Method for monitoring state of screw air compressor

Abstract

The invention relates to the technical field of data processing, in particular to a screw air compressor state monitoring method, which comprises the following steps: acquiring monitoring data of M dimensions of each monitoring object in an air compressor to obtain multi-dimensional real-time monitoring data and multi-dimensional historical monitoring data; screening suspected real multi-dimensional monitoring data from the multi-dimensional historical monitoring data, and obtaining the data real degree of each dimension in the suspected real multi-dimensional monitoring data; n reference dimensions of any monitoring object at the current moment are obtained, and the data authenticity of each reference dimension in each piece of suspected real multi-dimensional monitoring data and the difference between each piece of suspected real multi-dimensional monitoring data and the multi-dimensional real-time monitoring data are calculated according to the data authenticity of each reference dimension in each piece of suspected real multi-dimensional monitoring data. Obtaining the state voting weight of each piece of suspected real multi-dimensional monitoring data for any monitoring object; and according to the state voting weight of each piece of suspected real multi-dimensional monitoring data, a state evaluation result of any monitoring object at the current moment is obtained by utilizing a KNN algorithm, and the running state of the current air compressor is accurately obtained.

Description

Technical Field

[0001] The present invention relates to the technical field of data processing, and in particular to a method for monitoring the state of a screw air compressor. Background Art

[0002] A screw compressor is an air compressor that continuously compresses and discharges air through a pair of intermeshing screw rotors (male and female). It is widely used in machinery manufacturing, food and medicine, electronics, chemicals, building materials, and other fields. Screw compressors are key equipment in continuous production systems. Establishing an effective compressor condition monitoring system is crucial for predicting and maintaining equipment, ensuring production continuity, and ensuring equipment safety.

[0003] Existing air compressor status monitoring methods usually obtain the current air compressor status based on current monitoring data, combined with the air compressor status during historical monitoring and the changing trend of historical monitoring data. However, there are many factors that affect the air compressor status data. For example, if there is data monitored when the sensor is in an abnormal state in the historical monitoring data, it may also lead to misjudgment of the current air compressor status.

[0004] Therefore, how to accurately obtain the current status of the air compressor and improve the accuracy of the air compressor abnormal status warning has become an urgent problem to be solved. Summary of the Invention

[0005] In view of this, an embodiment of the present invention provides a method for monitoring the state of a screw air compressor to solve the problem of how to accurately obtain the current state of the air compressor and improve the accuracy of the abnormal state warning of the air compressor.

[0006] An embodiment of the present invention provides a method for monitoring the state of a screw air compressor, the method comprising the following steps: Obtain M-dimensional monitoring data of each monitoring object in the air compressor at any time to form multi-dimensional monitoring data at any time, M ≥ 2, and obtain multi-dimensional real-time monitoring data at the current time and multi-dimensional historical monitoring data within a preset historical period; Screening at least one suspected real multidimensional monitoring data from all multidimensional historical monitoring data, and obtaining the data authenticity of each dimension in each suspected real multidimensional monitoring data based on the period characteristics and dimension data correlation characteristics of all suspected real multidimensional monitoring data; For any monitored object, obtain N reference dimensions of the monitored object at the current moment. , according to the authenticity of each reference dimension in each suspected real multidimensional monitoring data, obtaining the influence weight of each suspected real multidimensional data on any of the monitoring objects; Obtaining a state voting weight of each suspected real multidimensional monitoring data on the any monitored object according to an influence weight of each suspected real multidimensional monitoring data on the any monitored object and a difference between each suspected real multidimensional monitoring data and the multidimensional real-time monitoring data; According to the state voting weight of each suspected real multidimensional monitoring data, the KNN algorithm is used to classify the multidimensional real-time monitoring data to obtain the state evaluation result of any monitoring object at the current moment. According to the state evaluation results of all monitoring objects at the current moment, the operating state of the air compressor at the current moment is determined.

[0007] Preferably, screening at least one suspected real multi-dimensional monitoring data from all multi-dimensional historical monitoring data includes: For any multidimensional historical monitoring data, taking the any multidimensional historical monitoring data as the last data, a window of preset length is constructed in the multidimensional historical monitoring data within a preset historical period, and the abnormality degree of each dimension monitoring data in the any multidimensional historical monitoring data is obtained based on the difference between the multidimensional historical monitoring data in the window; If the abnormality levels of all dimensional monitoring data in any multidimensional historical monitoring data are less than or equal to a preset abnormality level threshold, the any multidimensional historical monitoring data is recorded as suspected true multidimensional monitoring data.

[0008] Preferably, obtaining the abnormality degree of each dimension of monitoring data in any multidimensional historical monitoring data according to the difference between the multidimensional historical monitoring data in the window includes: For any dimension monitoring data in any multi-dimensional historical monitoring data, acquiring monitoring data belonging to the same dimension as the any dimension monitoring data in the window to form a monitoring sequence; Obtain the cumulative value of the absolute value of the difference between each monitoring data in the monitoring sequence and the monitoring data of any dimension, record it as the monitoring data difference cumulative value, normalize the monitoring data difference cumulative value, and obtain the abnormality degree of the monitoring data of any dimension in the any multidimensional historical monitoring data.

[0009] Preferably, obtaining the authenticity of each dimension of each suspected real multidimensional monitoring data based on the periodic characteristics and dimension data association characteristics of all suspected real multidimensional monitoring data includes: For any suspected real multidimensional monitoring data, a time window of a preset time length is established within a preset historical period, with the historical moment of the suspected real multidimensional monitoring data as the center, and the suspected real multidimensional monitoring data within the time window are combined into a suspected real multidimensional monitoring data sequence; For any dimension of any suspected real multidimensional monitoring data, from the abnormality degree of each dimension monitoring data in the suspected real multidimensional monitoring data sequence, obtain the abnormality degree under any dimension to form an abnormality degree subsequence, and fit the abnormality degree subsequence to obtain an abnormality degree fitting curve; Obtaining a periodic characteristic value of the data anomaly degree of any dimension in any suspected real multi-dimensional monitoring data according to the periodic characteristic of the anomaly degree fitting curve; The authenticity of the data in any dimension of the suspected real multidimensional monitoring data is obtained based on the periodic characteristic value of the data anomaly degree of any dimension in the suspected real multidimensional monitoring data and the anomaly degree correlation characteristics of the monitoring data of each dimension in the suspected real multidimensional monitoring data.

