A method and system for monitoring equipment status based on data analysis

Abstract

The present invention relates to the field of data processing technology, and in particular to a method and system for monitoring equipment status based on data analysis. The method comprises: collecting pressure drop data and air flow rate data of an air filter, filtering all the pressure drop data multiple times, and before each filtering, obtaining the final noise level of each pressure drop data, judging whether the pressure drop data needs to participate in this filtering based on the size of the final noise level of each pressure drop data, if necessary, adaptively adjusting the filtering window of the pressure drop data using the final noise level of each pressure drop data, if not, ending the filtering of the pressure drop data, filtering based on the adjusted filtering window of all the pressure drop data, evaluating whether the performance status of the air filter is abnormal based on all the filtered pressure drop data, thereby improving the accuracy of the entire filtering, ensuring the authenticity of the pressure drop data, and realizing accurate monitoring of the performance status of the air filter.

Description

Technical Field

[0001] The present invention relates to the technical field of data processing, and in particular to a method and system for monitoring equipment status based on data analysis. Background Art

[0002] As a key place to protect public health, the quality of air inside hospitals has become a top priority. In this context, air filters occupy an irreplaceable core position in the hospital environment system, and they lay a solid foundation for creating a safe and clean air environment inside the hospital. In order to ensure that the air filter always operates stably and efficiently throughout its life cycle, it is particularly important to accurately monitor the operating status. Among the many indicators for measuring the performance of air filters, the dynamic changes in pressure drop data can accurately reflect the working status and performance change trends inside the air filter. By continuously and accurately monitoring the fluctuations in pressure drop data, the performance status of the air filter can be analyzed, and potential problems can be discovered in time according to the performance changes of the air filter, and targeted maintenance measures can be taken.

[0003] In order to accurately control the status of the air filter, modern technology usually uses advanced sensor networks to collect real-time pressure drop data, and integrates complex data analysis algorithms and prediction models to achieve a comprehensive evaluation and forward-looking prediction of its performance. Due to the complex electromagnetic environment and signal transmission path inside the hospital, the collection of pressure drop data has brought great challenges. During the data collection process, external interference can easily cause noise data to mix into the original pressure drop data, causing some data to be distorted.

[0004] In order to eliminate these interferences, data preprocessing has become a key link. The existing technology usually uses SG filters to filter and denoise the data. For example, the existing Chinese patent application document with publication number CN118861553A discloses a method, device and electronic device for predicting oil health. The prediction method includes: obtaining initial return oil pressure data; the initial return oil pressure data is the return oil pressure time series data corresponding to the hydraulic system of the running tunnel boring machine; using the SG filter to smooth and filter the initial return oil pressure data to obtain the smoothed return oil pressure data; inputting the smoothed return oil pressure data into the trained long short-term memory model to obtain the predicted return oil pressure data of the hydraulic system in the future preset time period; inputting the predicted return oil pressure data into the oil health calculation model for calculation to obtain the predicted value of the oil health, and providing effective support for the operation and maintenance management of the hydraulic system of the tunnel boring machine based on the predicted value.

[0005] However, since the filtering strategy of the SG filter is to set a uniform-sized filtering window for all the data to be filtered, and the filtering strategy of the algorithm adopts a uniform single filtering operation for all data points, this rigid processing method is prone to misjudgment when facing the high-pressure drop data generated by the air filter itself. Since the high-pressure drop data may be very close to the noise data in terms of value, it is easy to be mistakenly identified as noise and smoothed out, which not only seriously damages the authenticity and integrity of the historical pressure drop data, but also makes the prediction results of the future pressure drop data output by the prediction model built based on these data deviate greatly from the actual situation, and cannot accurately monitor the performance status of the air filter. Summary of the invention

[0006] In order to solve the problem that when using SG filter to pre-process the pressure drop data required for monitoring during the performance status monitoring of the hospital air filter, the filtering strategy of the SG filter reduces the authenticity and reliability of the pressure drop data, thereby affecting the accuracy of subsequent performance status monitoring, the present invention proposes a device status monitoring method and system based on data analysis, which specifically adopts the following scheme:

[0007] In one aspect, the present invention provides a method for monitoring equipment status based on data analysis, comprising:

[0008] Collect all the pressure drop data and air flow rate data of the air filter in a specific time period, filter all the pressure drop data multiple times, and execute steps 1 to 2 before each filtering:

[0009] Step 1: Get the final noise level of each voltage drop data, including:

[0010] The initial noise level of the pressure drop data is evaluated by the deviation between each pressure drop data and the mean value of its neighboring data segment, and the irrelevance between the pressure drop data and the air flow rate data corresponding to the time of collection, wherein the initial noise level is positively correlated with the deviation and the irrelevance;

[0011] Calculate the final noise level of each pressure drop data according to the difference between the initial noise level of each pressure drop data and the initial noise level of the pressure drop data in its neighboring data segment, and the confidence level of the difference, wherein the final noise level is positively correlated with the difference and the confidence level;

[0012] Step 2: Based on the final noise level of each voltage drop data, determine whether the voltage drop data needs to participate in this filtering. If necessary, use the final noise level of each voltage drop data to adaptively adjust the filtering window of the voltage drop data. If not, the filtering of the voltage drop data ends.

[0013] Filtering is performed based on a filter window adjusted based on all the pressure drop data, and the status of the air filter is monitored based on all the filtered pressure drop data.

