Air compressor management method and system based on remote monitoring and fault self-healing and medium

Abstract

The invention provides an air compressor management method and system based on remote monitoring and fault self-healing and a medium, and the method comprises the steps: firstly monitoring the operation state of an air compressor through a multi-modal sensor, extracting key features based on feature processing, and combining the key features into multi-source heterogeneous data; secondly, a fault prediction model is adopted, modeling is conducted through a mechanical-thermal-electric coupling relation, and probability values of faults such as bearing abrasion and oil way blockage are output; then, when the fault probability value exceeds a threshold value, automatically generating a parameterized self-healing instruction based on a knowledge base driver, and driving to execute a self-healing operation; finally, after the self-healing operation is executed, the improvement rate is quantified through multi-source feedback data, model parameters of the fault prediction model are dynamically optimized based on the improvement rate, and the self-evolution capacity is formed. The predictive self-maintenance function is achieved, and the maintenance cost of the air compressor is reduced.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of air compressors, and more particularly, to an air compressor management method, system and medium based on remote monitoring and fault self-healing. BACKGROUND

[0002] Industrial air compressors, as key power sources, their running stability and energy efficiency management directly affect production safety and cost. The traditional operation and maintenance mode of air compressors has problems of fault response lag, insufficient energy efficiency optimization and high operation and maintenance cost.

[0003] At present, although the existing technology introduces remote monitoring, the cloud platform only realizes data visualization, and needs manual retrieval of historical data for comparison, lacking intelligent analysis capability; due to the lack of preventive maintenance decision, the equipment usually needs to adopt shutdown maintenance, resulting in production line interruption loss.

[0004] Therefore, an intelligent management technology of air compressors based on remote monitoring and fault self-healing is urgently needed. SUMMARY

[0005] In view of the above problems, the purpose of the present application is to provide an air compressor management method, system and medium based on remote monitoring and fault self-healing, through vibration, temperature and current three-domain collaborative analysis, realizing early diagnosis of air compressor faults, combining knowledge base matching corresponding self-healing operation instructions, to drive self-healing devices to realize online elimination of hidden dangers; building a predictive maintenance, automatic disposal and autonomous optimization model, realizing intelligent management of air compressors.

[0006] The first aspect of the present application provides an air compressor management method based on remote monitoring and fault self-healing, the method comprising:

[0007] Based on the preset sensor monitoring operation data, after the preset feature preprocessing, the multi-source heterogeneous data is fused;

[0008] The multi-source heterogeneous data is input into a pre-trained fault prediction model to obtain a fault probability of a preset fault;

[0009] If the fault probability exceeds a preset fault threshold, a self-healing control instruction is generated based on a preset knowledge base;

[0010] According to the self-healing control instruction, a self-healing device is driven to execute an instruction action;

[0011] In response to a preset delay time, multi-source feedback data is collected and processed;

[0012] According to the multi-source feedback data, an improvement rate is obtained;

[0013] The training weight is set based on the improvement rate, the fault probability and the self-recovery control instruction are combined as a training data set, and the weight parameters of the fault prediction model are trained and optimized.

[0014] In the scheme, the operation data includes mechanical vibration signals, thermal distribution images, and current waveform data, and the preset feature preprocessing is specifically:

[0015] The mechanical vibration signals are collected by a three-axis acceleration sensor, and environmental noise is eliminated by a preset filtering algorithm.

[0016] Based on the preset envelope demodulation logic, the mechanical impact features are obtained according to the mechanical vibration signals.

[0017] The thermal distribution image of the motor is photographed by an infrared thermal imager, and the temperature interval of the motor winding and bearing is identified and obtained.

[0018] Based on the preset region segmentation algorithm, the thermal distribution features of the winding and bearing are extracted according to the thermal distribution image.

[0019] The motor operating current is measured by a current transformer, and the current waveform is obtained.

[0020] Based on the preset FFT transform, the current harmonic features are obtained according to the current waveform.

[0021] In the scheme, the multi-source heterogeneous data is input into the pre-trained fault prediction model to obtain the fault probability of the preset fault, and the specific process is as follows:

[0022] The mechanical impact features are input into the preset time domain convolution branch, and the periodic vibration mode is obtained through the hollow convolution layer.

[0023] The thermal distribution features are input into the spatial attention branch to generate the weighted feature map of the temperature field key area.

[0024] The current harmonic features are input into the frequency domain analysis branch, and the transient harmonic energy is obtained through wavelet transform.

[0025] The periodic vibration mode, the weighted feature map, and the transient harmonic energy are input into the preset mechanical-thermal-electric coupling relationship model to obtain the probability distribution of the characteristic fault, wherein the characteristic fault includes bearing wear, oil line blockage, and coil short circuit.

[0026] In the scheme, the self-recovery control instruction is generated based on the preset knowledge base, and the specific process is as follows:

[0027] According to the fault type, the solution template of the preset self-recovery knowledge base is matched, wherein the self-recovery knowledge base includes a fault feature and execution parameter mapping table.

[0028] When the bearing wear is determined, an oil line purification instruction is generated;

[0029] When the oil line blockage is determined, a backwashing instruction group is generated;

[0030] When the coil short circuit is determined, a load transfer instruction is generated.

[0031] In the scheme, the driving self-healing device according to the self-healing control instruction executes the instruction action, and further comprises:

[0032] When the oil line purification instruction is determined, the gear pump start-stop and the heater temperature are controlled according to the instruction parameters, and the oil quality sensor data is monitored in real time;

[0033] When the backwashing instruction group is determined, the electromagnetic valve actuator is driven to perform opening and closing cycles according to the instruction sequence and specified time length and interval;

[0034] When the load transfer instruction is determined, the output power is adjusted based on the preset cloud platform scheduling.

[0035] In the scheme, the improvement rate is obtained according to the multi-source feedback data, and specifically:

[0036] The vibration frequency spectrum of the mechanical vibration signal before and after the self-healing operation is analyzed, and the characteristic frequency band energy attenuation rate is calculated;

[0037] The thermal distribution image before and after the self-healing operation is analyzed to obtain the area shrinkage rate of the preset high-temperature area and the maximum temperature difference;

[0038] The harmonic energy proportion change rate of the current waveform before and after the self-healing operation is analyzed;

[0039] The multi-source feedback data is obtained by combining the energy attenuation rate, the area shrinkage rate, the maximum temperature difference and the harmonic energy proportion change rate;

[0040] According to the fault type, a weighting coefficient is obtained;

[0041] Based on the preset linear weighting algorithm, the improvement rate is obtained according to the weighting coefficient and the multi-source feedback data.