[0010] Preferably, obtaining the data anomaly degree periodic characteristic value of any dimension in any suspected real multi-dimensional monitoring data according to the periodic characteristic of the anomaly degree fitting curve includes: Dividing the abnormality degree fitting curve into at least two sub-curves according to the air compressor cycle, uniformly obtaining a preset number of fitting values on each sub-curve, and the positions of the fitting values on each sub-curve correspond one to one; For any two adjacent sub-curves, respectively obtain the fitting value differences at corresponding positions in the two adjacent sub-curves to form a fitting value difference sequence, respectively obtain the absolute value of the difference between each two adjacent fitting value differences in the fitting value difference sequence, and obtain the corresponding accumulated value of the absolute value of the difference as the degree of difference between the two adjacent sub-curves; The difference between each two adjacent sub-curves is obtained respectively, and a corresponding difference degree cumulative value is obtained. The opposite of the difference degree cumulative value is normalized to obtain the data anomaly degree periodic characteristic value of any dimension.

[0011] Preferably, obtaining the authenticity of the data of any dimension in any suspected real multidimensional monitoring data according to the periodic characteristic value of the data anomaly degree of any dimension in the any suspected real multidimensional monitoring data and the anomaly degree correlation feature of the monitoring data of each dimension in the any suspected real multidimensional monitoring data includes: Obtaining a determination coefficient of the abnormality fitting curve, calculating the sum of the determination coefficient and the periodic characteristic value of the abnormality degree of the data in any dimension, and obtaining a periodic regularity characteristic value of the monitoring data in any dimension; Obtain a cumulative value of the absolute value of the difference between the degree of abnormality corresponding to each dimension in any suspected real multidimensional monitoring data and the any dimension, record it as the cumulative value of the abnormality degree difference, and use the inverse of the cumulative value of the abnormality degree difference as the dimensional correlation degree between the any dimension and each dimension in the any suspected real multidimensional monitoring data; The product of the periodic regularity characteristic value of the monitoring data of any dimension and the dimensional correlation degree between the any dimension and each dimension in the any suspected real multidimensional monitoring data is normalized to obtain the data authenticity of any dimension in the any suspected real multidimensional monitoring data.

[0012] Preferably, obtaining the influence weight of each suspected real multidimensional data on any monitoring object according to the authenticity of each reference dimension in each suspected real multidimensional monitoring data includes: For any suspected real multidimensional monitoring data, the average of the data authenticity of all reference dimensions is calculated according to the data authenticity of each dimension in the any suspected real multidimensional monitoring data, and recorded as the influence weight of the any suspected real multidimensional monitoring data on the any monitoring object.

[0013] Preferably, obtaining the state voting weight of each suspected real multidimensional monitoring data on the any monitored object according to the influence weight of each suspected real multidimensional monitoring data on the any monitored object and the difference between each suspected real multidimensional monitoring data and the multidimensional real-time monitoring data includes: An N-dimensional coordinate system is established with the N reference dimensions of any monitoring object at the current moment as coordinate axes, wherein the N coordinate axes of the N-dimensional coordinate system respectively represent the monitoring data of each reference dimension in the suspected real multidimensional monitoring data; For any suspected real multi-dimensional monitoring data, obtaining the Euclidean distance between the suspected real multi-dimensional monitoring data and the multi-dimensional real-time monitoring data in the N-dimensional coordinate system; The product of the reciprocal of the Euclidean distance and the influence weight of any suspected real multidimensional monitoring data on any monitored object is obtained to obtain the state voting weight of any suspected real multidimensional monitoring data on any monitored object.

[0014] Compared with the prior art, the embodiments of the present invention have the following beneficial effects: The present invention obtains monitoring data of M dimensions of each monitoring object in the air compressor at any time to form multidimensional monitoring data at any time, M≥2, obtains multidimensional real-time monitoring data at the current time and multidimensional historical monitoring data within a preset historical period; selects at least one suspected real multidimensional monitoring data from all multidimensional historical monitoring data, and obtains the data authenticity of each dimension in each suspected real multidimensional monitoring data based on the periodic characteristics and dimensional data association characteristics of all suspected real multidimensional monitoring data; for any monitoring object, obtains N reference dimensions of the monitoring object at the current time, Based on the data authenticity of each reference dimension in each suspected real multidimensional monitoring data, the influence weight of each suspected real multidimensional monitoring data on the monitoring object is obtained; based on the influence weight of each suspected real multidimensional monitoring data on the monitoring object and the difference between each suspected real multidimensional monitoring data and the multidimensional real-time monitoring data, the state voting weight of each suspected real multidimensional monitoring data on the monitoring object is obtained; based on the state voting weight of each suspected real multidimensional monitoring data, the multidimensional real-time monitoring data is classified using the KNN algorithm to obtain the state evaluation result of each monitoring object at the current moment; and based on the state evaluation results of all monitoring objects at the current moment, the operating state of the air compressor at the current moment is determined. The suspected real multidimensional monitoring data is first screened, and then the influence weight of each suspected real multidimensional monitoring data on any monitoring object in the air compressor is adjusted by obtaining the data authenticity of each dimension in each suspected real multidimensional monitoring data. This effectively reduces the impact of abnormal data caused by sensor abnormalities on the accuracy of the air compressor state monitoring results, facilitates timely detection of abnormal states of the air compressor and takes corresponding measures to ensure the continuous operation and working safety of the air compressor. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0016] Figure 1 This is a flow chart of a method for monitoring the state of a screw air compressor provided in Example 1 of the present invention. DETAILED DESCRIPTION

[0017] The embodiments of the present disclosure are described in detail below, and examples of the embodiments are shown in the accompanying drawings. The embodiments described below with reference to the accompanying drawings are exemplary and intended to be used to explain the present disclosure, but should not be understood as limiting the present disclosure.

[0018] It should be noted that the terms "first," "second," and the like in the specification of the present disclosure and the accompanying drawings are used to distinguish similar objects and are not necessarily used to describe a specific order or precedence. It should be understood that the terms used in this manner are interchangeable where appropriate, so that the embodiments of the present disclosure described herein can be implemented in an order other than those illustrated or described herein. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with the present disclosure. Instead, they are merely examples of devices and methods consistent with some aspects of the present disclosure.

