Data integration and acquisition method and system for equipment management platform

Abstract

The invention relates to the technical field of data processing, in particular to a data integration and collection method and system oriented to an equipment management platform, and the method comprises the steps: obtaining a reference time length according to the data change in each type of operation data time sequence, obtaining each type of operation data historical time sequence which takes the time of each historical fault repair which is the same as the fault type of the current fault repair as the termination moment and takes the duration as the reference duration, and determining the final extension demand degree of the operation data time sequence corresponding to the current fault repair so as to obtain the target duration; and obtaining all types of operation data updating time sequence sequences with the current fault repair time as the termination moment and the duration as the target duration as time domain matching data of the current fault repair, and completing data integration. According to the method, a data integration effect with a higher adaptation degree can be obtained by combining abnormal performance of equipment operation data, multi-source data relevance and influence conditions on data training in an actual scene, and the data integration efficiency is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of data processing, and in particular to a data integration and collection method and system for an equipment management platform. Background Art

[0002] The equipment service management platform collects equipment operation data and historical work order data, and uses machine learning and data mining technologies to build multi-dimensional portraits of equipment and users. The platform can then conduct demand forecasts and resource optimization based on the multi-dimensional portraits, and dispatch maintenance personnel and spare parts information in advance, which helps reduce resource waste and improve service efficiency.

[0003] The platform regularly imports equipment maintenance record data into the platform's local data center through data training to increase the prediction accuracy of the equipment record model. When training and importing data collected from multiple sources, data integration is required first. Specifically, discrete maintenance data is matched with continuous operating data of a certain length of time. For example, a single equipment failure record is matched with the operating data readings (such as temperature, vibration, etc.) collected by all relevant sensors within 16 hours before its occurrence to achieve data time series alignment and integration, thereby forming a wide table to prepare for subsequent data training.

[0004] Existing problem: Traditionally, when matching discrete maintenance record data of equipment with continuous operating data of corresponding duration, an operating data segment with a fixed duration preset before the maintenance record is often selected. However, in the actual equipment management platform corresponding to the time series alignment and integration scenario of multi-source data, if the continuous operating data matched with discrete maintenance data is too short, it will make it difficult for the operating data to reflect the complete abnormal performance of the equipment. If the matched continuous operating data is too long, it will lead to an increase in redundant computing resources consumed during data training, reducing data processing efficiency, and further resulting in a low data integration fit for the traditional use of fixed-duration operating data segments to time series align discrete maintenance data. Summary of the Invention

[0005] The present invention provides a data integration and collection method and system for an equipment management platform to solve the existing problems.

[0006] The data integration and collection method and system for the device management platform of the present invention adopt the following technical solutions: An embodiment of the present invention provides a data integration and collection method for a device management platform, the method comprising the following steps: Obtain a time series sequence of each type of operating data with the current fault report time as the end time and a duration of a preset period; obtain the operation mutation moment of each type of operating data based on the data changes in the time series sequence of each type of operating data; Obtain the reference duration based on the time intervals between the operational mutation moments of all types of operational data; obtain the historical time series of each type of operational data, with the time of each historical fault report of the same type as the current fault report as the end time and the duration as the reference duration; Determine the final expansion requirement of the operation data time series corresponding to the current fault report based on the difference between each type of operation data time series and each type of historical operation data time series; According to the final expansion demand of the operating data time series corresponding to the current fault report, the target duration is obtained; the updated time series of all types of operating data with the time of the current fault report as the end moment and the duration as the target duration are obtained as the time domain matching data of the current fault report to complete the data integration.

[0007] Furthermore, the specific steps of obtaining the operation mutation moment of each type of operation data include the following: In each type of operating data time series, obtain the mean of the data values ​​at all moments, recorded as the first mean, and obtain the inversely proportional normalized value of the absolute value of the difference between the data value at each moment and the first mean, recorded as the device durability performance at each moment; A threshold value N is preset, and a local period consisting of the N moments closest to the x-th moment and the x-th moment is obtained. The standard deviation of the data values ​​of all moments in the local period is recorded as the volatility of the x-th moment. Obtain the product of the inverse proportional value of the equipment durability performance at each moment and the volatility at each moment, and record it as the operation abnormality coefficient at each moment; According to the operational anomaly coefficients at all moments, the operational mutation moment of each type of operational data is determined.

[0008] Furthermore, the operation mutation moment of each type of operation data is determined based on the operation anomaly coefficients at all moments, including the following specific steps: In each type of operation data time series, with the time as the horizontal axis and the operation anomaly coefficient as the vertical axis, the least squares method is used for curve fitting to obtain the operation anomaly change curve; On the abnormal change curve, obtain the normalized value of the slope of the data point corresponding to the xth moment, record it as the trend value, obtain the mean of the data values ​​of all moments before the xth moment, record it as the second mean, obtain the mean of the data values ​​of all moments after the xth moment, record it as the third mean, obtain the product of the absolute value of the difference between the second mean and the third mean and the trend value, record it as the possibility of abnormal mutation at the xth moment; The moment when the possibility of abnormal mutation is the greatest is recorded as the operation mutation moment of each type of operation data.

