A medical oxygen air compressor fault prediction method and system

Abstract

The present invention relates to the field of data processing, and the present invention relates to a method and system for predicting faults of medical oxygen air compressors. The method comprises: collecting pressure data of the medical oxygen air compressor, and obtaining the noise performance degree of each pressure data point; generating a first-order cumulative sequence based on the collected pressure data, and calculating the credibility of each cumulative value in the first-order cumulative sequence based on the noise performance degree of the pressure data point, generating an adjacent mean sequence based on the first-order cumulative sequence, establishing a grey differential equation based on the adjacent mean sequence, and then obtaining the predicted value of the original pressure data based on the grey differential equation; in response to the predicted value exceeding the preset range of the pressure data, determining that the medical oxygen air compressor is abnormal. The method of the present invention can effectively improve the accuracy of fault prediction of medical oxygen air compressors.

Description

Technical Field

[0001] The present invention relates to the field of data processing, and more specifically, to a method and system for predicting a medical oxygen air compressor failure. Background Art

[0002] With the expansion of hospital scale and the advancement of medical technology, the importance of medical oxygen systems in clinical medicine has become increasingly prominent, especially in key links such as intensive care, surgery and emergency treatment. With the advancement of industrial technology, the diversity and precision of medical oxygen systems have increased, and the complexity of their constituent equipment, components, and system hardware and software has also increased. However, this complex system also brings potential failure risks. When an abnormality occurs in the medical oxygen system, if the fault diagnosis cannot be carried out quickly and targeted maintenance measures cannot be taken, serious medical accidents may occur and threaten the patient's life safety. In the medical oxygen system, the medical oxygen air compressor is the most important equipment. Therefore, it is crucial to predict the failure of the medical oxygen air compressor.

[0003] In the medical oxygen system, the main function of the medical oxygen air compressor is to pressurize the ambient air to the pressure required by the molecular sieve adsorption tower to ensure the normal operation of the molecular sieve. Therefore, by monitoring the pressure data of the medical oxygen air compressor for fault prediction, it can accurately reflect the potential fault hazards of the medical oxygen air compressor. The model is a commonly used grey prediction model, which is mainly used for the prediction and analysis of time series data. It converts the original data into a grey differential equation by performing a certain mathematical transformation on the original data. The equation can be used to analyze and predict the law of the pressure series data of the medical oxygen air compressor. However, due to the internal factors of the medical oxygen air compressor, there is noise in the original data. Even if it is filtered by the existing filtering algorithm, it is difficult to completely remove the noise. If the algorithm directly generates the adjacent mean sequence by the mean of the adjacent accumulated values, it will be affected by the false data in the sequence, resulting in the model being unable to accurately capture the real change trend of the data, which will lead to deviations in the prediction of the pressure data and the low accuracy of the medical oxygen air compressor failure prediction. (Xiong Manchen et al., Research on the Development of Civil Aviation Economy in the Post-epidemic Era Based on Grey Prediction Model, 2024) discloses a research method for the development of civil aviation economy in the post-epidemic era based on a grey prediction model. It first generates a first-order cumulative sequence based on the original data, and then takes the average of the adjacent cumulative values ​​in the first-order cumulative sequence to obtain an adjacent mean sequence, and then establishes a grey differential equation to predict the original data; however, due to the noise in the pressure data of the medical oxygen air compressor, for the pressure data of the air compressor, if the adjacent mean sequence is generated directly by the mean of adjacent cumulative values, the established grey differential equation will be inaccurate, which will lead to deviations in the prediction of the pressure data, and then lead to low accuracy in the prediction of medical oxygen air compressor failures. Summary of the invention

[0004] In order to solve the technical problem of low accuracy in predicting medical oxygen air compressor failures in the prior art, the present invention provides solutions in the following aspects.