[0019] In order to illustrate the technical solution of the present invention, specific embodiments are provided below.

[0020] See also Figure 1 , is a flow chart of a method for monitoring the state of a screw air compressor provided by the first embodiment of the present invention, such as Figure 1 As shown, the method may include: Step S101, obtain M-dimensional monitoring data of each monitoring object in the air compressor at any time, forming multidimensional monitoring data at any time, M≥2, obtain multidimensional real-time monitoring data at the current time and multidimensional historical monitoring data within a preset historical period.

[0021] Screw air compressor is a type of air compressor that is widely used in machinery manufacturing, food and medicine, electronics, chemicals, building materials and other fields. Screw air compressor is a key equipment in continuous production systems. Establishing an effective air compressor status monitoring system is of great significance for predicting and maintaining equipment, ensuring production continuity and equipment safety.

[0022] Existing air compressor status monitoring methods usually obtain the current air compressor status based on current monitoring data, combined with the air compressor status during historical monitoring and the changing trend of historical monitoring data. However, there are many factors that affect the air compressor status data. For example, if there is data monitored when the sensor is in an abnormal state in the historical monitoring data, it may also lead to misjudgment of the current air compressor status.

[0023] Therefore, this embodiment analyzes the data monitored by the sensor, screens the suspected real multi-dimensional monitoring data, and then adjusts the influence weight of each suspected real multi-dimensional monitoring data on any monitoring object in the air compressor by obtaining the data authenticity of each dimension in each suspected real multi-dimensional monitoring data, thereby effectively reducing the impact of abnormal data caused by sensor abnormality on the accuracy of the air compressor status monitoring results.

[0024] The air compressor mainly consists of the main body (spiral rotor, bearings, housing, etc.), drive system (motor, etc.), cooling and lubrication system (oil-gas separator, oil cooler, filter valve, etc.), and air circuit system (intake valve, air filter, exhaust pipe, etc.). Air compressor status monitoring includes multiple dimensions such as mechanical, electrical, thermal, and fluid. Therefore, in this embodiment, each part of the air compressor is taken as a monitoring object, and corresponding sensors are set in each monitoring object to monitor the status of each part of the air compressor from different dimensions. For example: on the bearing seat of the main body part, an acceleration sensor is installed to monitor the bearing vibration signal, and a thermocouple sensor is installed to monitor the bearing temperature; a current sensor is installed in the drive system to monitor the current data of the motor, a voltage sensor is installed to monitor the voltage data of the motor, and a thermistor sensor is installed at the motor position to monitor the motor temperature; a thermistor sensor is installed at the corresponding position of the cooling and lubrication system to monitor the oil temperature and cooling water temperature respectively, a pressure sensor is installed to monitor the oil pressure, and a gear flowmeter is installed at the oil pump outlet of the cooling and lubrication system to monitor the lubricating oil flow; a thermal flowmeter and a vortex flowmeter are installed in the intake and exhaust pipes of the air system to monitor the intake flow and exhaust flow respectively, a pressure sensor is installed to monitor the intake pressure and exhaust pressure, and a thermocouple sensor is installed on the exhaust pipe to monitor the exhaust temperature.

[0025] Since the abnormal bearing vibration is mainly reflected in the changes in the peak factor, kurtosis and RMS (root mean square) of the vibration signal, in this embodiment, the bearing vibration signal monitored by the acceleration sensor installed on the bearing seat is obtained once per second as the data monitored in the bearing. There is no restriction here and it can be set according to the specific implementation scenario. The peak factor , kurtosis , ,in, is the absolute maximum amplitude of the bearing vibration signal, is the vibration acceleration at the i-th sampling moment; is the mean value of vibration acceleration, is the standard deviation of vibration acceleration, n is the number of monitored data points, and the peak factor, kurtosis, and RMS belong to the existing technology and will not be described here.

[0026] Because the units and numerical ranges of different data obtained are different, in order to facilitate analysis, all the obtained data are first normalized to obtain monitoring data. Normalization processing belongs to the existing technology and will not be repeated here. The monitoring frequency of each sensor in this embodiment is 1Hz. In this embodiment, M-dimensional monitoring data of each monitored object in the air compressor are obtained once per second to form multidimensional monitoring data at each moment, M≥2, and the multidimensional monitoring data obtained at the current moment is recorded as multidimensional real-time monitoring data. The week before the current moment is used as the preset historical period, and the multidimensional monitoring data at each moment in the preset historical period is obtained and recorded as multidimensional historical monitoring data. There is no restriction here and it can be set according to the specific implementation scenario.

[0027] In addition, when obtaining the monitoring data of M dimensions of each monitoring object in the air compressor at any time, the corresponding operating status of each monitoring object at any time is obtained. The operating status includes: fault, moderate abnormality, mild abnormality and normal. For example, the operating status of the main engine part at any time is normal.

[0028] Step S102 : Screen at least one suspected real multidimensional monitoring data from all multidimensional historical monitoring data, and obtain the data authenticity of each dimension in each suspected real multidimensional monitoring data based on the periodic characteristics and dimension data association characteristics of all suspected real multidimensional monitoring data.

[0029] When the air compressor is in normal operation, the monitoring data of each dimension should be in a stable and regular change. When the air compressor has an abnormality, whether it is an abnormality in the air compressor itself or an abnormality in the sensor, the monitoring data will deviate from the original stable change law. Because the monitoring data of each dimension in the air compressor under normal operation will not change drastically in a short period of time, but will fluctuate within the normal range. Therefore, for any multidimensional historical monitoring data, any multidimensional historical monitoring data is taken as the last data, and a window with a preset length of 10 is constructed in the multidimensional historical monitoring data within the preset historical period (that is, the window includes 10 multidimensional historical monitoring data). There is no restriction here. It can be set according to the specific implementation scenario. According to the difference between the multidimensional historical monitoring data in the window, the abnormality degree of each dimension monitoring data in any multidimensional historical monitoring data is obtained, which is used for subsequent analysis of the operating status of each monitoring object in the air compressor.

