, the device comprising:
the first acquisition module is used for acquiring a research target of a mechanical sealing process and selecting a sample acquisition mode according to the research target of the mechanical sealing process;
the second acquisition module is used for determining sample data according to the sample acquisition mode, wherein the sample data at least comprises a process of acquiring an acoustic emission signal according to the sample acquisition mode and extracting a characteristic quantity of the acoustic emission signal;
and the determining module is used for obtaining a target parameter value according to the sample data.
A computer device comprising a memory storing a computer program and a processor implementing the steps of the method as claimed in any one of the above when the computer program is executed.
A computer-readable storage medium, on which a computer program is stored, which, when being executed by a processor, carries out the steps of the method according to any one of the preceding claims.
According to the mechanical sealing state judgment method and device based on machine learning, the computer equipment and the storage medium, a sample acquisition mode is selected by acquiring a research target of a mechanical sealing process and according to the research target of the mechanical sealing process; determining sample data according to the sample acquisition mode, wherein the sample data at least comprises a process of acquiring an acoustic emission signal according to the sample acquisition mode and extracting a characteristic quantity of the acoustic emission signal; and obtaining a target parameter value according to the sample data. By the method, the efficiency of judging the sealing state can be improved, manual participation is avoided, and automatic judgment is realized.
Drawings
Fig. 1 is an application environment diagram of a mechanical seal state determination method based on machine learning according to an embodiment;
FIG. 2 is a flow chart illustrating a method for determining a mechanical seal status based on machine learning according to an embodiment;
FIG. 3 is a flowchart illustrating step S2 according to an embodiment;
FIG. 4 is a schematic flow chart of a mechanical seal test process according to another embodiment;
FIG. 5 is a flowchart illustrating step S22 according to an embodiment;
FIG. 6 is a flowchart illustrating steps subsequent to step S23 in one embodiment;
FIG. 7 is a sample data structure diagram in accordance with an embodiment;
FIG. 8 is a flowchart illustrating steps subsequent to step S27 in one embodiment;
FIG. 9 is a flowchart illustrating step S29 according to an exemplary embodiment;
FIG. 10 is a flowchart illustrating step S3 according to an exemplary embodiment;
FIG. 11 is a flowchart illustrating step S32 according to an exemplary embodiment;
fig. 12 is a block diagram showing a structure of a mechanical seal state judgment device based on machine learning according to an embodiment;
FIG. 13 is a diagram illustrating an internal structure of a computer device according to an embodiment.
Detailed Description
In order to make the objects, technical solutions and advantages of the present application more apparent, the present application is described in further detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the present application and are not intended to limit the present application.
The mechanical sealing state judgment method based on machine learning can be applied to the application environment shown in fig. 1. Wherein the terminal 102 communicates with the server 104 via a network. The terminal 102 acquires a research target of the mechanical sealing process and transmits the research target to the server 104 through the network. The server 104 acquires a research target of a mechanical sealing process, and selects a sample acquisition mode according to the research target of the mechanical sealing process; determining sample data according to the sample acquisition mode, wherein the sample data at least comprises a process of acquiring an acoustic emission signal according to the sample acquisition mode and extracting a characteristic quantity of the acoustic emission signal; and obtaining a target parameter value according to the sample data. The terminal 102 may be, but not limited to, various personal computers, notebook computers, smart phones, tablet computers, and portable wearable devices, and the server 104 may be implemented by an independent server or a server cluster formed by a plurality of servers.
In one embodiment, as shown in fig. 2, a method for determining a mechanical seal state based on machine learning is provided, which is described by taking the method as an example applied to the server in fig. 1, and includes the following steps:
step S1: and acquiring a research target of the mechanical sealing process, and selecting a sample acquisition mode according to the research target of the mechanical sealing process.
Specifically, the mechanical seal is a shaft seal device of a rotating machine (such as a centrifugal pump, a centrifuge, a reaction kettle, a compressor and the like). Since the drive shaft extends through the inside and outside of the apparatus, there is a circumferential gap between the shaft and the apparatus through which the medium in the apparatus leaks out, and if the pressure in the apparatus is lower than atmospheric pressure, air leaks into the apparatus, so that a shaft seal device must be provided to prevent leakage. The shaft seal has many kinds, and the mechanical seal has the advantages of small leakage amount, long service life and the like, so the mechanical seal is the most important shaft sealing mode in the devices in the world. Mechanical seals, also called face seals, are defined in the relevant national standards as follows: "means for preventing fluid leakage, which is composed of at least one pair of end faces perpendicular to the rotation axis, and which are kept in fit and relatively slide under the action of fluid pressure and the elastic force (or magnetic force) of the compensation mechanism and the cooperation of the auxiliary seal. "
Depending on the type of seal, the objectives of the study are different in studying different types of seals. In one embodiment, the end jump of a moving ring of an aqueous medium contact type static ring compensation type flat end surface mechanical seal (working rotating speed of 6000rpm) is a main factor influencing the sealing performance, so the end jump is taken as a research target of the sealing process of the type. In another embodiment, a dry gas seal with a spiral groove (a static ring is of a compensation type, the moving ring is provided with a spiral groove, the working speed is 3000rpm), the problem that the leakage amount is too large is often found in use, slight intermittent contact (the seal is non-contact seal and should not be contacted in the stable working process) is found after acoustic emission monitoring is carried out on the dry gas seal, and by combining the two phenomena, preliminary judgment is possibly related to three problems of uneven installation of the static ring, excessive jump of the moving ring end and wave intensity of the static ring end face, so that overturning moment (uneven quantification of installation of the static ring), jump of the moving ring end face, double-peak wave intensity amplitude of the static ring end face and double-peak wave intensity phase of the static ring end face are all taken as research targets of the sealing process.
For different seal types and research objectives, it is known that the manner in which samples are taken varies for different research objectives. For example, in the process of mechanical sealing of an aqueous medium contact type static ring compensation type flat end face, because the end jump of a dynamic ring is a main factor influencing the sealing performance, but real-time measurement cannot be carried out on an actual machine, the seal is arranged on a special test bed to carry out real-time measurement and adjustment on the end jump of the dynamic ring; in a spiral groove dry gas seal process, the sample in this example is taken during the start and stop of the seal. Because the start-stop stage of the spiral groove dry gas seal operation is actually the process that the rotating speed is lower than the value of stable operation and the rotating speed changes, the lubricating of the spiral groove dry gas seal has stronger dependence on the dynamic pressure effect, and stronger contact friction can occur when the rotating speed is lower. This process is inherently unavoidable in the operation of spiral groove dry gas sealing, and when a sample is collected from this process, not only the contact is relatively strong (so that the acoustic emission signal carries rich information), but also a large number of differential samples can be collected in a short time because the speed is in rapid change.