[0014] The above technical solution firstly lays a solid foundation for accurately evaluating the performance of the air filter by collecting pressure drop data and air flow rate data. Pressure drop data, as a key indicator that directly reflects the degree of filter blockage, is combined with air flow rate data that can reflect the contextual information of the filter working status. The two complement each other and improve the accuracy of the filter status evaluation. Furthermore, in the data processing link, multiple filtering operations are performed on the pressure drop data. Before each filtering, a rigorous judgment is made on whether each pressure drop data participates in the filtering, and the filter window size is adaptively adjusted according to its final noise level. For pressure drop data with a higher final noise level, a larger filter window size can effectively remove noise, while for data with a lower final noise level, a smaller window size can remove slight noise while retaining the original change trend of the data intact. This flexible adjustment method based on the characteristics of the pressure drop data successfully strikes a balance between removing noise and retaining the characteristics of the real data. It can not only accurately remove noise, but also retain the authenticity and effectiveness of the data to the greatest extent, laying a solid foundation for reliable data for subsequent performance evaluation. Furthermore, the state of the pressure drop data is re-evaluated before each filtering, so that it can adapt to the dynamic changes of the data in real time. If a certain pressure drop data is filtered with a large window due to a high noise level in a certain filtering, the noise can be effectively removed after this filtering, and the final noise level is reduced in the subsequent filtering, then in the next round of filtering, the filter window size will be reduced according to the new situation. In this way, in the subsequent processing process, the real change characteristics of the data after preliminary filtering can be better preserved, and the risk of over-processing can be effectively avoided. Through this multiple dynamic judgment and adaptive adjustment for each pressure drop data, the accuracy of the entire filtering is improved, and the authenticity and effectiveness of the pressure drop data are effectively guaranteed, providing a more solid and reliable data support for the performance evaluation of the air filter based on the filtered data. Based on these data, the performance status of the air filter can be accurately monitored, and abnormalities can be found in time, so as to take appropriate maintenance measures. Furthermore, when obtaining the final noise level of each pressure drop data, the initial noise level is first calculated by quantifying the deviation of the pressure drop data from the mean of its neighborhood data segment and the correlation with the air flow rate data. When the deviation is larger and the correlation with the air velocity data is weaker, the data point is more likely to deviate from the normal range, and it is initially determined to be noise or an outlier. This method is effective in initially identifying abnormal data points caused by external interference or changes in the filter. Then, the initial noise level is further optimized by comprehensively considering the distribution of the degree of abnormality of each data point in the neighborhood and the confidence of these differences, so as to obtain a more accurate final noise level, so as to accurately distinguish between real data changes and noise, and more accurately identify those pressure drop data that do deviate from the normal range.

[0015] Furthermore, the filtering window of each voltage drop data is adaptively adjusted using the final noise level of the voltage drop data based on the following formula:

[0016]

[0017] In the formula, For the The pressure drop data was obtained in the When filtering, the size of the filter window is adaptively adjusted. is the number of filtering times, It is the size of the uniform filter window preset for all voltage drop data. It is The pressure drop data was obtained in the The final noise level before filtering, It is The pressure drop data was obtained in the The final noise level before filtering, is the floor symbol, is a natural exponential function.

[0018] The above technical solution adaptively adjusts the size of the filtering window of the pressure drop data according to the final noise level of each pressure drop data before each filtering. The adjusted filtering window size can better adapt to the dynamic characteristics of the pressure drop data, avoid the real high pressure drop data being mistaken for noise and removed, and enhance the accuracy of abnormal monitoring.

[0019] Furthermore, the final noise level of each voltage drop data satisfies the following relationship:

[0020] ;

[0021] In the formula, For the The final noise level of the voltage drop data is For the The number of pressure drop data other than itself in the neighborhood data segment of the pressure drop data, For the The sequence number of the pressure drop data other than itself in the neighboring data segment of the pressure drop data, For the The initial noise level of the voltage drop data is For the In the neighboring data segment of the pressure drop data, except for itself, The initial noise level of the voltage drop data is For the The initial noise level of the voltage drop data is The confidence level of the difference between the initial noise levels of the two pressure drop data.

[0022] The calculation of the final noise level in the above technical solution takes into account the difference between the initial noise level of each pressure drop data and the initial noise level of other pressure drop data in its neighborhood data segment, as well as the confidence of these differences. The larger the difference and the greater the confidence, the more likely the pressure drop data is to be noise. This helps to retain representative data features, reduce misjudgments, and prevent real data changes from being mistaken for noise.

[0023] Furthermore, the method for monitoring the status of the air filter based on all the filtered pressure drop data is as follows:

[0024] Input all filtered voltage drop data into In the model, the model makes a real-time prediction of the future pressure drop data after the specific time period. The model outputs the predicted value of the pressure drop data at each moment after the specific time period, and judges whether the performance of the air filter is abnormal based on the difference between the predicted value and the actual value of the pressure drop data at each moment after the specific time period, so as to realize the status monitoring of the air filter.

[0025] Furthermore, the confidence satisfies the following relationship:

[0026] ;

[0027] In the formula, For the The abnormality of the air velocity data corresponding to the collection time of the pressure drop data is For the The voltage drop data is compared with the neighboring data segment except itself. The time interval between the voltage drop data.

[0028] The above technical solution can evaluate the confidence of the difference in the degree of abnormality between the pressure drop data and other pressure drop data in its neighborhood data segment except itself by calculating the confidence, which helps to identify abnormal data points more accurately. The time interval in the formula takes into account the time interval between the pressure drop data and other pressure drop data in its neighborhood data segment. The shorter the time interval, the higher the confidence, which means that the above-mentioned difference in the degree of abnormality is more valuable for reference and helps to improve the accuracy of abnormal monitoring. The formula also combines the influence of air flow rate data. The degree of abnormality of air flow rate data will also affect the calculation of confidence. The smaller the degree of abnormality of air flow rate data, the higher the confidence.

[0029] Furthermore, the initial noise level of the voltage drop data satisfies the following relationship:

[0030] ;

[0031] In the formula, For the The initial noise level of the voltage drop data is For the The value of the pressure drop data, For the The mean of the neighborhood data segments of the pressure drop data, For the The pressure drop data is the same as the The irrelevance between the air velocity data corresponding to the collection time of the pressure drop data is is the normalization function;

[0032] The irrelevance satisfies the following relationship: ;

[0033] In the formula, For the The percentile of a pressure drop data in its neighborhood data segment, For the The percentile of the air velocity data corresponding to the collection time of the pressure drop data in its neighborhood data segment, Preset hyperparameters.