[0030] Among them, according to the difference between the multi-dimensional historical monitoring data in the window, the method for obtaining the abnormality degree of each dimension monitoring data in any multi-dimensional historical monitoring data is as follows: For any dimension monitoring data in any multi-dimensional historical monitoring data, acquiring monitoring data belonging to the same dimension as the any dimension monitoring data in the window to form a monitoring sequence; Obtain the cumulative value of the absolute value of the difference between each monitoring data in the monitoring sequence and the monitoring data of any dimension, record it as the monitoring data difference cumulative value, normalize the monitoring data difference cumulative value, and obtain the abnormality degree of the monitoring data of any dimension in the any multidimensional historical monitoring data.

[0031] In one embodiment, taking the c-th dimension monitoring data in the i-th multidimensional historical monitoring data as an example, the calculation formula for the abnormality degree of the c-th dimension monitoring data in the i-th multidimensional historical monitoring data is: in, is the abnormality degree of the c-th dimension monitoring data in the i-th multi-dimensional historical monitoring data; is the jth monitoring data in the monitoring sequence; is the c-th dimension monitoring data in the i-th multi-dimensional historical monitoring data; is the number of monitoring data in the monitoring sequence; norm() is the normalization function; is the absolute value symbol.

[0032] It should be noted that the greater the difference between each monitoring data in the monitoring sequence and the monitoring data of the cth dimension in the i-th multidimensional historical monitoring data, the more obvious the fluctuation of the monitoring data in the monitoring sequence, and the greater the degree of abnormality of the monitoring data of the c-th dimension in the i-th multidimensional historical monitoring data.

[0033] Similarly, the abnormality degree of each dimension monitoring data in each multidimensional historical monitoring data can be obtained.

[0034] When the sensor is in abnormal conditions such as poor contact or overtemperature, the monitored data value may deviate from the actual data of the air compressor, resulting in the monitoring data not being able to effectively reflect the operating status of the air compressor. The monitoring data may be abnormal while the air compressor is operating normally. Analyzing the current air compressor status based on the multi-dimensional historical monitoring data at this time will often cause the air compressor status analysis results to be incorrect. Therefore, it is necessary to remove the abnormal data generated by sensor abnormalities in the multi-dimensional historical monitoring data, and filter the actual operating data of the air compressor in the multi-dimensional historical monitoring data to reflect the operating status of the air compressor.

[0035] Abnormal data changes caused by air compressor abnormalities are usually gradually accumulated, that is, they change gradually according to a certain trend. When the sensor contact is poor or the transmission is abnormal, there will be obvious extreme changes in the data (such as sudden changes to 0 or the maximum upper limit value). Therefore, when the data suddenly shows a large abnormality, it is more likely to be caused by sensor abnormality. Therefore, if the abnormality level of all dimensional monitoring data in any multidimensional historical monitoring data is less than or equal to the preset abnormality level threshold, it is confirmed that any multidimensional historical monitoring data may be the real operating data of the air compressor, and any multidimensional historical monitoring data is recorded as suspected real multidimensional monitoring data. In this embodiment, the preset abnormality level threshold is 0.8, which is set according to historical experience. There is no restriction here and it can be set according to the specific implementation scenario. If the abnormality level of any dimensional monitoring data in any multidimensional historical monitoring data is greater than 0.8, it is confirmed that any multidimensional historical monitoring data is obviously abnormal data caused by sensor abnormality.

[0036] In addition to poor contact anomalies, sensors may also be affected by other interference factors, causing data drift anomalies in the monitored data. In this case, the above-mentioned preset anomaly threshold cannot completely eliminate the abnormal data caused by sensor anomalies. Since the changes in air compressor monitoring data are strongly correlated with the air compressor operating cycle, the abnormal performance of air compressor abnormal data will also show periodic changes. For example, bearing wear anomalies will show more obvious increases in characteristic values such as kurtosis and RMS during air compressor loading, and the characteristic values will be slightly relieved during unloading. Sensor anomalies generally manifest as irregular fluctuations or data drift, and do not have the periodic characteristics of air compressor operation. Therefore, based on the periodic characteristics of suspected real multi-dimensional monitoring data, it is possible to determine whether the suspected real multi-dimensional monitoring data is real air compressor data.

[0037] Since abnormal conditions such as sensor power supply abnormalities often cause irregular fluctuations in sensor monitoring data, the changes in the abnormality degree of the corresponding monitoring data are also irregular. The data abnormalities corresponding to air compressor abnormalities often increase gradually with a certain trend of change, and the data abnormalities closely related to the air compressor operation cycle show periodic changes.

[0038] Therefore, in order to analyze the periodic characteristics of suspected real multidimensional monitoring data, for any suspected real multidimensional monitoring data, with the historical moment of any suspected real multidimensional monitoring data as the center, a time window with a preset time length of 10 minutes is established within one week before the current moment. There is no restriction here. According to the specific implementation scenario setting, the suspected real multidimensional monitoring data within the time window can be composed into a suspected real multidimensional monitoring data sequence; for any dimension in any suspected real multidimensional monitoring data, in the abnormality degree of each dimension monitoring data in the suspected real multidimensional monitoring data sequence, the abnormality degree under any dimension is obtained to form an abnormality degree subsequence, and the abnormality degree subsequence is fitted to obtain an abnormality degree fitting curve. The fitting curve belongs to the prior art and will not be repeated here. According to the periodic characteristics of the abnormality degree fitting curve, the periodic characteristic value of the data abnormality degree of any dimension in any suspected real multidimensional monitoring data can be obtained.

[0039] Among them, according to the periodic characteristics of the anomaly degree fitting curve, the method for obtaining the periodic characteristic value of the data anomaly degree of any dimension in any suspected real multi-dimensional monitoring data is as follows: Since the data drift anomaly caused by sensor calibration and other problems will cause the monitoring data to increase a certain value relative to the real data as a whole or drift more and more over time, at this time, the data anomaly degree caused by data drift will not show periodic changes, and the air compressor anomaly will often make the relevant data more obvious during the loading process of the air compressor, while the anomaly is relatively mild during the unloading process, with a certain periodic regularity. Therefore, the anomaly degree fitting curve is divided into at least two sub-curves according to the air compressor cycle, that is, each sub-curve is a cycle period, and the air compressor cycle period depends on the load of the air compressor. In this embodiment, the reference value of the air compressor cycle period is set to 5 minutes, which is not limited here and can be set according to the specific implementation scenario. If the similarity of different sub-curves is high, it means that the anomaly degree change of the monitoring data corresponding to the sub-curve is periodic. In order to compare the similarity of each two curves, a preset number of fitting values are evenly obtained on each sub-curve, and the positions of the fitting values on each sub-curve correspond one to one. In this embodiment, the preset number is set to 20, which is not limited here and can be set according to the specific implementation scenario. For any two adjacent sub-curves, respectively obtain the fitting value differences at corresponding positions in the two adjacent sub-curves to form a fitting value difference sequence, respectively obtain the absolute value of the difference between each two adjacent fitting value differences in the fitting value difference sequence, and obtain the corresponding accumulated value of the absolute value of the difference as the degree of difference between the two adjacent sub-curves; The difference between each two adjacent sub-curves is obtained respectively, and a corresponding difference degree cumulative value is obtained. The opposite of the difference degree cumulative value is normalized to obtain the data anomaly degree periodic characteristic value of any dimension.