[0042] The second aspect of the present application provides an air compressor management system based on remote monitoring and fault self-healing, which comprises an air compressor management method program based on remote monitoring and fault self-healing.

[0043] Based on the preset sensor monitoring operation data, the multi-source heterogeneous data is fused after preset feature preprocessing;

[0044] inputting the multi-source heterogeneous data into a pre-trained fault prediction model to obtain a fault probability of a preset fault;

[0045] if the fault probability exceeds a preset fault threshold, generating a self-healing control instruction based on a preset knowledge base;

[0046] driving a self-healing device to perform an instruction action according to the self-healing control instruction;

[0047] in response to a preset delay time, collecting and processing multi-source feedback data;

[0048] obtaining an improvement rate according to the multi-source feedback data;

[0049] setting a training weight based on the improvement rate, combining the fault probability and the self-healing control instruction as a training data set, and training and optimizing the weight parameters of the fault prediction model.

[0050] In the scheme, the operation data includes mechanical vibration signals, thermal distribution images, and current waveform data, and the preset feature preprocessing is specifically:

[0051] The mechanical vibration signals are collected by a three-axis acceleration sensor, and environmental noise is eliminated by a preset filtering algorithm;

[0052] Based on a preset envelope demodulation logic, mechanical impact features are obtained according to the mechanical vibration signals;

[0053] The thermal distribution image of the motor is photographed by an infrared thermal imager, and the temperature interval of the motor winding and bearing is identified and obtained;

[0054] Based on a preset region segmentation algorithm, the thermal distribution features of the winding and bearing are extracted according to the thermal distribution image;

[0055] The motor operating current is measured by a current transformer to obtain the current waveform;

[0056] Based on a preset FFT transform, current harmonic features are obtained according to the current waveform.

[0057] In the scheme, the multi-source heterogeneous data is input into a pre-trained fault prediction model to obtain a fault probability of a preset fault, which is specifically:

[0058] The mechanical impact features are input into a preset time domain convolution branch to obtain a periodic vibration mode through a hollow convolution layer;

[0059] The thermal distribution features are input into a spatial attention branch to generate a weighted feature map of the temperature field key area;

[0060] The current harmonic features are input into a frequency domain analysis branch to obtain transient harmonic energy through wavelet transform;

[0061] inputting the periodic vibration mode, the weighted feature map and the transient harmonic energy into a preset mechanical-thermal-electric coupling relationship model to obtain a probability distribution of a feature fault, wherein the feature fault includes bearing wear, oil line blockage and coil short circuit.

[0062] The third aspect of the present application provides a computer readable storage medium, wherein the computer readable storage medium comprises a remote monitoring and fault self-healing based air compressor management method program, and the remote monitoring and fault self-healing based air compressor management method program is executed by a processor to realize the steps of the remote monitoring and fault self-healing based air compressor management method according to any one of the above.

[0063] The present application provides a remote monitoring and fault self-healing based air compressor management method, system and medium, which first monitors the running state of the air compressor through a multi-modal sensor, extracts key features based on feature processing, and combines them into multi-source heterogeneous data; secondly, a fault prediction model is used to output the probability values of bearing wear, oil line blockage and other faults through mechanical-thermal-electric coupling relationship modeling; then, when the fault probability value exceeds the threshold value, a parameterized self-healing instruction is automatically generated based on a knowledge base to drive the execution of the self-healing operation; finally, after the self-healing operation is executed, the improvement rate is quantified through multi-source feedback data, and the model parameters of the fault prediction model are dynamically optimized based on the improvement rate to form a self-evolution capability; the predictive self-maintenance function is realized, and the maintenance cost of the air compressor is reduced. BRIEF DESCRIPTION OF DRAWINGS

[0064] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the following will briefly introduce the drawings needed to be used in the embodiments. It should be understood that the following drawings only show some embodiments of the present application, and therefore should not be regarded as a limitation on the scope.

[0065] Figure 1 A flow chart of a remote monitoring and fault self-healing based air compressor management method of the present application is shown;

[0066] Figure 2 A feature preprocessing flow chart of running data provided by an embodiment of the present application is shown;

[0067] Figure 3 A running flow chart of a fault prediction model provided by an embodiment of the present application is shown;

[0068] Figure 4 A block diagram of a remote monitoring and fault self-healing based air compressor management system of the present application is shown. DETAILED DESCRIPTION

[0069] With reference to the accompanying drawings, the technical solutions in the embodiments of the present application will be clearly and completely described below, obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments of the present application, all other embodiments obtained by those skilled in the art without creative work fall within the scope of the present application.

[0070] Unless otherwise defined, all terms (including technical and scientific terms) used in the embodiments of the present application have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs. It will be further understood that terms, such as those defined in commonly used dictionaries, should be interpreted as having a meaning that is consistent with their meaning in the context of the relevant art and will not be interpreted in an idealized or overly formal sense unless expressly so defined in the embodiments of the present application.

[0071] The terms "first", "second", and similar terms used in the embodiments of the present application do not denote any order, quantity, or importance, but are only used to distinguish different constituent parts. The terms "one", "a", or "the" and similar terms do not denote a quantity restriction, but mean that at least one exists. Similarly, the terms "include" or "contain" and similar terms mean that the elements or objects before the terms encompass the elements or objects listed after the terms and their equivalents, and do not exclude other elements or objects.

[0072] The terms "connect" or "connected" and similar terms are not limited to physical or mechanical connections, but can include electrical connections, whether direct or indirect. The steps before or after the methods in the embodiments of the present application do not necessarily proceed in order. On the contrary, various steps can be processed in reverse order or simultaneously. In addition, other operations can be added to these processes, or one or more steps can be removed from these processes.

[0073] In addition, the functional modules in each embodiment of the present application can be integrated together to form an independent part, or each module can exist independently, or two or more modules can be integrated to form an independent part.

[0074] Figure 1 A flowchart of a kind of air compressor management method based on remote monitoring and fault self-recovery of the present application is shown.