The method for obtaining a sample according to the different embodiments described above can rapidly obtain a large number of samples having differences in operation with as few times as possible and in as short a time as possible.
Step S2: and determining sample data according to the sample acquisition mode, wherein the sample data at least comprises a process of acquiring an acoustic emission signal according to the sample acquisition mode and extracting a characteristic quantity of the acoustic emission signal.
Specifically, during operation of the mechanical seal, its friction pair generates acoustic emission signals, known to include those generated by leakage (if any) and those generated by contact friction (if any). By suitably mounting the acoustic emission sensor on the mechanical seal, the above-mentioned signal from the friction pair can be measured. This signal exhibits the following multi-scale characteristics:
(1) acoustic dimensions. The typical frequency is about 20KHZ to 1000KHZ, and under the relative position relation of the sealing rings at the moment, the distribution of leakage along the outlet of the friction pair, the form of contact friction, the distribution in the friction pair and the like can influence the characteristics of the acoustic emission signals on the scale;
(2) and (4) a dynamic scale. The typical frequency of which is related to the sealing rotational speed. Under the interaction of the periodic excitation of the rotation of the rotating shaft and the circumferential nonuniformity in the friction pair, the acoustic emission signal shows corresponding periodic variation on the scale;
(3) and (4) service scale. With the long-term service of the seal, the components in the mechanical seal, including the seal ring, undergo changes in performance, a characteristic on the service scale.
A characterizing form of the acoustic emission signal is established. It is represented by a "token vector" that carries the information contained in the acoustic emission signal. Although it is not excluded that the waveform sequence of the acoustic emission signal itself is directly taken as a characterization vector-in fact in some cases such a process is suitable, in other cases compressing its key information in a vector of lower dimensions may improve the performance of the method.
There are many parameters (the "parameter" in the present invention is a broad quantity: it can be presented in the form of a single scalar quantity, or in other various data structures such as time series) related to the working state of the mechanical seal, and some of the parameters cannot be directly measured in real time in a real machine. However, if one such model is built for a parameter of interest (called the target parameter) that cannot be directly measured in real time: indirect inference can be achieved by inputting directly measurable parameters and outputting an estimate of the target parameter. However, it is only possible to establish an effective model if the input parameters have to contain sufficiently rich information that is closely related to the operating state of the mechanical seal. The acoustic emission signal can be a major (or even the only) component of the input parameter, with the rich information it contains.
Step S3: and obtaining a target parameter value according to the sample data.
In particular, since the mechanism of operation of the mechanical seal and the mechanism of generation of the acoustic emission signal are extremely complex, it has not been possible to establish an inference method from the viewpoint of physical principles. The invention adopts another approach, namely, a machine learning method is adopted to train a model with the functions through data samples to obtain target parameter values.
According to the mechanical sealing state judgment method and device based on machine learning, the computer equipment and the storage medium, a sample acquisition mode is selected by acquiring a research target of a mechanical sealing process and according to the research target of the mechanical sealing process; determining sample data according to the sample acquisition mode, wherein the sample data at least comprises a process of acquiring an acoustic emission signal according to the sample acquisition mode and extracting a characteristic quantity of the acoustic emission signal; and obtaining a target parameter value according to the sample data. By the method, the efficiency of judging the sealing state can be improved, manual participation is avoided, and automatic judgment is realized.
In one embodiment, as shown in fig. 3, the step S2 includes:
step S21: acquiring at least one test parameter value within preset time in a test process, processing the at least one test parameter value, and determining the test parameter value of at least one sampling period within the preset time, wherein the test parameter value at least comprises an acoustic emission signal and an auxiliary parameter value;
step S22: processing the acoustic emission signal of at least one sampling period in the preset time to obtain a characteristic quantity of the acoustic emission signal;
step S23: and processing the auxiliary parameter values of at least one sampling period in the preset time, and determining the average value of the at least one auxiliary parameter value in the at least one sampling period.
Specifically, as shown in fig. 4, the test parameter values refer to real-time parameter values obtained during the test. The auxiliary parameter values refer to parameter values obtained during the test in addition to the acoustic emission signal. For example, in the process of researching an aqueous medium contact static ring compensation type flat end surface mechanical seal, the seal is arranged on a special test bed to carry out real-time measurement and adjustment on the end jump of a moving ring, and available auxiliary parameters comprise the rotating speed, the upstream pressure and the downstream pressure of the seal. And (3) operating and sealing at different rotating speeds, upstream pressure, downstream pressure and end jump by adopting an orthogonal test method, and carrying out a training sample acquisition test.
The preset time refers to the time during which the fluid remains in operation after each adjustment of the target parameter value during the test. The preset time can be 3s, 5s or 7 s. The sampling period refers to adjacent time periods within a preset time, and each sampling period is equal, wherein the sampling period may be 0.1, 0.2, 0.5, or the like. If the preset time is set to be 5s and the sampling period is set to be 0.5s, then the average value omega of the rotating speed and the average value p of the upstream pressure of the primary seal are obtained in each sampling period1And the mean value p of the downstream pressure2Then 10 training samples are obtained.
In one embodiment, as shown in fig. 5, the step S22 includes:
step S221: acquiring an acoustic emission signal of at least one sampling period in the preset time, and performing periodic secondary division on the acoustic emission signal of at least one sampling period to obtain the acoustic emission signal of at least one sampling period after periodic secondary division;
step S222: determining a signal energy value of at least one sampling period after the periodic secondary division according to the acoustic emission signal of at least one sampling period after the periodic secondary division;
step S223: and sequencing the signal energy values according to time, performing feature extraction on the sequenced signal energy values, and determining at least one characteristic quantity of at least one sampling period after the period is divided twice.
Specifically, the secondary division of the sampling period refers to re-dividing each sampling period into a plurality of sub-periods with equal time, and if the sampling period is 0.5s, the sampling period is further divided into 2500 sub-periods with length of 0.2ms, or 5000 sub-periods with length of 0.1ms, and the like.
In one embodiment, the processing of the acoustic emission signals is as follows: assuming that the preset time is 5s and the sampling period is 0.5s, the sampled 0.5s period is further divided into 2500 short-time waveforms with the length of 0.2ms, the RMS value (which represents the energy of the signal) of each short-time waveform is obtained, the 2500 RMS values are arranged into a time sequence according to the original sequence, and the following 3 characteristic quantities are extracted from the sequence:
(1) average value a0；
(2) The component on the frequency corresponding to the rotating speed of the rotating shaft obtained by Fourier decomposition is taken as the amplitude a1；
(3) Excluding the above-mentioned average value and the frequency component corresponding to the rotation speed of the rotating shaft, the statistical standard deviation sigma of the remaining partr。
In one embodiment, as shown in fig. 6, the step S23 is followed by:
step S24: acquiring an output parameter value of at least one sampling period within a preset time in the test process;
step S25: and summarizing the average value of the auxiliary parameter values of the at least one sampling period, the output parameter value of the at least one sampling period and the characteristic quantity of the transmission signal of the at least one sampling period, and taking the average value of the auxiliary parameter values of the at least one sampling period, the output parameter value of the at least one sampling period and the characteristic quantity of the transmission signal of the at least one sampling period as sample data.
Specifically, as shown in fig. 7, the output parameter value refers to a target parameter value. Wherein the output parameter is hγAnd (4) showing. General reference to the aboveObtaining the number and the characterization vector of the transmitted signal, obtaining 2560 sample data, each sample being mapped to a 7-dimensional vector [ omega, p ] according to the method1、p2、a0、a1、σr、hγ]。
In one embodiment, the step S2 further includes:
step S26: obtaining a test parameter value of at least one physical process in a test process, and matching the test parameter value of at least one physical process in the test process with a standard parameter value of at least one physical process in the test process;
step S27: and selecting the test parameter value of the physical process with the maximum correlation between the test parameter value of the at least one physical process in the test process and the standard parameter value of the at least one physical process in the test process.
Specifically, in one embodiment, a spiral groove dry gas seal development process includes a number of physical phases, such as a start-up phase and a stop phase. The process of time variation of various parameters is different for different physical phases. Taking the starting stage as an example, the spiral groove dry gas sealing starting process shows the following characteristics due to the dependence on the dynamic pressure effect: first rising sharply to a large height (called a sharp friction step) and then falling sharply (called a steady shift step). The stop phase exhibits characteristics similar to the start phase (except in reverse order: first through a steady shift phase and then through a severe friction phase). The acoustic emission signal is relatively stable during the steady speed change period, so that the sample is only collected during this period in the training and guessing.
And identifying and dividing start-stop stages. The following method is adopted for automatic stage division: several start phase signals were manually marked at 3 time points: starting begins, a severe friction section and a stable speed section are started, and the starting process is ended (the starting process is ended based on the fact that the acoustic emission signal does not drop down remarkably along with time). A DTW (dynamic time warping) algorithm is adopted to establish a correspondence between the marked starting process and the collected starting process (whether training or guessing), and 3 time points are mapped to the newly obtained signals, so that the newly obtained signals are automatically divided, as shown in fig. 3. In the subsequent steps, only the stable speed change section between two mark points of 'starting two section boundary' and 'starting process end' is used.
Because the acoustic emission signal does not contact in the normal stable working process, when the leakage amount of the seal is abnormally increased, the contact still probably does not occur or is very slight, and meanwhile, the acoustic signal generated by the leakage is mixed with the noise of a shaft system and the like in the same frequency band and cannot be distinguished, so that the effective information contained in the acoustic emission signal measured in the stable working process is very little. To address this problem, the sample in this example is taken during seal start-stop. The start-stop stage of the dry gas seal operation of the spiral groove is actually the process that the rotating speed is lower than the value of stable operation and the rotating speed changes, and because the lubrication of the dry gas seal of the spiral groove has stronger dependence on the dynamic pressure effect, stronger contact friction can occur when the rotating speed is lower. This process is inherently unavoidable in the operation of spiral groove dry gas sealing, and when a sample is collected from this process, not only the contact is relatively strong (so that the acoustic emission signal carries rich information), but also a large number of differential samples can be collected in a short time because the speed is in rapid change.
The test parameter values are different from the standard parameter values, the test parameter values refer to real-time parameter values of each test process and change along with different tests, and the standard parameter values refer to standardized parameter values in the test processes obtained through a plurality of tests or based on principles.
In one embodiment, as shown in fig. 8, the step S27 is followed by:
step S28: processing the test parameter value of the physical process with the maximum correlation degree to obtain an acoustic emission signal of at least one sampling period;
step S29: determining a characteristic quantity of the acoustic emission signal of the at least one sampling period according to the acoustic emission signal of the at least one sampling period;
step S30: and acquiring the average value of the auxiliary parameter values of at least one sampling period and the output parameter value of at least one sampling period, and taking the average value of the auxiliary parameter values of at least one sampling period, the output parameter value of at least one sampling period and the characteristic quantity of the transmission signal of at least one sampling period as sample data.
Specifically, in steps S28-S30, the sampling periods of the output parameter value, the auxiliary parameter value, and the characterization quantity of the transmission signal are all the same.
In one embodiment, as shown in fig. 9, the step S29 includes:
step S291: extracting signal energy values of the acoustic emission signals in the vicinity of at least one frequency band, and sorting the signal energy values by time;
step S292: and carrying out cycle division on the sequenced signal energy values for at least one time to obtain the signal energy value of at least one cycle, and taking the signal energy value of at least one cycle as a characterization quantity.
Specifically, for a stable speed change section of starting or stopping, a time sequence of energy composition of acoustic emission on +/-30 kHz frequency bands near 155kHz, 280kHz and 490kHz respectively is taken (energy composition time sequence of each frequency band is calculated every 0.1ms time period), each period is taken as a sample (samples are selected in an overlapping mode) according to a period corresponding to the instant rotating speed, and acoustic emission signals in the time period uniformly take three frequency band energies at 30 moments, so that a 90-dimensional vector is obtained as an acoustic emission characteristic quantity.
The above is a method of extracting samples from start-stop processes. With the above parameters, thousands of samples can be taken for each start-up or stop process as a "test" as described above (the number is only an example, and varies significantly depending on the start and stop speeds).
In one embodiment, as shown in fig. 10, the step S3 includes:
step S31: acquiring the sample data, and determining a trained model according to the sample data and the initial model;
step S32: and acquiring the average value of the auxiliary parameter values of at least one sampling period and the characteristic quantity of the transmitting signal of at least one sampling period in a preset time in the using process, inputting the average value of the auxiliary parameter values of at least one sampling period and the characteristic quantity of the transmitting signal of at least one sampling period into the trained model, and determining a target parameter value.
Specifically, the sample data is directly input into an initial model, and the initial model is trained to obtain a trained model.
The auxiliary parameters are different corresponding to different research processes, for example, in the end jump process of researching the mechanical seal, the auxiliary parameters are the rotating speed, the upstream pressure and the downstream pressure; and for the research on the spiral dry gas sealing process, the auxiliary parameter is the rotating speed or other parameters.