[0040] In one embodiment, taking the cth dimension in the bth suspected real multi-dimensional monitoring data as an example, the calculation formula for the data anomaly degree periodic characteristic value of the cth dimension in the bth suspected real multi-dimensional monitoring data is:

[0041] in, is the periodic characteristic value of the data anomaly degree of the cth dimension in the bth suspected real multi-dimensional monitoring data; is the ath fitting value in the uth sub-curve of the abnormality fitting curve; is the ath fitting value in the u+1th sub-curve of the abnormality fitting curve; is the a+1th fitting value in the uth sub-curve of the abnormality fitting curve; is the a+1th fitting value in the u+1th sub-curve of the abnormality fitting curve; is the number of fitting values in the u-th sub-curve of the abnormality fitting curve; is the preset number (i.e. the number of sub-curves in the abnormality fitting curve); norm() is the normalization function; is the absolute value symbol.

[0042] It should be noted that is the difference between the u-th sub-curve and the u+1-th sub-curve. The smaller the difference between the fitting values of the two adjacent corresponding positions in the u-th sub-curve and the u+1-th sub-curve, The smaller it is, the higher the similarity between the u-th segment sub-curve and the u+1-th segment sub-curve, and the more consistent with the periodic characteristics. The bigger it is.

[0043] If the sensor monitoring data drifts at a fixed value, meaning all monitoring data is a certain value added to the actual air compressor data, then the degree of anomaly in the monitoring data may still conform to the periodic characteristics of the air compressor. Therefore, it is not possible to determine whether the monitoring data is real air compressor data based solely on the periodic characteristic value of the data anomaly degree. However, sensor anomalies often occur individually and only affect the monitoring data of a certain dimension, while air compressor anomalies often cause anomalies in the monitoring data of multiple related dimensions to occur simultaneously. Therefore, it is necessary to determine the data authenticity of any dimension in any suspected real multidimensional monitoring data based on the periodic characteristic value of the data anomaly degree in any dimension, combined with the anomaly degree correlation characteristics of the monitoring data of each dimension in any suspected real multidimensional monitoring data, and determine whether the monitoring data of any dimension in any suspected real multidimensional monitoring data is real air compressor data.

[0044] Among them, according to the periodic characteristic value of the data anomaly degree of any dimension in any suspected real multidimensional monitoring data, combined with the anomaly degree correlation characteristics of the monitoring data of each dimension in any suspected real multidimensional monitoring data, the method for obtaining the data authenticity of any dimension in any suspected real multidimensional monitoring data is as follows: Obtaining the determination coefficient of the abnormality fitting curve, which belongs to the prior art and will not be described in detail here, calculating the sum of the determination coefficient and the periodic characteristic value of the abnormality degree of the data in any dimension to obtain the periodic regularity characteristic value of the monitoring data in any dimension; Obtain a cumulative value of the absolute value of the difference between the degree of abnormality corresponding to each dimension in any suspected real multidimensional monitoring data and the any dimension, record it as the cumulative value of the abnormality degree difference, and use the inverse of the cumulative value of the abnormality degree difference as the dimensional correlation degree between the any dimension and each dimension in the any suspected real multidimensional monitoring data; The product of the periodic regularity characteristic value of the monitoring data of any dimension and the dimensional correlation degree between the any dimension and each dimension in the any suspected real multidimensional monitoring data is normalized to obtain the data authenticity of any dimension in the any suspected real multidimensional monitoring data.

[0045] In one embodiment, taking the cth dimension in the bth suspected true multidimensional monitoring data as an example, the calculation formula for the data authenticity of the cth dimension in the bth suspected true multidimensional monitoring data is:

[0046] in, is the authenticity of the cth dimension in the bth suspected real multi-dimensional monitoring data; is the determination coefficient of the fitting curve of the degree of abnormality; is the periodic characteristic value of the data anomaly degree of the cth dimension in the bth suspected real multi-dimensional monitoring data; is the abnormality degree of the c-th dimension monitoring data in the b-th suspected real multi-dimensional monitoring data; is the abnormality degree of the monitoring data of the oth dimension in the bth suspected real multidimensional monitoring data; M is the number of dimensions in the bth suspected real multidimensional monitoring data; norm() is the normalization function; is the absolute value symbol.

[0047] It should be noted that is the periodic regularity characteristic value of the monitoring data in the cth dimension, The larger it is, the better the fitting goodness of the abnormality fitting curve is, that is, the stronger the regularity of the abnormality of the monitoring data in the data segment corresponding to the abnormality fitting curve is. The larger it is, the more the abnormality of the monitoring data in the cth dimension is consistent with the abnormality of the air compressor. the bigger it is; The larger it is, the stronger the regularity of the abnormality of the monitoring data in the data segment corresponding to the abnormality fitting curve is. The larger it is, the more the abnormality of the monitoring data in the cth dimension is consistent with the abnormality of the air compressor. the bigger it is; is the dimensional correlation degree between the cth dimension in the bth suspected real multidimensional monitoring data and each dimension in the ith multidimensional historical monitoring data. The smaller the difference in the abnormality degree between each dimension in the bth suspected real multidimensional monitoring data and the cth dimension, the greater the correlation degree of the monitoring data of each dimension in the bth suspected real multidimensional monitoring data, and the more the abnormality degree of the monitoring data of the cth dimension is caused by the abnormality of the air compressor. The bigger it is.