[0034] This formula reflects the degree of deviation of each pressure drop data from its neighboring data by calculating the deviation between each pressure drop data and the mean of the neighborhood data segment. The greater the deviation, the more likely the pressure drop data is to be abnormal noise data. Combined with the irrelevance of the pressure drop data and the air flow rate data, the greater the irrelevance, the more likely the pressure drop data is to be noise data. Combining the two can more accurately identify abnormal data points. In addition, irrelevance takes into account the strength of the connection between the pressure drop data and the air flow rate data. The greater the difference between the percentiles, the weaker the correlation between the pressure drop data and the air flow rate data, and the more likely it is an abnormal data point.

[0035] Furthermore, the method for judging whether the voltage drop data needs to participate in this filtering based on the size of the final noise level of each voltage drop data is: presetting a threshold of the final noise level; if the final noise level of the voltage drop data is greater than the threshold of the final noise level, the voltage drop data needs to participate in this filtering; if the final noise level of the voltage drop data is not greater than the threshold of the final noise level, the voltage drop data does not need to participate in this filtering.

[0036] Further, the method for determining whether the performance of the air filter is abnormal is:

[0037] Calculate the difference between the actual value and the predicted value of the pressure drop data at each moment after the specific time period; if the difference between the actual value and the predicted value of the pressure drop data at a certain moment is greater than a preset difference threshold, the pressure drop data at that moment is abnormal; if the difference does not exceed the difference threshold, the pressure drop data at that moment is not abnormal;

[0038] From the moment when the first abnormal pressure drop data appears, the number of abnormal pressure drop data within the preset time period is counted; if the proportion of the number of abnormal pressure drop data in the total number of all pressure drop data within the preset time period exceeds the preset proportion threshold, the performance of the air filter is in an abnormal state; if the proportion does not exceed the preset proportion threshold, the performance of the air filter is in a normal state.

[0039] Furthermore, the neighborhood data segments of the pressure drop data and the neighborhood data segments of the air flow rate data are divided according to the following method:

[0040] Each voltage drop data is selected before and after its collection time A data segment consisting of pressure drop data is used as a neighborhood data segment of the pressure drop data;

[0041] Each air velocity data is selected before and after its collection time The data segment consisting of air flow velocity data is used as the neighborhood data segment of the air flow velocity data. is the default value.

[0042] On the other hand, the present invention also provides an equipment status monitoring system for the entire life cycle of the equipment, the equipment status monitoring system includes a memory and a processor, the memory stores a computer program, and the processor executes the computer program to implement the steps of any equipment status monitoring method.

[0043] The present invention has the following effects:

[0044] The present invention innovatively starts from the intrinsic data characteristics of pressure drop data and air flow rate data, deeply explores the relationship and rules between the two, and accurately determines the suspected noise level of each data point in the selected pressure drop historical data segment through rigorous analysis and calculation. Based on this key indicator, a personalized filtering strategy is formulated for each data point, including flexibly determining the number of filtering times and the optimal window size under different filtering times, to achieve dynamic adaptive filtering. In this way, it is possible to effectively avoid the high pressure drop data from being mistakenly filtered out, thereby obtaining more accurate filtering results, improving the accuracy and authenticity of the pressure drop data prediction at future times, and achieving accurate monitoring of the equipment performance status. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] 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:

[0046] Figure 1 It is a schematic flow chart of the method of the present invention;

[0047] Figure 2 It is a schematic diagram of the method flow of step S21 of the present invention. DETAILED DESCRIPTION

[0048] 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.

[0049] The specific embodiments of the present invention are described in detail below with reference to the accompanying drawings.

[0050] Reference Figure 1 The device status monitoring method based on data analysis includes steps S1 to S3, which are as follows:

[0051] S1: Collect the pressure drop data and air flow rate data of the air filter within a specific time period.

[0052] Collect the pressure drop data and air flow rate data of the air filter in the hospital environment. These two data are of great significance for the performance evaluation of the air filter.

[0053] Pressure drop data is an important indicator to measure the filtration efficiency of air filters. As the air filter gradually becomes clogged, the pressure drop will increase, indicating that more particulate matter is captured. The higher the pressure drop data value, the more likely it is that the air filter needs to be cleaned or replaced, because excessive pressure drop will increase energy consumption and may even damage the equipment. By monitoring the changes in pressure drop data over time, the performance of the air filter can be predicted. When the pressure drop reaches a certain threshold, it usually indicates that the performance of the filter is abnormal and needs maintenance or replacement. Air flow rate is critical to maintaining system performance. Too low a flow rate may result in a decrease in filtration efficiency, while too high a flow rate may reduce the capture efficiency of particulate matter. Monitoring air flow rate data can also be used to analyze the operating status of the air filter.

[0054] Therefore, by deeply analyzing the relationship between pressure drop data and air flow rate data, we can better understand the working principle of air filters in the specific scenario of hospitals and their performance under different conditions. For example, when the air flow rate increases, the pressure drop usually increases. This relationship can help predict the behavior of air filters under different loads and provide a reference for performance status monitoring. Long-term monitoring of this data can preset the performance trend of air filters and plan maintenance work in advance.

[0055] In this step, pressure sensors are installed at the inlet and outlet of the air filter to collect pressure data and monitor the pressure changes caused by the air passing through the filter in real time. By comparing the pressure difference between the inlet and outlet, the pressure drop of the air filter can be indirectly calculated. The collection time is first set to half an hour, and the sampling frequency is once every two seconds. A flow meter (such as a hot wire anemometer or an ultrasonic anemometer) is used to measure the flow rate inside the air filter to ensure the accuracy and stability of the flow meter. The half-hour collection time and the sampling frequency of once every two seconds are also used to ensure the consistency and comparability of the data. The first collection time (the first half an hour) is used as a specific time period. Therefore, a pressure drop data and an air flow rate data can be obtained at each collection moment, ensuring the synchronization of all the collected pressure drop data and air flow rate data in the time series, so as to facilitate the subsequent correlation analysis of the two.

