Proximity sensor self-learning recognition system based on multi-dimensional signal processing

Abstract

The present invention discloses a proximity sensor self-learning recognition system based on multi-dimensional signal processing, which relates to the field of data analysis. The main scheme is: taking vibration data, power quality data, sound data, operation data, wear data and magnetic field data of the equipment in normal operating state and different fault type operating states as input, taking comprehensive indicators of the equipment and corresponding fault type labels as output, training a fault recognition self-learning model, solving the problem that the data collection type is single and the connection between different types of data is ignored, resulting in poor model training effect and reduced recognition accuracy. Whether the training data needs to be updated is judged according to the fault indicator deviation value and the fault indicator deviation threshold, solving the problem that the distribution of motor operation data may change with the passage of time and environmental changes, and the model performance will decline if it is not updated in time.

Description

Technical Field

[0001] The present invention relates to the technical field of data analysis, and in particular to a proximity sensor self-learning recognition system based on multi-dimensional signal processing. Background Art

[0002] In the field of motor fault identification, the proximity sensor self-learning identification system integrates data from multiple sensors and combines advanced signal processing technology and machine learning algorithms to effectively monitor and analyze the operating status of equipment, thereby realizing automatic identification of potential fault types. This not only provides strong support for equipment maintenance, but also helps to improve the safety and efficiency of industrial production.

[0003] The traditional method for the identification system to determine the fault type is to monitor the vibration signal of the motor equipment through a proximity sensor, input the collected vibration signal into a trained self-learning identification model, and the model automatically outputs the judgment result of the equipment fault type.

[0004] The existing technology still has the following shortcomings: the data collection type is single, and the connection between different types of data is ignored, resulting in poor model training effect, thereby reducing recognition accuracy. At the same time, with the passage of time and environmental changes, the distribution of motor operation data may change. If it is not updated in time, the model performance will decline. Summary of the invention

[0005] 1. Technical issues to be solved

[0006] In view of the shortcomings of the prior art, the present invention provides a proximity sensor self-learning recognition system based on multi-dimensional signal processing, which takes the vibration data, power quality data, sound data, operation data, wear data and magnetic field data of the equipment in normal operation state and different fault types as input, and converts the comprehensive index X r The corresponding fault type label is used as output to train the fault identification self-learning model, which solves the problem of single data collection type and ignoring the connection between different types of data, resulting in poor model training effect and reduced recognition accuracy. The fault indicator deviation value XD and the fault indicator deviation threshold XP are used to determine whether the training data of this fault type needs to be updated, which solves the problem that the distribution of motor operation data may change over time and environmental changes. If it is not updated in time, the model performance will decline.

[0007] (II) Technical solution

[0008] To achieve the above objectives, the present invention is implemented through the following technical solutions: A proximity sensor self-learning recognition system based on multi-dimensional signal processing, comprising:

[0009] A data acquisition module is used to collect vibration data, power quality data, sound data, operation data, wear data and magnetic field data of the equipment under normal operation and different fault types;

[0010] Index calculation module, which can calculate the vibration deviation index YA based on vibration data r , calculate the power quality deviation index YB based on the power quality data r , according to the vibration deviation index YA r And power quality deviation index YB r Calculation of electrical-vibration correlation index XA r ; Calculate the sound abnormality index YC based on the sound data r , according to the operation data and abnormal sound index YC r Calculation of sound-operation coupling index XB r ; Calculate the wear distribution index YD based on the wear data r , calculate the magnetic field-power quality deviation index YE based on the magnetic field data and power quality data r , according to the wear distribution index YD r And magnetic field-power quality deviation index YE r Calculate the performance degradation index XC r ; According to the electrical-vibration correlation index XA r , Sound-operation coupling index XB r and performance degradation index XC r Calculate the comprehensive index X r ;

[0011] The model judgment module can build a fault identification self-learning model through a support vector machine learning model, take the vibration data, power quality data, sound data, operation data, wear data and magnetic field data of the equipment in normal operation state and different fault type operation state as input, take the comprehensive index c and the corresponding fault type label as output, and train the fault identification self-learning model; input the current vibration data, power quality data, sound data, operation data, wear data and magnetic field data of the equipment into the trained fault identification self-learning model, and the model outputs the comprehensive index X and the corresponding fault type;

[0012] Update module to obtain the comprehensive index X output in the model judgment module and the corresponding comprehensive index X r , and according to comprehensive index X and comprehensive index X r Calculate the fault index deviation value XD r ; Preset fault indicator deviation threshold XP, according to the fault indicator deviation value XD r And the fault indicator deviation threshold XP determines whether the training data of this fault type needs to be updated.

[0013] In the preferred solution of the proximity sensor self-learning recognition system based on multi-dimensional signal processing: calculating the vibration deviation index YA r The method is:

[0014] Vibration data includes vibration acceleration signal AA ra , vibration peak AB r , RMS normal value AC and normal peak maximum value AD;

[0015] Calculate the vibration deviation index YA based on the vibration data r , the formula based on is:

[0016]

[0017] Among them, YA r is the vibration deviation index of the equipment in the rth operating state, r is the serial number corresponding to the different operating states of the equipment, and the value is a positive integer; AA ra is the vibration acceleration signal corresponding to the ath sampling point under the rth operating state of the equipment; a is the serial number corresponding to different sampling points, and its value is [1, b]; b is the total number of sampling points, and its value is a positive integer; AB r is the vibration peak value of the equipment in the rth operating state.

[0018] In the preferred solution of the proximity sensor self-learning identification system based on multi-dimensional signal processing: calculating the power quality deviation index YB r The method is:

[0019] Power quality data including active power BA r , reactive power BB r , rated power factor BC, harmonic voltage effective value BD rd , Fundamental voltage effective value BE r , Maximum voltage per unit time BF r , minimum voltage value per unit time BG r And rated voltage value BH;

[0020] Calculate the power quality deviation index YB based on the power quality data r , the formula based on is:

[0021]

[0022] Among them, YB r BA is the power quality deviation index of the equipment in the rth operating state; r is the active power of the equipment in the rth operating state; BB r is the reactive power of the equipment in the rth operating state; BE rBF is the effective value of the fundamental voltage of the equipment in the rth operating state; r BG is the maximum voltage per unit time in the rth operating state of the equipment; r BD is the minimum voltage per unit time in the rth operating state of the equipment; rd is the effective value of the dth harmonic voltage under the rth operating state of the equipment, d is the serial number corresponding to different harmonic voltages, and its value is [2, f r ]; f r It is the total number of harmonic voltages in the rth operating state of the equipment, and its value is a positive integer greater than 1.