[0075] As shown in Figure 1 The first aspect of the present application discloses an air compressor management method based on remote monitoring and fault self-recovery, the method comprises:

[0076] S102, based on the preset sensor monitoring operation data, through the preset feature preprocessing, fusion is obtained multi-source heterogeneous data;

[0077] S104, inputting the multi-source heterogeneous data into a pre-trained fault prediction model to obtain a fault probability of a preset fault;

[0078] S106, if the fault probability exceeds a preset fault threshold, generating a self-healing control instruction based on a preset knowledge base;

[0079] S108, driving a self-healing device to execute an instruction action according to the self-healing control instruction;

[0080] S110, in response to a preset delay time, collecting and processing multi-source feedback data;

[0081] S112, obtaining an improvement rate according to the multi-source feedback data;

[0082] S114, setting a training weight value based on the improvement rate, combining the fault probability and the self-healing control instruction as a training data set, training and optimizing a weight parameter of the fault prediction model.

[0083] It should be noted that, in the air compressor unit network deployment phase, a three-axis vibration sensor is installed on the motor bearing seat, an infrared thermal imager is arranged on the motor shell, and a current transformer is connected in series in the power cable, forming a holographic monitoring network. During equipment operation, raw sensor signals are collected based on a preset collection frequency. As an implementation manner, vibration data are first subjected to a preset Butterworth low-pass filter to eliminate high-frequency noise, and then subjected to Hilbert transform to extract impact features; the thermal imaging data are segmented into winding and bearing regions by using a region growing algorithm, and the maximum temperature difference gradient between the regions is calculated; the current signal is subjected to a preset FFT transform to generate a frequency spectrum graph, and the 3 / 5 / 7 harmonic amplitude ratios are extracted. The preprocessed feature vectors and the raw data are combined to form multi-source heterogeneous data, which are uploaded to a cloud platform, and based on a cloud pre-trained fault prediction model, probability values of three types of faults, i.e., bearing wear, oil line blockage and coil short circuit, are output. When any fault probability exceeds a set threshold, a preset knowledge base is automatically matched based on the fault type to obtain a self-healing control instruction; and a corresponding self-healing execution device is driven based on the self-healing control instruction. After the self-healing operation is executed, sensor data and oil viscosity data are re-collected, and a total improvement rate is generated by using a linear weighting algorithm. Based on the improvement rate, together with the original fault probability and the execution parameter, a training sample is formed, a neural network full connection layer weight coefficient is optimized by using a back propagation algorithm, and a complete closed loop from state perception to model evolution is formed.

[0084] Figure 2 A feature preprocessing flowchart of operation data provided by an embodiment of the application is shown.

[0085] According to an embodiment of the application, as Figure 2As shown, the operation data includes mechanical vibration signals, thermal distribution images and current waveform data, and the preset feature preprocessing is specifically:

[0086] S202, collect mechanical vibration signals through a three-axis acceleration sensor, and eliminate environmental noise through a preset filtering algorithm;

[0087] S204, based on a preset envelope demodulation logic, obtain mechanical impact features according to the mechanical vibration signals;

[0088] S206, capture thermal distribution images of the motor through an infrared thermal imager, and identify and obtain temperature intervals of motor windings and bearings;

[0089] S208, based on a preset region segmentation algorithm, extract thermal distribution features of windings and bearings according to the thermal distribution images;

[0090] S210, measure motor operating current through a current transformer to obtain a current waveform;

[0091] S212, based on a preset FFT transform, obtain current harmonic features according to the current waveform.

[0092] It should be noted that mechanical vibration signal processing starts from collecting X, Y and Z direction original waveforms through a three-axis acceleration sensor, the signal is first filtered through a low-pass filter with a cutoff frequency of 5kHz to eliminate environmental high-frequency interference, and then envelope demodulation technology based on Hilbert transform is used to perform empirical mode decomposition on the filtered signal to obtain an envelope spectrum, and the kurtosis index of the bearing outer ring fault feature band in the envelope spectrum is extracted as the mechanical impact feature. In the thermal distribution processing link, the infrared thermal imager captures the global thermal map of the motor at a set interval, and an improved watershed algorithm is used to segment the image into four regions of stator winding, rotor core, front and rear bearings, and the average temperature of each region and the maximum temperature difference between adjacent regions are calculated. The current waveform analysis obtains the instantaneous values of three-phase current through the coil transformer, and performs windowed FFT transform on A, B and C phase currents to generate 0-1kHz frequency spectrum, and calculates the energy proportion of specific harmonics using harmonic subgroup analysis method. Finally, the mechanical impact features, thermal zone temperature difference and harmonic distortion rate are standardized and packaged as a data packet.

[0093] Figure 3 The operation flowchart of the fault prediction model provided by the embodiment of the application is shown.

[0094] According to the embodiment of the application, as Figure 3 shown, the multi-source heterogeneous data is input into the pre-trained fault prediction model to obtain the fault probability of the preset fault, and the specific process is as follows:

[0095] S302, input the mechanical impact feature into a preset time domain convolution branch, and obtain a periodic vibration mode through a cavity convolution layer;

[0096] S304, input the thermal distribution feature into a spatial attention branch, and generate a weighted feature map of a temperature field key area;

[0097] S306, input the current harmonic feature into a frequency domain analysis branch, and obtain transient harmonic energy through wavelet transform;

[0098] S308, input the periodic vibration mode, the weighted feature map and the transient harmonic energy into a preset mechanical-thermal-electric coupling relationship model, and obtain a probability distribution of a characteristic fault, wherein the characteristic fault includes bearing wear, oil line blockage and coil short circuit.

[0099] It should be noted that the multi-modal fault prediction model provided in the embodiment adopts a three-branch parallel architecture. As an implementation, the vibration feature is input into the time domain convolution branch, and the periodic mode of bearing impact is extracted through 3 layers of cavity convolution with an expansion rate of 2, and vibration state data containing fault characteristic frequency is output. As an implementation, the thermal distribution feature is input into the spatial attention branch, and after extracting the basic feature, the bearing area feature map is given N times weight, and the background thermal interference is suppressed, wherein N is greater than 2. As an implementation, the current feature is input into the frequency domain analysis branch, and a wavelet transform function is used for multi-scale decomposition, and wavelet packet energy entropy of multiple subbands is calculated, and harmonic energy is quantified. The three features are input into a graph neural network in a fusion layer, and a topology structure with vibration nodes as the core, temperature nodes as the auxiliary and current nodes as the verification is constructed. Then, the coupling relationship that the vibration impact is enhanced when the bearing wears, the local temperature rises and the current harmonic increases is modeled through a message passing mechanism. The multi-dimensional fusion features of the graph neural network are received by the full connection layer, and three types of prediction results of bearing wear probability, oil line blockage probability and coil short circuit probability are output, and when any probability value exceeds a set threshold, an early warning is triggered.