The initial model refers to a support vector machine for machine learning or various artificial neural networks and the like. The trained model refers to a model obtained by inputting sample data into an initial model and training the initial model. Support Vector Machines (SVMs) are a mature and widely used method for modeling in a supervised learning (supervised learning) manner, and are conventionally used for classification (classification) of data and then further used for regression (regression). SVMs can be applied to nonlinear problems by kernel methods, which are one of the common kernel learning (kernel learning) methods.
In one embodiment, a programming implementation for regression model training using a support vector machine is provided as follows:
Figure BDA0001991185270000131
Figure BDA0001991185270000141
Figure BDA0001991185270000151
Figure BDA0001991185270000161
in one embodiment, as shown in fig. 11, the step S32 includes:
step S321: inputting the average value of the auxiliary parameter values of the at least one sampling period and the characteristic quantity of the transmitting signal of the at least one sampling period into the trained model to obtain target parameter values corresponding to different sample data, and calculating the standard deviation of the target parameter values corresponding to the different sample data;
step S322: the method comprises the steps of obtaining an average value of target parameter values of at least one sampling period within preset time in a test process, and determining a standard deviation corresponding to the average value of the target parameter values of the at least one sampling period according to the average value of the target parameter values of the at least one sampling period.
Specifically, in steps S321-S322, by analyzing the standard deviation of the obtained average value of the target parameter value and the average value of the target parameter value, the judgment accuracy is improved, and whether the method is functioning normally is judged by using the inference results of a plurality of samples in the same inference test, so that misjudgment is avoided without the user knowing.
In one embodiment, the step S322 is followed by:
step S323: and if the standard deviations of the target parameter values corresponding to the different sample data are all smaller than the standard deviation corresponding to the average value of the target parameter values of at least one sampling period, and the standard deviations of the target parameter values corresponding to the different sample data have no statistical correlation, calculating the average value of the target parameter values corresponding to the different sample data, and taking the average value of the target parameter values corresponding to the different sample data as the target parameter value.
Specifically, the sample is input into the trained model, and the output end jump is the end jump estimated by the sample. An inferential test is made up of samples taken over a short period of time.
If the distribution of the end hops estimated from each sample in the same inference test satisfies the following condition:
(1) is represented as normal distribution, and the standard deviation is less than h of training testγStandard deviation of (d);
(2) there is no statistical correlation. Specifically, let the estimated end-jump of each sample be
Figure BDA0001991185270000162
Then
Figure BDA0001991185270000163
And
Figure BDA0001991185270000164
are statistically independent.
The inference is deemed valid and the average of the end-hops estimated for each sample is taken as the end-hop value inferred for that test. In other cases, further manual analysis of the inference method is required.
It should be understood that, although the steps in the flowchart of fig. 2 are shown in order as indicated by the arrows, the steps are not necessarily performed in order as indicated by the arrows. The steps are not performed in the exact order shown and described, and may be performed in other orders, unless explicitly stated otherwise. Moreover, at least a portion of the steps in fig. 2 may include multiple sub-steps or multiple stages that are not necessarily performed at the same time, but may be performed at different times, and the order of performance of the sub-steps or stages is not necessarily sequential, but may be performed in turn or alternately with other steps or at least a portion of the sub-steps or stages of other steps.
In one embodiment, as shown in fig. 12, there is provided a machine learning-based mechanical seal state determination device including: a first obtaining module 10, a second obtaining module 20 and a determining module 30, wherein:
the first acquisition module 10 is used for acquiring a research target of a mechanical sealing process and selecting a sample acquisition mode according to the research target of the mechanical sealing process;
a second obtaining module 20, configured to determine sample data according to the sample obtaining manner, where the sample data at least includes a process of obtaining an acoustic emission signal according to the sample obtaining manner, and extracting a characteristic quantity of the acoustic emission signal;
and the determining module 30 is configured to obtain a target parameter value according to the sample data.
In one embodiment, the second obtaining module 20 includes:
the parameter acquiring module 201 is configured to acquire at least one test parameter value within a preset time in a test process, process the at least one test parameter value, and determine a test parameter value of at least one sampling period within the preset time, where the test parameter value at least includes an acoustic emission signal and an auxiliary parameter value;
the first characteristic quantity acquisition model 202 is used for processing the acoustic emission signal of at least one sampling period within the preset time to obtain the characteristic quantity of the acoustic emission signal;
an average value obtaining module 203 for the auxiliary parameter value, configured to process the auxiliary parameter value in at least one sampling period within the preset time, and determine an average value of the at least one auxiliary parameter value in the at least one sampling period.
In one embodiment, the token obtaining model 202 includes:
the periodic secondary division module 2021 is configured to obtain the acoustic emission signal of at least one sampling period within the preset time, and perform periodic secondary division on the acoustic emission signal of at least one sampling period to obtain an acoustic emission signal of at least one sampling period after the periodic secondary division;
the signal energy value acquiring module 2022 is configured to determine a signal energy value of at least one sampling period after the periodic secondary division according to the acoustic emission signal of at least one sampling period after the periodic secondary division;
the sorting module 2023 is configured to sort the signal energy values according to time, perform feature extraction on the sorted signal energy values, and determine at least one characterization quantity of at least one sampling period after the period is divided twice.
In one embodiment, the average value obtaining module 203 of the auxiliary parameter values comprises:
an output parameter value obtaining module 204, configured to obtain an output parameter value of at least one sampling period within a preset time in a test process;
the first sample data obtaining module 205 is configured to summarize the average value of the auxiliary parameter values of the at least one sampling period, the output parameter value of the at least one sampling period, and the characteristic quantity of the transmission signal of the at least one sampling period, and use the average value of the auxiliary parameter values of the at least one sampling period, the output parameter value of the at least one sampling period, and the characteristic quantity of the transmission signal of the at least one sampling period as sample data.
In one embodiment, the second obtaining module 20 further includes:
the matching module 206 is configured to obtain a test parameter value of at least one physical process in a test process, and match the test parameter value of the at least one physical process in the test process with a standard parameter value of the at least one physical process in the test process;
and the correlation calculation module 207 is configured to select a test parameter value of the physical process with the largest correlation between the test parameter value of the at least one physical process in the test process and the standard parameter value of the at least one physical process in the test process.
In one embodiment, the correlation calculation module 207 further includes:
the acoustic emission signal acquisition module 208 is configured to process the test parameter value of the physical process with the largest correlation, and acquire an acoustic emission signal of at least one sampling period;