[0048] Similarly, the authenticity of each dimension of each suspected real multi-dimensional monitoring data is obtained.

[0049] Step S103: for any monitored object, obtain N reference dimensions of the monitored object at the current moment. , according to the authenticity of each reference dimension in each suspected real multidimensional monitoring data, obtain the influence weight of each suspected real multidimensional data on any of the monitoring objects.

[0050] Since the operating status of each monitored object in the air compressor is reflected by multiple monitoring data, and the same degree of fluctuation of different monitoring data reflects different abnormalities in the state of the air compressor, that is, the normal fluctuation range of different monitoring data is different. Therefore, the more similar the abnormal performance of each dimension of the suspected real multidimensional monitoring data is to the multidimensional real-time monitoring data at the current moment, the closer the air compressor state corresponding to the suspected real multidimensional monitoring data is to the operating state of the air compressor at the current moment, and the corresponding suspected real multidimensional monitoring data has a greater influence weight on the analysis result of the operating state of the air compressor at the current moment. Therefore, it is necessary to obtain the influence weight of each suspected real multidimensional monitoring data on the operating state of the air compressor at the current moment to determine the operating state of the air compressor at the current moment.

[0051] Because each monitoring object of the air compressor will have multiple monitoring data, and the monitoring data with the most obvious abnormal performance is the monitoring data that can best represent the current operating status of the air compressor, for any monitoring object in the air compressor, the abnormality level of each dimension of the monitoring data of any monitoring object at the current moment is sorted from high to low (that is, the abnormality level of the monitoring data of each dimension corresponding to any monitoring object in the multi-dimensional real-time monitoring data is sorted), and the first N reference dimensions of any monitoring object at the current moment are selected, 2≤N≤M. In this embodiment, N is set to 3, which is not limited here. It can be set according to the specific implementation scenario, and the monitoring data of the first three reference dimensions are used as a reference basis for evaluating the operating status of any monitoring object at the current moment. Accordingly, the closer the monitoring data of the reference dimension corresponding to the suspected real multi-dimensional monitoring data is to the real data of the air compressor, the greater its reference significance, that is, the greater the weight of its influence on the analysis result of the current operating status of any monitoring object. Therefore, according to the data authenticity of each reference dimension in each suspected real multi-dimensional monitoring data, the influence weight of each suspected real multi-dimensional data on any monitoring object is obtained.

[0052] Among them, according to the authenticity of each reference dimension in each suspected real multidimensional monitoring data, the method for obtaining the influence weight of each suspected real multidimensional data on any monitoring object is as follows: For any suspected real multidimensional monitoring data, the average of the data authenticity of all reference dimensions is calculated according to the data authenticity of each dimension in the any suspected real multidimensional monitoring data, and recorded as the influence weight of the any suspected real multidimensional monitoring data on the any monitoring object.

[0053] In one embodiment, taking the qth monitoring object and the bth suspected real multi-dimensional monitoring data as an example, the calculation formula for the influence weight of the bth suspected real multi-dimensional monitoring data on the qth monitoring object is:

[0054] in, is the influence weight of the bth suspected true multi-dimensional monitoring data on the qth monitoring object; is the authenticity of the data of the dth reference dimension in the bth suspected true multidimensional monitoring data; N is the number of reference dimensions of each monitoring object.

[0055] It should be noted that The larger the value is, the closer the monitoring data of the dth reference dimension in the bth suspected real multi-dimensional monitoring data is to the real data of the air compressor, and the greater its reference significance is. The bigger it is.

[0056] Similarly, obtain the influence weight of each suspected real multidimensional data on the qth monitoring object.

[0057] Step S104, obtaining the state voting weight of each suspected real multidimensional monitoring data on the any monitored object according to the influence weight of each suspected real multidimensional monitoring data on the any monitored object and the difference between each suspected real multidimensional monitoring data and the multidimensional real-time monitoring data.

[0058] After obtaining the influence weight of each suspected real multidimensional data on the qth monitoring object, the operating status of the qth monitoring object at the current moment can be judged based on the influence weight of each suspected real multidimensional data on the qth monitoring object and the monitoring data of each reference dimension corresponding to any monitoring object at the current moment.

[0059] In this embodiment, the KNN algorithm is used to classify the operating status of the qth monitoring object at the current moment, and a three-dimensional coordinate system is established with the three reference dimensions of the qth monitoring object at the current moment as coordinate axes. The three coordinate axes of the three-dimensional coordinate system respectively represent the monitoring data of the three reference dimensions in the suspected real multidimensional monitoring data. The K value in the KNN algorithm is set to 50. There is no restriction here. It can be set according to the specific implementation scenario. 50 suspected real multidimensional monitoring data are obtained from all suspected real multidimensional monitoring data and mapped to the three-dimensional coordinate system to obtain 50 reference data points.

[0060] The multi-dimensional real-time monitoring data at the current moment is mapped to the three-dimensional coordinate system to obtain the real-time data point. The closer the reference data point is to the real-time data point of the current qth monitoring object in the three-dimensional coordinate system, the closer the reference data point is to the operating status of the current qth monitoring object, and the greater its voting weight should be. Therefore, for any reference data point, the voting weight of any reference data point on the status of the qth monitoring object can be obtained based on the distance difference between any reference data point and the real-time data point in the three-dimensional coordinate system, combined with the influence weight of the suspected real multi-dimensional monitoring data corresponding to any reference data point on the qth monitoring object, and then the voting weight of each reference data point on the status of the qth monitoring object can be used to obtain the operating status of the qth monitoring object at the current moment.

[0061] Among them, according to the distance difference between any reference data point and the real-time data point in the three-dimensional coordinate system, combined with the influence weight of the suspected real multi-dimensional monitoring data corresponding to any reference data point on the qth monitoring object, the method for obtaining the state voting weight of any reference data point on the qth monitoring object is as follows: Obtain the Euclidean distance between any reference data point and the real-time data point in the three-dimensional coordinate system. The Euclidean distance belongs to the prior art and will not be described in detail here. The product of the reciprocal of the Euclidean distance and the influence weight of the suspected real multidimensional monitoring data corresponding to any reference data point on the qth monitoring object is obtained to obtain the state voting weight of any reference data point on the qth monitoring object.