[0100] According to the embodiment of the present application, the self-healing control instruction is generated based on a preset knowledge base, specifically:

[0101] According to the fault type, a solution template of a preset self-healing knowledge base is matched, wherein the self-healing knowledge base includes a fault feature and an execution parameter mapping table;

[0102] When it is determined that the bearing is worn, an oil line purification instruction is generated;

[0103] When it is determined that the oil line is blocked, a backwashing instruction set is generated;

[0104] When it is determined that the coil is short-circuited, a load transfer instruction is generated.

[0105] It should be noted that the knowledge base driven instruction generation provided in the embodiment comprises three-level operation logic. First, according to the highest probability fault type, the solution library is retrieved, for example, bearing wear of 0.2-0.5mm corresponds to oil line purification of 2 cycles and viscosity adjustment to 46cSt. Then, after the solution matching is successful, the parameter optimization stage is entered, for example, for the oil line blockage case, the backwashing parameters are dynamically set according to the blockage degree index, for example, for mild blockage, a single cycle of 5-second pulse is executed, and for severe blockage, three cycles of 10-second pulse are executed; for the bearing wear case, the purification duration is determined in combination with the vibration energy value, for example, 10 minutes of purification corresponds to 100mV of vibration amount. Finally, a machine-readable instruction set is generated, the oil line purification instruction comprises a centrifugal pump speed curve, a heater temperature set value or a target viscosity threshold value; the backwashing instruction specifies the opening and closing timing of the electromagnetic valve; the load transfer instruction specifies the standby unit startup delay and power switching gradient.

[0106] According to the embodiment of the present application, the self-healing device is driven to execute the instruction action according to the self-healing control instruction, which further comprises:

[0107] When it is judged as an oil line purification instruction, the gear pump start-stop and the heater temperature are controlled according to the instruction parameters, and the oil quality sensor data is monitored in real time;

[0108] When it is judged as a backwashing instruction set, the electromagnetic valve actuator is driven to open and close in cycles according to the instruction sequence and specified time and interval;

[0109] When it is judged as a load transfer instruction, the output power is adjusted based on the preset cloud platform scheduling.

[0110] It should be noted that when the oil line purification instruction is executed, the gear pump is accelerated to the target speed with an S-shaped speed curve, and at the same time, the heater power is dynamically adjusted according to the oil temperature sensor feedback through the PID regulator; in addition, the online viscometer measures the value every 30 seconds, and when the continuous 3 readings meet the standard, the purification is automatically stopped. When the backwashing instruction is executed, the electromagnetic valve is driven according to the preset timing, the preset compressed air pressure is maintained during the opening stage to continuously blow, and the drain valve is opened to release pressure during the closing stage; the pressure transmitter data is read after each cycle to judge the unblocking state. The load transfer operation is written into the frequency converter register through the preset communication protocol, first, the frequency of the fault unit is lowered, for example, to 50% at a rate of 5% per second; at the same time, the standby unit frequency is increased to take over the load, and the pressure sensor monitors the pipe network fluctuation in real time during the transfer process, and if the pressure change rate is out of limit, the PID feedforward compensation algorithm is enabled.

[0111] According to the embodiment of the present application, the improvement rate is obtained according to the multi-source feedback data, specifically:

[0112] The vibration frequency spectrum of the mechanical vibration signal before and after the self-healing operation is analyzed, and the characteristic frequency band energy attenuation rate is calculated;

[0113] analyze the thermal distribution images before and after the self-recovery operation to obtain a preset area shrinkage rate of a high-temperature area and a maximum temperature difference;

[0114] analyze the harmonic energy proportion change rate of the current waveform before and after the self-recovery operation;

[0115] The multi-source feedback data is obtained by combining the energy decay rate, the area shrinkage rate, the maximum temperature difference and the harmonic energy proportion change rate.

[0116] According to the fault type, a weighting coefficient is obtained.

[0117] Based on a preset linear weighting algorithm, an improvement rate is obtained according to the weighting coefficient and the multi-source feedback data.

[0118] It should be noted that the improvement rate quantitative evaluation provided in this embodiment executes a four-step analysis process. First, for the mechanical vibration signals before and after the self-recovery operation, the power spectrum density integral ratio before and after execution is calculated in the preset bearing characteristic frequency band to obtain the energy decay rate. Second, for the thermal distribution images before and after the self-recovery operation, the high-temperature area with a temperature set temperature threshold is extracted through image binarization processing, and the pixel area reduction ratio and the maximum temperature difference reduction value in the area after execution are calculated. Third, for the current waveform before and after the self-recovery operation, the 10-second current waveforms before and after execution are analyzed, and the harmonic content proportion is calculated for 5 times to obtain the harmonic distortion rate reduction value. Finally, a weight coefficient is assigned according to the fault type. As an implementation manner, the vibration weight is 0.6, the temperature weight is 0.3, and the current weight is 0.1 for the bearing wear scenario; the vibration weight is 0.3, the temperature weight is 0.5, and the current weight is 0.2 for the oil line blockage scenario; and the improvement rate is calculated according to a preset linear weighting algorithm.

[0119] It is worth mentioning that it also includes optimizing the fault prediction model based on a federated learning framework, specifically:

[0120] Each air compressor terminal locally trains a model subnetwork and uploads weight parameters to a preset aggregation cloud platform;

[0121] The weight parameters are fused by a security averaging algorithm of the cloud platform, and the weight parameters of the global model are distributed to each air compressor terminal;

[0122] When a new fault mode is added, an incremental learning mechanism is triggered.

[0123] It should be noted that the embodiment adopts a federated learning framework, and implements three links including local training, secure aggregation, and incremental update. Each air compressor terminal locally loads a basic model, fine-tunes the full connection layer parameters using recent 30-day operation data, and uploads the encrypted weight parameters after training. After the aggregation server of the cloud platform receives the weight parameters of at least 10 terminals, the secure multi-party computing technology is used to perform weighted average to generate a new global model. When a new fault mode is detected, for example, motor brush wear, the incremental learning mechanism is triggered to append a new classification branch after the original full connection layer, and a small amount of labeled samples are used for transfer learning. The updated model is compressed by a preset compression operation to generate a lightweight version and is pushed to the air compressor terminal.