a characteristic quantity determining module 209, configured to determine a characteristic quantity of the acoustic emission signal of the at least one sampling period according to the acoustic emission signal of the at least one sampling period;
the second sample data obtaining module 210 is configured to obtain an average value of the auxiliary parameter values of at least one sampling period and an output parameter value of at least one sampling period, and use the average value of the auxiliary parameter values of the at least one sampling period, the output parameter value of the at least one sampling period, and the characterization quantity of the transmission signal of the at least one sampling period as sample data.
In one embodiment, the token determination module 209 includes:
an extraction module 2091 configured to extract signal energy values of the acoustic emission signal around at least one frequency band, and sort the signal energy values by time;
the second token quantity obtaining module 2092 is configured to perform at least one cycle division on the sorted signal energy values, obtain a signal energy value of at least one cycle, and use the signal energy value of the at least one cycle as a token quantity.
In one embodiment, the determining module 30 includes:
a training module 301, configured to obtain the sample data, and determine a trained model according to the sample data and the initial model;
and a target parameter determining module 302, configured to obtain an average value of the auxiliary parameter values in at least one sampling period and a characteristic quantity of the transmission signal in at least one sampling period within a preset time in a use process, input the average value of the auxiliary parameter values in at least one sampling period and the characteristic quantity of the transmission signal in at least one sampling period into the trained model, and determine a target parameter value.
In one embodiment, the target parameter determination module 302 includes:
a calculating module 3021, configured to input the average value of the auxiliary parameter values in the at least one sampling period and the characterization quantity of the transmission signal in the at least one sampling period into the trained model, to obtain target parameter values corresponding to different sample data, and calculate a standard deviation of the target parameter values corresponding to the different sample data;
the standard deviation obtaining module 3022 is configured to obtain an average value of the target parameter value in at least one sampling period within a preset time in the test process, and determine a standard deviation corresponding to the average value of the target parameter value in the at least one sampling period according to the average value of the target parameter value in the at least one sampling period.
In one embodiment, the standard deviation obtaining module 3022 includes:
an analyzing module 3023, configured to calculate an average value of the target parameter values corresponding to the different sample data if the standard deviations of the target parameter values corresponding to the different sample data are all smaller than the standard deviation corresponding to the average value of the target parameter values in at least one sampling period, and there is no statistical correlation between the standard deviations of the target parameter values corresponding to the different sample data, and use the average value of the target parameter values corresponding to the different sample data as the target parameter value.
For specific limitations of a mechanical sealing state judgment device based on machine learning, refer to the above limitations of a mechanical sealing state judgment method based on machine learning, which are not described herein again. The modules in the mechanical sealing state judgment device based on machine learning can be wholly or partially realized by software, hardware and a combination thereof. The modules can be embedded in a hardware form or independent from a processor in the computer device, and can also be stored in a memory in the computer device in a software form, so that the processor can call and execute operations corresponding to the modules.
In one embodiment, a computer device is provided, which may be a server, and its internal structure diagram may be as shown in fig. 13. The computer device includes a processor, a memory, a network interface, and a database connected by a system bus. Wherein the processor of the computer device is configured to provide computing and control capabilities. The memory of the computer device comprises a nonvolatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program, and a database. The internal memory provides an environment for the operation of an operating system and computer programs in the non-volatile storage medium. The database of the computer device is used for storing mechanical sealing state judgment data based on machine learning. The network interface of the computer device is used for communicating with an external terminal through a network connection. The computer program is executed by a processor to implement a machine learning-based mechanical seal state determination method.
Those skilled in the art will appreciate that the architecture shown in fig. 13 is merely a block diagram of some of the structures associated with the disclosed aspects and is not intended to limit the computing devices to which the disclosed aspects apply, as particular computing devices may include more or less components than those shown, or may combine certain components, or have a different arrangement of components.
In one embodiment, a computer device is provided, comprising a memory and a processor, the memory having a computer program stored therein, the processor implementing the following steps when executing the computer program:
acquiring a research target of a mechanical sealing process, and selecting a sample acquisition mode according to the research target of the mechanical sealing process;
determining sample data according to the sample acquisition mode, wherein the sample data at least comprises a process of acquiring an acoustic emission signal according to the sample acquisition mode and extracting a characteristic quantity of the acoustic emission signal;
and obtaining a target parameter value according to the sample data.
In one embodiment, a computer-readable storage medium is provided, having a computer program stored thereon, which when executed by a processor, performs the steps of:
acquiring a research target of a mechanical sealing process, and selecting a sample acquisition mode according to the research target of the mechanical sealing process;
determining sample data according to the sample acquisition mode, wherein the sample data at least comprises a process of acquiring an acoustic emission signal according to the sample acquisition mode and extracting a characteristic quantity of the acoustic emission signal;
and obtaining a target parameter value according to the sample data.
It will be understood by those skilled in the art that all or part of the processes of the methods of the embodiments described above can be implemented by hardware instructions of a computer program, which can be stored in a non-volatile computer-readable storage medium, and when executed, can include the processes of the embodiments of the methods described above. Any reference to memory, storage, database, or other medium used in the embodiments provided herein may include non-volatile and/or volatile memory, among others. Non-volatile memory can include read-only memory (ROM), Programmable ROM (PROM), Electrically Programmable ROM (EPROM), Electrically Erasable Programmable ROM (EEPROM), or flash memory. Volatile memory can include Random Access Memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in a variety of forms such as Static RAM (SRAM), Dynamic RAM (DRAM), Synchronous DRAM (SDRAM), Double Data Rate SDRAM (DDRSDRAM), Enhanced SDRAM (ESDRAM), Synchronous Link DRAM (SLDRAM), Rambus Direct RAM (RDRAM), direct bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM).
The technical features of the above embodiments can be arbitrarily combined, and for the sake of brevity, all possible combinations of the technical features in the above embodiments are not described, but should be considered as the scope of the present specification as long as there is no contradiction between the combinations of the technical features.
The above-mentioned embodiments only express several embodiments of the present application, and the description thereof is more specific and detailed, but not construed as limiting the scope of the invention. It should be noted that, for a person skilled in the art, several variations and modifications can be made without departing from the concept of the present application, which falls within the scope of protection of the present application. Therefore, the protection scope of the present patent shall be subject to the appended claims.