[0062] In one embodiment, taking the fth reference data point as an example, the calculation formula for the voting weight of the fth reference data point on the state of the qth monitoring object is:

[0063] in, is the voting weight of the f-th reference data point on the state of the q-th monitoring object; is the Euclidean distance between the fth reference data point and the real-time data point in the three-dimensional coordinate system; is the influence weight of the suspected real multidimensional monitoring data corresponding to the f-th reference data point on the q-th monitoring object.

[0064] It should be noted that The smaller it is, the closer the fth reference data point is to the operating status of the current qth monitoring object. the bigger it is; The larger the value is, the greater the influence weight of the fth reference data point on the analysis result of the operating status of the current qth monitoring object is. The bigger it is.

[0065] Similarly, the state voting weight of each reference data point for the qth monitoring object is obtained, and the state voting weight of each reference data point for the qth monitoring object is used as the state voting weight of the suspected real multidimensional monitoring data corresponding to each reference data point for the qth monitoring object.

[0066] In step S105, the multidimensional real-time monitoring data is classified using the KNN algorithm according to the state voting weight of each suspected real multidimensional monitoring data to obtain the state evaluation result of any monitoring object at the current moment. Based on the state evaluation results of all monitoring objects at the current moment, the operating state of the air compressor at the current moment is determined.

[0067] After obtaining the voting weight of each reference data point on the state of the qth monitored object, the voting weight of each reference data point on the state of the qth monitored object is used as the voting weight of the data point in the KNN algorithm. Each reference data point corresponds to a historical moment. The operating status of each monitored object at each moment has been obtained in step S101. Therefore, the operating status of the qth monitored object at the historical moment corresponding to each reference data point and the voting weight of each reference data point on the state of the qth monitored object can be combined to use the KNN algorithm to classify the real-time data points of the qth monitored object in the multi-dimensional real-time monitoring data to obtain the state evaluation result of the qth monitored object at the current moment. The classification using the KNN algorithm belongs to the existing technology and will not be repeated here.

[0068] Similarly, the status evaluation results of each monitoring object in the air compressor at the current moment are obtained, and early warnings are issued based on the status evaluation results: If the status evaluation results of each monitoring object in the air compressor are normal, the operating status of the air compressor at the current moment is normal, and no early warning is required at the current moment; If the status assessment result of any monitoring object in the air compressor is slightly abnormal, it is necessary to issue an early warning to the air compressor and adjust the parameters corresponding to the monitoring object in the air compressor to restore normal operation; If the status assessment result of any monitoring object in the air compressor is moderately abnormal, the air compressor needs to be warned immediately, the abnormality in the air compressor needs to be promptly checked and adjustment measures need to be taken; If the status assessment result of any monitoring object in the air compressor is faulty, it means that the air compressor has an obvious fault problem and needs to be stopped immediately and repaired.

[0069] In summary, the present invention obtains monitoring data of M dimensions of each monitoring object in the air compressor at any time to form multidimensional monitoring data at any time, M≥2, obtains multidimensional real-time monitoring data at the current time and multidimensional historical monitoring data within a preset historical period; selects at least one suspected real multidimensional monitoring data from all multidimensional historical monitoring data, and obtains the data authenticity of each dimension in each suspected real multidimensional monitoring data based on the periodic characteristics and dimensional data association characteristics of all suspected real multidimensional monitoring data; for any monitoring object, obtains N reference dimensions of the monitoring object at the current time, Based on the data authenticity of each reference dimension in each suspected real multidimensional monitoring data, the influence weight of each suspected real multidimensional monitoring data on the monitoring object is obtained; based on the influence weight of each suspected real multidimensional monitoring data on the monitoring object and the difference between each suspected real multidimensional monitoring data and the multidimensional real-time monitoring data, the state voting weight of each suspected real multidimensional monitoring data on the monitoring object is obtained; based on the state voting weight of each suspected real multidimensional monitoring data, the multidimensional real-time monitoring data is classified using the KNN algorithm to obtain the state evaluation result of each monitoring object at the current moment; and based on the state evaluation results of all monitoring objects at the current moment, the operating state of the air compressor at the current moment is determined. The suspected real multidimensional monitoring data is first screened, and then the influence weight of each suspected real multidimensional monitoring data on any monitoring object in the air compressor is adjusted by obtaining the data authenticity of each dimension in each suspected real multidimensional monitoring data. This effectively reduces the impact of abnormal data caused by sensor abnormalities on the accuracy of the air compressor state monitoring results, facilitates timely detection of abnormal states of the air compressor and takes corresponding measures to ensure the continuous operation and working safety of the air compressor.

[0070] The above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit the same. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. These modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the various embodiments of the present invention, and should all be included in the scope of protection of the present invention.

Claims

1. A method for monitoring the state of a screw air compressor, characterized in that: The method comprises: Obtain M-dimensional monitoring data of each monitoring object in the air compressor at any time to form multi-dimensional monitoring data at any time, M ≥ 2, and obtain multi-dimensional real-time monitoring data at the current time and multi-dimensional historical monitoring data within a preset historical period; Screening at least one suspected real multidimensional monitoring data from all multidimensional historical monitoring data, and obtaining the data authenticity of each dimension in each suspected real multidimensional monitoring data based on the period characteristics and dimension data correlation characteristics of all suspected real multidimensional monitoring data; For any monitored object, obtain N reference dimensions of the monitored object at the current moment. , according to the authenticity of each reference dimension in each suspected real multidimensional monitoring data, obtaining the influence weight of each suspected real multidimensional data on any of the monitoring objects; Obtaining a state voting weight of each suspected real multidimensional monitoring data on the any monitored object according to an influence weight of each suspected real multidimensional monitoring data on the any monitored object and a difference between each suspected real multidimensional monitoring data and the multidimensional real-time monitoring data; According to the state voting weight of each suspected real multidimensional monitoring data, the KNN algorithm is used to classify the multidimensional real-time monitoring data to obtain the state evaluation result of any monitoring object at the current moment. According to the state evaluation results of all monitoring objects at the current moment, the operating state of the air compressor at the current moment is determined.