[0056] S2: Perform multiple filtering on all voltage drop data, and before each filtering, adaptively adjust the filtering window of the voltage drop data.

[0057] The step of adaptively adjusting the filter window of the voltage drop data includes steps S21 and S22:

[0058] S21: Obtain the final noise level of each voltage drop data. For specific methods, refer to Figure 2 As shown in , it includes steps S211 and S212.

[0059] S211: Evaluate the initial noise level of the pressure drop data by the deviation between each pressure drop data and the mean value of its neighboring data segment, and the irrelevance between the pressure drop data and the air flow rate data corresponding to the time of collection.

[0060] For any voltage drop data, the evaluation method of its initial noise level is as follows: first select the neighborhood data segment of the voltage drop data, specifically, take the voltage drop data as the center, and select the neighborhood data segment before and after the acquisition time. A pressure drop data, together with itself, constitutes the neighborhood data segment of the pressure drop data. is the default value. is 100, that is, the voltage drop data, the first 100 voltage drop data at the time of collecting the voltage drop data, and the last 100 voltage drop data at the time of collecting the voltage drop data, a total of 201 voltage drop data are used as the neighborhood data segment of the voltage drop data. It is particularly noted that if there are less than 100 voltage drop data before a certain voltage drop data, the number of differences is selected from after the voltage drop data. For example, there are 50 voltage drop data before the voltage drop data, and there are still 50 differences, then 50 more voltage drop data are taken from after the voltage drop data, and a total of 150 data are taken after the voltage drop data. If the data after the voltage drop data is less than 100, according to the same method, a few more data are taken from before the voltage drop data.

[0061] The initial noise level of each voltage drop data is specifically expressed by the formula:

[0062]

[0063] In the formula, For the The initial noise level of the voltage drop data is For the The value of the pressure drop data, For the The mean of the neighborhood data segments of the pressure drop data, For the The pressure drop data is the same as the The irrelevance between the air flow rate data corresponding to the collection time of the pressure drop data.

[0064] In this formula, Indicates The amount by which a pressure drop data exceeds the mean pressure drop data of its neighboring data segment. The larger the value, the The more prominent the pressure drop data is, the more likely it is abnormal data of real high pressure drop caused by the performance change of the air filter itself, or it may be noise data. In order to further distinguish, analyze the The irrelevance between the pressure drop data and the air velocity data corresponding to the time of collection According to Poiseuille's law, when the fluid passes through the filter, the increase in flow rate will lead to an increase in pressure drop. Therefore, when the air flow rate increases, it will usually cause the pressure drop inside the air filter to increase. If the correlation between a pressure drop data and the air flow rate data corresponding to the sampling time is lower (the greater the irrelevance), it means that the pressure drop data is more likely to be noise, rather than real abnormal data caused by changes in the performance of the air filter itself.

[0065] In this formula, the The irrelevance between the pressure drop data and the air velocity data corresponding to the time of collection The calculation formula is:

[0066]

[0067] In the formula, For the The percentile of a pressure drop data in its neighborhood data segment, For the The percentile of the air velocity data corresponding to the collection time of the pressure drop data in its neighborhood data segment, To preset hyperparameters, set , is to prevent the two percentiles from being equal. is the absolute value symbol. is the difference between the percentiles of the pressure drop data and the air velocity data in their respective neighborhood data segments, The irrelevance between the two types of data can be evaluated. The larger the difference between the percentiles, the weaker the correlation between the pressure drop data and its corresponding air velocity data. The greater the irrelevance, the more likely the pressure drop data is to be noise. Adding hyperparameters can prevent calculation problems caused by equal percentiles and improve the accuracy of subsequent evaluations.

[0068] Furthermore, for any air flow rate data, the neighborhood data segments are divided in the same way as the pressure drop data, and then the initial noise level of each air flow rate data is evaluated in the same way.

[0069] The initial noise level of the air flow rate data is evaluated by the deviation between each air flow rate data and the mean value of its neighboring data segment, and the irrelevance between the air flow rate data and the pressure drop data corresponding to the acquisition time. The initial noise level is positively correlated with the deviation and the irrelevance, which is specifically expressed by the formula:

[0070]

[0071] In the formula, For the The initial noise level of the air velocity data is For the The value of the air velocity data, For the The mean of the neighborhood data segments of air velocity data, For the The air velocity data is the same as the The irrelevance between the pressure drop data corresponding to the collection time of the air flow rate data.

[0072] In this formula, Indicates The excess of the air velocity data of each segment relative to the mean of the air velocity data of its neighboring data segment. The larger the value, the The more prominent the air flow rate data is, the more likely it is the real high flow rate abnormal data caused by the performance change of the air filter itself, or it may be abnormal data caused by noise. In order to further distinguish, analyze the The irrelevance between the air velocity data and the corresponding pressure drop data at the time of collection ,and Similarly, according to Poiseuille's law, if the correlation between an air flow rate data and the pressure drop data corresponding to its sampling time is lower (the irrelevance is greater), it means that the air flow rate data is more likely to be noise, rather than real abnormal data caused by changes in the performance of the air filter itself.

[0073] The initial noise levels of both the pressure drop data and the air velocity data are based on a comprehensive consideration of the prominence of the data and the correlation between the two types of data. The greater the initial noise level of a data, the more likely it is that the data is noise.

[0074] S212: Calculate the final noise level of each voltage drop data according to the difference between the initial noise level of each voltage drop data and the initial noise level of the voltage drop data in its neighboring data segment, and the confidence level of the difference.

