A real-time data-driven air compressor remaining life prediction method and system

Abstract

The present invention relates to the field of medical equipment maintenance, and the present invention relates to a real-time data-driven method and system for predicting the remaining life of an air compressor, wherein the method comprises: collecting multiple groups of operating data of the air compressor, each group of operating data including the temperature, vibration amplitude and pressure of the air compressor at each moment in the operation; performing feature processing on each group of operating data to obtain feature data; inputting the temperature characteristic value, vibration amplitude characteristic value and pressure characteristic value of each group of operating data into a life prediction model to obtain the life estimation value corresponding to each group of operating data; calculating the weight coefficient of each group of operating data; and performing weighted summation on the life estimation value according to the weight coefficient to obtain the remaining life of the air compressor. The present invention collects data of multiple operations of the air compressor, combines indicators such as pressure stability, vibration amplitude and temperature stability, accurately evaluates the operating status and fault degree of the air compressor, and thus effectively predicts its remaining life.

Description

Technical Field

[0001] The present invention generally relates to the field of medical equipment maintenance. More specifically, the present invention relates to a real-time data-driven air compressor remaining life prediction method and system. Background Art

[0002] In the medical oxygen production system, the air compressor is the core equipment. The operating status of the air compressor directly affects the stability and safety of the oxygen production system. However, when the air compressor is in operation for a long time, it is prone to wear, aging and other faults, which will cause the oxygen production system to shut down. Therefore, predicting the life of the air compressor and performing maintenance in time are crucial to ensure the safe and stable operation of the oxygen production system.

[0003] Usually, the life prediction of air compressors mainly relies on manual experience and regular maintenance, and relies on workers' comprehensive judgment of multiple factors such as equipment wear and tear patterns, historical failure data, maintenance data, and equipment brands. Therefore, the existing methods for predicting the life of air compressors have problems such as high labor costs and poor timeliness. And due to the subjectivity of manual experience in predicting the life of air compressors, there is a situation where there is a large deviation between the manual prediction results and the actual situation.

[0004] With the rapid development of artificial intelligence technology, life prediction methods based on machine learning have gradually emerged, such as recurrent neural networks (RNN) and long short-term memory networks (LSTM) based on deep learning. The existing technology also combines artificial intelligence technology with air compressor life prediction methods to reduce the impact of subjective factors on the accuracy of air compressor life prediction. The Chinese patent application document with publication number CN112990542A discloses a life prediction method for air compressors. The method performs feature engineering on the vibration data (total energy value OA) and current data of the air compressor to generate training data; uses the training data to train the convolutional neural network (CNN) and long short-term memory network (LSTM) model; saves the model as an H5 file; and finally uses the Flask framework to package the model as a Web service for easy application and access. This method can predict the life of the air compressor in advance to notify maintenance personnel to perform maintenance.

[0005] However, the failure of the air compressor itself is not necessarily reflected in the data collected during the operation of the air compressor. For example: insufficient lubrication of the air compressor causes the temperature of the air compressor to continue to rise with the number of times the air compressor is used during continuous operation; or the bearings, seals and other components of the air compressor are damaged, causing a specific vibration frequency and a large amplitude vibration. If the frequency of the vibration is small, there will be no large frequency vibration in the operation of the air compressor. Since these failures are the main reasons that affect the service life of the air compressor (shortening the life of the air compressor), the judgment of the remaining life of the air compressor based on the data of a single operation of the air compressor may not be accurate enough. Summary of the invention

[0006] In order to solve the above-mentioned technical problem that the judgment of the remaining life of the air compressor through the data of a single operation of the air compressor may not be accurate enough, the present invention provides solutions in the following aspects.