2. A method for monitoring the state of a screw air compressor according to claim 1, characterized in that: The step of screening at least one suspected real multi-dimensional monitoring data from all multi-dimensional historical monitoring data includes: For any multidimensional historical monitoring data, taking the any multidimensional historical monitoring data as the last data, a window of preset length is constructed in the multidimensional historical monitoring data within a preset historical period, and the abnormality degree of each dimension monitoring data in the any multidimensional historical monitoring data is obtained based on the difference between the multidimensional historical monitoring data in the window; If the abnormality levels of all dimensional monitoring data in any multidimensional historical monitoring data are less than or equal to a preset abnormality level threshold, the any multidimensional historical monitoring data is recorded as suspected true multidimensional monitoring data.

3. A method for monitoring the state of a screw air compressor according to claim 2, characterized in that: The obtaining, based on the difference between the multidimensional historical monitoring data in the window, the abnormality degree of each dimension monitoring data in any multidimensional historical monitoring data includes: For any dimension monitoring data in any multi-dimensional historical monitoring data, obtaining monitoring data belonging to the same dimension as the any dimension monitoring data in the window to form a monitoring sequence; Obtain the cumulative value of the absolute value of the difference between each monitoring data in the monitoring sequence and the monitoring data of any dimension, record it as the monitoring data difference cumulative value, normalize the monitoring data difference cumulative value, and obtain the abnormality degree of the monitoring data of any dimension in the any multidimensional historical monitoring data.

4. A method for monitoring the state of a screw air compressor according to claim 2, characterized in that: The step of obtaining the authenticity of each dimension of each suspected real multidimensional monitoring data based on the periodic characteristics and dimension data association characteristics of all suspected real multidimensional monitoring data includes: For any suspected real multidimensional monitoring data, a time window of a preset time length is established within a preset historical period, with the historical moment of the suspected real multidimensional monitoring data as the center, and the suspected real multidimensional monitoring data within the time window are combined into a suspected real multidimensional monitoring data sequence; For any dimension of any suspected real multidimensional monitoring data, from the abnormality degree of each dimension monitoring data in the suspected real multidimensional monitoring data sequence, obtain the abnormality degree under any dimension to form an abnormality degree subsequence, and fit the abnormality degree subsequence to obtain an abnormality degree fitting curve; Obtaining a periodic characteristic value of the data anomaly degree of any dimension in any suspected real multi-dimensional monitoring data according to the periodic characteristic of the anomaly degree fitting curve; The authenticity of the data in any dimension of the suspected real multidimensional monitoring data is obtained based on the periodic characteristic value of the data anomaly degree of any dimension in the suspected real multidimensional monitoring data and the anomaly degree correlation characteristics of the monitoring data of each dimension in the suspected real multidimensional monitoring data.

5. The method for monitoring the state of a screw air compressor according to claim 4, characterized in that: The step of obtaining a data anomaly degree periodic characteristic value of any dimension in any suspected real multi-dimensional monitoring data according to the periodic characteristic of the anomaly degree fitting curve includes: Dividing the abnormality degree fitting curve into at least two sub-curves according to the air compressor cycle, uniformly obtaining a preset number of fitting values on each sub-curve, and the positions of the fitting values on each sub-curve correspond one to one; For any two adjacent sub-curves, respectively obtain the fitting value differences at corresponding positions in the two adjacent sub-curves to form a fitting value difference sequence, respectively obtain the absolute value of the difference between each two adjacent fitting value differences in the fitting value difference sequence, and obtain the corresponding accumulated value of the absolute value of the difference as the degree of difference between the two adjacent sub-curves; The difference between each two adjacent sub-curves is obtained respectively, and a corresponding difference degree cumulative value is obtained. The opposite of the difference degree cumulative value is normalized to obtain the data anomaly degree periodic characteristic value of any dimension.

6. A method for monitoring the state of a screw air compressor according to claim 4, characterized in that: The obtaining, based on the periodic characteristic value of the data anomaly degree of any dimension in the any suspected real multidimensional monitoring data and the anomaly degree correlation feature of the monitoring data of each dimension in the any suspected real multidimensional monitoring data, the authenticity of the data in any dimension in the any suspected real multidimensional monitoring data includes: Obtaining a determination coefficient of the abnormality fitting curve, calculating the sum of the determination coefficient and the periodic characteristic value of the abnormality degree of the data in any dimension, and obtaining a periodic regularity characteristic value of the monitoring data in any dimension; Obtain a cumulative value of the absolute value of the difference between the degree of abnormality corresponding to each dimension in any suspected real multidimensional monitoring data and the any dimension, record it as the cumulative value of the abnormality degree difference, and use the inverse of the cumulative value of the abnormality degree difference as the dimensional correlation degree between the any dimension and each dimension in the any suspected real multidimensional monitoring data; The product of the periodic regularity characteristic value of the monitoring data of any dimension and the dimensional correlation degree between the any dimension and each dimension in the any suspected real multidimensional monitoring data is normalized to obtain the data authenticity of any dimension in the any suspected real multidimensional monitoring data.

7. The method for monitoring the state of a screw air compressor according to claim 1, characterized in that: The step of obtaining the influence weight of each suspected real multidimensional data on any one of the monitoring objects according to the authenticity of each reference dimension in each suspected real multidimensional monitoring data includes: For any suspected real multidimensional monitoring data, the average of the data authenticity of all reference dimensions is calculated according to the data authenticity of each dimension in the any suspected real multidimensional monitoring data, and recorded as the influence weight of the any suspected real multidimensional monitoring data on the any monitoring object.

8. The method for monitoring the state of a screw air compressor according to claim 1, characterized in that: Obtaining a state voting weight of each suspected real multidimensional monitoring data on the any monitored object according to an influence weight of each suspected real multidimensional monitoring data on the any monitored object and a difference between each suspected real multidimensional monitoring data and the multidimensional real-time monitoring data includes: An N-dimensional coordinate system is established with the N reference dimensions of any monitoring object at the current moment as coordinate axes, wherein the N coordinate axes of the N-dimensional coordinate system respectively represent the monitoring data of each reference dimension in the suspected real multidimensional monitoring data; For any suspected real multi-dimensional monitoring data, obtaining the Euclidean distance between the suspected real multi-dimensional monitoring data and the multi-dimensional real-time monitoring data in the N-dimensional coordinate system; The product of the reciprocal of the Euclidean distance and the influence weight of any suspected real multidimensional monitoring data on any monitored object is obtained to obtain the state voting weight of any suspected real multidimensional monitoring data on any monitored object.

Images