[0009] Furthermore, the specific steps of obtaining the reference duration are as follows: For all types of running data, the time interval between any two running mutation moments is used as the clustering distance. The DBSCAN density clustering algorithm is used to perform clustering operations to obtain several clusters and the cluster center time of each cluster. Obtain the cluster with the largest number of running mutation moments and record it as the target cluster; In the target cluster, obtain the inversely proportional normalized value of the time interval between each running mutation moment and the cluster center moment, record it as the first interval value, and multiply the first interval value by the number of running mutation moments, record it as the abnormal trend factor of each running mutation moment in the target cluster; Determine the high-probability abnormal moment based on the magnitude of the abnormal trend factor at each operation mutation moment in the target cluster; Get the time interval between the high-probability abnormality moment and the current moment, and record it as the reference duration.

[0010] Furthermore, the specific steps of determining the high-probability abnormal situation moment include the following: In the target cluster, the operation mutation moment with the largest abnormal trend factor is recorded as the high-probability abnormal moment.

[0011] Furthermore, the step of determining the final expansion requirement of the operation data time series corresponding to the current fault repair report includes the following specific steps: In the historical time series of the yth type of operating data corresponding to all historical fault reports of the same fault type as the current fault report, the data means at the same time are sequentially obtained to form the historical time series of the yth type of operating data means; In the yth type of operation data time series sequence corresponding to the current fault report, obtain the yth type of operation data time series sequence segment between the high probability abnormality moment and the current moment; Determine the value of the operation data time series analysis corresponding to the current fault report based on the difference between the historical time series sequence of the mean value of each type of operation data and the time series sequence segment of each type of operation data; Obtain the middle moment in the time series of each type of operating data corresponding to the current fault report, record it as the target moment, obtain the normalized value of the difference between the target moment and the high-probability abnormal moment, and record it as the expansion coefficient corresponding to the current fault report; The product of the time series analysis value of the operating data corresponding to the current fault report and the expansion coefficient is recorded as the time series expansion demand of the operating data corresponding to the current fault report; According to the time series expansion requirement of the operating data corresponding to the current fault repair report, the final time series expansion requirement of the operating data corresponding to the current fault repair report is determined.

[0012] Furthermore, the specific steps of determining the time series analysis value of the operating data corresponding to the current fault report are as follows: Obtain the inversely proportional normalized value of the Pearson correlation coefficient between the historical time series sequence of the mean value of the y-th type of operating data and the time series sequence segment of the y-th type of operating data, and record it as the first uncorrelated value; In the y-th type of operation data mean historical time series sequence and the y-th type of operation data time series sequence segment, the sum of the absolute values ​​of the differences between all two elements with the same ordinal value is recorded as the second unrelated value; The product of the first irrelevant value and the second irrelevant value is recorded as the irrelevant value between the current and historical data of the yth type of operation data; The sum of the current and historical irrelevance of all types of operating data is recorded as the operating data time series analysis value corresponding to the current fault report.

[0013] Furthermore, the method of determining the final expansion requirement of the operation data time series corresponding to the current fault repair report according to the expansion requirement of the operation data time series corresponding to the current fault repair report includes the following specific steps: Use the historical time series of all types of operating data corresponding to all historical fault reports of the same fault type as the current fault report as the input of the pre-trained model to obtain the data packet ratio and gradient norm during the model training and evaluation process; Obtain the normalized value of the ratio of the data packet ratio to the gradient norm, record it as the first ratio, and record the normalized value of the product of the first ratio and the operating data timing expansion requirement corresponding to the current fault report as the final operating data timing expansion requirement corresponding to the current fault report.

[0014] Furthermore, the specific steps of obtaining the target duration are as follows: Obtain the product of the preset basic cycle and the final expansion demand degree of the operation data time series corresponding to the current fault repair report, and record it as the first product; When the target time is greater than or equal to the high-probability abnormal time, the sum of the preset basic period and the first product is recorded as the target duration; When the target time is less than the high-probability abnormal time, the difference between the preset basic period and the first product is recorded as the target duration.

[0015] The present invention also proposes a data integration and collection system for a device management platform, comprising a memory, a processor, and a computer program stored in the memory and runnable on the processor. The processor executes the computer program stored in the memory to implement the steps of the aforementioned data integration and collection method for a device management platform.