[0124] It is worth mentioning that before performing the self-healing operation, a virtual verification is also performed, specifically:

[0125] Simulate the execution of the self-healing instruction based on the preset digital twin model to predict the change curve of the key parameters;

[0126] When the prediction shows that the pressure fluctuation is out of limit or the temperature rises sharply, the execution parameters are automatically adjusted;

[0127] When the change curve of the key parameters is determined to be a high-risk operation, the preset backup scheme is automatically enabled;

[0128] Record the deviation value of the virtual verification result and the actual execution data for calibrating the accuracy of the twin model.

[0129] It should be noted that the embodiment also provides a virtual verification mechanism based on a digital twin architecture. First, a device physical model is constructed on the cloud platform, including motor electromagnetic characteristic equations, bearing dynamics equations, pipeline fluid equations, etc. When receiving the self-healing instruction, the instruction parameters are injected into the virtual space for real-time simulation, for example, inputting backwash pressure 0.6 MPa, predicting the pipeline pressure fluctuation curve through the computational fluid dynamics model; inputting the load transfer parameter, solving the motor torque-speed characteristic curve change. If the simulation shows that the pressure peak value exceeds 120% of the rated value or the temperature gradient is the preset temperature change rate, the system automatically generates a parameter optimization suggestion. In addition, when it is determined to be a high-risk operation, for example, emergency shutdown in short circuit, the system preloads multiple sets of disposal plans, for example, the main scheme executes device hard shutdown, and the alternative scheme executes power soft landing combined with mechanical braking operation. During actual execution, the twin model synchronously receives sensor data for online verification, and when the pressure prediction error or temperature error exceeds the set range, the turbulent flow coefficient or thermal conductivity coefficient of the fluid model is automatically corrected to realize model self-calibration.

[0130] Figure 4 A block diagram of an air compressor management system based on remote monitoring and fault self-healing is shown.

[0131] AsFigure 4 As shown, the second aspect of the present application discloses a remote monitoring and fault self-healing based air compressor management system 4, comprising a memory 41 and a processor 42, wherein the memory comprises a remote monitoring and fault self-healing based air compressor management method program, and the remote monitoring and fault self-healing based air compressor management method program is executed by the processor to implement the following steps:

[0132] Based on the preset sensor monitoring operation data, after the preset feature preprocessing, the multi-source heterogeneous data is fused;

[0133] The multi-source heterogeneous data is input into a pre-trained fault prediction model to obtain a fault probability of a preset fault;

[0134] If the fault probability exceeds a preset fault threshold, a self-healing control instruction is generated based on a preset knowledge base;

[0135] According to the self-healing control instruction, a self-healing device is driven to execute an instruction action;

[0136] In response to a preset delay time, multi-source feedback data is collected and processed;

[0137] According to the multi-source feedback data, an improvement rate is obtained;

[0138] Based on the improvement rate, a training weight is set, the fault probability and the self-healing control instruction are combined as a training data set, and the weight parameters of the fault prediction model are trained and optimized.

[0139] It should be noted that, in the air compressor unit network deployment stage, a three-axis vibration sensor is installed on the motor bearing seat, an infrared thermal imager is arranged on the motor shell, and a current transformer is connected in series in the power cable, forming a holographic monitoring network. During equipment operation, the original sensing signals are collected based on a preset collection frequency. As an implementation manner, the vibration data are first subjected to a preset Butterworth low-pass filter to eliminate high-frequency noise, and then subjected to Hilbert transform to demodulate the envelope curve and extract impact features; the thermal imaging data are subjected to a region growing algorithm to segment the winding and bearing regions, and the maximum temperature difference gradient between the regions is calculated; the current signal is subjected to a preset FFT transform to generate a frequency spectrum graph, and the 3 / 5 / 7 harmonic amplitude ratios are extracted. The preprocessed feature vectors and the original data are uploaded to a cloud platform as multi-source heterogeneous data, and based on a pre-trained fault prediction model on the cloud, probability values of three types of faults, i.e., bearing wear, oil line blockage and coil short circuit, are output. When the probability of any fault exceeds a preset threshold, a preset knowledge base is automatically matched based on the fault type to obtain a self-healing control instruction, and the corresponding self-healing execution device is driven based on the self-healing control instruction. After the self-healing operation is executed, the sensor data and the oil viscosity data are re-collected, and a total improvement rate is generated through a linear weighting algorithm. Based on the improvement rate, the original fault probability and the execution parameters together constitute a training sample, and the neural network full connection layer weight coefficients are optimized through a back propagation algorithm to form a complete closed loop from state perception to model evolution.

[0140] According to the embodiment of the application, the operation data include mechanical vibration signals, thermal distribution images and current waveform data, and the preprocessed features are specifically:

[0141] The mechanical vibration signals are collected by a three-axis acceleration sensor, and environmental noise is eliminated through a preset filtering algorithm;

[0142] Based on a preset envelope demodulation logic, mechanical impact features are obtained from the mechanical vibration signals;

[0143] The thermal distribution images of the motor are photographed by an infrared thermal imager, and the temperature intervals of the motor winding and bearing are identified and obtained;

[0144] Based on a preset region segmentation algorithm, thermal distribution features of the winding and bearing are extracted from the thermal distribution images;

[0145] The motor operating current is measured by a current transformer, and a current waveform is obtained;

[0146] Based on a preset FFT transform, current harmonic features are obtained from the current waveform.

[0147] It should be noted that mechanical vibration signal processing starts from three-axis acceleration sensor collecting X, Y and Z direction original waveform, the signal is first passed through a low-pass filter with a cutoff frequency of 5 kHz to eliminate environmental high-frequency interference, and then envelope demodulation technology based on Hilbert transform is used to obtain envelope spectrum by empirical mode decomposition of the filtered signal, and the kurtosis index of the bearing outer ring fault characteristic frequency band in the envelope spectrum is extracted as the mechanical impact feature. In the heat distribution processing link, the infrared thermal imager shoots the global thermal map of the motor at a set interval, and the improved watershed algorithm is used to segment the image into four regions of stator winding, rotor core, front and rear bearings, and the average temperature of each region and the maximum temperature difference between adjacent regions are calculated. Current waveform analysis obtains three-phase current instantaneous value through coil mutual inductor, and generates 0-1 kHz frequency spectrum by windowed FFT transformation of A, B and C phase currents respectively, and the harmonic subgroup analysis method is used to calculate the energy proportion of specific harmonic. Finally, the multi-dimensional feature vectors such as mechanical impact feature, thermal zone temperature difference and harmonic distortion rate are standardized and packaged into data packets.