[0007] In a first aspect, a real-time data-driven method for predicting the remaining life of an air compressor is provided, the method comprising: collecting multiple groups of operating data of the air compressor, each group of operating data comprising the temperature, vibration amplitude and pressure of the air compressor at each moment in the current operation; performing feature processing on each group of operating data to obtain feature data, wherein the feature data comprises a temperature feature value, a vibration amplitude feature value and a pressure feature value; inputting the temperature feature value, the vibration amplitude feature value and the pressure feature value of each group of operating data into a life prediction model to obtain a life estimate corresponding to each group of operating data; calculating a weight coefficient for each group of operating data, wherein the weight coefficient of the first i Weight coefficient of group operation data η i The calculation formula is: , exp() is an exponential function with real number e as base. i Group operation data, A i is the characteristic value of its vibration amplitude, T i is its temperature characteristic value, Q i is its pressure characteristic value, T i-1 For the i -1 set of temperature characteristic values ​​of operating data, T i+1 For the i +1 set of temperature characteristic values ​​of operating data, Q S It is the standard pressure characteristic value of the preset size;

[0008] The estimated lifespan is weighted and summed according to the weight coefficient to obtain the remaining lifespan of the air compressor. R : ,in n is the total number of times the air compressor runs. r i For the i Life estimates for group operating data.

[0009] The beneficial effects of the present invention are:

[0010] The present invention collects data from multiple operations of the air compressor, and obtains the weight coefficient of the abnormality degree (fault degree) of the operating state of the air compressor in each operation based on the stability of the pressure generated and output by the air compressor, the amplitude of the air compressor vibration, and the stability of the temperature of the air compressor during each operation. Among them, the larger the weight coefficient, the greater the abnormality degree of the operating state of the air compressor in the corresponding operation. At this time, this group of operating data can better reflect the remaining life of the air compressor (the life of the air compressor is usually related to its own fault degree). Multiple life estimates for the remaining life of the air compressor are obtained through each group of operating data and the remaining life prediction model, and then each life estimate is weighted and summed by the weight coefficient, so as to obtain a more accurate result of the air compressor life prediction.

[0011] Preferably, a real-time data-driven air compressor remaining life prediction method further includes: in response to the temperature value of the air compressor at the end of any operation being greater than a temperature threshold, determining a weight coefficient of operation data corresponding to the operation of the air compressor to be 1.

[0012] The effect is: when the temperature of the air compressor is too high at a certain moment in an operation, it means that the probability of failure is relatively high. At this time, the weight coefficient of the operating data corresponding to the air compressor operation is increased to increase the weight of the corresponding operating data in predicting the remaining life of the air compressor, thereby increasing the accuracy of the results of the air compressor life prediction.

[0013] Preferably, a real-time data-driven method for predicting the remaining life of an air compressor further includes: taking the moment when the vibration amplitude in each group of operating data is greater than a preset vibration amplitude threshold as the target moment, obtaining the time difference between all the target moments adjacent in time sequence; in response to the variance corresponding to all the time differences being less than a preset variance threshold, modifying the weight coefficient of the group of operating data, and using the modified weight coefficient as the weight coefficient of the group of operating data; wherein for the first j The formula for correcting the weight coefficient of group operation data is: ; For the j The weight coefficient of the group operation data before correction, For the j The corrected weight coefficient of the group operation data, For the j The number of target moments corresponding to the group running data, For the j The standard deviation of the vibration amplitude in the group operation data, norm () represents the standard normalization function.

[0014] The effect is: when the air compressor has a periodic vibration amplitude that is too high during one operation, it means that the probability of damage to the bearings, seals and other parts of the air compressor is high. At this time, the weight coefficient of the operation data corresponding to the air compressor operation is increased to increase the weight of the corresponding operation data in predicting the remaining life of the air compressor, thereby increasing the accuracy of the results of the air compressor life prediction. The increase in the weight coefficient amplitude is proportional to the standard deviation corresponding to the vibration amplitude of the group of operation data (that is, the degree of fluctuation of the vibration amplitude).