[0023] In the preferred solution of the proximity sensor self-learning recognition system based on multi-dimensional signal processing: calculating the electrical-vibration correlation index XA r The method is:

[0024] According to the vibration deviation index YA r And power quality deviation index YB r Calculation of electrical-vibration correlation index XA r , the formula based on is:

[0025] XA r =α 1r ×YA r +α r2 ×YB r

[0026] Among them, XA r is the electrical-vibration correlation index of the equipment under the rth operating state; α 1r YA is the vibration deviation index of the equipment in the rth operating state r The weight coefficient is 0.3~0.7; α r2 YB is the power quality deviation index of the equipment in the rth operating state r The weight coefficient is 0.3 to 0.7; and α r1 +α r2 =1.

[0027] In the preferred solution of the proximity sensor self-learning recognition system based on multi-dimensional signal processing: calculating the sound abnormality index YC r The method is:

[0028] Sound data includes actual sound pressure CA r , reference sound pressure CB, sound signal fundamental frequency amplitude CC r , Sound signal harmonic amplitude CD rh , frequency band signal-to-noise ratio CE rk , signal-to-noise ratio normal range upper limit CE max and the lower limit of the normal range of signal-to-noise ratio CEmin ;

[0029] Calculate the sound abnormality index YC based on the sound data r , the formula based on is:

[0030]

[0031] Among them, YC r The sound abnormality index of the equipment in the rth operating state; the actual sound pressure CA r is the actual sound pressure of the equipment in the rth operating state; CC r CD is the fundamental frequency amplitude of the sound signal in the rth operating state of the equipment; rh is the amplitude of the hth harmonic in the sound signal under the rth operating state of the equipment, h is the serial number corresponding to different harmonics in the sound signal, and its value is [2, i r ];i r is the total number of harmonics in the sound signal of the equipment in the rth operating state, and its value is a positive integer greater than 1; CE rk is the signal-to-noise ratio value of the kth frequency band in the rth operating state of the device, k is the serial number corresponding to different frequency bands, and its value is [1, l r ]; l r It is the total number of frequency bands in the rth operating state of the device, and its value is a positive integer.

[0032] In the preferred solution of the proximity sensor self-learning recognition system based on multi-dimensional signal processing: calculating the sound-operation coupling index XB r The method is:

[0033] Operating data includes actual operating temperature DA r , Normal operating temperature DB r , actual speed DC r and rated speed DD;

[0034] According to the operation data and abnormal sound index YC r Calculation of sound-operation coupling index XB r , the formula based on is:

[0035]

[0036] Among them, XB r DA is the sound-operation coupling index of the equipment in the rth operating state; r is the actual operating temperature of the equipment in the rth operating state; DB r is the normal operating temperature of the equipment in the rth operating state; DC r is the actual speed of the equipment in the rth operating state.

[0037] In the preferred solution of the proximity sensor self-learning recognition system based on multi-dimensional signal processing: calculating the wear distribution index YD r The method is:

[0038] Wear data including average wear particle diameter EA rm , Wear particle quantity EB rm and the total number of wear particles EC under normal conditions;

[0039] Calculate the wear distribution index YD based on the wear data r , the formula based on is:

[0040]

[0041] Among them, YD r is the wear distribution index of the equipment under the rth operating state; EA rm is the average diameter of wear particles in the mth size range under the rth operating state of the equipment, m is the serial number corresponding to different size ranges, and its value is [1, n]; n is the total number of different size ranges; EB rm is the number of wear particles in the mth size range under the rth operating state of the equipment; β 1r is the weight coefficient of the wear particle size index under the equipment's rth operating state, with a value of 0.4 to 0.6; β 2r is the weight coefficient of the wear particle distribution entropy under the equipment's rth operating state, ranging from 0.4 to 0.6; and β 1r +β 2r =1.

[0042] In the preferred solution of the proximity sensor self-learning recognition system based on multi-dimensional signal processing: the performance degradation index XC is calculated. r The method is:

[0043] Magnetic field data includes actual magnetic field strength FA r and normal magnetic field strength FB;

[0044] The power quality data also includes the actual insulation resistance GA r And normal insulation resistance GB;

[0045] Calculate the magnetic field-power quality deviation index YE based on the magnetic field data and power quality data r , the formula based on is:

[0046]

[0047] Among them, YE r FA is the magnetic field-power quality deviation index of the equipment in the rth operating state; ris the actual magnetic field strength of the equipment in the rth operating state; GA r is the actual insulation resistance of the equipment in the rth operating state;

[0048] According to the wear distribution index YD r And magnetic field-power quality deviation index YE r Calculate the performance degradation index XC r , the formula based on is:

[0049]

[0050] Among them, XC r It is the performance degradation index of the equipment in the rth operating state.

[0051] In the preferred solution of the proximity sensor self-learning recognition system based on multi-dimensional signal processing: calculating the comprehensive index X r The method is:

[0052] According to the electrical-vibration correlation index XA r , Sound-operation coupling index XB r and performance degradation index XC r Calculate the comprehensive index X r , the formula based on is:

[0053]

[0054] Among them, X r is the comprehensive index of the equipment under the rth operating state; γ 1r XA is the electrical-vibration correlation index of the equipment under the rth operating state r The weight coefficient is 0.3~0.5; γ 2r The sound-operation coupling index XB of the equipment in the rth operating state r The weight coefficient is 0.2~0.5; γ 3r XC is the performance degradation index of the equipment in the rth operating state r The weight coefficient is 0.2 to 0.5; and γ 1r +γ 2r +γ 3r =1.