[0016] The beneficial effects of the technical solution of the present invention are: In an embodiment of the present invention, a time series sequence of each type of operating data is obtained, with the time of the current fault report as the end moment and the duration as a preset period. A reference duration is obtained based on data changes in each type of operating data time series sequence. A historical time series sequence of each type of operating data is obtained, with the time of each historical fault report of the same fault type as the current fault report as the end moment and the duration as the reference duration. Based on the difference between each type of operating data time series sequence and each type of historical time series sequence of operating data, the final expansion demand of the operating data time series corresponding to the current fault report is determined. Based on the equipment durability and local fluctuation performance analysis, the operation anomaly coefficient at a single moment is obtained, and then an operation anomaly change curve is constructed. Operation mutation analysis points corresponding to each type of operating data are obtained from the anomaly change curve. High-probability anomaly points are screened out by combining the correlation performance of multiple types of operating data. These points are then collaboratively analyzed with other multiple types of operating data under the same historical equipment maintenance conditions to obtain the time series expansion demand. The time series expansion demand is then corrected based on the training effect of historical data, ensuring the accuracy of the subsequent selection of time-domain matching data for the current fault report. The target duration is obtained, and a time series of all types of operating data updates with the current fault report as the end time and the target duration as the target duration is obtained as the time domain matching data for the current fault report, completing data integration. Thus, the present invention can combine the abnormal performance of equipment operating data in actual scenarios, the correlation of multi-source data, and the impact on data training to achieve a more adaptable data integration effect, thereby improving data integration efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0018] Figure 1 This is a flowchart of the steps of a data integration and collection method for a device management platform according to the present invention; Figure 2 Select a schematic diagram for the analysis time of fault repair; Figure 3 A schematic diagram of an operation mutation analysis point on an abnormal operation change curve of a single type of operation data of a device; Figure 4 It is a schematic diagram of the change curve of the historical time series sequence of the mean value of the same type of operating data and the time series sequence segment of the operating data. DETAILED DESCRIPTION

[0019] To further illustrate the technical means and effectiveness of the present invention to achieve its intended purpose, the following, in conjunction with the accompanying drawings and preferred embodiments, describes in detail a data integration and collection method and system for a device management platform, including its specific implementation, structure, features, and effectiveness. In the following description, references to different "one embodiment" or "another embodiment" do not necessarily refer to the same embodiment. Furthermore, specific features, structures, or characteristics of one or more embodiments may be combined in any suitable manner.

[0020] Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs.

[0021] The following describes in detail a data integration and collection method and system for a device management platform provided by the present invention with reference to the accompanying drawings.

[0022] See also Figure 1 , which shows a flowchart of a data integration and collection method for a device management platform provided by an embodiment of the present invention, the method comprising the following steps: Step S001: Obtain a time series sequence of each type of operation data with the current fault report time as the end time and a duration of a preset period; obtain the operation mutation time of each type of operation data according to the data changes in the time series sequence of each type of operation data.

[0023] In this embodiment, the platform primarily performs time-series alignment on the discrete maintenance data and continuous operation data of devices, enabling data integration of multi-source data collected from these devices. Because power equipment is critical infrastructure in today's rapidly developing industrial environment, the device management platform for power equipment integrates multiple cutting-edge technologies. IoT technology enables real-time data collection, big data technology deeply explores the potential value of device operation data, and artificial intelligence technology enables demand forecasting and resource optimization. When processing multi-source data, the platform first performs time-series alignment and integration on the discrete maintenance data and continuous operation data to improve the accuracy of subsequent data training. Traditionally, continuous time-domain data with a fixed operating time period preceding the discrete maintenance data is used as alignment data, for example, matching a single device fault record to the operating data readings collected by all relevant sensors within the 16 hours prior to its occurrence. Traditional time-series alignment methods that use a pre-set fixed operating data period have low integration accuracy. Therefore, this embodiment combines actual device management platform scenarios to obtain time-series alignment results between discrete maintenance record data and continuous multi-type operation data to improve data integration efficiency.

[0024] First, preliminary preparations and data collection are required. Specifically, auxiliary analysis data is retrieved from various modules within the equipment management platform, including the multi-sensor storage module, the work order data storage module, and the training parameter storage module. The work order data storage module retrieves real-time maintenance record data awaiting time-series alignment, including fault reporting time and fault detection type information. The multi-sensor storage module retrieves various types of operational data awaiting time-series alignment, including but not limited to vibration, pressure, temperature, and current data. The training parameter storage module retrieves a pre-trained model trained using historical time-series-aligned maintenance record data and operational data. This model is a Long Short-Term Memory (LSTM) prediction model, a prediction model based on a long short-term memory network (LSTM) network that can process and learn complex patterns and long-term dependencies in time series data. By integrating data from different sources and systems, the LSTM model can obtain more comprehensive historical information, thereby improving the accuracy of future event predictions. This is a well-known technique. The model input consists of a historical single equipment fault record matched to operational data readings collected by all relevant sensors within the 16 hours prior to the occurrence. The read data is uploaded to the data acquisition system for subsequent analysis and processing.

[0025] This allows real-time acquisition of each type of operating data (such as vibration, pressure, temperature, current, etc.), as well as the time and type of each fault report, and pre-trained models during equipment operation.

[0026] The frequency of collecting each type of operating data is once per second, and the collection period is the most recent month. Each type of operating data is normalized using the maximum and minimum norm method to unify the dimensions. This is a well-known technique, and the specific method will not be described here.