[0075] In the previous steps, the initial noise level of each pressure drop data has been calculated. However, the initial noise level is only based on the analysis of the neighborhood data segment in a short period of time, which is not enough to fully reflect the overall characteristics of the pressure drop data. If you only look at the pressure drop data (neighborhood data segment) of a short period of time near a certain pressure drop data, you may miss the overall trend in a longer period of time, resulting in inaccurate analysis results. For example, a certain pressure drop data has a large deviation compared with its neighborhood data segment, which may be an abnormal noise point, but compared with all the pressure drop data, the overall deviation is not large and cannot be determined as a noise point. Therefore, it is necessary to further optimize the above initial noise level to obtain a more accurate final noise level for each pressure drop data.

[0076] In order to improve the accuracy of the abnormality assessment of each pressure drop data, it is necessary to further optimize the calculation method of the abnormality degree. The initial noise degree of all pressure drop data in the neighborhood data segment of the pressure drop data is calculated according to the method in step S211, and then the difference between the initial noise degree of the pressure drop data and the initial noise degrees of other pressure drop data in its neighborhood data segment is calculated. The larger the difference, the more prominent the abnormality degree of the pressure drop data point is compared with the abnormality degrees of other data points, and the confidence of the difference is analyzed. If the difference between the initial noise degree of a pressure drop data and other pressure drop data in its neighborhood data segment is larger, and the confidence of the difference is higher, the pressure drop data is more likely to be real noise.

[0077] Therefore, the final noise level of the pressure drop data is calculated by combining the difference between the initial noise level of each pressure drop data and other pressure drop data in its neighborhood data segment, as well as the confidence level of the difference. The calculation formula is:

[0078]

[0079] In the formula, For the The final noise level of the voltage drop data is For the The number of pressure drop data other than itself in the neighborhood data segment of the pressure drop data, For the The sequence number of the pressure drop data other than itself in the neighboring data segment of the pressure drop data, For the The initial noise level of the voltage drop data is For the In the neighboring data segment of the pressure drop data, except for itself, The initial noise level of the voltage drop data is Representative The initial noise level of a voltage drop data is related to the The difference in the initial noise level of the voltage drop data is greater. The greater the difference between the abnormal degree of a pressure drop data and the initial noise degree of other pressure drop data in its neighborhood, the greater the difference between the abnormal degree of a pressure drop data and the initial noise degree of other pressure drop data in its neighborhood. The more likely a voltage drop data is to be noise or an outlier, the greater its final noise level. for The confidence level is calculated as:

[0080]

[0081] In the formula, For the The abnormality of the air velocity data corresponding to the collection time of the pressure drop data is For the The voltage drop data is compared with the neighboring data segment except itself. The time interval between the pressure drop data, the larger the confidence value, the The initial noise level of the voltage drop data is related to the The greater the reference value of the difference between the initial noise levels of the pressure drop data.

[0082] Among them, the confidence is considered from the following two aspects:

[0083] On the one hand, considering the The voltage drop data and its neighboring data segment The time interval of the voltage drop data, the smaller the time interval, the The voltage drop data and its neighboring data segment The closer the voltage drop data are in time, the greater the difference in the initial noise level between the two. The more credible, For two voltage drop data, the smaller the time interval between them, it means that the two voltage drop data are closer in time, and the changes between them are more continuous and consistent. Therefore, if the time interval between two voltage drop data is small, the difference in the initial noise level between them can be considered more reliable, because this difference reflects the actual data difference rather than accidental fluctuations. Therefore, this difference has a greater reference value.

[0084] On the other hand, considering The initial noise level of the air velocity data corresponding to the collection time of the pressure drop data is smaller, indicating that The more accurate the calculation result, If the initial noise level of the air flow rate data corresponding to the pressure drop data collection time is small, it means that the air flow rate data changes more steadily near the collection time, so the calculated The more accurate. In this case, It can better reflect the real anomaly rather than the false anomaly caused by the fluctuation of the air velocity data itself. Therefore, the reference value of this difference is greater.

[0085] Since the neighborhood data segment of each pressure drop data contains itself and 100 adjacent pressure drop data before and after, when calculating the initial noise level of all pressure drop data in the neighborhood data segment of the pressure drop data, it is actually based on a larger range of pressure drop data. This means that when analyzing the initial noise level of all pressure drop data in the neighborhood data segment of a certain pressure drop data point, more relevant pressure drop data are considered, which makes the calculated final noise level more comprehensive. Reflect the overall characteristics of the pressure drop data, and the calculation of the final noise level takes into account the difference in abnormality and its credibility, which helps to more accurately identify noise data points and improve the reliability of subsequent prediction and evaluation.

[0086] S22: Based on the size of the final noise level of each voltage drop data, determine whether the voltage drop data needs to participate in this filtering. If necessary, use the final noise level of each voltage drop data to adaptively adjust the filtering window of the voltage drop data. If not, the filtering of the voltage drop data ends.

[0087] Before adjusting the filter window of each voltage drop data, given that different voltage drop data present different levels of suspected noise, the number of filtering times required is also different. In this step, the first task is to accurately determine the number of filtering times for each voltage drop data.

[0088] In this process, the threshold of the final noise level needs to be set. Here, it is set to 0.25, which is an empirical value. Before each filtering, each pressure drop data is compared with the final noise level of the pressure drop data and 0.25. If the final noise level of the pressure drop data is greater than 0.25, it indicates that the noise of the pressure drop data has not been fully filtered out, and the filtering effect is poor. It is still necessary to carry out the next filtering operation to further improve the purity and authenticity of the data. On the contrary, if the final noise level of the pressure drop data is not greater than 0.25, it means that the filtering effect of this pressure drop data has reached the expected level and can reflect the true state of the pressure drop data. At this time, the pressure drop data does not need to participate in this filtering to avoid excessive processing.