[0005] In a first aspect, the present invention provides a method for predicting a medical oxygen air compressor failure, comprising:

[0006] Collecting pressure data of the medical oxygen air compressor and obtaining the noise performance degree of each pressure data point; the noise performance degree is calculated based on the regularity within the local range of the pressure data point, and the higher the regularity, the lower the noise performance degree;

[0007] Generate a first-order cumulative sequence based on the collected pressure data, and calculate the credibility of each cumulative value in the first-order cumulative sequence based on the noise performance of the pressure data point, and the credibility of the cumulative value is negatively correlated with the noise performance of the corresponding pressure data point; generate an adjacent mean sequence based on the first-order cumulative sequence, wherein the kth mean in the adjacent mean sequence The calculation expression is:

[0008] ;

[0009] In the formula, represents the credibility of the k-1th accumulated value, represents the credibility of the kth accumulated value, represents the kth accumulated value, Indicates Accumulated values;

[0010] Establishing a grey differential equation according to the adjacent mean value sequence, and then obtaining a predicted value of the original pressure data according to the grey differential equation;

[0011] In response to the predicted value exceeding a preset range of the pressure data, it is determined that an abnormality occurs in the medical oxygen air compressor.

[0012] Its beneficial effect is: since the pressure data of the medical oxygen air compressor during operation is easily affected by noise, when using the gray GM (1, 1) model to predict its pressure data, if the adjacent mean sequence obtained by directly using the mean of adjacent accumulated values ​​to establish the gray differential equation, the model accuracy will be low, resulting in inaccurate pressure data prediction values.

[0013] The medical oxygen air compressor fault prediction method of the present invention takes into account the credibility of the two cumulative values ​​corresponding to the mean when generating an adjacent mean sequence based on a first-order cumulative sequence, assigns a lower weight to the cumulative value with a lower credibility, and assigns a higher weight to the cumulative value with a higher credibility, so that the calculated mean is more accurate; thereby, the established grey differential equation more accurately reflects the nonlinear relationship in the pressure data and the influence of external factors, thereby making the predicted value of the acquired pressure data more accurate; in addition, when calculating the credibility of the cumulative value, the influence of the noise of the pressure data point on the cumulative value is taken into account, and firstly, the noise performance degree of the corresponding pressure data point is averaged, so that the credibility of the calculated cumulative value is inversely proportional to the average of the noise performance degree of the corresponding pressure data point, and the greater the noise performance degree of the pressure data point, the lower the credibility of the calculated cumulative value, thereby more accurately calculating the credibility of the cumulative value, and the method of the present invention can effectively improve the accuracy of fault prediction of the medical oxygen air compressor.

[0014] Preferably, the calculation expression of the credibility of the accumulated value is:

[0015] ;

[0016] In the formula, represents the credibility of the kth accumulated value, Indicates the qth pressure data point The noise performance level, Represents an exponential function with the natural constant e as the base.

[0017] Preferably, for the i-th data point of the pressure data sequence, the method for obtaining the noise performance degree thereof includes:

[0018] Calculate the Euclidean distances between two adjacent data points in the local range of the data point, and find the average of the calculated Euclidean distances;

[0019] Obtaining the Euclidean distance associated with the data point, where the Euclidean distance associated with the data point refers to the Euclidean distance between the data point and its adjacent data points;

[0020] The noise performance degree of the data point is calculated according to the deviation between the mean and the Euclidean distance associated with the data point, and the noise performance degree is positively correlated with the deviation.

[0021] The beneficial effect is as follows: in this embodiment, the difference between the Euclidean distance between a data point and its adjacent data points and the Euclidean distance between two adjacent data points in the local range of the data point is used to measure the regularity of the change of the pressure data in the local range of the i-th data point. The larger the difference, the more it means that the appearance of the i-th data point causes the data in its local range to change irregularly, and the worse its regularity. The noise performance degree acquisition method of this embodiment can more accurately calculate the noise performance degree of the i-th data point of the pressure data sequence.

[0022] Preferably, for the i-th data point in the pressure data sequence, its related Euclidean distance includes the Euclidean distance between the data point and its previous data point, and the Euclidean distance between the data point and its next data point, and the calculation expression of the noise performance degree is:

[0023] ;

[0024] In the formula, Indicates the degree of noise performance of the ith data point, represents the number of data points in the pressure data sequence P, Represents the i-th pressure data point The number of data points in the local area; Represents the i-th pressure data point The relative distance between the jth and j+1th data points in the local range, represents the Euclidean distance between the ith data point and its previous data point, Represents the Euclidean distance between the i-th data point and the next data point.

[0025] The beneficial effect is as follows: in this embodiment, when the i-th data point is not the first or last data point in the pressure data sequence, the Euclidean distance between the i-th data point and its previous data point and the average of the Euclidean distance between the i-th data point and its subsequent data point are used to characterize the Euclidean distance between the i-th data point and its adjacent data points, thereby making the calculation of the Euclidean distance related to the i-th data point more accurate; and thus making the calculation result of the noise performance degree of the data point more accurate.