[0055] In the preferred solution of the proximity sensor self-learning recognition system based on multi-dimensional signal processing, the method for determining whether the training data needs to be updated is:

[0056] According to comprehensive index X and comprehensive index X r Calculate the fault index deviation value XD r , the formula based on is:

[0057]

[0058] Among them, XD r is the fault index deviation value of the rth fault type;

[0059] According to the fault index deviation value XD r The fault indicator deviation threshold XP determines whether the training data of this fault type needs to be updated, based on the following criteria:

[0060]

[0061] (III) Beneficial effects

[0062] The present invention provides a proximity sensor self-learning recognition system based on multi-dimensional signal processing, which has the following beneficial effects:

[0063] (1) By collecting vibration data, power quality data, sound data, operation data, wear data, and magnetic field data under normal operating conditions and different fault types of equipment, multi-dimensional data can be provided for training the model to improve model accuracy.

[0064] (2) Calculate the electrical-vibration correlation index XA based on vibration data and power quality data r , which reflects the coordinated working of the electrical and mechanical parts of the equipment, is conducive to identifying potential faults, and calculates the sound-operation coupling index XB based on sound data and operation data r , which can more comprehensively reflect the operating status of the equipment, provide a basis for optimizing and adjusting the equipment operating parameters, and calculate the performance degradation index XC based on wear data, magnetic field data and power quality data r , can more keenly detect early signs of faults, provide a basis for fault location and diagnosis, according to the electrical-vibration correlation index XA r , Sound-operation coupling index XB r and performance degradation index XC r Calculate the comprehensive index X r , which can evaluate the equipment from multiple dimensions and reflect the equipment operation characteristics more comprehensively.

[0065] (3) The vibration data, power quality data, sound data, operation data, wear data and magnetic field data of the equipment under normal operation and different fault types are used as input, and the comprehensive index X is calculated. r And the corresponding fault type label is used as the output to train the fault recognition self-learning model, which solves the problem of single data collection type and ignoring the connection between different types of data, resulting in poor model training effect and thus reduced recognition accuracy.

[0066] (4) According to the fault index deviation value XD r The fault indicator deviation threshold XP is used to determine whether the training data of this fault type needs to be updated, which solves the problem that the distribution of motor operation data may change over time and due to environmental changes. If it is not updated in time, the model performance will decline. BRIEF DESCRIPTION OF THE DRAWINGS

[0067] Figure 1 It is a structural schematic diagram of the proximity sensor self-learning recognition system based on multi-dimensional signal processing of the present invention.

[0068] Figure 2 The electrical-vibration correlation index XA is calculated in the proximity sensor self-learning recognition system based on multi-dimensional signal processing of the present invention. r Parameter relationship structure diagram.

[0069] Figure 3 The performance degradation index XC is calculated in the proximity sensor self-learning recognition system based on multi-dimensional signal processing of the present invention. r Parameter relationship structure diagram. DETAILED DESCRIPTION

[0070] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0071] See also Figure 1-Figure 3 The present invention provides a proximity sensor self-learning recognition system based on multi-dimensional signal processing, comprising:

[0072] The data acquisition module is used to collect vibration data, power quality data, sound data, operation data, wear data and magnetic field data under normal operation status and different fault type operation status of the equipment.

[0073] In the above scheme, by collecting vibration data, power quality data, sound data, operation data, wear data and magnetic field data under normal operating conditions and different fault types of equipment, multi-dimensional data can be provided for the training model to improve the accuracy of the model.

[0074] Index calculation module, which can calculate the vibration deviation index YA based on vibration data r , calculate the power quality deviation index YB based on the power quality data r , according to the vibration deviation index YA r And power quality deviation index YB rCalculation of electrical-vibration correlation index XA r ; Calculate the sound abnormality index YC based on the sound data r , according to the operation data and abnormal sound index YC r Calculation of sound-operation coupling index XB r ; Calculate the wear distribution index YD based on the wear data r , calculate the magnetic field-power quality deviation index YE based on the magnetic field data and power quality data r , according to the wear distribution index YD r And magnetic field-power quality deviation index YE r Calculate the performance degradation index XC r ; According to the electrical-vibration correlation index XA r , Sound-operation coupling index XB r and performance degradation index XC r Calculate the comprehensive index X r .

[0075] Specifically, the vibration deviation index YA is calculated r The method is:

[0076] Vibration data includes vibration acceleration signal AA ra , vibration peak AB r , RMS normal value AC and normal peak maximum value AD.

[0077] It should be noted that the vibration acceleration signal AA ra It is a physical quantity that describes the vibration state of the equipment. It reflects the speed of the equipment during the vibration process and is measured by an acceleration sensor. Vibration peak value AB r It refers to the maximum value reached by the vibration acceleration signal during the vibration of the equipment, reflecting the strongest instantaneous vibration intensity during the vibration process. The acquisition method is: the vibration acceleration signal AA obtained by measuring the acceleration sensor ra Import the data into MATLAB software, write a program to traverse the data sequence, and find the maximum value in the data as the vibration peak AB r . The root mean square normal value AC is a measure of the energy of the vibration acceleration signal. It is obtained by collecting the vibration acceleration signal within a unit time, such as hours, days or weeks, during the normal operation of the equipment. The unit time depends on the operation cycle and stability of the equipment. Then, the signal value of each sampling point is squared, and the square values ​​are added and divided by the number of sampling points to obtain the average value. Finally, the average value is squared to obtain the root mean square normal value AC. The normal peak maximum value AD refers to the maximum value of the vibration peak value that may occur during the normal operation of the equipment. It is used to distinguish normal vibration fluctuations from abnormal vibrations that may indicate a fault. It is obtained by referring to the equipment manual.

[0078] Calculate the vibration deviation index YA based on the vibration data r , the formula based on is:

[0079]

[0080] Among them, YA r is the vibration deviation index of the equipment in the rth operating state, r is the serial number corresponding to the different operating states of the equipment, and the value is a positive integer; AA ra is the vibration acceleration signal corresponding to the ath sampling point under the rth operating state of the equipment; a is the serial number corresponding to different sampling points, and its value is [1, b]; b is the total number of sampling points, and its value is a positive integer; AB r is the vibration peak value of the equipment in the rth operating state.