[0089] Before each filtering, this rigorous judgment logic is continuously used for each voltage drop data to determine the most appropriate number of filtering times, ensuring that the original characteristics and value of the voltage drop data are retained to the greatest extent while effectively removing noise.

[0090] Furthermore, before each filtering, for all the voltage drop data that need to participate in this filtering, the respective filtering windows are adaptively adjusted according to the magnitude of their respective final noise levels, so as to realize a dynamic adaptive filtering process.

[0091] Specifically, when a voltage drop data is filtered for the first time, if its final noise level is high, it means that it is more likely to be interfered by noise. In this case, in order to suppress noise more effectively, a larger filtering window is required. Because a larger window can include more data points to participate in the fitting process, thereby enhancing the ability to smooth noise and improving the purity and reliability of the data. When a voltage drop data is in the stage of multiple filtering, it also follows the principle that the larger the final noise level, the larger the required filtering window. In addition, it is also necessary to consider the filtering effect of the voltage drop data after the last filtering. If the filtering effect of the voltage drop data after the last filtering is not good, for example, after the last filtering, the final noise level of the voltage drop data is still high, and the change is not obvious compared with the previous one, this indicates that the last filtering failed to fully remove the noise, then in this filtering, a larger window is needed for a deeper and larger filtering operation. In this way, the actual effect of each filtering can be accurately evaluated, and the window size adjustment for the next filtering can be provided.

[0092] Based on the above logical analysis, the window size of each voltage drop data at different filtering times is adaptively adjusted to ensure that in the entire filtering process, the appropriate window size can be flexibly selected according to the noise characteristics of the voltage drop data to effectively remove the noise, and the real change characteristics of the data can be retained to the greatest extent, providing a high-quality data foundation for subsequent data analysis.

[0093] This step provides the calculation formula for adaptively adjusting the filter window:

[0094]

[0095] In the formula, is the number of filtering times, which is the key variable for determining which calculation method to use. Different filtering times correspond to different calculation logics, reflecting the dynamic adaptability of the data processing process. For the The pressure drop data was obtained in the During the secondary filtering, the size of the filter window is adaptively adjusted. It is the core goal of the entire calculation, and its accurate determination directly affects the filtering effect. It is the size of the uniform filter window preset for all voltage drop data. The empirical value of is 10, which provides an initial benchmark scale for the calculation and ensures that all data have a common starting filter window reference when there are no other factors affecting it. It is The pressure drop data was obtained in the The final noise level before the second filtering. This parameter reflects the noise condition of the current data point when the filtering operation is about to be performed, and is one of the important bases for determining the adjustment of the filter window size. It is The pressure drop data was obtained in the The final noise level before the last filtering can be compared with the current noise level to evaluate the filtering effect of the previous filtering. and The calculation method is the same, the only difference is the number of filtering times, To round down the sign, by giving Adding 0.5 and rounding down implements the rounding operation, so that the size of the filter window can be determined in a more reasonable integer form, which meets the accuracy requirements for data processing in actual application scenarios. It is a natural exponential function, which plays a key quantitative role in the formula and is used to adjust the weight of the filter window size according to the change of noise level.

[0096] In this formula, the specific description of rounding is as follows: When the decimal part of is less than 0.5, the decimal part after adding 0.5 is still less than 1, and the result remains unchanged after rounding down. When the decimal part of is greater than or equal to 0.5, adding 0.5 will make the decimal part exceed 1, and the result will increase by 1 after rounding down.

[0097] In this formula, Reflects the The pressure drop data is shown in The filtering effect of the second filtering is The better the filtering effect of the secondary filter, The larger the value of The size of the filter window should be reduced when filtering for the second time. The worse the filtering effect of the secondary filter, the The value should be smaller, The size of the filter window should be increased when filtering again. Build The filtering effect of the second filter is The negative correlation between the size of the filter window during the secondary filtering.

[0098] In this formula, Taking into account the The voltage drop data is in The filtering effect of the second filter is The voltage drop data is at the current final noise level, specifically:

[0099] when If it is larger, the The pressure drop data was obtained in the The final noise level before the second filtering is high, that is, the current voltage drop data is seriously interfered by noise. From the perspective of noise removal, a larger filtering window is needed to better smooth the data and reduce the impact of noise.

[0100] If at the same time Smaller, The larger the value, the poorer the filtering effect of the last filtering. The final noise level of the voltage drop data has not been significantly reduced. The result is larger, which strongly suggests that The filter window needs to be greatly enlarged during the secondary filtering to more effectively remove noise. Larger, Smaller means the filtering effect of the last filtering is better. The final noise level of the voltage drop data has been significantly reduced. Larger, but The result will not be too large, which means that considering the good effect of the previous filtering, there is no need to increase the filtering window excessively to avoid over-smoothing the real change characteristics of the data.

[0101] when When it is smaller, the The pressure drop data was obtained in the The final noise level before the second filtering is low. From the perspective of preserving the authenticity of the data, it is hoped that the filtering window will not be too large to avoid over-filtering. Smaller, The larger the value, the poorer the filtering effect of the last filtering. The final noise level of the voltage drop data did not decrease significantly, but due to It is smaller, at this time The result will not be too large, which means that considering the good filtering effect last time, there is no need to increase the filtering window too much to avoid over-smoothing the real change characteristics of the data. Larger, Smaller means the filtering effect of the last filtering is better. The resulting noise level of the voltage drop data has been significantly reduced, and It is smaller, at this time This further indicates that when the current noise level is low and the previous filtering effect is good, a smaller filtering window should be maintained to maximize the retention of the true change characteristics of the data.