Claims (11)

1. A mechanical seal state judgment method based on machine learning is characterized by comprising the following steps:
acquiring a research target of a mechanical sealing process, and selecting a sample acquisition mode according to the research target of the mechanical sealing process;
determining sample data according to the sample acquisition mode, wherein the sample data at least comprises a process of acquiring an acoustic emission signal according to the sample acquisition mode and extracting a characteristic quantity of the acoustic emission signal;
obtaining a target parameter value according to the sample data;
if the standard deviations of the target parameter values corresponding to different sample data are smaller than the standard deviation corresponding to the average value of the target parameter values of at least one sampling period, and the standard deviations of the target parameter values corresponding to different sample data have no statistical correlation, calculating the average value of the target parameter values corresponding to different sample data, and taking the average value of the target parameter values corresponding to different sample data as a new target parameter value;
wherein, according to the sample obtaining mode, determining sample data comprises:
acquiring at least one test parameter value within preset time in a test process, processing the at least one test parameter value, and determining the test parameter value of at least one sampling period within the preset time, wherein the test parameter value at least comprises an acoustic emission signal and an auxiliary parameter value;
processing the acoustic emission signal of at least one sampling period in the preset time to obtain a characteristic quantity of the acoustic emission signal;
processing the auxiliary parameter values of at least one sampling period in the preset time, and determining the average value of at least one auxiliary parameter value in at least one sampling period;
the obtaining of the target parameter according to the sample data comprises:
acquiring the sample data, and determining a trained model according to the sample data and the initial model;
and acquiring the average value of the auxiliary parameter values of at least one sampling period and the characteristic quantity of the acoustic emission signal of at least one sampling period in a preset time in the using process, inputting the average value of the auxiliary parameter values of at least one sampling period and the characteristic quantity of the acoustic emission signal of at least one sampling period into the trained model, and determining a target parameter value.
2. The method of claim 1, wherein the processing the acoustic emission signal of at least one sampling period within the preset time to obtain the characteristic quantity of the acoustic emission signal comprises:
acquiring an acoustic emission signal of at least one sampling period in the preset time, and performing periodic secondary division on the acoustic emission signal of at least one sampling period to obtain the acoustic emission signal of at least one sampling period after periodic secondary division;
determining a signal energy value of at least one sampling period after the periodic secondary division according to the acoustic emission signal of at least one sampling period after the periodic secondary division;
and sequencing the signal energy values according to time, performing feature extraction on the sequenced signal energy values, and determining at least one characteristic quantity of at least one sampling period after the period is divided twice.
3. The method of claim 1, wherein the processing the auxiliary parameter value for at least one sampling period within the predetermined time period, and determining the average value of the at least one auxiliary parameter value for the at least one sampling period comprises:
acquiring an output parameter value of at least one sampling period within a preset time in the test process;
and summarizing the average value of the auxiliary parameter values of the at least one sampling period, the output parameter value of the at least one sampling period and the characteristic quantity of the acoustic emission signal of the at least one sampling period, and taking the average value of the auxiliary parameter values of the at least one sampling period, the output parameter value of the at least one sampling period and the characteristic quantity of the acoustic emission signal of the at least one sampling period as sample data.
4. A mechanical seal state judgment method based on machine learning is characterized by comprising the following steps:
acquiring a research target of a mechanical sealing process, and selecting a sample acquisition mode according to the research target of the mechanical sealing process;
determining sample data according to the sample acquisition mode, wherein the sample data at least comprises a process of acquiring an acoustic emission signal according to the sample acquisition mode and extracting a characteristic quantity of the acoustic emission signal;
obtaining a target parameter value according to the sample data;
if the standard deviations of the target parameter values corresponding to different sample data are smaller than the standard deviation corresponding to the average value of the target parameter values of at least one sampling period, and the standard deviations of the target parameter values corresponding to different sample data have no statistical correlation, calculating the average value of the target parameter values corresponding to different sample data, and taking the average value of the target parameter values corresponding to different sample data as a new target parameter value;
wherein, according to the sample obtaining mode, determining the sample data further comprises:
obtaining a test parameter value of at least one physical process in a test process, and matching the test parameter value of at least one physical process in the test process with a standard parameter value of at least one physical process in the test process;
selecting a test parameter value of the physical process with the largest correlation between the test parameter value of at least one physical process in the test process and the standard parameter value of at least one physical process in the test process;
the obtaining of the target parameter according to the sample data comprises:
acquiring the sample data, and determining a trained model according to the sample data and the initial model;
and acquiring the average value of the auxiliary parameter values of at least one sampling period and the characteristic quantity of the acoustic emission signal of at least one sampling period in a preset time in the using process, inputting the average value of the auxiliary parameter values of at least one sampling period and the characteristic quantity of the acoustic emission signal of at least one sampling period into the trained model, and determining a target parameter value.
5. The method of claim 4, wherein selecting the test parameter value of the physical process having the greatest correlation between the test parameter value of the at least one physical process and the standard parameter value of the at least one physical process during the test process comprises: processing the test parameter value of the physical process with the maximum correlation degree to obtain an acoustic emission signal of at least one sampling period;
determining a characteristic quantity of the acoustic emission signal of the at least one sampling period according to the acoustic emission signal of the at least one sampling period;