[0015] Preferably, feature extraction for each group of operating data to obtain feature data includes: feature processing the temperature at each moment in the group of data to obtain the average value of all temperature values ​​of the group of operating data, and taking the average value as the temperature characteristic value of the group of operating data; feature processing the vibration amplitude at each moment in the group of data to obtain the maximum value of all vibration amplitude values ​​of the group of operating data, and taking the maximum value as the vibration amplitude characteristic value of the group of operating data; feature processing the pressure at each moment in the group of data to obtain the variance of all pressure values ​​of the group of operating data, and taking the variance as the pressure characteristic value of the group of operating data; and determining the temperature characteristic value, vibration amplitude characteristic value and pressure characteristic value of the group of operating data as the characteristic data of the group of operating data.

[0016] Preferably, constructing the prediction model includes: constructing a training set based on historical data of a predetermined number of damaged air compressors, wherein the historical data of any damaged air compressor includes the temperature, vibration amplitude, pressure of the damaged air compressor at each moment in an operation and the life of the damaged air compressor during the operation; constructing a deep learning model to be trained based on an attention mechanism, wherein the deep learning model to be trained is a Transforme model including an encoder and a decoder; training the deep learning model to be trained through the training set, so that the trained deep learning model to be trained is the life prediction model.

[0017] Preferably, the PyTorch framework is used to construct the deep learning model to be trained based on the attention mechanism.

[0018] The effect is: Since PyTorch natively supports GPU acceleration, using the PyTorch framework to build a deep learning model to be trained can significantly increase the speed of training the deep learning model to be trained in subsequent steps.

[0019] Preferably, training the initial prediction model through any one of the historical data in the training set includes: performing feature processing on the temperature, vibration amplitude and pressure of the damaged air compressor corresponding to the historical data at each moment in an operation to obtain feature data corresponding to the historical data; inputting the feature data corresponding to the historical data into the deep learning model to be trained, and taking the output of the deep learning model to be trained as the predicted value; taking the life of the damaged air compressor corresponding to the historical data during the operation as the true value, determining the difference between the predicted value and the true value through the cross entropy function, and updating the parameters of the deep learning model to be trained through the stochastic gradient descent algorithm.

[0020] Preferably, before characteristic processing is performed on any set of operating data or the temperature, vibration amplitude and pressure of any damaged air compressor at each moment in an operation, any set of operating data or the temperature, vibration amplitude and pressure of any damaged air compressor at each moment in an operation are filtered and cleaned.

[0021] The effect is: by filtering and cleaning the data, the data quality can be improved, decision support can be enhanced, and the accuracy of the results obtained after subsequent data analysis can be improved.

[0022] In a second aspect, a real-time data-driven air compressor remaining life prediction system includes a processor and a memory, wherein the memory stores a computer program, and the processor executes the computer program to implement a real-time data-driven air compressor remaining life prediction method as described in any one of the above invention contents.

[0023] The beneficial effects of the present invention are:

[0024] The present invention collects data from multiple operations of the air compressor, combines indicators such as pressure stability, vibration amplitude and temperature stability, and accurately evaluates the operating status and fault degree of the air compressor, thereby effectively predicting its remaining life. BRIEF DESCRIPTION OF THE DRAWINGS

[0025] By reading the following detailed description with reference to the accompanying drawings, the above and other objects, features and advantages of the exemplary embodiments of the present invention will become readily understood. In the accompanying drawings, several embodiments of the present invention are shown in an exemplary and non-limiting manner, and the same or corresponding reference numerals represent the same or corresponding parts, wherein:

[0026] Figure 1 is a flowchart schematically illustrating the steps of a real-time data-driven air compressor remaining life prediction method according to an embodiment of the present invention;

[0027] Figure 2 is a flowchart schematically illustrating the steps of constructing a life prediction model according to an embodiment of the present invention;

[0028] Figure 3 It is a structural block diagram schematically showing a real-time data-driven air compressor remaining life prediction system according to this embodiment. DETAILED DESCRIPTION

[0029] 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 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 those skilled in the art without creative work are within the scope of protection of the present invention.

[0030] The specific embodiments of the present invention are described in detail below with reference to the accompanying drawings.

[0031] Figure 1 The figure schematically shows a flowchart of the steps of a real-time data-driven air compressor remaining life prediction method according to an embodiment of the present invention.