[0027] It should be noted that the purpose of this embodiment is to perform time-series alignment and integration of discrete equipment maintenance record data and continuous operation data, that is, for a single discrete fault repair report timestamp to be integrated, a continuous time period is selected within the total optional time range of its corresponding operation data to perform time-series alignment with the discrete repair report data. Specifically, the abnormal performance of the operation data at each moment within the total optional time range is first analyzed, and the correlation between multiple types of operation data is analyzed on the basis of the abnormal performance to obtain high-probability abnormal points within the total optional time range. Then, the historical data under the same fault detection type and the training influence performance analysis of the historical data are combined to obtain the time-series expansion requirement, and the time-series expansion requirement is used to realize the time-series alignment and integration process of discrete and continuous data.

[0028] It should be further explained that: it is necessary to first determine the operation mutation analysis point in the operation data corresponding to the equipment maintenance record. Considering that the operation period aligned with the discrete maintenance data time series should include the complete abnormal performance of the equipment as much as possible, the durability performance and local fluctuation of the operation data at a single moment are analyzed within the total optional time range of the time axis to obtain the anomaly coefficient, and then the operation abnormal change curve is constructed, and the operation mutation analysis point corresponding to the single type of operation data in the time domain is determined according to the steepness of the change curve. The operation mutation analysis point may be the time point when the equipment has an abnormal condition. Since the traditional selection of operation data with a preset fixed length of 16 hours is used as the time series alignment data of the fault record data, and directly selecting 16 hours of data for alignment may have poor adaptability, in this embodiment, twice the length of the preset fixed value is used as the total optional time range on the time axis, and the time length of the operation data for time series alignment is determined within the total optional time range through the analysis in this embodiment to improve adaptability. Schematic diagram of the selection of analysis time length for fault maintenance, as shown in Figure 2 As shown, Figure 2 From left to right in the middle is the time axis direction. Starting from the discrete maintenance record point, in reverse chronological order, the traditional preset selection time (16 hours) and the total optional time range (32 hours) are obtained respectively.

[0029] Preferably, in one embodiment of the present invention, the method for obtaining the operation mutation moment of each type of operation data includes: The preset period (total optional duration range) is 32 hours, which is used as an example for description.

[0030] Get the time series of each type of operation data with the current fault report time as the end time and the duration as the preset period.

[0031] In each type of running data time series, obtain the mean of the data values ​​at all times, record it as the first mean, and obtain the absolute value of the difference between the data value at each moment and the first mean The inverse proportional normalized value is recorded as the equipment durability performance at each moment.

[0032] It should be noted that: in this embodiment, As The inverse normalized value of This is a linear normalization function used to normalize data values ​​to a value between 0 and 1. For each type of operating data, the smaller the difference between the data value at one moment and the data values ​​at other moments, the better the device's durability. For example, for device temperature parameters, when there are no abnormal conditions, the device temperature should differ little from the historical temperature.

[0033] The preset quantity threshold N is 6, which is used as an example for description.

[0034] In each type of operating data time series, obtain the N moments closest to the x-th moment in time and the local period formed by the x-th moment, and record the standard deviation of the data values ​​​​of all moments in the local period as the volatility of the x-th moment.

[0035] It should be noted that when an abnormal condition occurs in the equipment, its operating parameters will show unstable fluctuations, such as the current fluctuation of electrical equipment before current overload. Therefore, for each moment, if the standard deviation of the current type of operating data in the local time period is larger, the fluctuation is more severe. At the same time, the lower the equipment durability performance at each moment, the greater the abnormal performance of the equipment.

[0036] In each type of operating data time series, the product of the inverse proportional value of the equipment durability performance at each moment and the volatility at each moment is obtained, and recorded as the operating anomaly coefficient at each moment.

[0037] The inverse proportional value of the device durability performance at each moment is obtained by subtracting the difference between 1 and the device durability performance at each moment.

[0038] In each type of operation data time series, the time is used as the horizontal axis and the operation anomaly coefficient is used as the vertical axis. The least squares method is used for curve fitting to obtain the operation anomaly change curve.

[0039] Among them, the least square method is a well-known technology, and the specific method will not be introduced here.

[0040] On the abnormal change curve, get the slope of the data point corresponding to the xth moment The normalized value of is recorded as the trend value, the mean of the data values ​​of all moments before the x-th moment (including the x-th moment) is obtained, recorded as the second mean, the mean of the data values ​​of all moments after the x-th moment (excluding the x-th moment) is obtained, recorded as the third mean, and the product of the absolute value of the difference between the second mean and the third mean and the trend value is obtained, recorded as the possibility of abnormal mutation at the x-th moment.

[0041] What needs to be explained is: As The higher the slope of the data point on the abnormal operation change curve corresponding to the current type of operation data at each moment, the greater the possibility that the equipment abnormality occurred at that moment. At the same time, the greater the difference in the abnormal operation coefficient before and after the moment, the higher the possibility that the moment was an abnormal condition.

[0042] On the abnormal operation change curve, the moment with the greatest possibility of abnormal mutation is recorded as the operation mutation moment of each type of operation data, and the data point at the operation mutation moment is the operation mutation analysis point.

[0043] It should be noted that if there are multiple moments with the highest probability of abnormal sudden change on the abnormal operation change curve, the first one will be selected for subsequent analysis in chronological order. Figure 3 As shown, Figure 3 The horizontal axis of the abnormal operation change curve is time (t), the vertical axis is the abnormal operation coefficient, and the dots marked on the abnormal operation change curve are the mutation analysis points.