[0026] Preferably, the local range of the i-th data point is a range of w data points in length centered on the current data point.

[0027] Preferably, the method further includes correcting the noise performance of the pressure data points, including:

[0028] While collecting pressure data, temperature data and vibration data of the medical oxygen air compressor are collected to obtain a first temperature data point and a first vibration data point corresponding to the i-th pressure data point collection moment; obtain a positive correlation between a change of the pressure data point within a local range of the i-th pressure data point and a change of the temperature data point within a local range of the first temperature data point, recorded as a first positive correlation; and a positive correlation between a change of the pressure data point within a local range of the i-th pressure data point and a change of the vibration data point within a local range of the first vibration data point, recorded as a second positive correlation;

[0029] The corrected noise performance degree is calculated according to the first positive correlation and the second positive correlation, and the corrected noise performance degree is negatively correlated with both the first positive correlation and the second positive correlation.

[0030] The beneficial effects are as follows: Under normal circumstances, during the operation of the medical oxygen air compressor, there is a close relationship between the pressure data, the temperature data and the vibration data, because when the gas is compressed, the collisions between the gas molecules increase, resulting in an increase in temperature, that is, the pressure data and the temperature data are positively correlated; at the same time, changes in pressure data will also cause the mechanical parts inside the compressor to be subjected to uneven forces, resulting in changes in vibration data, so there is also a positive correlation between the pressure data and the vibration data. Therefore, the noise performance degree of the pressure data point can be corrected according to the first positive correlation and the second positive correlation, and the corrected noise performance degree is negatively correlated with both the first positive correlation and the second positive correlation; the method of this embodiment can be used to more accurately correct the noise performance degree of the pressure data point, so that the calculation result of the noise performance degree of the pressure data point is more accurate, thereby improving the accuracy of the prediction result when predicting faults of the medical oxygen air compressor.

[0031] Preferably, the first positive correlation and the second positive correlation are both obtained by using the Pearson correlation coefficient method, and the calculation expression of the corrected noise performance degree is:

[0032] ;

[0033] In the formula, Indicates the degree of noise performance of the ith data point, represents the first positive correlation, represents the second positive correlation, Indicates the corrected noise performance of the ith data point.

[0034] Preferably, the preset range of the pressure data is 0.45 MPa to 0.64 MPa.

[0035] Preferably, the method further comprises: in response to a failure of the medical oxygen air compressor, inputting the current operation data of the medical oxygen air compressor into the deep reinforcement learning model, so as to obtain the failure type of the medical oxygen air compressor; the operation data comprises temperature data, pressure data and vibration data, and the training method of the deep reinforcement learning model comprises:

[0036] Input the operation data into the policy network of the deep reinforcement learning model to generate failure mode diagnosis results;

[0037] Inputting the diagnosis result into the value network of the deep reinforcement learning model to evaluate the accuracy of the diagnosis result;

[0038] According to the accuracy of the diagnosis results and the reward mechanism of the reinforcement learning algorithm, update the parameters of the policy network and the value network, and determine whether the deep reinforcement learning model converges;

[0039] In response to the deep reinforcement learning model converging, the training is stopped; otherwise, the fault mode diagnosis results are iteratively generated, the accuracy of the diagnosis results is evaluated, and the parameters of the policy network and the value network are updated until the deep reinforcement learning model converges.

[0040] Its beneficial effects are: when it is determined that the medical oxygen air compressor has a fault, the specific fault type of the medical oxygen air compressor can be further obtained, making the fault prediction of the medical oxygen air compressor more accurate; the deep reinforcement learning model can learn complex fault modes and improve the accuracy of fault diagnosis. Through deep learning technology, the model can capture subtle changes in equipment operation and accurately identify complex fault types; by training the deep reinforcement learning model, the deep reinforcement learning model can accurately predict the fault type of the medical oxygen air compressor.

[0041] In a second aspect, the present invention provides a medical oxygen air compressor fault prediction system, comprising a processor and a memory, wherein the memory stores computer program instructions, and when the computer program instructions are executed by the processor, the medical oxygen air compressor fault prediction method of the present invention is implemented.