[0102] By using the method of adaptively adjusting the filter window size in this step, the size of the filter window of each pressure drop data can be dynamically adjusted before each filtering based on the specific change characteristics of the pressure drop data, such as the noise level and its change trend. This dynamic adjustment enables the subsequent filtering process to closely fit the actual situation of the data, greatly improving the accuracy of the filtering. For example, in the performance status monitoring of the air filter, the pressure drop data collected at different times may be subject to noise interference of different degrees and types. Through this adaptive filter window adjustment, appropriate filtering strategies can be adopted for different data conditions, so that the filtered data can more accurately reflect the actual working status of the air filter. Based on these data, the performance status of the air filter can be more accurately monitored so that equipment maintenance can be carried out in a timely manner.

[0103] S3: Filtering is performed based on the filter window adjusted based on all the pressure drop data, and the status of the air filter is monitored based on all the filtered pressure drop data.

[0104] The SG filter is used to filter all the pressure drop data participating in each filtering. Each time the filtering operation is performed, the filtering and denoising is performed according to the size of the filter window adjusted according to each pressure drop data. This can not only effectively remove the noise, but also retain the trend and important details of the data. Finally, the denoised pressure drop data is obtained. The denoised pressure drop data contains the real data values ​​of the performance changes of the air filter itself. Based on these real data values, the performance status of the air filter can be accurately monitored in the future.

[0105] Among them, the method for monitoring the status of the air filter based on all the pressure drop data after filtering is:

[0106] Input all filtered voltage drop data into In the model, the model makes a real-time prediction of the future pressure drop data after the specific time period, and the model outputs the predicted value of the pressure drop data at each moment after the specific time period. The Autoregressive-Integrated-Moving-Average (ARM) model is a mature time series forecasting method that has been applied in many fields, including industrial production and environmental monitoring. It is to predict future data based on historical data. The model includes three parameters: autoregressive order , difference order and the moving average order , this step is preset according to the experience value , , Then, the pressure drop data collected in step S1 for a period of time after denoising in steps S2-S3 is input as the historical pressure drop data. The model is used to obtain the predicted value of the pressure drop data at the next collection time of the historical pressure drop data. Let the next collection time after the specific time period be T1, and the next collection time after T1 be T2.

[0107] Whether the performance of the air filter is abnormal is judged according to the difference between the predicted value and the actual value of the pressure drop data at each moment after the specific time period, so as to realize the status monitoring of the air filter, specifically including:

[0108] Calculate the difference between the actual value and the predicted value of the pressure drop data at each moment after the specific time period; if the difference between the actual value and the predicted value of the pressure drop data at a certain moment is greater than a preset difference threshold, the pressure drop data at that moment is abnormal; if the difference does not exceed the difference threshold, the pressure drop data at that moment is not abnormal. In this embodiment, the difference threshold between the actual value and the predicted value of the pressure drop data is set to 2 Pascals. If the difference between the predicted value and the actual value of the pressure drop data at time T1 is greater than 2 Pascals, it is considered that the pressure drop data at time T1 is abnormal. Otherwise, the pressure drop data at time T1 is not abnormal. For the next acquisition time of T1 is T2, All the pressure drop data collected at and before the time are used as new historical pressure drop data. Steps S1 to S3 are repeated to calculate the final noise level and filter and denoise. The denoised historical pressure drop data is input into Model, predicts Then, the predicted value of the pressure drop data at time T1 is determined to determine whether the pressure drop data at time T2 is abnormal. The next collection time T3 at the moment All the pressure drop data collected at and before the time are used as new historical pressure drop data. Steps S1 to S3 are repeated to calculate the final noise level and filter and denoise. The denoised historical pressure drop data is input into Model, predicts The predicted value of the pressure drop data at time T1 is obtained, and then whether the pressure drop data at time T2 is abnormal is determined. This is repeated continuously to achieve real-time monitoring of whether the pressure drop data is abnormal.

[0109] Statistics are collected at each moment after the specific time period to determine whether the pressure drop data is abnormal. After the specific time period, starting from the moment when the first abnormal pressure drop data appears, the number of abnormal pressure drop data within the preset time length is counted; if the proportion of the number of abnormal pressure drop data in the total number of all pressure drop data within the preset time length exceeds the preset proportion threshold, the performance of the air filter is in an abnormal state; if the proportion does not exceed the preset proportion threshold, the performance of the air filter is in a normal state. In this embodiment, the preset time length is set to 1 minute, and the preset proportion threshold is set to 15%. Both values ​​are empirical values. For example, from Starting from the moment, all pressure drop data within 1 minute are continuously collected. Within this minute, if the number of abnormal pressure drop data (pressure drop data with a difference between the predicted value and the actual value greater than 2 Pascals) exceeds 15% of the total number of all pressure drop data within this minute, it is considered that the performance of the air filter is in an abnormal state and there may be hidden faults inside, so that corresponding measures can be taken in time, such as further checking the air filter and cleaning or replacing it if necessary. If the number of abnormal pressure drop data within this minute does not exceed 15% of the total number of all pressure drop data within this minute, it is considered that the performance of the air filter is in a normal state.

[0110] At any time thereafter, one minute of pressure drop data can be continuously collected starting from that time, and the same method can be used to evaluate whether the performance of the air filter is in an abnormal state, thereby achieving real-time monitoring of the performance status of the air filter.

[0111] Therefore, this embodiment can more accurately evaluate the final noise level of each pressure drop data and improve the accuracy of abnormality monitoring through the operations of step S1 to step S3. The calculation of the final noise level takes into account the difference in abnormality levels and its credibility, which helps to more accurately identify noise data points and improve the reliability of predictive evaluation. By performing real-time prediction of the denoised pressure drop data and performing performance abnormality evaluation based on the difference between the predicted value and the actual value, the performance changes of the air filter can be effectively monitored, and possible problems can be discovered and resolved in a timely manner.

[0112] The present invention also provides an equipment status monitoring system based on data analysis, the equipment status monitoring system comprises a memory and a processor, the memory stores a computer program, and the processor executes the computer program to implement the steps of any equipment status monitoring method.