[0044] Step S002: Obtain a reference duration based on the time intervals between the operational mutation moments of all types of operational data; obtain a historical time series sequence of each type of operational data with the time of each historical fault report of the same fault type as the current fault report as the end moment and the duration being the reference duration.

[0045] It should be noted that: further, it is necessary to collaboratively analyze the correlation of multi-source data and its impact on data training to obtain the degree of time series expansion requirements. Considering that when an abnormality occurs in the equipment, various types of operating data should be correlated, that is, the differences in the operation mutation analysis points corresponding to each type of operating data in the time domain should be small. Therefore, based on this rule, high-probability abnormality points that can most accurately express the moment when the abnormality occurred are selected. Furthermore, for the same fault detection type, the changes reflected by the operating data may be similar. The less the operating data in the current fault record meets this similarity feature, the higher the analysis value of the operating data. In this case, the selection range of time domain operating data can be appropriately expanded.

[0046] It is also important to note that selecting an excessively long period for operational data will affect the accuracy and efficiency of subsequent data training. Therefore, the time-scale expansion results of the time series alignment selection used in the above analysis were adjusted based on historical data training performance. To further determine the moment of abnormality, a combined analysis of the correlations between multiple types of operational data was performed. When an equipment abnormality occurs, it should have a correlated impact on all types of operational data. Therefore, the operational mutation analysis points for each type of operational data should have minimal differences on the timeline.

[0047] Preferably, in one embodiment of the present invention, the method for obtaining the reference duration includes: For the running mutation moments of all types of running data, the time interval between any two running mutation moments is used as the clustering distance, and the DBSCAN density clustering algorithm is used for clustering operation to obtain several clusters and the cluster center moment of each cluster.

[0048] DBSCAN (Density-Based Spatial Clustering of Applications with Noise) is a density-based clustering algorithm. This is a well-known technique, and the specific method is not described here. The cluster center time of each cluster is the mean of all the running mutation times in each cluster.

[0049] Count the number of running mutation moments in each cluster, obtain the cluster with the largest number of running mutation moments, and record it as the target cluster.

[0050] Among them, if there are multiple clusters with the largest number of running mutation moments, any one of them will be selected for subsequent analysis.

[0051] In the target cluster, get the time interval between each running mutation moment and the cluster center moment The inverse proportional normalized value of is recorded as the first interval value, and the product of the first interval value and the number of running mutation moments in the target cluster is recorded as the abnormal trend factor of each running mutation moment in the target cluster.

[0052] What needs to be explained is: As The inversely proportional normalized value of , the more the number of running mutation moments in the target cluster, the more it conforms to the clustering law of the running mutation moments of various types of running data, and the greater the possibility of it being the moment of abnormal occurrence. At the same time, the closer the running mutation moment is to the cluster center, the greater the probability that the running mutation moment is the moment of abnormal occurrence.

[0053] In the target cluster, the operation mutation moment with the largest abnormal trend factor is recorded as the high-probability abnormal moment. That is, the operation mutation analysis point corresponding to the high-probability abnormal moment is the high-probability abnormal point.

[0054] Get the time interval between the high-probability abnormality moment and the current moment, and record it as the reference duration.

[0055] Obtain a historical time series sequence of each type of operation data with the time of each historical fault report of the same fault type as the current fault report as the end time and the duration as the reference duration.

[0056] Step S003: Determine the final expansion requirement of the operation data time series corresponding to the current fault report according to the difference between each type of operation data time series sequence and each type of operation data historical time series sequence.

[0057] Preferably, in one embodiment of the present invention, a method for obtaining the final expansion requirement of the operation data time series corresponding to the current fault repair report includes: Taking the yth type of operating data as an example, in the historical time series of the yth type of operating data corresponding to all historical fault reports with the same fault type as the current fault report, the data mean at the same time is obtained in sequence to form the historical time series of the yth type of operating data mean.

[0058] In the yth type of operation data time series sequence corresponding to the current fault report, the yth type of operation data time series sequence segment between the high probability abnormality moment and the current moment is obtained.

[0059] It should be noted that the historical time series sequence of the mean value of the y-th type of operating data is equal to the time series sequence segment of the y-th type of operating data. The change curve diagram of the historical time series sequence of the mean value of the same type of operating data and the time series sequence segment of the operating data is as follows: Figure 4 shown. Figure 4 In the figure, the horizontal axis is time (t) and the vertical axis is the operating data value. The operating data change curve of the current maintenance record and the historical operating mean change curve are in different time periods, but the duration is the time interval from the high probability abnormal point (high probability abnormal moment) to the discrete maintenance time point (current moment). Figure 4 The two changing curves are aligned and displayed.

[0060] Get the Pearson correlation coefficient between the historical time series of the mean value of the y-th type of operating data and the time series segment of the y-th type of operating data The inversely proportional normalized value of is recorded as the first irrelevant value.