[0032] like Figure 1 As shown, a real-time data-driven air compressor remaining life prediction method includes steps S101 to S105.

[0033] Step S101: Collect multiple groups of operating data of the air compressor.

[0034] Among them, air compressor i The corresponding run i The group operation data includes the temperature, vibration amplitude and pressure of the air compressor at each moment during the operation. i is a positive integer. The duration of each operation of the air compressor is the same, and the frequency of collecting the air compressor operation data is the same, so the number of data points in each set of operation data is the same. The data points include the temperature, vibration amplitude and pressure of the air compressor at a moment.

[0035] It should be noted that the working principle of an air compressor is to compress a large amount of air into a closed space to increase the pressure and density of the air, so that the air can be stored and used more efficiently. The pressure at any moment in the operating data is the pressure of the gas generated and output by the air compressor at that moment.

[0036] Step S102: extracting features from each set of operation data to obtain feature data.

[0037] The characteristic data include temperature characteristic values, vibration amplitude characteristic values ​​and pressure characteristic values.

[0038] It should be noted that, during one operation of the air compressor, the temperature of the air compressor at each moment is subjected to characteristic processing to obtain the temperature characteristic value corresponding to the current operation of the air compressor; the vibration amplitude of the air compressor at each moment is subjected to characteristic processing to obtain the vibration amplitude characteristic value corresponding to the current operation of the air compressor; and the pressure of the air compressor at each moment is subjected to characteristic processing to obtain the pressure characteristic value corresponding to the current operation of the air compressor.

[0039] In one embodiment, performing feature processing on any group of operating data to obtain feature data includes: performing feature processing on the temperature at each moment in the group of data to obtain the average value of all temperature values ​​of the group of operating data, and using the average value as the temperature feature value of the group of operating data; performing feature processing on the vibration amplitude at each moment in the group of data to obtain the maximum value of all vibration amplitude values ​​of the group of operating data, and using the maximum value as the vibration amplitude feature value of the group of operating data; performing feature processing on the pressure at each moment in the group of data to obtain the variance of all pressure values ​​of the group of operating data, and using the variance as the pressure feature value of the group of operating data; and determining the temperature feature value, vibration amplitude feature value and pressure feature value of the group of operating data as the feature data of the group of operating data.

[0040] Step S103: inputting the temperature characteristic value, vibration amplitude characteristic value and pressure characteristic value of each set of operating data into the life prediction model to obtain the life estimation value corresponding to each set of operating data.

[0041] It should be noted that the life prediction model is used to evaluate the remaining life of the air compressor based on the characteristic data of the air compressor (i.e., the temperature characteristic value, vibration amplitude characteristic value, and pressure characteristic value corresponding to each set of operating data). The life estimate value output by the life prediction model is an estimated value of the remaining life of the air compressor. In one embodiment, the life prediction model outputs 1.45 (years), and the life prediction model estimates the remaining life of the air compressor to be 1.45 years.

[0042] Step S104: Calculate the weight coefficient of each group of operating data.

[0043] Among them, i Weight coefficient of group operation data η i The calculation formula is:

[0044] .

[0045] exp() is an exponential function with real number e as base. i Group operation data, A i is the characteristic value of its vibration amplitude, T i is its temperature characteristic value, Q iis its pressure characteristic value, T i-1 For the i -1 set of temperature characteristic values ​​of operating data, T i+1 For the i +1 set of temperature characteristic values ​​of operating data, Q S It is the standard pressure characteristic value of the preset size. Standard pressure characteristic value Q S It is the pressure generated and output by the air compressor under the most ideal conditions, the standard pressure characteristic value Q S It is also the pressure that the air compressor is expected to produce and output.