[0148] According to the embodiment of the present application, the multi-source heterogeneous data is input into the pre-trained fault prediction model to obtain the fault probability of the preset fault, specifically:

[0149] The mechanical impact feature is input into the preset time domain convolution branch to obtain the periodic vibration mode through the hollow convolution layer;

[0150] The thermal distribution feature is input into the spatial attention branch to generate the weighted feature map of the key area of the temperature field;

[0151] The current harmonic feature is input into the frequency domain analysis branch to obtain the transient harmonic energy through wavelet transform;

[0152] The periodic vibration mode, the weighted feature map and the transient harmonic energy are input into the preset mechanical-thermal-electric coupling relationship model to obtain the probability distribution of the characteristic fault, wherein the characteristic fault includes bearing wear, oil line blockage and coil short circuit.

[0153] It should be noted that the multi-modal fault prediction model provided in the embodiment adopts a three-branch parallel architecture. As an implementation, the vibration feature input time domain convolution branch extracts the periodic pattern of bearing impact through 3 layers of dilated rate 2 hollow convolution, and outputs vibration state data containing fault characteristic frequency. As an implementation, the heat distribution feature input spatial attention branch gives N times weight to the bearing area feature map after extracting the basic features, and suppresses the background heat interference, where N is greater than 2. As an implementation, the current feature is input into the frequency domain analysis branch, and a wavelet transform function is used for multi-scale decomposition to calculate the wavelet packet energy entropy of multiple subbands to quantify the harmonic energy. The three features are input into the graph neural network in the fusion layer to construct a topological structure with vibration nodes as the core, temperature nodes as the auxiliary, and current nodes as the verification. Then, the message passing mechanism is used to model the coupling relationship that the vibration impact is enhanced when the bearing wears, causing local temperature rise, and then the current harmonic increases. The multi-dimensional fusion features of the graph neural network are received by the full connection layer, and three types of prediction results of bearing wear probability, oil line blockage probability and coil short circuit probability are output, and when any probability value exceeds the set threshold, an early warning is triggered.

[0154] According to the embodiment of the present application, the self-healing control instruction is generated based on the preset knowledge base, specifically:

[0155] According to the fault type, the solution template of the preset self-healing knowledge base is matched, wherein the self-healing knowledge base includes a fault feature and an execution parameter mapping table;

[0156] When it is determined that the bearing is worn, an oil line purification instruction is generated;

[0157] When it is determined that the oil line is blocked, a backwashing instruction set is generated;

[0158] When it is determined that the coil is short-circuited, a load transfer instruction is generated.

[0159] It should be noted that the knowledge base driven instruction generation provided in the embodiment includes three-level operation logic. First, according to the highest probability fault type, the solution library is searched, for example, bearing wear of 0.2-0.5mm corresponds to oil line purification of 2 cycles and viscosity adjustment to 46cSt. Then, after the solution is matched, the parameter optimization stage is entered, for example, for the oil line blockage case, the backwashing parameters are dynamically set according to the blockage degree index, for example, single cycle 5 second pulse for mild blockage and three cycles 10 second pulse cycle for severe blockage; for the bearing wear case, the purification duration is determined in combination with the vibration energy value, for example, 10 minutes of purification for every 100mV vibration quantity. Finally, a set of machine readable instructions is generated, the oil line purification instruction includes the centrifugal pump speed curve, the heater temperature set value or the target viscosity threshold; the backwashing instruction clearly specifies the opening and closing timing of the electromagnetic valve; the load transfer instruction specifies the standby unit start delay and power switching gradient.

[0160] According to the embodiment of the present application, the driving self-healing device to execute instruction actions according to the self-healing control instruction further comprises:

[0161] When determining the oil line purification instruction, the gear pump start-stop and the heater temperature are controlled according to the instruction parameters, and the oil quality sensor data is monitored in real time;

[0162] When determining the backwashing instruction group, the electromagnetic valve actuator is driven to open and close in a cycle according to the instruction sequence and specified time length and interval;

[0163] When determining the load transfer instruction, the output power is adjusted based on the preset cloud platform scheduling.

[0164] It should be noted that when executing the oil line purification instruction, the gear pump is accelerated to the target speed with an S-shaped speed curve, and at the same time, the heater power is dynamically adjusted according to the oil temperature sensor feedback through the PID regulator; in addition, the online viscometer measures the value every 30 seconds, and when the continuous three readings meet the standard, the purification is automatically stopped. When executing the backwashing instruction, the electromagnetic valve is driven according to the preset timing, the preset compressed air pressure is maintained during the opening stage to continuously blow, and the drain valve is opened to release pressure during the closing stage; the pressure transmitter data is read after each cycle to determine the unblocking state. The load transfer operation is written into the frequency converter register through the preset communication protocol, first, the frequency of the faulty unit is lowered, for example, to 50% at a rate of 5% per second; at the same time, the standby unit synchronously increases the frequency to take over the load, and the pressure sensor monitors the pipe network fluctuation in real time during the transfer process, and if the pressure change rate exceeds the limit, the PID feedforward compensation algorithm is enabled.

[0165] According to the embodiment of the present application, the improvement rate is obtained according to the multi-source feedback data, specifically:

[0166] The vibration frequency spectrum of the mechanical vibration signal before and after the self-healing operation is analyzed, and the characteristic frequency band energy attenuation rate is calculated;

[0167] The thermal distribution image before and after the self-healing operation is analyzed to obtain the area shrinkage rate of the preset high temperature area and the maximum temperature difference;

[0168] The harmonic energy proportion change rate of the current waveform before and after the self-healing operation is analyzed;

[0169] The multi-source feedback data is obtained by combining the energy attenuation rate, the area shrinkage rate, the maximum temperature difference, and the harmonic energy proportion change rate;

[0170] According to the fault type, a weighting coefficient is obtained;

[0171] Based on the preset linear weighting algorithm, the improvement rate is obtained according to the weighting coefficient and the multi-source feedback data.