[0061] Among them, As The inverse proportional normalized value of , the acquisition of the Pearson correlation coefficient is a well-known technology, and the specific method is not introduced here. The smaller the Pearson correlation coefficient, the less similar the two data sequences are.

[0062] In the historical time series sequence of the mean value of the y-th type of operating data and the time series sequence segment of the y-th type of operating data, the absolute value of the difference between two elements with the same ordinal value is obtained, and the sum of the absolute values ​​of the difference between two elements with all the same ordinal values ​​is recorded as the second unrelated value.

[0063] The product of the first irrelevant value and the second irrelevant value is recorded as the current and historical irrelevant value of the y-th type of operating data.

[0064] The sum of the current and historical irrelevance of all types of operating data is recorded as the operating data time series analysis value corresponding to the current fault report.

[0065] What needs to be explained is that: the larger the first irrelevant value is, the greater the difference between the changing patterns of various types of operating data currently analyzed and the historical data under the same fault is. At the same time, the larger the second irrelevant value is, that is, the greater the difference in the operating data values, it further indicates that the operating data corresponding to the maintenance record data to be aligned in the current analysis does not conform to the general fault characteristics, and the more important it is to analyze its operating data, that is, the higher the scalability value of the operating data is.

[0066] Get the middle moment in the time series of each type of operating data corresponding to the current fault report, record it as the target moment, and get the difference between the target moment and the high probability abnormal moment The normalized value of is recorded as the expansion coefficient corresponding to the current fault report.

[0067] The product of the time series analysis value of the operating data corresponding to the current fault report and the expansion coefficient is recorded as the time series expansion demand of the operating data corresponding to the current fault report.

[0068] What needs to be explained is: As Since the total selectable duration is twice the traditional preset selection duration, when the high-probability abnormal moment is less than the target moment, the traditional preset selection duration needs to be extended. Conversely, when the high-probability abnormal moment is greater than the target moment, the traditional preset selection duration needs to be shortened.

[0069] The historical time series of all types of operating data corresponding to all historical fault reports of the same fault type as the current fault report are used as the input of the pre-trained model to obtain the data packet ratio and gradient norm during the model training and evaluation process.

[0070] The packet ratio and gradient norm can be calculated and analyzed during the training and evaluation of the LSTM prediction model. This is a well-known technique, and the specific method will not be introduced here.

[0071] What needs to be explained is that: when a fault alarm occurs in the equipment, it is necessary to select continuous operating data of a certain length before the fault alarm for fault analysis. It is known that the duration of each type of historical time series sequence of operating data corresponding to historical fault reports is a traditional preset selection duration. Therefore, in this embodiment, all types of historical time series sequences of operating data corresponding to all historical fault reports with the same fault type as the current fault report are taken to form a data set, which is input into a pre-trained model to obtain the data packet ratio and gradient norm during the model training and evaluation process. Among them, in time series prediction, the data packet ratio refers to the proportion of data volume in different time periods to the total data volume. The data packet ratio can reflect the diversity and representativeness of the data set. A reasonable data packet ratio helps to improve the generalization ability of the model. By ensuring that the model is exposed to sufficiently diverse data during the training process, the deviation of the model in practical applications can be reduced. The gradient norm refers to the length or size of the gradient vector during the training of a deep learning model, which can help adjust the learning rate or use gradient clipping technology to stabilize the training process. Therefore, both packet ratio and gradient norm play an important role in the training and evaluation of LSTM prediction models. The packet ratio is related to data quality and distribution, while the gradient norm is related to the stability and efficiency of model training. Therefore, by analyzing the packet ratio and gradient norm of all types of historical operational data time series corresponding to all historical fault reports of the same fault type as the current fault report during model training, we can determine whether the dataset obtained using the traditional preset selection duration for historical fault reports of the same fault type as the current fault report is suitable. If not, it is necessary to adjust the traditional preset selection duration, that is, adjust the required degree of expansion of the operational data time series corresponding to the current fault report, and obtain the final required degree of expansion of the operational data time series corresponding to the current fault report.

[0072] Get the ratio of packet rate to gradient norm The normalized value of is recorded as the first ratio, and the product of the first ratio and the time series expansion requirement of the operation data corresponding to the current fault repair is The normalized value of is recorded as the final expansion demand degree of the operation data time series corresponding to the current fault repair report.

[0073] What needs to be explained is: and As and The normalized value of . If the gradient norm of the operating data within the reference time period before the time of all historical fault reports of the same fault type as the current fault report is larger during the model training and evaluation process, it means that the training effect is worse. If the proportion of data packets during the model training and evaluation process is larger, it means that the specific type of data packets accounted for a larger proportion within the reference time period before the time of the historical fault report. That is, the confidence parameter of the data packet proportion is used as the inverse proportional value (i.e., the reciprocal) of the gradient norm. Therefore, the first ratio is used as the correction amplitude to adjust the operating data time series expansion requirement to obtain the final operating data time series expansion requirement.

[0074] Step S004: finally expand the demand degree according to the time series of the operation data corresponding to the current fault report and obtain the target duration; obtain the updated time series of all types of operation data with the time of the current fault report as the end moment and the duration as the target duration as the time domain matching data of the current fault report to complete the data integration.