[0113] 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.

[0114] 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 device status monitoring method based on data analysis, characterized in that: include: Collect all the pressure drop data and air flow rate data of the air filter in a specific time period, filter all the pressure drop data multiple times, and before each filtering, perform steps 1 to 2: Step 1: Get the final noise level of each voltage drop data, including: The initial noise level of the pressure drop data is evaluated by the deviation between each pressure drop data and the mean value of its neighboring data segment, and the irrelevance between the pressure drop data and the air flow rate data corresponding to the time of collection, wherein the initial noise level is positively correlated with the deviation and the irrelevance; Calculate the final noise level of each pressure drop data according to the difference between the initial noise level of each pressure drop data and the initial noise level of the pressure drop data in its neighboring data segment, and the confidence level of the difference, wherein the final noise level is positively correlated with the difference and the confidence level; Step 2: Based on the final noise level of each voltage drop data, determine whether the voltage drop data needs to participate in this filtering. If necessary, use the final noise level of each voltage drop data to adaptively adjust the filtering window of the voltage drop data. If not, the filtering of the voltage drop data ends. Filtering is performed based on the filter window adjusted based on all the pressure drop data, and the status of the air filter is monitored based on all the pressure drop data after filtering; Adaptively adjusting the filter window of each voltage drop data using the final noise level of the voltage drop data is based on the following formula: In the formula, For the The pressure drop data was obtained in the When filtering, the size of the filter window is adaptively adjusted. is the number of filtering times, It is the size of the uniform filter window preset for all voltage drop data. It is The pressure drop data was obtained in the The final noise level before filtering, It is The pressure drop data was obtained in the The final noise level before filtering, is the floor symbol, is a natural exponential function.

2. The device status monitoring method based on data analysis according to claim 1 is characterized in that: The final noise level of each voltage drop data is calculated as: ； In the formula, For the The final noise level of the voltage drop data is For the The number of pressure drop data other than itself in the neighborhood data segment of the pressure drop data, For the The sequence number of the pressure drop data other than itself in the neighboring data segment of the pressure drop data, For the The initial noise level of the voltage drop data is For the In the neighboring data segment of the pressure drop data, except for itself, The initial noise level of the voltage drop data is For the The initial noise level of the voltage drop data is The confidence level of the difference between the initial noise levels of the two pressure drop data.

3. The device status monitoring method based on data analysis according to claim 1 is characterized in that: The method for monitoring the status of the air filter based on all the pressure drop data after filtering is: Input all filtered voltage drop data into In the model, the model makes a real-time prediction of the future pressure drop data after the specific time period. The model outputs the predicted value of the pressure drop data at each moment after the specific time period, and judges whether the performance of the air filter is abnormal based on the difference between the predicted value and the actual value of the pressure drop data at each moment after the specific time period, so as to realize the status monitoring of the air filter.

4. The device status monitoring method based on data analysis according to claim 2 is characterized in that: The confidence level satisfies the following relationship: ； In the formula, For the The abnormality of the air velocity data corresponding to the collection time of the pressure drop data is For the The voltage drop data is compared with the neighboring data segment except itself. The time interval between the voltage drop data.

5. The device status monitoring method based on data analysis according to claim 2 is characterized in that: The initial noise level of the voltage drop data satisfies the following relationship: ； In the formula, For the The initial noise level of the voltage drop data is For the The value of the pressure drop data, For the The mean of the neighborhood data segments of the pressure drop data, For the The pressure drop data is the same as the The irrelevance between the air velocity data corresponding to the collection time of the pressure drop data is is the normalization function; The irrelevance satisfies the following relationship: ; In the formula, For the The percentile of a pressure drop data in its neighborhood data segment, For the The percentile of the air velocity data corresponding to the collection time of the pressure drop data in its neighborhood data segment, Preset hyperparameters.

6. The device status monitoring method based on data analysis according to claim 2 is characterized in that: The method for judging whether the voltage drop data needs to participate in this filtering based on the final noise level of each voltage drop data is: Preset the threshold of the final noise level; If the final noise level of the voltage drop data is greater than the threshold of the final noise level, the voltage drop data needs to participate in this filtering; if the final noise level of the voltage drop data is not greater than the threshold of the final noise level, the voltage drop data does not need to participate in this filtering.

7. The device status monitoring method based on data analysis according to claim 3 is characterized in that: The method to judge whether the performance of the air filter is abnormal is: Calculate the difference between the actual value and the predicted value of the pressure drop data at each moment after the specific time period; if the difference between the actual value and the predicted value of the pressure drop data at a certain moment is greater than a preset difference threshold, the pressure drop data at that moment is abnormal; If the difference does not exceed the difference threshold, there is no abnormality in the pressure drop data at this moment; From the moment when the first abnormal pressure drop data appears, the number of abnormal pressure drop data within the preset time period is counted; if the proportion of the number of abnormal pressure drop data in the total number of all pressure drop data within the preset time period exceeds the preset proportion threshold, the performance of the air filter is in an abnormal state; if the proportion does not exceed the preset proportion threshold, the performance of the air filter is in a normal state.

8. The device status monitoring method based on data analysis according to claim 5 is characterized in that: The neighborhood data segments of the pressure drop data and the neighborhood data segments of the air velocity data are divided according to the following method: Each voltage drop data is selected before and after its collection time A data segment consisting of pressure drop data is used as a neighborhood data segment of the pressure drop data; Each air velocity data is selected before and after its collection time The data segment consisting of air flow velocity data is used as the neighborhood data segment of the air flow velocity data. is the default value.

9. A device status monitoring system based on data analysis, characterized in that: The equipment status monitoring system comprises a memory and a processor, wherein a computer program is stored in the memory, and the processor executes the computer program to implement the steps of the equipment status monitoring method according to any one of claims 1 to 8.

Images