[0046] It should be noted that for i Group operation data, Indicates that the air compressor is i The difference between the actual output pressure and the expected result in the run, and The larger the difference, the worse the performance of the air compressor. The bigger; Indicates that the air compressor is i The intensity of vibration generated during each operation can be used to assess the health of the air compressor. The larger the value, the worse the performance of the air compressor. Indicates that the air compressor is i The first run, i -1st run and i +1 run compared to the extent of temperature fluctuations, The larger the value, the greater the deviation of the temperature from the ideal value (under ideal conditions, the temperature of the air compressor remains basically unchanged or changes slightly when it runs continuously for many times). Based on this, the weight coefficient The larger the i The operating data of the air compressor during the first operation can better reflect the abnormality of the air compressor. Since the remaining life of the air compressor is usually related to its failure, the weight coefficient The larger the value is, the operating data of the air compressor in this operation can better reflect the abnormality degree (or fault degree) of the air compressor. Therefore, the operating data of the air compressor in this operation is more valuable for reference.

[0047] Furthermore, a real-time data-driven air compressor remaining life prediction method also includes: in response to the temperature value of the air compressor at the end of any operation being greater than a temperature threshold, determining the weight coefficient of the operation data corresponding to the operation of the air compressor to be 1.

[0048] It should be noted that when the temperature of the air compressor at a certain moment in an operation is greater than the temperature threshold, it means that the sensor is very likely to have failed. At this time, in order to increase the weight of the operating data collected by the air compressor in this operation in the subsequent calculation of the estimated value of the remaining life of the air compressor, the weight coefficient of the operating data corresponding to this operation of the air compressor is determined to be 1 to increase the weight coefficient corresponding to this operation of the air compressor.

[0049] Furthermore, a real-time data-driven air compressor remaining life prediction method also includes: taking the moment when the vibration amplitude in each group of operating data is greater than a preset vibration amplitude threshold as the target moment, obtaining the time difference between all the target moments adjacent in time sequence; in response to the variance corresponding to all the time differences being less than a preset variance threshold, modifying the weight coefficient of the group of operating data, and using the modified weight coefficient as the weight coefficient of the group of operating data; wherein j The formula for correcting the weight coefficient of group operation data is: ; For the j The weight coefficient of the group operation data before correction, For the j The corrected weight coefficient of the group operation data, For the j The number of target moments corresponding to the group running data, For the j The standard deviation of the vibration amplitude in the group operation data, norm () represents the standard normalization function.

[0050] It should be noted that when the bearings, seals and other parts of the air compressor are damaged, it is usually manifested as abnormal vibrations with periodic large vibration amplitudes in the air compressor. Therefore, when a periodic large vibration amplitude is detected in a set of data, the operating group data can better reflect the degree of failure of the air compressor and thus reflect the remaining life of the air compressor. At this time, the weight coefficient of the group of operating data is corrected to increase the weight coefficient of the group of operating data. In one operation of the air compressor corresponding to the group of operating data, the more times the air compressor has a large vibration amplitude, the larger the corrected weight coefficient corresponding to the group of operating data; the larger the standard deviation of the vibration amplitude value at each moment corresponding to the group of operating data, the greater the discrete degree of the vibration amplitude value at each moment, that is, the vibration amplitude of the air compressor is unstable, and at this time, the larger the corrected weight coefficient corresponding to the group of operating data.

[0051] Step S105: performing weighted summation on the estimated life values ​​according to the weight coefficient to obtain the remaining life of the air compressor.

[0052] Among them, calculate the remaining life of the air compressor R The formula is: , n is the total number of times the air compressor runs. r i For the i Life estimates for group operating data.

[0053] Furthermore, if the remaining life of the air compressor is less than a preset life threshold, an alarm is issued to enable the staff to perform maintenance and / or repair on the air compressor. In one embodiment, the life threshold is 0.5 years.

[0054] Figure 2 FIG. 4 is a flowchart schematically illustrating the steps of constructing a life prediction model according to an embodiment of the present invention.

[0055] like Figure 2 As shown, constructing the life prediction model includes steps S201 to S203.

[0056] Step S201: constructing a training set based on historical data of a predetermined number of damaged air compressors.

[0057] The historical data of any damaged air compressor includes the temperature, vibration amplitude, pressure of the damaged air compressor at each moment during one operation and the life of the damaged air compressor during the operation.