[0081] It should be noted that the operating principle of this formula is: It is used to measure the deviation between the actual RMS value and the normal RMS value. It is used to measure the deviation between the actual value of the vibration peak and the normal peak maximum value AD. This formula comprehensively considers the influence of the above two aspects to obtain the vibration deviation index YA r .

[0082] Specifically, the power quality deviation index YB is calculated r The method is:

[0083] Power quality data including active power BA r , reactive power BB r , rated power factor BC, harmonic voltage effective value BD rd , Fundamental voltage effective value BE r , Maximum voltage per unit time BF r , minimum voltage value per unit time BC r And rated voltage value BH.

[0084] It should be noted that the active power BA r Reactive power refers to the electrical energy actually used to do work in a device per unit time. It is a physical quantity used to measure the rate at which electrical energy is converted into other forms of energy. It can be measured using a power analyzer. r It is the electrical energy used to establish and maintain the electric and magnetic fields in the AC circuit. It can be measured using a power analyzer. The rated power factor BC refers to the ratio of active power to apparent power of the equipment under rated operating conditions. It reflects the efficiency of the equipment in utilizing electrical energy. It can be obtained by checking the equipment nameplate or product manual. Harmonic voltage effective value BD rdIt refers to the effective value of the harmonic voltage component. Harmonics refer to other frequency components besides the fundamental frequency component. These frequencies are integer multiples of the fundamental frequency and are called harmonics. The effective value of the harmonic voltage is BD rd Obtained by measuring with a spectrum analyzer. Fundamental voltage effective value BE r It refers to the effective value of the fundamental frequency voltage component. It is an important indicator for measuring the size of AC voltage and is measured using a spectrum analyzer. The maximum voltage per unit time BF r It refers to the maximum amplitude of the voltage signal in a specific time period, and the minimum voltage per unit time BG r It refers to the minimum amplitude of the voltage signal in the same time period, and the maximum voltage per unit time BF r And the minimum voltage per unit time BG r The method for determining is: use voltage sensors, power parameter testers and other data acquisition equipment to collect voltage signals, import the collected voltage data sequence into MATLAB software, write a program to traverse the data sequence, find the maximum and minimum values ​​in the data, and use them as the maximum voltage per unit time BF r And the minimum voltage per unit time BG r The rated voltage value BH refers to the standard voltage value when the equipment is in normal operation. It represents the voltage range that the equipment can adapt to under normal working conditions. It can be obtained by checking the equipment nameplate or product manual.

[0085] Calculate the power quality deviation index YB based on the power quality data r , the formula based on is:

[0086]

[0087] Among them, YB r BA is the power quality deviation index of the equipment in the rth operating state; r is the active power of the equipment in the rth operating state; BB r is the reactive power of the equipment in the rth operating state; BE r BF is the effective value of the fundamental voltage of the equipment in the rth operating state; r BG is the maximum voltage per unit time in the rth operating state of the equipment; r BD is the minimum voltage per unit time in the rth operating state of the equipment; rd is the effective value of the dth harmonic voltage under the rth operating state of the equipment, d is the serial number corresponding to different harmonic voltages, and its value is [2, f r ]; f r It is the total number of harmonic voltages in the rth operating state of the equipment, and its value is a positive integer greater than 1.

[0088] It should be noted that the operating principle of this formula is: It is used to measure the deviation between the actual power factor and the rated power factor. Used to measure the degree of deviation of harmonic voltage. Used to measure voltage fluctuations, this formula comprehensively considers the influence of three aspects to obtain the power quality deviation index YB r .

[0089] Specifically, the electrical-vibration correlation index XA is calculated r The method is:

[0090] According to the vibration deviation index YA r And power quality deviation index YB r Calculation of electrical-vibration correlation index XA r , the formula based on is:

[0091] XA r =α 1r ×YA r +α r2 ×YB r

[0092] Among them, XA r is the electrical-vibration correlation index of the equipment under the rth operating state; α 1r YA is the vibration deviation index of the equipment in the rth operating state r The weight coefficient is 0.3 to 0.7, according to the vibration deviation index YA under different operating conditions. r Electrical-vibration correlation index XA r The degree of influence is determined; α r2 YB is the power quality deviation index of the equipment in the rth operating state r The weight coefficient is 0.3 to 0.7, according to the power quality deviation index YB under different operating conditions. r Electrical-vibration correlation index XA r The degree of influence of is determined; and α r1 +α r2 =1.

[0093] It should be noted that this formula takes into account the vibration deviation index YA through a weighted method. r And power quality deviation index YB r The influence of electrical-vibration correlation index XA is obtained r .

[0094] Specifically, the sound abnormality index YC is calculated r The method is:

[0095] Sound data includes actual sound pressure CAr , reference sound pressure CB, sound signal fundamental frequency amplitude CC r , Sound signal harmonic amplitude CD rh , frequency band signal-to-noise ratio CE rk , signal-to-noise ratio normal range upper limit CE max and the lower limit of the normal range of signal-to-noise ratio CE min .

[0096] It should be noted that the actual sound pressure CA r It refers to the pressure fluctuation value corresponding to the sound waves actually existing in the surrounding air due to equipment vibration and other reasons during the operation of the equipment. It reflects the intensity of the sound generated by the equipment during operation. It is measured by a sound pressure meter and the unit is pa. The reference sound pressure CB is the benchmark value used to compare and measure the sound intensity. The sound pressure value corresponding to the weakest sound that can be heard by the human ear is usually defined as the reference sound pressure, and its value is 2×10 -5 pa. Sound signal fundamental frequency amplitude CC r It refers to the amplitude corresponding to the fundamental frequency in the sound signal, which reflects the main vibration frequency and intensity of the equipment. The fundamental frequency is the basic frequency of the vibration and the main frequency in the sound frequency component. The fundamental frequency amplitude of the sound signal CC r Obtained by measuring with a spectrum analyzer. Sound signal harmonic amplitude CD rh It refers to the amplitude corresponding to the harmonic frequency. The change of harmonic amplitude can reflect whether the equipment is abnormal. It is measured by spectrum analyzer. rk It refers to the ratio of signal power to noise power within a certain frequency bandwidth. It is used to measure the degree to which the device sound signal is interfered with by noise within a specific frequency range. It is measured using professional acoustic measuring instruments such as audio analyzers or signal-to-noise ratio testers. max and the lower limit of the normal range of signal-to-noise ratio CE min It is a reference value used to judge whether the quality of the sound signal of the equipment is normal. The acquisition method is: when the equipment is operating normally, use the signal-to-noise ratio tester to measure the signal-to-noise ratio data of the equipment multiple times, calculate the average value and standard deviation of these signal-to-noise ratio data, and calculate the value of the average value plus two times the standard deviation as the upper limit of the normal range of the signal-to-noise ratio value CE max Calculate the value of the mean minus two times the standard deviation as the lower limit of the normal range of the signal-to-noise ratio CE min .