[0172] It should be noted that the improvement rate quantitative evaluation provided in this embodiment executes a four-step analysis process. First, for the mechanical vibration signals before and after the self-recovery operation, the power spectrum density integral ratio before and after the execution is calculated in the preset bearing characteristic frequency band, and the energy decay rate is obtained. Second, for the thermal distribution images before and after the self-recovery operation, the high-temperature region with the temperature set temperature threshold is extracted through image binarization processing, and the pixel area reduction ratio and the maximum temperature difference reduction value in the region after the execution are calculated. Then, for the current waveform before and after the self-recovery operation, the 10-second current waveforms before and after the execution are analyzed, and the 5th harmonic content ratio is calculated respectively, and the harmonic distortion rate reduction value is obtained. Finally, according to the fault type, a weight coefficient is assigned; as an implementation, the bearing wear scenario takes the vibration weight as 0.6, the temperature weight as 0.3, and the current weight as 0.1; the oil line blockage scenario takes the vibration weight as 0.3, the temperature weight as 0.5, and the current weight as 0.2; and the improvement rate is calculated according to the preset linear weighting algorithm.

[0173] It is worth mentioning that it also includes optimizing the fault prediction model based on the federated learning framework, specifically:

[0174] Each air compressor terminal locally trains a model subnetwork and uploads the weight parameters to a preset aggregation cloud platform;

[0175] The cloud platform fuses the weight parameters through a secure averaging algorithm and distributes the weight parameters of the global model to each air compressor terminal;

[0176] When a new fault mode is added, an incremental learning mechanism is triggered.

[0177] It should be noted that this embodiment adopts a federated learning framework and implements three links including local training, secure aggregation, and incremental update. Each air compressor terminal locally loads a basic model, fine-tunes the fully connected layer parameters using recent 30-day operation data, and uploads the encrypted weight parameters after training. After receiving the weight parameters of at least 10 terminals, the aggregation server of the cloud platform performs weighted averaging using secure multi-party computation technology to generate a new global model. When a new fault mode is detected, such as motor brush wear, an incremental learning mechanism is triggered to append a new classification branch after the original fully connected layer, and a small amount of labeled samples are used for transfer learning. The updated model is compressed by a preset compression operation to generate a lightweight version and is pushed to the air compressor terminal.

[0178] It is worth mentioning that before executing the self-recovery operation, a virtual verification is also included, specifically:

[0179] Simulate the execution of the self-recovery instruction based on the preset digital twin model to predict the change curve of the key parameters;

[0180] When the predicted pressure fluctuation exceeds the limit or the temperature rises sharply, the execution parameters are automatically adjusted;

[0181] When it is determined that the operation is high-risk operation according to the change curve of the key parameter, a preset backup scheme is automatically enabled;

[0182] The deviation value of the virtual verification result and the actual execution data is recorded for calibrating the precision of the twin model.

[0183] It should be noted that the embodiment also provides a virtual verification mechanism based on a digital twin architecture. First, a physical model of the device is constructed on a cloud platform, including motor electromagnetic characteristic equations, bearing dynamics equations, pipeline fluid equations, etc. When a self-healing instruction is received, the instruction parameters are injected into the virtual space for real-time simulation, for example, inputting backwash pressure 0.6 MPa, the pipeline pressure fluctuation curve is predicted through the computational fluid dynamics model; inputting load transfer parameters, the motor torque-speed characteristic curve change is solved. If the simulation shows that the pressure peak value exceeds 120% of the rated value or the temperature gradient exceeds the preset temperature change rate, the system automatically generates parameter optimization suggestions. In addition, when it is determined that the operation is high-risk operation, for example, emergency shutdown in short circuit, the system preloads multiple sets of disposal plans, for example, the main scheme executes device hard shutdown, and the alternative scheme executes power slow-down combined with mechanical brake operation. During actual execution, the twin model synchronously receives sensor data for online checking, and when the pressure prediction error or the temperature error exceeds the set range, the turbulent flow coefficient or the heat conduction coefficient of the fluid model is automatically corrected to realize model self-calibration.

[0184] The third aspect of the present application provides a computer readable storage medium, wherein the computer readable storage medium comprises a remote monitoring and fault self-healing based air compressor management method program, and the remote monitoring and fault self-healing based air compressor management method program is executed by a processor to realize the steps of the remote monitoring and fault self-healing based air compressor management method according to any one of the above.

[0185] In summary, the present application provides a remote monitoring and fault self-healing based air compressor management method, system and medium. First, the running state of the air compressor is monitored by a multi-modal sensor, and key features are extracted based on feature processing and combined into multi-source heterogeneous data. Second, a fault prediction model is used to model the mechanical-thermal-electric coupling relationship, and output the probability values of bearing wear, oil line blockage and other faults. Third, when the fault probability value exceeds the threshold value, a parameterized self-healing instruction is automatically generated based on the knowledge base to drive the execution of the self-healing operation. Finally, after the self-healing operation is executed, the improvement rate is quantified through multi-source feedback data, and the model parameters of the fault prediction model are dynamically optimized based on the improvement rate to form a self-evolution ability. The predictive self-maintenance function is realized, and the maintenance cost of the air compressor is reduced.

[0186] If the functions are implemented in the form of software function modules and sold or used as independent products, they can be stored in a computer readable storage medium. Based on this understanding, the technical solutions of the present application or the parts of the technical solutions that essentially contribute to the prior art or the parts of the technical solutions can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes a plurality of instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the method described in the various embodiments of the present application. The aforementioned storage medium includes a U disk, a mobile hard disk, a read-only memory (ROM, Read-Only Memory), a random access memory (RAM, Random Access Memory), a magnetic disk or an optical disk, and various media that can store program codes.

[0187] The above only describes the preferred embodiments of the present application and is not intended to limit the present application. For those skilled in the art, the present application can have various modifications and changes. Any modification, equivalent replacement, improvement, etc. made within the spirit and principles of the present application shall be included in the protection scope of the present application.

Claims

1. An air compressor management method based on remote monitoring and fault self-healing, characterized in that: The method comprises: Based on the preset sensor monitoring operation data, after the preset feature preprocessing, multi-source heterogeneous data is fused; Inputting the multi-source heterogeneous data into a pre-trained fault prediction model to obtain a failure probability of a preset fault; If the failure probability exceeds a preset failure threshold, generating a self-healing control instruction based on a preset knowledge base; driving the self-healing device to execute the command action according to the self-healing control command; In response to a preset delay time, multi-source feedback data is collected and processed; Obtaining an improvement rate based on the multi-source feedback data; The training weights are set based on the improvement rate, and the fault probability and the self-healing control instructions are used as a training data set to train and optimize the weight parameters of the fault prediction model.