[0058] Step S202: Construct a deep learning model to be trained based on the attention mechanism.

[0059] In one embodiment, the PyTorch framework is used to construct the deep learning model to be trained based on the attention mechanism.

[0060] Step S203: training the deep learning model to be trained by using the training set, and using the trained deep learning model to be trained as the life prediction model.

[0061] In one embodiment, training the initial prediction model through any of the historical data in the training set includes: performing feature processing on the temperature, vibration amplitude and pressure of the damaged air compressor corresponding to the historical data at each moment in an operation to obtain feature data corresponding to the historical data; inputting the feature data corresponding to the historical data into the deep learning model to be trained, and taking the output of the deep learning model to be trained as the predicted value; taking the life of the damaged air compressor corresponding to the historical data during the operation as the true value, determining the difference between the predicted value and the true value through a cross entropy function, and updating the parameters of the deep learning model to be trained through a stochastic gradient descent algorithm.

[0062] Figure 3 It is a structural block diagram schematically showing a real-time data-driven air compressor remaining life prediction system according to this embodiment.

[0063] The present invention also provides a real-time data-driven air compressor remaining life prediction system. Figure 3 As shown, the system includes a processor and a memory, wherein the memory stores computer program instructions, and when the computer program instructions are executed by the processor, a real-time data-driven air compressor remaining life prediction method according to the first aspect of the present invention is implemented.

[0064] In the present invention, the aforementioned memory may be any tangible medium containing or storing a program that can be used by or in combination with an instruction execution system, apparatus or device. For example, a computer-readable storage medium may be any appropriate magnetic storage medium or magneto-optical storage medium, such as a resistive random access memory RRAM (Resistive Random Access Memory), a dynamic random access memory DRAM (Dynamic Random Access Memory), a static random access memory SRAM (Static Random-Access Memory), an enhanced dynamic random access memory EDRAM (Enhanced Dynamic Random Access Memory), a high-bandwidth memory HBM (High-Bandwidth Memory), a hybrid memory cube HMC (Hybrid Memory Cube), etc., or any other medium that can be used to store the required information and can be accessed by an application, a module, or both. Any such computer storage medium may be part of a device or accessible or connectable to a device. Any application or module described in the present invention may be implemented using computer-readable / executable instructions that may be stored or otherwise maintained by such a computer-readable medium.

[0065] In the description of this specification, "plurality" or "several" means at least two, such as two, three or more, etc., unless otherwise clearly and specifically defined.

[0066] Although this specification has shown and described a number of embodiments of the present invention, it will be apparent to those skilled in the art that such embodiments are provided by way of example only. Those skilled in the art will conceive of many modifications, changes and alternatives without departing from the ideas and spirit of the present invention. It should be understood that in the practice of the present invention, various alternatives to the embodiments of the present invention described herein may be employed.

Claims

1. A real-time data-driven air compressor remaining life prediction method, characterized in that: include: Collect multiple sets of operating data of the air compressor, each set of operating data includes the temperature, vibration amplitude and pressure of the air compressor at each moment during this operation; Performing feature processing on each set of operating data to obtain feature data, wherein the feature data includes a temperature feature value, a vibration amplitude feature value, and a pressure feature value; Inputting the temperature characteristic value, the vibration amplitude characteristic value and the pressure characteristic value of each set of operating data into the life prediction model to obtain the life estimation value corresponding to each set of operating data; Calculate the weight coefficient of each group of running data, where i Weight coefficient of group operation data η i The calculation formula is: , exp() is an exponential function with real number e as base. i Group operation data, A i is the characteristic value of its vibration amplitude, T i is its temperature characteristic value, Q i is its pressure characteristic value, T i-1 For the i -1 set of temperature characteristic values ​​of operating data, T i+1 For the i +1 set of temperature characteristic values ​​of operating data, Q S It is the standard pressure characteristic value of the preset size; The estimated lifespan is weighted and summed according to the weight coefficient to obtain the remaining lifespan of the air compressor. R : ,in n is the total number of times the air compressor runs. r i For the i Life estimates for group operating data.