and acquiring an average value of the auxiliary parameter values of at least one sampling period and an output parameter value of at least one sampling period, and taking the average value of the auxiliary parameter values of at least one sampling period, the output parameter value of at least one sampling period and the characteristic quantity of the acoustic emission signal of at least one sampling period as sample data.
6. The method of claim 5, wherein determining the characteristic quantity of the acoustic emission signal for the at least one sampling period from the acoustic emission signal for the at least one sampling period comprises:
extracting signal energy values of the acoustic emission signals in the vicinity of at least one frequency band, and sorting the signal energy values by time;
and carrying out cycle division on the sequenced signal energy values for at least one time to obtain the signal energy value of at least one cycle, and taking the signal energy value of at least one cycle as a characterization quantity.
7. The method of claim 1 or 4, wherein the inputting the average value of the auxiliary parameter values for the at least one sampling period and the characterizing quantity of the acoustic emission signal for the at least one sampling period into the trained model, the determining the target parameter values comprises:
inputting the average value of the auxiliary parameter values of the at least one sampling period and the characteristic quantity of the acoustic emission signal of the at least one sampling period into the trained model to obtain target parameter values corresponding to different sample data, and calculating the standard deviation of the target parameter values corresponding to the different sample data;
the method comprises the steps of obtaining an average value of target parameter values of at least one sampling period within preset time in a test process, and determining a standard deviation corresponding to the average value of the target parameter values of the at least one sampling period according to the average value of the target parameter values of the at least one sampling period.
8. A machine learning-based mechanical seal state determination device, characterized by comprising:
the first acquisition module is used for acquiring a research target of a mechanical sealing process and selecting a sample acquisition mode according to the research target of the mechanical sealing process;
the second acquisition module is used for determining sample data according to the sample acquisition mode, wherein the sample data at least comprises a process of acquiring an acoustic emission signal according to the sample acquisition mode and extracting a characteristic quantity of the acoustic emission signal;
the determining module is used for obtaining a target parameter value according to the sample data; if the standard deviations of the target parameter values corresponding to different sample data are smaller than the standard deviation corresponding to the average value of the target parameter values of at least one sampling period, and the standard deviations of the target parameter values corresponding to different sample data have no statistical correlation, calculating the average value of the target parameter values corresponding to different sample data, and taking the average value of the target parameter values corresponding to different sample data as a new target parameter value;
wherein the second obtaining module is further configured to:
acquiring at least one test parameter value within preset time in a test process, processing the at least one test parameter value, and determining the test parameter value of at least one sampling period within the preset time, wherein the test parameter value at least comprises an acoustic emission signal and an auxiliary parameter value;
processing the acoustic emission signal of at least one sampling period in the preset time to obtain a characteristic quantity of the acoustic emission signal;
processing the auxiliary parameter values of at least one sampling period in the preset time, and determining the average value of at least one auxiliary parameter value in at least one sampling period;
the determination module is further to:
acquiring the sample data, and determining a trained model according to the sample data and the initial model;
and acquiring the average value of the auxiliary parameter values of at least one sampling period and the characteristic quantity of the acoustic emission signal of at least one sampling period in a preset time in the using process, inputting the average value of the auxiliary parameter values of at least one sampling period and the characteristic quantity of the acoustic emission signal of at least one sampling period into the trained model, and determining a target parameter value.
9. A machine learning-based mechanical seal state determination device, characterized by comprising:
the first acquisition module is used for acquiring a research target of a mechanical sealing process and selecting a sample acquisition mode according to the research target of the mechanical sealing process;
the second acquisition module is used for determining sample data according to the sample acquisition mode, wherein the sample data at least comprises a process of acquiring an acoustic emission signal according to the sample acquisition mode and extracting a characteristic quantity of the acoustic emission signal;
the determining module is used for obtaining a target parameter value according to the sample data; if the standard deviations of the target parameter values corresponding to different sample data are smaller than the standard deviation corresponding to the average value of the target parameter values of at least one sampling period, and the standard deviations of the target parameter values corresponding to different sample data have no statistical correlation, calculating the average value of the target parameter values corresponding to different sample data, and taking the average value of the target parameter values corresponding to different sample data as a new target parameter value;
wherein the second obtaining module is further configured to:
obtaining a test parameter value of at least one physical process in a test process, and matching the test parameter value of at least one physical process in the test process with a standard parameter value of at least one physical process in the test process;
selecting a test parameter value of the physical process with the largest correlation between the test parameter value of at least one physical process in the test process and the standard parameter value of at least one physical process in the test process;
the determination module is further to:
acquiring the sample data, and determining a trained model according to the sample data and the initial model;
and acquiring the average value of the auxiliary parameter values of at least one sampling period and the characteristic quantity of the acoustic emission signal of at least one sampling period in a preset time in the using process, inputting the average value of the auxiliary parameter values of at least one sampling period and the characteristic quantity of the acoustic emission signal of at least one sampling period into the trained model, and determining a target parameter value.
10. A computer device comprising a memory and a processor, the memory storing a computer program, wherein the processor implements the steps of the method of any one of claims 1 to 7 when executing the computer program.
11. A computer-readable storage medium, on which a computer program is stored, which, when being executed by a processor, carries out the steps of the method of any one of claims 1 to 7.
CN201910180532.0A 2019-03-11 2019-03-11 Mechanical sealing state judgment method and device based on machine learning Active CN109991314B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201910180532.0A CN109991314B (en) 2019-03-11 2019-03-11 Mechanical sealing state judgment method and device based on machine learning