2. The air compressor management method based on remote monitoring and fault self-healing according to claim 1 is characterized in that: The operation data includes mechanical vibration signals, thermal distribution images and current waveform data, which are pre-processed by preset features, specifically: The mechanical vibration signal is collected by a three-axis acceleration sensor and the environmental noise is eliminated through a preset filtering algorithm; Obtaining a mechanical shock feature according to the mechanical vibration signal based on a preset envelope demodulation logic; Use an infrared thermal imager to capture the thermal distribution image of the motor and identify and obtain the temperature range of the motor windings and bearings; Based on a preset region segmentation algorithm, extracting thermal distribution characteristics of the winding and the bearing according to the thermal distribution image; The motor operating current is measured through a current transformer to obtain the current waveform; Based on the preset FFT transformation, the current harmonic characteristics are obtained according to the current waveform.

3. The air compressor management method based on remote monitoring and fault self-healing according to claim 2 is characterized in that: The multi-source heterogeneous data is input into the pre-trained fault prediction model to obtain the failure probability of the preset fault, specifically: The mechanical shock feature is input into a preset time domain convolution branch, and a periodic vibration pattern is obtained through a dilated convolution layer; Inputting the thermal distribution features into the spatial attention branch to generate a weighted feature map of the key areas of the temperature field; Inputting the current harmonic characteristics into the frequency domain analysis branch, and obtaining transient harmonic energy through wavelet transformation; The periodic vibration mode, the weighted characteristic map, and the transient harmonic energy are input into a preset mechanical-thermal-electrical coupling relationship model to obtain a probability distribution of characteristic faults, wherein the characteristic faults include bearing wear, oil circuit blockage, and coil short circuit.

4. The air compressor management method based on remote monitoring and fault self-healing according to claim 1 is characterized in that: The self-healing control instructions are generated based on the preset knowledge base, specifically: Matching a solution template from a preset self-healing knowledge base based on the fault type, wherein the self-healing knowledge base includes a mapping table between fault characteristics and execution parameters; When it is determined that the bearing is worn, an oil circuit purification instruction is generated; When it is determined that the oil circuit is blocked, a backwash instruction group is generated; When the coil is determined to be short-circuited, a load transfer command is generated.

5. The air compressor management method based on remote monitoring and fault self-healing according to claim 4 is characterized in that: Driving the self-healing device to execute the command action according to the self-healing control command also includes: When it is judged as an oil circuit purification instruction, the gear pump start and stop and heater temperature are controlled according to the instruction parameters, and the oil quality sensor data is monitored in real time; When it is determined to be a backwash instruction group, the solenoid valve actuator is driven to perform an open and close cycle at the specified duration and interval according to the instruction sequence; When it is determined to be a load transfer instruction, the output power is adjusted based on the preset cloud platform scheduling.

6. The air compressor management method based on remote monitoring and fault self-healing according to claim 1 is characterized in that: The improvement rate is obtained according to the multi-source feedback data, specifically: Analyze the vibration spectrum of the mechanical vibration signal before and after the self-healing operation and calculate the energy attenuation rate of the characteristic frequency band; Analyze the thermal distribution images before and after the self-healing operation to obtain the area shrinkage rate and maximum temperature difference of the preset high-temperature area; Analyze the change rate of harmonic energy proportion in the current waveform before and after self-healing operation; The multi-source feedback data is obtained by combining the energy attenuation rate, the area shrinkage rate, the maximum temperature difference and the harmonic energy proportion change rate; According to the fault type, the weighting coefficient is obtained; Based on a preset linear weighted algorithm, an improvement rate is obtained according to the weighted coefficient and the multi-source feedback data.

7. An air compressor management system based on remote monitoring and fault self-healing, characterized in that: The system includes a memory and a processor. The memory includes an air compressor management method program based on remote monitoring and fault self-healing. When the air compressor management method program based on remote monitoring and fault self-healing is executed by the processor, the following steps are implemented: Based on the preset sensor monitoring operation data, after the preset feature preprocessing, multi-source heterogeneous data is fused; Inputting the multi-source heterogeneous data into a pre-trained fault prediction model to obtain a failure probability of a preset fault; If the failure probability exceeds a preset failure threshold, generating a self-healing control instruction based on a preset knowledge base; driving the self-healing device to execute the command action according to the self-healing control command; In response to a preset delay time, multi-source feedback data is collected and processed; Obtaining an improvement rate based on the multi-source feedback data; The training weights are set based on the improvement rate, and the fault probability and the self-healing control instructions are used as a training data set to train and optimize the weight parameters of the fault prediction model.

8. The air compressor management system based on remote monitoring and fault self-healing according to claim 7 is characterized in that: The operation data includes mechanical vibration signals, thermal distribution images and current waveform data, which are pre-processed by preset features, specifically: The mechanical vibration signal is collected by a three-axis acceleration sensor and the environmental noise is eliminated through a preset filtering algorithm; Obtaining a mechanical shock feature according to the mechanical vibration signal based on a preset envelope demodulation logic; Use an infrared thermal imager to capture the thermal distribution image of the motor and identify and obtain the temperature range of the motor windings and bearings; Based on a preset region segmentation algorithm, extracting thermal distribution characteristics of the winding and the bearing according to the thermal distribution image; The motor operating current is measured through a current transformer to obtain the current waveform; Based on the preset FFT transformation, the current harmonic characteristics are obtained according to the current waveform.

9. The air compressor management system based on remote monitoring and fault self-healing according to claim 8 is characterized in that: The multi-source heterogeneous data is input into the pre-trained fault prediction model to obtain the failure probability of the preset fault, specifically: The mechanical shock feature is input into a preset time domain convolution branch, and a periodic vibration pattern is obtained through a dilated convolution layer; Inputting the thermal distribution features into the spatial attention branch to generate a weighted feature map of the key areas of the temperature field; Inputting the current harmonic characteristics into the frequency domain analysis branch, and obtaining transient harmonic energy through wavelet transformation; The periodic vibration mode, the weighted characteristic map, and the transient harmonic energy are input into a preset mechanical-thermal-electrical coupling relationship model to obtain a probability distribution of characteristic faults, wherein the characteristic faults include bearing wear, oil circuit blockage, and coil short circuit.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: The computer-readable storage medium includes an air compressor management method program based on remote monitoring and fault self-healing. When the air compressor management method program based on remote monitoring and fault self-healing is executed by a processor, the steps of the air compressor management method based on remote monitoring and fault self-healing as described in any one of claims 1 to 6 are implemented.

Images