[0042] In summary, the beneficial effect of the present invention is that the method of the present invention can effectively improve the accuracy of fault prediction of medical oxygen air compressors. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] 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:

[0044] Figure 1is a flowchart schematically showing a method for predicting a medical oxygen air compressor failure according to an embodiment of the present invention;

[0045] Figure 2 is a schematic diagram schematically showing Euclidean distances associated with data points according to an embodiment of the present invention;

[0046] Figure 3 The figure schematically shows the structure of a medical oxygen air compressor fault prediction system according to an embodiment of the present invention. DETAILED DESCRIPTION

[0047] 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.

[0048] The specific embodiments of the present invention are described in detail below with reference to the accompanying drawings.

[0049] Medical oxygen air compressor fault prediction method embodiment:

[0050] like Figure 1 As shown, the medical oxygen air compressor fault prediction method of the present invention comprises:

[0051] S101, collecting pressure data of the medical oxygen air compressor, and obtaining the noise performance degree of each pressure data point, wherein the noise performance degree is calculated based on the regularity within a local range of the pressure data point, and the higher the regularity, the lower the noise performance degree;

[0052] A pressure sensor may be used to collect pressure data of the medical oxygen air compressor. In this embodiment, the pressure data collection frequency is 5 Hz. In other embodiments, the collection frequency may also be other suitable frequencies.

[0053] Because the pressure sensor may be affected by the surrounding environment or the sensor itself may have problems when collecting pressure data of the medical oxygen air compressor, resulting in noise in the collected pressure data sequence P. Such noise manifests itself as irregular changes in the pressure data within a local range. Therefore, the regularity of the pressure data within the local range of a certain pressure data point can be used to calculate the noise performance degree of the pressure data point. The higher the regularity, the lower the noise performance degree.

[0054] S102, generating an adjacent mean sequence, specifically: generating a first-order cumulative sequence based on the collected pressure data, and calculating the credibility of each cumulative value in the first-order cumulative sequence based on the noise performance of the pressure data point, wherein the credibility of the cumulative value is negatively correlated with the noise performance of the corresponding pressure data point; generating an adjacent mean sequence based on the first-order cumulative sequence, wherein the kth mean in the adjacent mean sequence is The calculation expression is:

[0055] ;

[0056] In the formula, represents the credibility of the k-1th accumulated value, represents the credibility of the kth accumulated value, represents the kth accumulated value, Indicates Accumulated values;

[0057] The mean in the adjacent mean sequence is obtained by weighted averaging the corresponding two accumulated values. The traditional way to calculate the mean in the adjacent mean sequence is to average the corresponding two accumulated values, that is, to assign a weight of 0.5 to each of the two accumulated values. In this embodiment, when calculating the mean in the adjacent mean sequence using the corresponding two accumulated values, a lower weight is assigned to the accumulated value with lower credibility, and a higher weight is assigned to the accumulated value with higher credibility, so that the calculated mean is more accurate.

[0058] In this embodiment, the calculation expression of the credibility of the accumulated value is:

[0059] ;

[0060] In the formula, represents the credibility of the kth accumulated value, Indicates the qth pressure data point The noise performance level, It represents the exponential function with the natural constant e as the base;

[0061] Since the accumulated value is obtained by accumulating the corresponding pressure data points, the noise of the pressure data points will be transmitted to the corresponding accumulated value to a certain extent. When the noise of the accumulated value is greater, the credibility of the accumulated value is lower. Therefore, the credibility of the accumulated value can be calculated according to the noise performance degree of the pressure data points. When calculating the credibility of the accumulated value, the present embodiment first averages the noise performance degree of the corresponding pressure data points, so that the credibility of the calculated accumulated value is inversely proportional to the average of the noise performance degree of the corresponding pressure data points, thereby more accurately calculating the credibility of the accumulated value.

[0062] S103, establishing a grey differential equation according to the adjacent mean value sequence, and then obtaining a predicted value of the original pressure data according to the grey differential equation;

[0063] After the grey differential equation is established, the least square method is used to estimate the parameters of the grey differential equation, including the growth coefficient and initial value in the model, and further the predicted value of the original pressure data can be obtained.

[0064] S104: In response to the predicted value exceeding a preset range of the pressure data, it is determined that an abnormality occurs in the medical oxygen air compressor.