[0097] Calculate the sound abnormality index YC based on the sound data r , the formula based on is:

[0098]

[0099] Among them, YC r The sound abnormality index of the equipment in the rth operating state; the actual sound pressure CA r is the actual sound pressure of the equipment in the rth operating state; CC r CD is the fundamental frequency amplitude of the sound signal in the rth operating state of the equipment; rh is the amplitude of the hth harmonic in the sound signal under the rth operating state of the equipment, h is the serial number corresponding to different harmonics in the sound signal, and its value is [2, i r ];i r is the total number of harmonics in the sound signal of the equipment in the rth operating state, and its value is a positive integer greater than 1; CE rk is the signal-to-noise ratio value of the kth frequency band in the rth operating state of the device, k is the serial number corresponding to different frequency bands, and its value is [1, l r ]; l r It is the total number of frequency bands in the rth operating state of the device, and its value is a positive integer.

[0100] It should be noted that the operating principle of this formula is: Used to measure the sound pressure level of equipment. It is used to measure the degree of deviation of harmonics in sound signals. Used to measure sound clarity, this formula takes three factors into consideration to get the sound abnormality index YC r , where e is the base of the natural logarithm in the exponential function.

[0101] Specifically, the sound-operation coupling index XB is calculated r The method is:

[0102] Operating data includes actual operating temperature DA r , Normal operating temperature DB r , actual speed DC r and rated speed DD.

[0103] It should be noted that the actual operating temperature DA r It refers to the actual temperature reached by the equipment during operation. The method of obtaining it is: use an infrared thermometer to measure the temperature values ​​of different parts of the equipment, such as transformers and windings, and accumulate the temperature values ​​of different parts and divide them by the number of data to obtain the actual operating temperature DA r . Normal operating temperature DB r It refers to the temperature range in which the equipment can operate stably for a long time under ideal working conditions without adversely affecting the performance and life of the equipment. The method of obtaining it is: refer to the equipment's operating manual, technical specification or product manual to obtain the normal operating temperature range, and calculate (temperature range upper limit + temperature range lower limit) / 2 to obtain the actual operating temperature DA r . Actual speed DCr It refers to the actual number of revolutions per minute or per second of rotating parts such as the motor rotor or machine tool spindle during the operation of the equipment, which reflects the operating speed of the equipment. The acquisition method is: the measuring head of the tachometer is in contact with the rotating shaft of the equipment, and the measuring head transmits the rotation speed of the equipment to the meter head through transmission mechanisms such as gears and friction wheels. The meter head displays the speed value mechanically or electronically, and this value is the actual speed DC r The rated speed DD refers to the rotation speed that the equipment should reach under rated load and rated working conditions, which can be obtained by checking the nameplate of the equipment.

[0104] According to the operation data and abnormal sound index YC r Calculation of sound-operation coupling index XB r , the formula based on is:

[0105]

[0106] Among them, XB r DA is the sound-operation coupling index of the equipment in the rth operating state; r is the actual operating temperature of the equipment in the rth operating state; DB r is the normal operating temperature of the equipment in the rth operating state; DC r is the actual speed of the equipment in the rth operating state.

[0107] It should be noted that the operating principle of this formula is: The operating data deviation index is obtained by calculating the deviation between the actual operating temperature of the equipment and the normal operating temperature, as well as the deviation between the actual speed and the rated speed. This formula comprehensively considers the abnormal sound index YC r and the running data deviation index, and obtain the sound-running coupling index XB r .

[0108] Specifically, the wear distribution index YD is calculated r The method is:

[0109] Wear data including average wear particle diameter EA rm , Wear particle quantity EB rm and the total number of wear particles EC under normal conditions.

[0110] It should be noted that the average diameter of wear particles EA rmIt refers to the average size of wear particles in different size ranges generated by friction between parts during equipment operation. The size ranges include 0-10μm, 0-50μm, 50-100μm, etc. The acquisition method is: collect equipment lubricating oil samples, use centrifugal separation to extract wear particles from the samples, use the measurement software provided by the electron microscope to measure the diameter of the observed wear particles, record the diameter of each wear particle and classify them according to the size range, for each size range, add up the diameters of all particles in the range, and then divide by the number of particles to obtain the average wear particle diameter EA. rm . Wear particle number EB rm It refers to the number of particles in different size ranges among the wear particles generated by the equipment. By counting the number of particles in different size ranges, we can understand the specific situation of equipment wear in more detail. The acquisition method is: collect equipment lubricant samples, use centrifugal separation to extract wear particles from the samples, use the measurement software provided by the electron microscope to measure the diameter of the observed wear particles, record the diameter of each wear particle and classify them according to the size range, and count the number of particles in each size range as the wear particle number EB. rm The total number of wear particles EC under normal conditions refers to the range of the number of wear particles contained in a certain volume of lubricating oil or lubricating medium under normal operating conditions of the equipment. It is used to determine whether the equipment is in a normal wear state. The acquisition method is: in the early stage of equipment commissioning or when the equipment is operating normally, lubricating oil samples are collected regularly, and the number of wear particles in the lubricating oil samples is measured using the measurement software provided by the electron microscope. The number of wear particles in all samples is accumulated and divided by the number of samples to obtain the total number of wear particles EC under normal conditions.