[0075] It should be noted that: further, it is necessary to achieve time sequence alignment of equipment discrete maintenance data and continuous operation data according to the final expansion requirement of the operation data time sequence.

[0076] Preferably, in one embodiment of the present invention, the method for obtaining the target duration includes: The default basic period (traditional default selection duration) is 16 hours, which is used as an example for description.

[0077] The product of the preset basic cycle and the final expansion demand degree of the operation data time series corresponding to the current fault repair report is obtained, and recorded as the first product.

[0078] When the target time is greater than or equal to the high-probability abnormal situation time, the sum of the preset basic period and the first product is recorded as the target duration.

[0079] When the target time is less than the high-probability abnormal time, the difference between the preset basic period and the first product is recorded as the target duration.

[0080] Obtain the updated time series of all types of operating data with the current fault report as the end time and the target duration as the time domain matching data of the current fault report, thereby achieving the timestamp alignment of the equipment's discrete maintenance data and continuous operation data and completing data integration.

[0081] It should be noted that the specific implementation of the complete process for improving data integration and collection for the equipment management platform is as follows: First, for each fault maintenance record, the time domain alignment data (i.e., the time domain matching data for each fault report) is associated with the equipment operation data. Then, a data set is formed by all the maintenance data to be trained and the matched aligned operation data, where half of the data set is selected as the training set and the other half as the validation set. The loss function during training is the mean square error (MSE) or mean absolute error (MAE). The Adam optimizer is used, with a learning rate setting range of 0.001 to 0.01, and an impulse factor (Momentum) is used to avoid local optimality. The training set data is imported into the local database of the equipment management platform prediction model, and training is terminated when the error parameter no longer decreases to prevent overfitting. This is a well-known technology. Finally, a data integration and collection method with higher adaptability for equipment management platforms is implemented.

[0082] The present invention also provides a data integration and collection system for a device management platform, comprising a memory, a processor, and a computer program stored in the memory and runnable on the processor. The processor executes the computer program stored in the memory to implement the steps of the aforementioned data integration and collection method for a device management platform.

[0083] So far, the present invention is completed.

[0084] In summary, in an embodiment of the present invention, a time series sequence of each type of operating data with the time of the current fault report as the end moment and the duration as the preset period is obtained, and a reference duration is obtained based on the data changes in each type of operating data time series sequence, and a historical time series sequence of each type of operating data with the time of each historical fault report of the same fault type as the current fault report as the end moment and the duration as the reference duration is obtained, and based on the difference between each type of operating data time series sequence and each type of operating data historical time series sequence, the final expansion demand of the operating data time series corresponding to the current fault report is determined, thereby obtaining the target duration, and obtaining the updated time series sequence of all types of operating data with the time of the current fault report as the end moment and the duration as the target duration as the time domain matching data of the current fault report, thereby completing data integration. The present invention can combine the abnormal performance of equipment operating data in actual scenarios, the correlation of multi-source data, and the impact on data training to obtain a data integration effect with a higher degree of adaptability, thereby improving data integration efficiency.

[0085] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. A data integration and collection method for a device management platform, characterized in that: The method comprises the following steps: Obtain a time series sequence of each type of operating data with the current fault report time as the end time and a duration of a preset period; obtain the operation mutation moment of each type of operating data based on the data changes in the time series sequence of each type of operating data; Obtain the reference duration based on the time intervals between the operational mutation moments of all types of operational data; obtain the historical time series of each type of operational data, with the time of each historical fault report of the same type as the current fault report as the end time and the duration as the reference duration; Determine the final expansion requirement of the operation data time series corresponding to the current fault report based on the difference between each type of operation data time series and each type of historical operation data time series; According to the final expansion demand of the operating data time series corresponding to the current fault report, the target duration is obtained; the updated time series of all types of operating data with the time of the current fault report as the end moment and the duration as the target duration are obtained as the time domain matching data of the current fault report to complete the data integration.

2. The data integration and collection method for a device management platform according to claim 1, characterized in that: The specific steps of obtaining the operation mutation moment of each type of operation data are as follows: In each type of operating data time series, obtain the mean of the data values ​​at all moments, recorded as the first mean, and obtain the inversely proportional normalized value of the absolute value of the difference between the data value at each moment and the first mean, recorded as the device durability performance at each moment; A threshold value N is preset, and a local period consisting of the N moments closest to the x-th moment and the x-th moment is obtained. The standard deviation of the data values ​​of all moments in the local period is recorded as the volatility of the x-th moment. Obtain the product of the inverse proportional value of the equipment durability performance at each moment and the volatility at each moment, and record it as the operation abnormality coefficient at each moment; According to the operational anomaly coefficients at all moments, the operational mutation moment of each type of operational data is determined.