[0065] Normally, when the medical oxygen air compressor is operating normally, the predicted value of its pressure data will be within a range. When the predicted value of the pressure data of the medical oxygen air compressor exceeds the range, it means that the medical oxygen air compressor is likely to have a fault. Therefore, when the predicted value of the pressure data exceeds the preset range of the pressure data, it can be determined that the medical oxygen air compressor has an abnormality. When the predicted value is less than the lower limit of the preset range of the pressure data, an ultra-low alarm is triggered; when the predicted value is greater than the upper limit of the preset range of the pressure data, an ultra-high alarm is triggered.

[0066] Since the pressure data of the medical oxygen air compressor during operation is easily affected by noise, when the grey GM (1, 1) model is used to predict its pressure data, if the adjacent mean sequence obtained by directly using the mean of adjacent accumulated values ​​is used to establish the grey differential equation, the model accuracy will be low, resulting in inaccurate pressure data prediction values.

[0067] The medical oxygen air compressor fault prediction method of the present invention takes into account the credibility of the two cumulative values ​​corresponding to the mean when generating an adjacent mean sequence based on a first-order cumulative sequence, assigns a lower weight to the cumulative value with a lower credibility, and assigns a higher weight to the cumulative value with a higher credibility, so that the calculated mean is more accurate; thereby, the established grey differential equation more accurately reflects the nonlinear relationship in the pressure data and the influence of external factors, thereby making the predicted value of the acquired pressure data more accurate; in addition, when calculating the credibility of the cumulative value, the influence of the noise of the pressure data point on the cumulative value is taken into account, and firstly, the noise performance degree of the corresponding pressure data point is averaged, so that the credibility of the calculated cumulative value is inversely proportional to the average of the noise performance degree of the corresponding pressure data point, and the greater the noise performance degree of the pressure data point, the lower the credibility of the calculated cumulative value, thereby more accurately calculating the credibility of the cumulative value, and the method of the present invention can effectively improve the accuracy of fault prediction of the medical oxygen air compressor.

[0068] It can be seen from the above embodiments that the regularity of pressure data in a local range of a certain pressure data point can be used to calculate the noise performance degree of the pressure data point, wherein the higher the regularity, the lower the noise performance degree. In one embodiment, for the i-th data point of the pressure data sequence, the method for obtaining the noise performance degree includes:

[0069] S201, calculating the Euclidean distances between two adjacent data points within the local range of the data point, and averaging the calculated Euclidean distances;

[0070] In this embodiment, the local range of the ith data point is a range of w data points centered on the current data point. The value of w may be 11. In other embodiments, w may also be other suitable values.

[0071] For data points at the edge of the data sequence, if there is a missing part in the local range, the missing part is directly discarded. For example, for the first data point of the pressure data sequence , the missing part is the left side points ( is the floor function) is discarded, then the data The local range size is the non-missing part, that is, the right side The point and itself have w- points).

[0072] The Euclidean distance between two adjacent data points in the local range can be recorded as ,in express The Euclidean distance between the jth and j+1th data points in the local range, That is to say The number of data points in the local range.

[0073] The Euclidean distance between two data points is calculated as follows: a rectangular coordinate system is established with the acquisition time of the data point as the horizontal axis and the pressure value of the data point as the vertical axis; in the rectangular coordinate system, the two data points are plotted according to the acquisition time and pressure value of the two data points respectively to obtain the corresponding two coordinate points, and then the distance between the two coordinate points is calculated according to the horizontal and vertical coordinates of the two coordinate points.

[0074] S202, obtaining the Euclidean distance associated with the data point, where the Euclidean distance associated with the data point refers to the Euclidean distance between the data point and its adjacent data points;

[0075] S203, calculating the noise performance degree of the data point according to the deviation between the mean and the Euclidean distance related to the data point, wherein the noise performance degree is positively correlated with the deviation.

[0076] In this embodiment, the difference between the Euclidean distance between a data point and its adjacent data points and the Euclidean distance between two adjacent data points in the local range of the data point is used to measure the regularity of the change of the pressure data in the local range of the i-th data point. The larger the difference, the more it means that the appearance of the i-th data point causes the data in its local range to change irregularly, and the worse its regularity. The noise performance degree acquisition method of this embodiment can more accurately calculate the noise performance degree of the i-th data point of the pressure data sequence.