[0111] Calculate the wear distribution index YD based on the wear data r , the formula based on is:

[0112]

[0113] Among them, YD r is the wear distribution index of the equipment under the rth operating state; EA rm is the average diameter of wear particles in the mth size range under the rth operating state of the equipment, m is the serial number corresponding to different size ranges, and its value is [1, n]; n is the total number of different size ranges; EB rm is the number of wear particles in the mth size range under the rth operating state of the equipment; β 1r is the weight coefficient of the wear particle size index under the rth operating state of the equipment, with a value of 0.4 to 0.6. The wear distribution index YD is calculated based on the wear particle size index under different operating states. r The importance of determining;2r is the weight coefficient of the wear particle distribution entropy under the rth operating state of the equipment, with a value of 0.4 to 0.6. The wear distribution index YD is calculated based on the wear particle distribution entropy under different operating states. r The importance of β is determined; and 1r +β 2r =1.

[0114] It should be noted that the operating principle of this formula is: It is the wear particle size index under the equipment's rth operating state, which is used to measure the effect of the size and quantity of wear particles on the distribution of wear particles. is the wear particle distribution entropy under the rth operating state of the equipment, which is used to measure the uniformity of wear particle distribution. This formula uses a weighted method to comprehensively consider the wear particle size index and the wear particle distribution entropy to obtain the wear distribution index YD r .

[0115] Specifically, the performance degradation index XC is calculated r The method is:

[0116] Magnetic field data includes actual magnetic field strength FA r and normal magnetic field strength FB.

[0117] It should be noted that the actual magnetic field strength FA r It refers to the strength of the real magnetic field generated by the device when it is running. It is measured using a Hall effect sensor. The specific method is: place the Hall effect sensor at the location where the magnetic field strength needs to be measured, connect the magnetic field strength meter to the Hall effect sensor, and obtain the actual magnetic field strength FA by reading the value displayed by the magnetic field strength meter. r The normal magnetic field strength FB refers to the range of magnetic field strength that the equipment should generate under normal operating conditions, which can be obtained by referring to the equipment's operating manual, technical specifications or product instructions.

[0118] The power quality data also includes the actual insulation resistance GA r And normal insulation resistance GB.

[0119] It should be noted that the actual insulation resistance GA r It refers to the resistance value of the insulation part of the equipment under actual operating conditions, reflecting the ability of the equipment insulation material to prevent current leakage, and is measured by a digital insulation resistance tester. Normal insulation resistance GB refers to the insulation resistance value range that the equipment should have under ideal, trouble-free operating conditions. It is an important reference standard for judging whether the insulation state of the equipment is good, and is obtained by referring to the equipment's operating manual, technical specification or product manual.

[0120] Calculate the magnetic field-power quality deviation index YE based on the magnetic field data and power quality data r , the formula based on is:

[0121]

[0122] Among them, YE r FA is the magnetic field-power quality deviation index of the equipment in the rth operating state; r is the actual magnetic field strength of the equipment in the rth operating state; GA r It is the actual insulation resistance of the equipment in the rth operating state.

[0123] It should be noted that the operating principle of this formula is: Used to measure the actual magnetic field strength FA r The degree of deviation from the normal magnetic field strength FB, Used to measure the actual insulation resistance GA r The formula comprehensively considers the degree of deviation of the magnetic field and the degree of deviation of the insulation resistance to obtain the magnetic field-power quality deviation index YE. r .

[0124] According to the wear distribution index YD r And magnetic field-power quality deviation index YE r Calculate the performance degradation index XC r , the formula based on is:

[0125]

[0126] Among them, XC r It is the performance degradation index of the equipment in the rth operating state.

[0127] It should be noted that the operating principle of this formula is: By taking the square root, the effect of the smaller value of the two indicators is magnified, ln(1+YD r ×YE r ) By taking the natural logarithm, the influence of the larger value of the two indicators is magnified. This formula comprehensively considers the wear distribution indicator YD r And magnetic field-power quality deviation index YE r The performance degradation index XC is obtained by considering the influence of two aspects. r . Where ln is the natural logarithm function.

[0128] Specifically, the calculation of the comprehensive index X r The method is:

[0129] According to the electrical-vibration correlation index XA r , Sound-operation coupling index XBr and performance degradation index XC r Calculate the comprehensive index E r , the formula based on is:

[0130]

[0131] Among them, X r is the comprehensive index of the equipment under the rth operating state; γ 1r XA is the electrical-vibration correlation index of the equipment under the rth operating state r The weight coefficient is 0.3 to 0.5, according to the electrical-vibration correlation index XA under different operating conditions. r For comprehensive index X r Determine the degree of influence; 2r The sound-operation coupling index XB of the equipment in the rth operating state r The weight coefficient is 0.2 to 0.5, according to the sound-operation coupling index XB under different operating conditions. r For comprehensive index X r Determine the degree of influence; 3r XC is the performance degradation index of the equipment in the rth operating state r The weight coefficient is 0.2 to 0.5, according to the performance degradation index XC under different operating conditions. r For comprehensive index X r The degree of influence of γ is determined; and 1r +γ 2r +γ 3r =1.

[0132] It should be noted that this formula takes into account the electrical-vibration correlation index XA through a weighted method. r , Sound-operation coupling index XB r and performance degradation index XC r The three aspects of influence are used at the same time, and the sin function is used to map the final result into a bounded range to obtain the comprehensive index X r .

[0133] In the above scheme, the electrical-vibration correlation index XA is calculated based on the vibration data and power quality data. r , which reflects the coordinated working of the electrical and mechanical parts of the equipment, is conducive to identifying potential faults, and calculates the sound-operation coupling index XB based on sound data and operation data r , which can more comprehensively reflect the operating status of the equipment, provide a basis for optimizing and adjusting the equipment operating parameters, and calculate the performance degradation index XC based on wear data, magnetic field data and power quality data r, can more keenly detect early signs of faults, provide a basis for fault location and diagnosis, according to the electrical-vibration correlation index XA r , Sound-operation coupling index XB r and performance degradation index XC r Calculate the comprehensive index X r , which can evaluate the equipment from multiple dimensions and reflect the equipment operation characteristics more comprehensively.