3. The data integration and collection method for a device management platform according to claim 2, characterized in that: The specific steps of determining the operation mutation moment of each type of operation data based on the operation anomaly coefficient at all moments are as follows: In each type of operation data time series, with the time as the horizontal axis and the operation anomaly coefficient as the vertical axis, the least squares method is used for curve fitting to obtain the operation anomaly change curve; On the abnormal change curve, obtain the normalized value of the slope of the data point corresponding to the xth moment, record it as the trend value, obtain the mean of the data values ​​of all moments before the xth moment, record it as the second mean, obtain the mean of the data values ​​of all moments after the xth moment, record it as the third mean, obtain the product of the absolute value of the difference between the second mean and the third mean and the trend value, record it as the possibility of abnormal mutation at the xth moment; The moment when the possibility of abnormal mutation is the greatest is recorded as the operation mutation moment of each type of operation data.

4. The data integration and collection method for a device management platform according to claim 1, characterized in that: The specific steps of obtaining the reference duration are as follows: For all types of running data, the time interval between any two running mutation moments is used as the clustering distance. The DBSCAN density clustering algorithm is used to perform clustering operations to obtain several clusters and the cluster center time of each cluster. Obtain the cluster with the largest number of running mutation moments and record it as the target cluster; In the target cluster, obtain the inversely proportional normalized value of the time interval between each running mutation moment and the cluster center moment, record it as the first interval value, and multiply the first interval value by the number of running mutation moments, record it as the abnormal trend factor of each running mutation moment in the target cluster; Determine the high-probability abnormal moment based on the magnitude of the abnormal trend factor at each operation mutation moment in the target cluster; Get the time interval between the high-probability abnormality moment and the current moment, and record it as the reference duration.

5. The data integration and collection method for a device management platform according to claim 4, characterized in that: The specific steps of determining the high-probability abnormal situation moment are as follows: In the target cluster, the operation mutation moment with the largest abnormal trend factor is recorded as the high-probability abnormal moment.

6. The data integration and collection method for a device management platform according to claim 4, characterized in that: The specific steps of determining the final expansion requirement of the operation data time series corresponding to the current fault repair report are as follows: In the historical time series of the yth type of operating data corresponding to all historical fault reports of the same fault type as the current fault report, the data means at the same time are sequentially obtained to form the historical time series of the yth type of operating data means; In the yth type of operation data time series sequence corresponding to the current fault report, obtain the yth type of operation data time series sequence segment between the high probability abnormality moment and the current moment; Determine the value of the operation data time series analysis corresponding to the current fault report based on the difference between the historical time series sequence of the mean value of each type of operation data and the time series sequence segment of each type of operation data; Obtain the middle moment in the time series of each type of operating data corresponding to the current fault report, record it as the target moment, obtain the normalized value of the difference between the target moment and the high-probability abnormal moment, and record it as the expansion coefficient corresponding to the current fault report; The product of the time series analysis value of the operating data corresponding to the current fault report and the expansion coefficient is recorded as the time series expansion demand of the operating data corresponding to the current fault report; According to the time series expansion requirement of the operating data corresponding to the current fault repair report, the final time series expansion requirement of the operating data corresponding to the current fault repair report is determined.

7. The data integration and collection method for a device management platform according to claim 6, characterized in that: The specific steps of determining the time series analysis value of the operating data corresponding to the current fault report are as follows: Obtain the inversely proportional normalized value of the Pearson correlation coefficient between the historical time series sequence of the mean value of the y-th type of operating data and the time series sequence segment of the y-th type of operating data, and record it as the first uncorrelated value; In the y-th type of operation data mean historical time series sequence and the y-th type of operation data time series sequence segment, the sum of the absolute values ​​of the differences between all two elements with the same ordinal value is recorded as the second unrelated value; The product of the first irrelevant value and the second irrelevant value is recorded as the irrelevant value between the current and historical data of the yth type of operation data; The sum of the current and historical irrelevance of all types of operating data is recorded as the operating data time series analysis value corresponding to the current fault report.

8. The data integration and collection method for a device management platform according to claim 6, characterized in that: The method of determining the final expansion requirement of the time series of the operating data corresponding to the current fault repair report according to the expansion requirement of the time series of the operating data corresponding to the current fault repair report includes the following specific steps: Use the historical time series of all types of operating data corresponding to all historical fault reports of the same fault type as the current fault report as the input of the pre-trained model to obtain the data packet ratio and gradient norm during the model training and evaluation process; Obtain the normalized value of the ratio of the data packet ratio to the gradient norm, record it as the first ratio, and record the normalized value of the product of the first ratio and the operating data timing expansion requirement corresponding to the current fault report as the final operating data timing expansion requirement corresponding to the current fault report.

9. The data integration and collection method for a device management platform according to claim 6, characterized in that: The specific steps of obtaining the target duration are as follows: Obtain the product of the preset basic cycle and the final expansion demand degree of the operation data time series corresponding to the current fault repair report, and record it as the first product; When the target time is greater than or equal to the high-probability abnormal time, the sum of the preset basic period and the first product is recorded as the target duration; When the target time is less than the high-probability abnormal time, the difference between the preset basic period and the first product is recorded as the target duration.

10. A data integration and collection system for a device management platform, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the computer program is executed by a processor, the steps of the data integration and collection method for a device management platform as described in any one of claims 1 to 9 are implemented.

Images