[0077] In one embodiment, for the i-th data point in the pressure data sequence, its related Euclidean distance includes the Euclidean distance between the data point and its previous data point, and the Euclidean distance between the data point and its next data point, and the calculation expression of the noise performance degree is:

[0078] ;

[0079] In the formula, Indicates the degree of noise performance of the ith data point, represents the number of data points in the pressure data sequence P, Represents the i-th pressure data point The number of data points in the local area; Represents the i-th pressure data point The relative distance between the jth and j+1th data points in the local range, represents the Euclidean distance between the ith data point and its previous data point, Represents the Euclidean distance between the i-th data point and the next data point.

[0080] When When , it indicates the current pressure data point If it is not the first or last data point in the data series, then the pressure data point There are two adjacent data points, one on each side, and the Euclidean distances between the adjacent data points on the left and right sides are and , This means the current pressure data point The relevant Euclidean distance is That is, it represents the deviation between the mean and the Euclidean distance associated with the data point.

[0081] when When , it indicates the current pressure data point is the first data point of the data sequence, and the pressure data point There is only one relevant data point, , represents the deviation between the mean and the Euclidean distance associated with the data point; when When , it indicates the current pressure data point is the last data point of the data series, and the pressure data point There is only one relevant data point, ,but represents the deviation between the mean and the Euclidean distance associated with the data point; because the deviation is negatively correlated with regularity, and regularity is negatively correlated with the degree of noise representation, the gap is positively correlated with the degree of noise representation.

[0082] like Figure 2 As shown, assuming that the i-th data point The coordinate point in the rectangular coordinate system is A, the coordinate point of the i-1th data point in the rectangular coordinate system is B, and the coordinate point of the i+1th data point in the rectangular coordinate system is C, then is the length of line segment AB, is the length of line segment AC.

[0083] In this embodiment, when the i-th data point is not the first or last data point in the pressure data sequence, the Euclidean distance between the i-th data point and its previous data point and the average of the Euclidean distance between the i-th data point and its subsequent data point are used to characterize the Euclidean distance between the i-th data point and its adjacent data points, thereby making the calculation of the Euclidean distance related to the i-th data point more accurate; when the i-th data point is the first or last data point in the pressure data sequence, only the Euclidean distance between the i-th data point and its adjacent data point is used to calculate the Euclidean distance related to the i-th data point; thereby making the calculation result of the noise performance degree of the data point more accurate.

[0084] In one embodiment, the method further includes correcting the noise level of the pressure data points, including:

[0085] S301. While collecting pressure data, temperature data and vibration data of the medical oxygen air compressor are collected to obtain the first temperature data point and the first vibration data point corresponding to the i-th pressure data point collection moment; obtain the positive correlation between the change of the pressure data point within the local range of the i-th pressure data point and the change of the temperature data point within the local range of the first temperature data point, recorded as the first positive correlation; and the positive correlation between the change of the pressure data point within the local range of the i-th pressure data point and the change of the vibration data point within the local range of the first vibration data point, recorded as the second positive correlation;

[0086] The value of the first temperature data point corresponding to the i-th pressure data point collection moment is the temperature data collected at that collection moment; similarly, the value of the first vibration data point corresponding to the i-th pressure data point collection moment is the vibration data collected at that collection moment.

[0087] S302: Calculate a corrected noise performance degree according to the first positive correlation and the second positive correlation, wherein the corrected noise performance degree is negatively correlated with both the first positive correlation and the second positive correlation.

[0088] Normally, during the operation of the medical oxygen air compressor, there is a close relationship between the pressure data, temperature data and vibration data, because when the gas is compressed, the collisions between the gas molecules increase, resulting in an increase in temperature, that is, the pressure data and the temperature data are positively correlated; at the same time, changes in pressure data will also cause the mechanical parts inside the compressor to be subjected to uneven forces, resulting in changes in vibration data, so there is also a positive correlation between the pressure data and the vibration data. Therefore, the noise performance level of the pressure data point can be corrected based on the first positive correlation and the second positive correlation, and the corrected noise performance level is negatively correlated with both the first positive correlation and the second positive correlation; the method of this embodiment can be used to more accurately correct the noise performance level of the pressure data point, so that the calculation result of the noise performance level of the pressure data point is more accurate, thereby improving the accuracy of the prediction result when predicting faults of the medical oxygen air compressor.

[0089] In one embodiment, the first positive correlation and the second positive correlation are both obtained by using the Pearson correlation coefficient method, and the calculation expression of the corrected noise performance degree is:

[0090] ;

[0091] In the formula, Indicates the degree of noise performance of the ith data point, represents the first positive correlation, represents the second positive correlation, Indicates the corrected noise performance of the ith data point.