[0134] The model judgment module can build a fault identification self-learning model through the support vector machine learning model, taking the vibration data, power quality data, sound data, operation data, wear data and magnetic field data of the equipment under normal operation and different fault types as input, and converting the comprehensive index X r and the corresponding fault type label as output to train the fault identification self-learning model; the current vibration data, power quality data, sound data, operation data, wear data and magnetic field data of the equipment are input into the trained fault identification self-learning model, and the model outputs the comprehensive index X and the corresponding fault type.

[0135] It should be noted that different fault types include stator winding fault, rotor winding fault, electrical connection fault, bearing fault, rotor imbalance fault and ventilation fault, etc.

[0136] In the above scheme, the vibration data, power quality data, sound data, operation data, wear data and magnetic field data of the equipment under normal operation and different fault types are taken as input, and the comprehensive index X r And the corresponding fault type label is used as the output to train the fault recognition self-learning model, which solves the problem of single data collection type and ignoring the connection between different types of data, resulting in poor model training effect and thus reduced recognition accuracy.

[0137] Update module to obtain the comprehensive index X output in the model judgment module and the corresponding comprehensive index X r , and according to comprehensive index X and comprehensive index X r Calculate the fault index deviation value XD r ; Preset fault indicator deviation threshold XP, according to the fault indicator deviation value XD r And the fault indicator deviation threshold XP determines whether the training data of this fault type needs to be updated.

[0138] Specifically, the method to determine whether the training data needs to be updated is:

[0139] According to comprehensive index X and comprehensive index X r Calculate the fault index deviation value XD r , the formula based on is:

[0140]

[0141] It should be noted that this formula is calculated by calculating the comprehensive index X and the comprehensive index X r The distance between them is used to obtain the fault index deviation value XD r .

[0142] Among them, XD r is the fault index deviation value of the rth fault type.

[0143] According to the fault index deviation value XD r The fault indicator deviation threshold XP is used to determine whether the training data of this fault type needs to be updated. The criteria are:

[0144]

[0145] It should be noted that the method for determining the fault index deviation threshold XP is as follows: obtaining the historical comprehensive index value output by the fault identification self-learning model, calculating the historical comprehensive index value and the comprehensive index X r The fault indicator deviation values ​​between the two are selected, the lower 50% data are screened out and their average value is calculated as the reference value of the fault indicator deviation threshold XP.

[0146] In the above scheme, according to the fault indicator deviation value XD r The fault indicator deviation threshold XP is used to determine whether the training data of this fault type needs to be updated, which solves the problem that the distribution of motor operation data may change over time and due to environmental changes. If it is not updated in time, the model performance will decline.

[0147] The above embodiments may be implemented in whole or in part by software, hardware, firmware or any other combination thereof. When implemented using software, the above embodiments may be implemented in whole or in part in the form of a computer program product. A person of ordinary skill in the art may appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein may be implemented in electronic hardware or in combination with computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution.

[0148] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0149] The above description is only a specific implementation manner of the present application, but the protection scope of the present application is not limited thereto. Any technician familiar with the technical field can easily think of changes or substitutions within the technical scope disclosed in the present application, which should be included in the protection scope of the present application.

Claims

1. A proximity sensor self-learning recognition system based on multi-dimensional signal processing, characterized by: include: A data acquisition module is used to collect vibration data, power quality data, sound data, operation data, wear data and magnetic field data of the equipment under normal operation and different fault types; Index calculation module, which can calculate the vibration deviation index YA based on vibration data r , calculate the vibration deviation index YA r The method is: Vibration data includes vibration acceleration signal AA ra , vibration peak AB r , RMS normal value AC and normal peak maximum value AD; Calculate the vibration deviation index YA based on the vibration data r , the formula based on is: Among them, YA r is the vibration deviation index of the equipment in the rth operating state, r is the serial number corresponding to the different operating states of the equipment, and the value is a positive integer; AA ra is the vibration acceleration signal corresponding to the ath sampling point under the rth operating state of the equipment; a is the serial number corresponding to different sampling points, and its value is [1, b]; b is the total number of sampling points, and its value is a positive integer; AB r is the vibration peak value of the equipment in the rth operating state; Calculation of power quality deviation index YB based on power quality data r , calculate the power quality deviation index YB r The method is: Power quality data including active power BA r , reactive power BB r , rated power factor BC, harmonic voltage effective value BD rd , Fundamental voltage effective value BE r , Maximum voltage per unit time BF r , minimum voltage value per unit time BC r And rated voltage value BH; Calculate the power quality deviation index YB based on the power quality data r , the formula based on is: Among them, YB r BA is the power quality deviation index of the equipment in the rth operating state; r is the active power of the equipment in the rth operating state; BB r is the reactive power of the equipment in the rth operating state; BE r BF is the effective value of the fundamental voltage of the equipment in the rth operating state; r BG is the maximum voltage per unit time in the rth operating state of the equipment; r BD is the minimum voltage per unit time in the rth operating state of the equipment; rd is the effective value of the dth harmonic voltage under the rth operating state of the equipment, d is the serial number corresponding to different harmonic voltages, and its value is [2, f r ]; f r is the total number of harmonic voltages in the equipment's rth operating state, and is a positive integer greater than 1; According to the vibration deviation index YA r And power quality deviation index YB r Calculation of electrical-vibration correlation index XA r ; Calculate the sound abnormality index YC based on the sound data r , according to the operation data and abnormal sound index YC r Calculation of sound-operation coupling index XB r ; Calculate the wear distribution index YD based on the wear data r , calculate the magnetic field-power quality deviation index YE based on the magnetic field data and power quality data r , according to the wear distribution index YD r And magnetic field-power quality deviation index YE r Calculate the performance degradation index XC r ; According to the electrical-vibration correlation index XA r , Sound-operation coupling index XB r and performance degradation index XC r Calculate the comprehensive index E r ; The model judgment module can build a fault identification self-learning model through a support vector machine learning model, take the vibration data, power quality data, sound data, operation data, wear data and magnetic field data of the equipment in normal operation state and different fault type operation state as input, take the comprehensive index c and the corresponding fault type label as output, and train the fault identification self-learning model; input the current vibration data, power quality data, sound data, operation data, wear data and magnetic field data of the equipment into the trained fault identification self-learning model, and the model outputs the comprehensive index X and the corresponding fault type; Update module to obtain the comprehensive index X output in the model judgment module and the corresponding comprehensive index X r , and according to comprehensive index X and comprehensive index E r Calculate the fault index deviation value XD r ; Preset fault indicator deviation threshold XP, according to the fault indicator deviation value XD r And the fault indicator deviation threshold XP determines whether the training data of this fault type needs to be updated.