2. A real-time data-driven air compressor remaining life prediction method according to claim 1, characterized in that: Also includes: In response to the temperature value of the air compressor at the end of any operation being greater than the temperature threshold, the weight coefficient of the operation data corresponding to the operation of the air compressor is determined to be 1.

3. A real-time data-driven air compressor remaining life prediction method according to claim 1, characterized in that: Also includes: The moment when the vibration amplitude in each group of operation data is greater than a preset vibration amplitude threshold is taken as the target moment, and the time difference between all the target moments adjacent in time sequence is obtained; In response to the variance corresponding to all time differences being less than a preset variance threshold, the weight coefficient of the group of operation data is modified, and the modified weight coefficient is used as the weight coefficient of the group of operation data; j The formula for correcting the weight coefficient of group operation data is: ; For the j The weight coefficient of the group operation data before correction, For the j The corrected weight coefficient of the group operation data, For the j The number of target moments corresponding to the group running data, For the j The standard deviation of the vibration amplitude in the group operation data, norm () represents the standard normalization function.

4. A real-time data-driven air compressor remaining life prediction method according to claim 1, characterized in that: The feature processing of any set of operating data to obtain feature data includes: Performing feature processing on the temperature at each moment in the set of operating data to obtain an average value of all temperature values ​​of the set of operating data, and taking the average value as the temperature feature value of the set of operating data; performing feature processing on the vibration amplitude at each moment in the set of data to obtain a maximum value of all vibration amplitude values ​​of the set of operating data, and taking the maximum value as the vibration amplitude feature value of the set of operating data; performing feature processing on the pressure at each moment in the set of data to obtain a variance of all pressure values ​​of the set of operating data, and taking the variance as the pressure feature value of the set of operating data; The temperature characteristic value, the vibration amplitude characteristic value and the pressure characteristic value of the group of operating data are determined as the characteristic data of the group of operating data.

5. The real-time data-driven air compressor remaining life prediction method according to claim 1 is characterized in that: Constructing the prediction model includes: A training set is constructed based on the historical data of a predetermined number of damaged air compressors, wherein the historical data of any damaged air compressor includes the temperature, vibration amplitude, pressure of the damaged air compressor at each time during one operation and the life of the damaged air compressor during the one operation; Constructing a deep learning model to be trained based on an attention mechanism, wherein the deep learning model to be trained is a Transforme model including an encoder and a decoder; The deep learning model to be trained is trained using the training set, and the trained deep learning model to be trained is used as the life prediction model.

6. A real-time data-driven air compressor remaining life prediction method according to claim 5, characterized in that: The PyTorch framework is used to build the deep learning model to be trained based on the attention mechanism.

7. A real-time data-driven air compressor remaining life prediction method according to claim 5, characterized in that: Training the deep learning model to be trained by using any of the historical data in the training set includes: Performing feature processing on the temperature, vibration amplitude, and pressure of the damaged air compressor corresponding to the historical data at each moment in one operation to obtain feature data corresponding to the historical data; Inputting the feature data corresponding to the historical data into the deep learning model to be trained, and taking the output of the deep learning model to be trained as the predicted value; The life of the damaged air compressor corresponding to the historical data during this operation is taken as the true value, the difference between the predicted value and the true value is determined by the cross entropy function, and the parameters of the deep learning model to be trained are updated by the stochastic gradient descent algorithm.

8. A real-time data-driven air compressor remaining life prediction method according to claim 7, characterized in that: Before performing feature processing on any set of operating data or the temperature, vibration amplitude and pressure of any damaged air compressor at each moment in an operation, any set of operating data or the temperature, vibration amplitude and pressure of any damaged air compressor at each moment in an operation are filtered and cleaned.

9. A real-time data-driven air compressor remaining life prediction system, comprising a processor and a memory, wherein the memory stores a computer program, characterized in that: The processor executes the computer program to implement a real-time data-driven air compressor remaining life prediction method as described in any one of claims 1-8.

Images