[0092] Because the Pearson correlation coefficient ranges from -1 to 1), The purpose of is to limit the positive correlation between pressure data points and temperature data points to positive numbers, The purpose of is to limit the positive correlation between the pressure data point and the vibration data point to a positive number, so that the correction factor The value range is 0-1.

[0093] In one embodiment, the preset range of the pressure data is 0.45 MPa to 0.64 MPa.

[0094] In one embodiment, the method further includes: in response to a failure of the medical oxygen air compressor, inputting the current operation data of the medical oxygen air compressor into a deep reinforcement learning model, thereby obtaining the failure type of the medical oxygen air compressor; the operation data includes temperature data, pressure data and vibration data, and the training method of the deep reinforcement learning model includes:

[0095] S401, inputting the operation data into the policy network of the deep reinforcement learning model to generate a fault mode diagnosis result;

[0096] In this embodiment, the policy network of the deep reinforcement learning model adopts a convolutional neural network or a recurrent neural network. In other embodiments, the policy network may also adopt other suitable neural networks.

[0097] S402, inputting the diagnosis result into the value network of the deep reinforcement learning model to evaluate the accuracy of the diagnosis result;

[0098] In this embodiment, the value network adopts a fully connected neural network or a convolutional neural network. In other embodiments, the value network may also adopt other suitable neural networks.

[0099] S403. Update the parameters of the policy network and the value network according to the accuracy of the diagnosis results and the reward mechanism of the reinforcement learning algorithm, and determine whether the deep reinforcement learning model converges;

[0100] In this embodiment, the reinforcement learning algorithm adopts Q-learning or deep Q network. In other embodiments, the reinforcement learning algorithm may also adopt other suitable networks.

[0101] S404. In response to the deep reinforcement learning model converging, stop training; otherwise, iteratively generate fault mode diagnosis results, evaluate the accuracy of the diagnosis results, and update the parameters of the policy network and the value network until the deep reinforcement learning model converges.

[0102] When the medical oxygen air compressor is judged to be faulty, the specific fault type of the medical oxygen air compressor is further obtained, making the fault prediction of the medical oxygen air compressor more accurate; the deep reinforcement learning model can learn complex fault modes and improve the accuracy of fault diagnosis. Through deep learning technology, the model can capture subtle changes in equipment operation and accurately identify complex fault types; by training the deep reinforcement learning model, the deep reinforcement learning model can accurately predict the fault type of the medical oxygen air compressor.

[0103] Example of medical oxygen air compressor fault prediction system:

[0104] The present invention also provides a medical oxygen compressor fault prediction system. Figure 3As shown, the medical oxygen air compressor fault prediction system includes a processor and a memory, and the memory stores computer program instructions. When the computer program instructions are executed by the processor, the medical oxygen air compressor fault prediction method described in the above embodiment is implemented.

[0105] The medical oxygen air compressor fault prediction system also includes other components familiar to those skilled in the art, such as a communication bus and a communication interface, whose settings and functions are known in the art and will not be described in detail here.

[0106] 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.

[0107] 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.

[0108] 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 method for predicting faults of a medical oxygen air compressor, characterized in that: include: Collecting pressure data of the medical oxygen air compressor and obtaining the noise performance degree of each pressure data point, wherein the noise performance degree is calculated based on the regularity within the local range of the pressure data point, and the higher the regularity, the lower the noise performance degree; Generate a first-order cumulative sequence based on the collected pressure data, and calculate the credibility of each cumulative value in the first-order cumulative sequence based on the noise performance of the pressure data point, and the credibility of the cumulative value is negatively correlated with the noise performance of the corresponding pressure data point; generate an adjacent mean sequence based on the first-order cumulative sequence, wherein the kth mean in the adjacent mean sequence The calculation expression is: ； In the formula, Indicates The credibility of the accumulated value, Indicates The credibility of the accumulated value, Indicates The accumulated value, Indicates Accumulated values; Establishing a grey differential equation according to the adjacent mean value sequence, and then obtaining a predicted value of the original pressure data according to the grey differential equation; In response to the predicted value exceeding a preset range of the pressure data, it is determined that an abnormality occurs in the medical oxygen air compressor.