2. The proximity sensor self-learning recognition system based on multi-dimensional signal processing according to claim 1, characterized in that: Calculation of electrical-vibration correlation index XA r The method is: According to the vibration deviation index YA r And power quality deviation index YB r Calculation of electrical-vibration correlation index XA r , the formula based on is: XA r =a 1r ×YA r +a r2 ×YB r Among them, XA r is the electrical-vibration correlation index of the equipment under the rth operating state; α 1r YA is the vibration deviation index of the equipment in the rth operating state r The weight coefficient is 0.3~0.7; α r2 YB is the power quality deviation index of the equipment in the rth operating state r The weight coefficient is 0.3 to 0.7; and α r1 +α r2 =1.

3. The proximity sensor self-learning recognition system based on multi-dimensional signal processing according to claim 2 is characterized in that: Calculate the sound abnormality index YC r The method is: Sound data includes actual sound pressure CA r , reference sound pressure CB, sound signal fundamental frequency amplitude CC r , Sound signal harmonic amplitude CD rh , frequency band signal-to-noise ratio CE rk , signal-to-noise ratio normal range upper limit CE max and the lower limit of the normal range of signal-to-noise ratio CE min ; Calculate the sound abnormality index YC based on the sound data r , the formula based on is: Among them, YC r The sound abnormality index of the equipment in the rth operating state; the actual sound pressure CA r is the actual sound pressure of the equipment in the rth operating state; CC r CD is the fundamental frequency amplitude of the sound signal in the rth operating state of the equipment; rh is the amplitude of the hth harmonic in the sound signal under the rth operating state of the equipment, h is the serial number corresponding to different harmonics in the sound signal, and its value is [2, i r ];i r is the total number of harmonics in the sound signal of the equipment in the rth operating state, and its value is a positive integer greater than 1; CE rk is the signal-to-noise ratio value of the kth frequency band in the rth operating state of the device, k is the serial number corresponding to different frequency bands, and its value is [1, l r ]; l r It is the total number of frequency bands in the rth operating state of the device, and its value is a positive integer.

4. The proximity sensor self-learning recognition system based on multi-dimensional signal processing according to claim 3 is characterized in that: Calculation of sound-operation coupling index XB r The method is: Operating data includes actual operating temperature DA r , Normal operating temperature DB r , actual speed DC r and rated speed DD; According to the operation data and abnormal sound index YC r Calculation of sound-operation coupling index XB r , the formula based on is: Among them, XB r DA is the sound-operation coupling index of the equipment in the rth operating state; r is the actual operating temperature of the equipment in the rth operating state; DB r is the normal operating temperature of the equipment in the rth operating state; DC r is the actual speed of the equipment in the rth operating state.

5. The proximity sensor self-learning recognition system based on multi-dimensional signal processing according to claim 4 is characterized in that: Calculate the wear distribution index YD r The method is: Wear data including average wear particle diameter EA rm , Wear particle quantity EB rm and the total number of wear particles EC under normal conditions; Calculate the wear distribution index YD based on the wear data r , the formula based on is: Among them, YD r is the wear distribution index of the equipment under the rth operating state; EA rm is the average diameter of wear particles in the mth size range under the rth operating state of the equipment, m is the serial number corresponding to different size ranges, and its value is [1, n]; n is the total number of different size ranges; EB rm is the number of wear particles in the mth size range under the rth operating state of the equipment; β 1r is the weight coefficient of the wear particle size index under the equipment's rth operating state, with a value of 0.4 to 0.6; β 2r is the weight coefficient of the wear particle distribution entropy under the equipment's rth operating state, ranging from 0.4 to 0.6; and β 1r +β 2r =1.

6. The proximity sensor self-learning recognition system based on multi-dimensional signal processing according to claim 5, characterized in that: Calculate the performance degradation index XC r The method is: Magnetic field data includes actual magnetic field strength FA r and normal magnetic field strength FB; The power quality data also includes the actual insulation resistance GA r And normal insulation resistance GB; Calculate the magnetic field-power quality deviation index YE based on the magnetic field data and power quality data r , the formula based on is: Among them, YE r FA is the magnetic field-power quality deviation index of the equipment in the rth operating state; r is the actual magnetic field strength of the equipment in the rth operating state; GA r is the actual insulation resistance of the equipment in the rth operating state; According to the wear distribution index YD r And magnetic field-power quality deviation index YE r Calculate the performance degradation index XC r , the formula based on is: Among them, XC r It is the performance degradation index of the equipment in the rth operating state.

7. The proximity sensor self-learning recognition system based on multi-dimensional signal processing according to claim 6, characterized in that: Calculate the comprehensive index X r The method is: According to the electrical-vibration correlation index XA r , Sound-operation coupling index XB r and performance degradation index XC r Calculate the comprehensive index E r , the formula based on is: Among them, E r is the comprehensive index of the equipment under the rth operating state; γ 1r XA is the electrical-vibration correlation index of the equipment under the rth operating state r The weight coefficient is 0.3~0.5; γ 2r The sound-operation coupling index XB of the equipment in the rth operating state r The weight coefficient is 0.2~0.5; γ 3r XC is the performance degradation index of the equipment in the rth operating state r The weight coefficient is 0.2 to 0.5; and γ 1r +γ 2r +γ 3r =1.

8. The proximity sensor self-learning recognition system based on multi-dimensional signal processing according to claim 7, characterized in that: The method to determine whether the training data needs to be updated is: According to comprehensive index X and comprehensive index E r Calculate the fault index deviation value XD r , the formula based on is: Among them, XD r is the fault index deviation value of the rth fault type; According to the fault index deviation value XD r The fault indicator deviation threshold XP is used to determine whether the training data of this fault type needs to be updated. The criteria are:

Images