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201910180532.0A CN109991314B (en) 2019-03-11 2019-03-11 Mechanical sealing state judgment method and device based on machine learning

Publications (2)

Publication Number Publication Date
CN109991314A CN109991314A (en) 2019-07-09
CN109991314B true CN109991314B (en) 2020-12-29

Family

ID=67130241

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201910180532.0A Active CN109991314B (en) 2019-03-11 2019-03-11 Mechanical sealing state judgment method and device based on machine learning

Country Status (1)

Country Link
CN (1) CN109991314B (en)

Families Citing this family (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN111289621A (en) * 2019-12-27 2020-06-16 清华大学 Mechanical seal signal analysis and fault diagnosis method, device and equipment
DE102020134365A1 (en) * 2020-12-21 2022-06-23 Eagleburgmann Germany Gmbh & Co. Kg Method for monitoring a mechanical seal assembly and mechanical seal assembly
CN113947115B (en) * 2021-09-30 2024-09-13 清华大学 Mechanical seal fault early warning method, device, computer equipment and storage medium

Family Cites Families (13)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US5433245A (en) * 1993-08-16 1995-07-18 Westinghouse Electric Corporation Online valve diagnostic monitoring system having diagnostic couplings
JP2004157091A (en) * 2002-11-08 2004-06-03 Nittetsu Drum Kk Ae sensor
GB0310425D0 (en) * 2003-05-07 2003-06-11 Leuven K U Res & Dev Inspection system for sealed packages
US7748256B2 (en) * 2007-07-03 2010-07-06 Caterpillar Inc. Nondestructive testing of torsional vibration dampers
US8365599B2 (en) * 2010-04-06 2013-02-05 Varel Europe S.A.S. Acoustic emission toughness testing for PDC, PCBN, or other hard or superhard materials
US9453780B2 (en) * 2012-04-24 2016-09-27 Aktiebolaget Skf Module for determining an operating characteristic of a bearing
CN103837303B (en) * 2014-03-25 2016-03-30 清华大学 A kind of fine motion reciprocation sealing dynamic characteristic experiment platform
EP3420254B1 (en) * 2016-02-23 2021-11-24 John Crane UK Limited Systems and methods for predictive diagnostics for mechanical systems
EP4293244A3 (en) * 2017-05-22 2024-04-10 Waukesha Bearings Corporation Bearing monitoring/analysis system
CN107991165B (en) * 2017-11-03 2020-07-14 中国石油大学(北京) Cement sheath air tightness failure judgment method
CN108591451B (en) * 2018-07-02 2020-06-09 清华大学 Intelligent adjusting system and method for mechanical seal
CN108956043B (en) * 2018-07-03 2020-05-22 清华大学 Mechanical seal multi-scale real-time monitoring and analyzing method
CN109325629B (en) * 2018-10-10 2022-01-07 中国石油化工股份有限公司 Method for predicting leakage fault of mechanical seal of in-service rotating equipment

Non-Patent Citations (1)

* Cited by examiner, † Cited by third party
Title
Modelling Acoustic Emissions Generated by Tribological Behaviour of Mechanical Seals for Condition Monitoring and Fault Detection;Hossein Towsyfyan 等;《Tribology International》;20180901;第125卷;第46-58页 *

Also Published As

Publication number Publication date
CN109991314A (en) 2019-07-09

Similar Documents

Publication Publication Date Title
CN109991314B (en) Mechanical sealing state judgment method and device based on machine learning
Pichler et al. Fault detection in reciprocating compressor valves under varying load conditions
US10393621B2 (en) Method for assessing the condition of rotating machinery connected to an electric motor
Yu Health condition monitoring of machines based on hidden Markov model and contribution analysis
EP3327419A1 (en) Engine health diagnostic apparatus and method
CN110579354B (en) Bearing detection method based on convolutional neural network
CN108709635B (en) Method and device for determining main frequency component information of rotary mechanical vibration signal
CN111523683B (en) Method and system for predicting technological parameters in tobacco processing
CN113839825B (en) Equipment fault detection method, system, device, computer equipment and storage medium
CN111027260A (en) Reciprocating compressor fault diagnosis method based on neural network
CN112525749B (en) Tribology state online identification method based on friction signal recursion characteristic
CN110441584A (en) Equipment running state monitoring method, device, storage medium and system
CN113110961B (en) Equipment abnormality detection method and device, computer equipment and readable storage medium
Zhang et al. Applying improved multi-scale entropy and support vector machines for bearing health condition identification
Pichler et al. Fault detection in reciprocating compressor valves for steady-state load conditions
Fang et al. Reciprocating compressors intelligent fault diagnosis under multiple operating conditions based on adaptive variable scale morphological filter
Dhakar et al. Fault diagnosis of air compressor set-up using decision tree based J48 classification algorithm
CN111538755A (en) Equipment operation state anomaly detection method based on normalized cross correlation and unit root detection
Wu et al. Instantaneous frequency estimation method for the vibration signal of rotating machinery based on STFTSC algorithm
CN105866360A (en) A method of detecting water holding capacity of meat
CN113790130B (en) Method for evaluating rotating speed fluctuation stability of wind generating set
JP2016145739A (en) Method for monitoring bearing state and device for monitoring bearing state
Liu et al. GARCH family models oriented health indicators for bearing degradation monitoring
Fan et al. Remaining useful life prediction of rolling bearings by the particle filter method based on degradation rate tracking
Ying et al. Rolling bearing vibration signal analysis based on dual-entropy, holder coefficient and gray relation theory

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination
GR01 Patent grant
GR01 Patent grant