2. The medical oxygen air compressor fault prediction method according to claim 1, characterized in that: The calculation expression of the credibility of the accumulated value is: ； In the formula, Indicates The credibility of the accumulated value, Indicates the qth pressure data point The noise performance level, Represents an exponential function with the natural constant e as the base.

3. The medical oxygen air compressor fault prediction method according to claim 1, characterized in that: For the i-th data point of the pressure data sequence, the method for obtaining the noise performance degree includes: Calculate the Euclidean distances between two adjacent data points in the local range of the data point, and average the calculated Euclidean distances; Obtaining the Euclidean distance associated with the data point, where the Euclidean distance associated with the data point refers to the Euclidean distance between the data point and its adjacent data points; The noise performance degree of the data point is calculated according to the deviation between the mean and the Euclidean distance associated with the data point, and the noise performance degree is positively correlated with the deviation.

4. The medical oxygen air compressor fault prediction method according to claim 2, characterized in that: For the i-th data point in the pressure data sequence, its related Euclidean distance includes the Euclidean distance between the data point and its previous data point, and the Euclidean distance between the data point and its next data point. The calculation expression of its noise performance degree is: ； In the formula, Indicates the degree of noise performance of the ith data point, represents the number of data points in the pressure data sequence P, Represents the i-th pressure data point The number of data points in the local area; Represents the i-th pressure data point The relative distance between the jth and j+1th data points in the local range, represents the Euclidean distance between the ith data point and its previous data point, Represents the Euclidean distance between the i-th data point and the next data point.

5. The medical oxygen air compressor fault prediction method according to claim 2, characterized in that: The local range of the i-th data point is a range of w data points in length centered on the current data point.

6. The medical oxygen air compressor fault prediction method according to claim 2, characterized in that: It also includes corrections for the noise level of the pressure data points, including: While collecting pressure data, temperature data and vibration data of the medical oxygen air compressor are collected to obtain a first temperature data point and a first vibration data point corresponding to the i-th pressure data point collection moment; obtain a positive correlation between a change of the pressure data point within a local range of the i-th pressure data point and a change of the temperature data point within a local range of the first temperature data point, recorded as a first positive correlation; and a positive correlation between a change of the pressure data point within a local range of the i-th pressure data point and a change of the vibration data point within a local range of the first vibration data point, recorded as a second positive correlation; The corrected noise performance degree is calculated according to the first positive correlation and the second positive correlation, and the corrected noise performance degree is negatively correlated with both the first positive correlation and the second positive correlation.

7. The method for predicting failure of a medical oxygen air compressor according to claim 6, characterized in that: The first positive correlation and the second positive correlation are both obtained by using the Pearson correlation coefficient method, and the calculation expression of the corrected noise performance degree is: ； In the formula, Indicates the degree of noise performance of the ith data point, represents the first positive correlation, represents the second positive correlation, Indicates the corrected noise performance of the ith data point.

8. The method for predicting failure of a medical oxygen air compressor according to claim 1, characterized in that: The preset range of the pressure data is 0.45Mpa to 0.64Mpa.

9. The method for predicting a medical oxygen compressor failure according to any one of claims 1 to 8, characterized in that: Also includes: In response to a failure of the medical oxygen air compressor, the operation data of the medical oxygen air compressor at the current moment is input into the deep reinforcement learning model, so as to obtain the failure type of the medical oxygen air compressor; The operating data includes temperature data, pressure data and vibration data, and the training method of the deep reinforcement learning model includes: Input the operation data into the policy network of the deep reinforcement learning model to generate failure mode diagnosis results; Inputting the diagnosis result into the value network of the deep reinforcement learning model to evaluate the accuracy of the diagnosis result; According to the accuracy of the diagnosis results and the reward mechanism of the reinforcement learning algorithm, update the parameters of the policy network and the value network, and determine whether the deep reinforcement learning model converges; In response to the deep reinforcement learning model converging, the training is stopped; otherwise, the fault mode diagnosis results are iteratively generated, the accuracy of the diagnosis results is evaluated, and the parameters of the policy network and the value network are updated until the deep reinforcement learning model converges.

10. A medical oxygen compressor fault prediction system, comprising a processor and a memory, wherein the memory stores computer program instructions, characterized in that: When the computer program instructions are executed by the processor, the medical oxygen air compressor fault prediction method according to any one of claims 1 to 9 is implemented.

Images